Using administrative data to explore the employment and benefit receipt outcomes of vocational rehabilitation applicants years after program exit

Abstract

BACKGROUND: Vocational rehabilitation (VR) helps people with disabilities achieve employment. VR administrative data only capture whether VR service recipients were employed at program exit, making it difficult to measure whether employment is sustained.

OBJECTIVE: This study used linked administrative data from the Social Security Administration (SSA) and Rehabilitation Services Administration (RSA) to explore the employment, Supplemental Security Income (SSI) payment receipt, and Social Security Disability Insurance (SSDI) benefit receipt outcomes of VR applicants during the first seven calendar years after program exit.

METHODS: The analysis sample included all VR case closures from 2004 through 2006. We linked the RSA-911 file to SSA’s Disability Analysis File and Master Earnings File to measure outcomes. Regression analysis controlled for observable characteristics.

RESULTS: Applicants exiting with employment were most likely to be employed or have SSI or SSDI benefits suspended over the subsequent seven years. Those who did not receive services had better outcomes than those who received services but exited without employment. Interestingly, SSDI non-beneficiaries who were working at program exit were more likely than others to eventually receive SSDI.

CONCLUSIONS: The correlation between employment status at closure and future outcomes provides an opportunity to target further assistance to VR customers as they leave the program.

Keywords

Vocational rehabilitation outcomes employment disability insurance Supplemental Security Income

1 Introduction

Despite the importance to policymakers of understanding vocational rehabilitation (VR) outcomes over time, the longer-term outcomes of program participants are not well understood. Because agencies do not need to track VR applicants after they stop receiving services, the VR administrative data do not contain measures of outcomes after program exit. For applicants who receive VR services, records maintained by the agency that provided services indicate which applicants were working for a 90-day period before they exited the program. Other outcomes—including certain types of benefit receipt—are also only measured by VR agencies at the time of program exit. Yet, recording outcomes at program exit only provides a limited perspective of how VR may have influenced outcomes. For applicants who did not receive services, VR does not collect any information about employment outcomes at exit. Although there are efforts to improve how VR tracks and reports program-relevant outcomes, such as the Workforce Innovation and Opportunity Act (WIOA) requirement state VR agencies eventually report standardized performance measures in the first and third quarters after program exit, linking VR administrative data to other data sources provides the best opportunity to better understand the outcomes of VR applicants after they leave the program.

This study links VR administrative data and Social Security administrative data to examine the longer-term employment and disability benefit receipt outcomes of VR applicants after they exit from VR. The Social Security data enrich our knowledge of VR applicants by providing administratively measured earnings, Supplemental Security Income (SSI) payment receipt, and Social Security Disability Insurance (SSDI) benefit receipt information long after VR program exit. For VR applicants who exited the program in calendar years 2004 through 2006, we examine three sets of outcomes up to seven calendar years after program exit: annual employment, SSI and SSDI benefit receipt for those who were not receiving benefits at program exit, and benefit suspension or termination due to work among those who received SSI or SSDI benefits at program exit.

For each set of outcomes, we use multivariate regression analysis to explore how they are correlated with whether the applicant received services and whether the applicant was employed at program exit. Although we do not estimate the causal impact of VR on these outcomes, the correlations estimated from our regression models reveal how these outcomes are associated with VR service receipt and employment status at exit after controlling for several policy relevant characteristics.

This study answers four research questions regarding VR applicant outcomes:

What are the longer-term employment, SSI payment receipt, and SSDI benefit receipt outcomes of VR applicants?

How do those outcomes vary by individual-level and state-level characteristics?

How is VR service receipt associated with longer-term outcomes?

Are the relationships between individual-level characteristics, state-level characteristics, and longer-term outcomes different once we account for VR service receipt status?

Our findings suggest that among those who received VR services, employment status at program exit is strongly correlated with longer-term employment and disability benefit receipt outcomes. Among VR service recipients who were SSI or SSDI beneficiaries, those who were employed at program exit were most likely to still be employed or experience suspension or termination of SSI or SSDI benefits in the longer term. Conversely, those who exited without employment after receiving services were less likely—even relative to VR applicants who did not receive VR services—to work or work at substantive levels. Interestingly, relative to non-beneficiaries who did not receive services or who received services and exited without employment, SSDI receipt after exiting from VR increased among non-beneficiaries who received services and exited with employment. Together, these findings suggest that opportunities may exist to provide targeted services and supports to VR applicants based on their status at program exit.

2 Background

2.1 The federal and state VR program

Nationally, VR received 578,488 applications from eligible applicants in federal fiscal year 2014 (Rehabilitation Services Administration [RSA], 2016a). However, not all eligible applicants receive services. Application to the program is voluntary.

