Bounding the return on investment and projecting the costs of expanding PROMISE services and activities: Initial insights from PROMISE for policymakers

3.1Life expectancy

Although multiple sources were available for life expectancy, few studies explicitly took disability into account (Chetty et al., 2016; Kegler, Baldwin, Rudd, & Walsh, 2017; Sasson, 2016). Those that did do so discussed how many years of life before the expected start of a disabling health condition (Crimmins, Zhang, & Saito, 2016; Laditka & Laditka, 2016). Two sources included in this analysis were based on actuarial data from the SSA Office of the Chief Actuary, one for the general population (Bell and Miller, 2005) and one for individuals age 26 who had lived 10 or more years with a disability (Zayatz, 2005). A third source of life expectancy estimates for the analysis were based on results from the Wisconsin Longitudinal Study (WLS), which explicitly accounted for disability in life expectancy estimates (Ng, Shaw & Peeters, 2018). The WLS estimates are based on a sample who lived to age 60, so these estimates skewed toward a higher average. All three sources provided life expectancy estimates by sex. Glei and Horiuchi (2007) point to the gap in life expectancy that started in the early 1900’s, reconfirmed in their analysis that covered mortality rates from 1751 to 2004 across 29 countries. The life expectancy estimates from these three sources broken out by sex is shown in Figure 1.

Fig.1

Life expectancy for people in the general population and those with disabilities. General Population estimates from 2014 and With Disabilities 1996-2000 are based on actuarial estimates from the Social Security Administration. The Wisconsin Longitudinal Study data was collected from 1957–2011.

As one would expect, the life expectancy for the general population, which includes people with and without disabilities, contains the highest values for life expectancy when compared descriptively to the other two sources (Bell & Miller, 2005). Approximately 12.7% of the general population (Erickson, Lee, & von Schrader, 2017) has a disability, so those individuals with disabilities are represented in this sample. The impact of disability on life expectancy is unclear.

The lowest life expectancy estimates were based on data collected from 1996 to 2000 (SSA, 2005), approximately 20 years old at the time of this analysis. These estimates were based on someone age 26 who had lived 10 or more years with a disability, a scenario that matches PROMISE youth life experience, as they were only eligible for recruitment before their 17th birthday. As a point of comparison to the 1996 to 2000 SSA estimates (and not shown in Figure 1), the general population life expectancy in 2000 was 79.5 years for females and 74.1 years for males (Centers for Disease Control, 2002). The drawbacks of using older life expectancy estimates may be little (GBD 2015 DALYs and HALE Collaborators, 2016), but the difference between these disability-based estimates and general population estimates is 19.7% fewer years for those with disabilities.

A third source, the WLS, returned estimates from data collected as late as 2011 that included disability status (Ng et al., 2018). Unfortunately, Ng and colleagues (2018) did not distinguish between disability from childhood and disability acquired as an adult, but they provided estimates in years of how life expectancy was impacted for those with a disability, which resulted in life expectancy for women at 79.4 years and for men at 76.6 years. The WLS male life expectancy is higher than the general population estimate because the WLS sample excluded individuals who died before age 60. Conditional on living to age 60, the male-female life expectancy gap decreases but still exists in the WLS. That said, estimates from all three sources aligned with results for males with intellectual disability in Westphalia-Lippe, Germany, although males with intellectual disability in Baden-Wuerttemberg lived on average 65.3 years. A similar pattern as shown in Figure 1 was observed – females experience lower life expectancy than the general population but higher than males (Dieckmann, Giovis & Offergeld, 2015).

Aside from the WLS (Ng et al., 2018) that examined people with disabilities as a single group, most other research has focused on specific disability statuses like intellectual disability (Dieckman et al., 2015), cerebral palsy (Brooks et al., 2014), and traumatic brain injury (Harrison-Felix, et al., 2012). In the WLS, other factors such as smoking and obesity were also included. For the estimate of the upper bound of lifetime SSI and Medicaid receipt, three values were calculated for each for the six life expectancy estimates shown in Figure 1.