The VR program is administered at the state level by 80 state vocational rehabilitation agencies (SVRAs) and overseen at the federal level by RSA in the Department of Education. Along with five agencies serving U.S. territories, 24 states have two SVRAs—one for people who are blind and another for those with any other condition type (“general”). The remaining 26 states and the District of Columbia have a single SVRA (“combined”) that serves all residents.

The federal government provides the majority of VR program funding, with state governments matching at a rate of roughly $1 for every $4 from the federal government. In fiscal year 2014, $3 billion in grant funds were administered across the 80 state SVRAs that covered 578,488 eligible individuals (RSA, 2016b).

2.2 The VR service path

Applicants can take several paths through the VR system (Fig. 1). Within 60 days of application, a SVRA counselor assesses whether an applicant is eligible for VR services. To be considered eligible for services, an applicant must have a health condition that impedes employment, be able to benefit from VR services, and require services to prepare for, secure, retain, or regain employment. SSI and SSDI beneficiaries are typically presumed to be eligible for VR services because they have met the Social Security Administration’s (SSA) definition of disability. If an applicant is found ineligible for VR services or does not provide sufficient information to complete an application, the case is administratively closed. Almost one in five applicants is found ineligible for services.

After an applicant is deemed eligible for services, a VR counselor collects information about the person needed to develop an individualized plan for employment (IPE). WIOA specifies that an IPE be developed within 90 days after eligibility, but past research has shown wide variation across agencies in the time between application and IPE (Schimmel Hyde et al., 2014). An IPE contains the applicant’s specific employment goal, the services that the VR agency will provide to achieve that goal, and the responsibilities of both the applicant and the agency (29 USC §722. Pub. L. 114-38. 129 Stat. 437).

Approximately one-quarter of VR applicants exit the VR program after being found eligible but before receiving services. The most common reasons for program exit at this stage include the applicant not being located, refusing services, or failing to cooperate with VR staff during this process. From the applicants’ perspective, those who exit at this stage state that their service expectations were not met, they did not adhere to counselor expectations of them (such as keeping appointments), or they experienced poor counselor relationships, health, and job opportunities, according to a study involving qualitative interviews of those who exited from VR early (Rigles et al., 2011).

Individuals with completed IPEs can begin receiving services. Those services may be provided by a variety of sources, as determined by service availability and the person’s needs. An applicant has two exit points after receiving services. A successful closure is an individual who is employed (for some number of hours at any wage level); after maintaining employment for 90 days, the applicant’s case is closed. Less than one in three applicants exits with employment. Employment could be in an integrated or non-integrated setting and with or without supports. About one in four applicants exits after receiving services, but without an employment outcome. These types of closures can occur because of refusal to accept services or inability to find appropriate placement, among others.

2.3 Existing literature on SSA and earnings outcomes for VR clients

The literature on VR outcomes largely relies on the information contained in the RSA-911 case service report (Institute for Community Inclusion, 2010). The RSA-911 file indicates which individuals, of those who received services, exited with employment—a statistic commonly referred to as the rehabilitation rate. For researchers, the RSA-911 file has the advantage of being readily available. However, analyses of these data also come with a strong disadvantage: for individuals who did not receive services, employment information is not available at the time of program exit, despite the fact that many will have been employed. As a result, researchers might choose to limit the sample to VR applicants who received services rather than all applicants regardless of service receipt.

Studies that have exclusively used RSA-911 data in a non-experimental approach have found some consistent patterns in service access, exiting with employment, and outcomes such as wages for select individual characteristics. VR clients who are white or male, who have relatively high educational attainment, or who have sensory, cognitive, or physical disabilities tend to have better employment outcomes upon VR exit than people who are Black or female, have relatively low educational attainment, have psychiatric disabilities, or receive Social Security disability benefits (Capella, 2002; Hayward & Schmidt-Davis, 2003; Mwachofi et al., 2009; Nazarov, 2013; U.S. Government Accountability Office [GAO], 2005). An expanding area of research demonstrates that agency- or state-level factors are associated with outcome patterns (e.g., Chan et al., 2014; U.S. GAO, 2005). For example, Nord et al. (2013) examined several state- and agency-level factors associated with the employment rates of VR applicants with cognitive disabilities, finding that agencies with higher proportions of applicants from the general population or in states with higher employment participation rates had better employment rates for the individuals with cognitive disabilities they served, whereas those with higher agency costs-per-employment outcome or higher participant-to-counselor ratios had lower employment rates.