3.2SSI and Medicaid costs

The ROI analysis requires estimates of the present discounted value of the stream of SSI benefits and Medicaid costs. The starting value for SSI was $8,520, based on 2013 SSI payment amounts (Social Security Administration, 2019a). The expected annual SSI increase was 3.0%, the median historical cost-of-living adjustment made to SSI by SSA (Social Security Administration, 2019b). The initial value for annual Medicaid cost, $16,859, was based on 2014 estimates for individuals with disabilities (Kaiser Family Foundation, 2014). Projections for growth in Medicaid spending, 3.6% annually, were based on numbers from the Medicaid and CHIP Payment and Access Commission (2016). All calculations started at age 18 and each year’s expenditures are based on the previous year’s value multiplied by one plus the increase and then discounted to present dollars using a 2.0% inflation rate (Federal Open Market Committee, 2019). This formula was applied across the years until the final life expectancy year, rounded to the whole year shown in Figure 1, and then summed across the life span.

3.4Upper bound of lifetime benefit receipt

ROI estimates were built starting at age 18 through the average life expectancy by sex. Three life expectancy values were used: general population based on 2014 life tables, 1996–2000 estimates, and WLS. The results from those calculations, listed in Table 2, provided the expected individual total lifetime receipt of SSI, total lifetime Medicaid expenditures, and the sum of those values. These totals could be as high as $2.3 million for males and $2.6 million for females if 2014 general population estimates are used, estimates that by definition include individuals with and without disabilities. If individuals with disabilities live on average as long as the general population and the annual increase remains approximately the same, these are the amounts that could be paid out over 59–64 years to individuals with disabilities who never enter the workforce. If the life expectancy is shortened by 19.7% to match the 1996–2000 estimates, the lifetime cost would be 1.5 million for men and $1.7 million for women. The 1996–2000 life expectancy estimates are the most relevant to this population as they account for disability and do not presume that the population lives until age 60.

These values provide the upper bound of program costs for each person under the most extreme scenario where a youth on SSI never enters the workforce. Given the total expenditure of $229 million to implement PROMISE to 13,444 youth, these estimates indicate that an equivalent amount would have been spent in SSI benefits and Medicaid costs for between 17 and 25 males or 15 and 22 females who never engage in paid work.

3.5Labor market attachment

Beyond working-aged SSI recipients, individuals with disabilities in the United States face poorer labor market outcomes than their peers without disabilities. Only 37.3% of working-aged individuals with disabilities are employed compared to 79.4% of those without disabilities. When they do work, they are employed less intensively, 23.9% versus 60.3% working full-time and full-year. Consequently, working-aged individuals with disabilities face higher rates of poverty, 26.1% compared to 10.4% (Erickson, Lee, & von Schrader, 2017). Hence, individuals with disabilities no longer receiving SSI benefits are reasonably modeled under the scenarios of part-time and full-time employment at minimum wage.

The fact that SSI eligibility has income restrictions implies that the observed labor market attachment of SSI recipients will be weaker than the general population. Among working-aged SSI recipients in 2017, only 4.8% engaged in paid work (SSI ASR, 2017). Looking at the employment patterns of adult SSI recipients in the five years prior to benefit receipt reveals that roughly a quarter of them were employed in each year (Daly, 1998). Thus, the two modeled scenarios, no and limited workforce participation, for individuals still receiving SSI benefits align with the evidence for this population.

3.6Tax revenue

Expected taxes paid were obtained from TAXSIM, tax simulation software that utilizes federal and state tax law information by state to generate estimates of expected tax liability (Feenberg & Coutts, 1993). Tax estimates were obtained for Arizona, Arkansas, California, Colorado, Maryland, Montana, New York, North Dakota, South Dakota, Utah, and Wisconsin, the states involved in PROMISE. Tax estimates assumed offsetting income growth and inflation and did not assume any increase in income or change in future tax rates, resulting in conservative estimates of an individual’s tax liability across a lifetime. Paid taxes did assume that the person started working and paying taxes at age 18. To remain conservative in the estimation of the lower bound, the lowest life expectancies were selected.

Tax estimates, unchanging across the lifetime, were obtained for three levels of employment. Prior evidence shows that youth on SSI who continue on SSI after the age 18 redetermination earn $1,500 annually while those determined ineligible at 18 earn $4,720 annually, both adjusted to 2018 dollars (Deshpande, 2016). The former group are represented under the limited employment label in Table 1 while the latter are represented by the part-time label. An individual employed at minimum wage would earn these amounts by working roughly 3 and 10 hours per week, respectively. The part-time group is modeled under the assumption that they did not maintain SSI eligibility following the age 18 redetermination, which aligns with the evidence in Deshpande (2016). The final group considered in this paper is those employed full-time, full-year in a minimum-wage position, and like the tax estimates, the income estimates did not change over time.

Due to variation in the minimum wage across the PROMISE states, the annual earnings for the full-time group range between $15,080 and $23,088, as seen in Table 3. States also vary in their income eligibility requirements for Medicaid. Income relative to the federal poverty level and number of dependents are key determinants of Medicaid eligibility. Table 3 shows how the projected earnings level and four different household scenarios correspond to Medicaid eligibility in each state.

For three levels of employment, TAXSIM accounted for a set of essential input variables. Estimates were prepared of each of the 11 PROMISE states, the three levels of employment described above, and estimated gross social security benefit. Only individuals with limited employment were assumed to remain in SSI. Marital status was allowed to vary. The number of dependents ranged between zero and one, which is representative of the SSI receiving population –70.8% have no children and 14.4% have one child (Stegman & Hemmeter, 2014).

3.7Lower bound of lifetime benefit receipt

For the individual with a disability who never enters the workforce, we calculated expected lifetime receipt of SSI and Medicaid expenditures as shown in Table 2. We now consider three alternate scenarios where individuals engage in different levels of paid work. For these employed individuals, money flows back into the system in forms of federal tax liability, state tax liability, and the U.S. federal payroll tax referred as FICA. For each state that was part of PROMISE, estimated tax liabilities and FICA payments were summed and then multiplied by the expected working life by sex, i.e., 44 years for men and 48 years for women. The resulting lifetime estimates of tax liability and FICA payments are reported in Table 4.

Across all states, the tax liabilities were non-positive values for the limited and part-time employment levels, but with FICA added to tax liabilities the lifetime totals were all positive except for people with dependents working under limited employment in New York. Unsurprisingly, the limited employment scenario, which earns only $1,500 annually, yields little back into the government in terms of direct revenue. Individuals modeled working full-time at minimum wage are expected to generate tax revenue between $82,775 and $269,376 across their lifetimes.

The estimate of the lower bound for the ROI offsets SSI and Medicaid costs against tax revenue. Table 5 reports this offset amount for each of the three employment levels and by sex and state. For those individuals receiving SSI with limited employment, the amount paid in taxes and FICA is minimal relative to the amount received from SSI and Medicaid. Therefore, the lower bound of benefits does not differ much from the upper bound of benefits, ranging between $1.5 million and $1.7 million. Individuals with part-time employment receive a net $1.0 million to $1.2 million in benefits, which represents a reduction in lifetime benefits by approximately a third. The full-time employment scenario varies considerably by state and household structure from net benefit receipt of $1.1 million to net revenue of $269,376. For example, a single female living in Arizona with no dependents would receive $1.7 million in benefits over the course of her life under limited employment but would net contribute $256,248 back in tax revenue under a full-time employment scenario. This alternative employment path represents a $2.0 million benefit to the government.

The PROMISE intervention aimed to improve the employment outcomes for youth on SSI. This analysis has examined the lifetime net benefit receipt for three different employment outcomes with increasing attachment to the labor market. Table 6 reports, by state, the number of youth in that state who would need to work over their lifetime to offset the $229 million spent nationally on PROMISE. These numbers are not based on the recruited PROMISE population but rather are a function of the cost of PROMISE relative to the reduction in benefits and increase in tax revenue resulting from increased earnings. Both part-time and full-time employment have the ability to substantially offset the cost of the project. On average, 477 youth, i.e., a number equivalent to 7.1% of the roughly 6,700 intervention youth, would need to move from never working to part-time employment to completely cover the cost of PROMISE. In other words, 477 of the intervention youth being employed part-time over their entire working lives would cumulatively offset the cost of PROMISE through reductions in benefit receipt and increases in tax revenue. As Table 6 shows, the benefit of work is so significant that in New York only 129 male youth working full-time over their lives (assuming they also are single with no dependents) could cover the full cost of the national PROMISE initiative.