Linking RSA-911 data to other individual-level data sources, especially those that contain earnings, can expand the knowledge of outcomes for VR applicants and have the advantage of containing outcome information for all VR applicants. State unemployment records, for example, contain quarterly earnings for VR applicants and have been used for understanding an agency’s potential return on investment (Northwest Economic Research Center, 2013; Wilhelm & Robinson, 2012) and longitudinal outcomes for a cohort of different types of applicants to specific VR agencies (Dean et al. 1999; Dean et al., 2014; Tremblay et al., 2006). These studies are usually small in scale, program or agency specific, tend to have limited follow-up periods, and typically exclude individuals who move out of state.

An emerging strand of research merges RSA-911 data to SSA and other federal data because of the advantages in tracking long-term outcomes, particularly regarding SSA disability benefits (Martin & Stapleton, 2016). GAO staff, for example, linked RSA and SSA earnings records to examine annual earnings of SSA disability beneficiaries involved with VR (U.S. GAO, 2007), finding that about 40% of beneficiaries experienced an increase in their annual earnings between the year before VR application and the year after VR exit, and that the proportion of applicants with earnings decreased over time. Other studies have focused on the experiences of SSA beneficiaries involved with VR (including using SSA beneficiaries not involved with VR as a comparison group, as in O’Neill, Mamun, Potamites, Chan, and da Silva Cordoso’s [2015] study) or the long-term SSA benefit receipt outcomes for all applicants, not just SSA disability beneficiaries, to better observe the interaction between SSA benefits and VR services, such as for SSDI benefit receipt (Stapleton & Martin, 2012) and for the experiences of transition-age youth (Honeycutt et al., 2016).

3 Data and methods

3.1 Data

To construct the variables needed for the analysis, we linked records in three administrative data sources—the RSA-911 file, SSA’s Disability Analysis File (DAF) version 12, and SSA’s Master Earnings File (MEF)—by Social Security Number. All three data sources are stored on SSA’s mainframe, but only SSA staff with the necessary security clearance are permitted to access the MEF. Michelle Stegman Bailey, a SSA employee with permission to access the MEF, conducted all parts of this analysis that involved manipulating MEF data. The analysis sample includes all VR cases that closed for any reason during calendar years 2004 through 2006, including those that were not eligible for services and those that closed before receiving an IPE. 1 Most individuals in the file experienced only one case closure during the period. The vast majority (94.5%) of VR clients in the analysis sample have just one case closure observation from 2004 to 2006. Of the remaining 5.5% of sample members, 5.2 percentage points have two case closures during the study period. The remaining 0.3 percentage points of people have three or more closures from 2004 to 2006. The finding that most VR clients in the sample have one case closure is important because it minimizes concerns that correlation at the VR customer level is significant and should be controlled for as part of the analysis.

The RSA-911 file contains case-level data for all closed VR cases in the United States and its territories. Only closed VR cases are described in the RSA-911 files used for the analysis. We relied on the RSA-911 to gather demographic, primary disabling condition, VR service receipt, and employment status at VR application information for the analysis sample. We used the DAF12 to obtain program information for all SSI and SSDI beneficiaries and the MEF for annual earnings information collected from W-2 forms for all people with a Social Security number. Using these data, we constructed measures of SSI payment receipt, SSDI benefit receipt, and earnings-based outcomes from the calendar year of program exit to six full calendar years after program exit.

Our sample contains about 1.78 million case closures, representing VR clients who varied by demographic characteristics, program characteristics, and SSA benefit receipt status (Table 1). Including concurrent beneficiaries, approximately 18% of analysis sample members were SSI recipients and 24% were SSDI beneficiaries. Relative to SSDI-only beneficiaries, non- beneficiaries, SSI-only recipients, and concurrent beneficiaries tended to be younger, probably because SSDI eligibility depends in part on a person (or in some cases a person’s parent or deceased spouse) having a sufficient work history. Those who received an SSI payment were more likely to be non-white than SSDI-only or non-beneficiaries. Although the high school completion rate was similar by SSA disability benefit receipt status (about 36%), non-beneficiaries and SSDI-only beneficiaries were more likely than SSI recipients or concurrent beneficiaries to have obtained some postsecondary education. Whereas the primary impairment distribution varied substantively across SSA disability benefit categories, the largest impairment group overall was mental health.

3.2 Outcome measures

For the analysis, we examined six outcomes across three areas:

Employment

During the first six full calendar years after VR program exit, the number of years an individual earned at least one quarter of coverage for Social Security benefits.

During the first six full calendar years after VR program exit, the number of years an individual earned more than the annualized substantial gainful activity (SGA) amount.

Benefit suspension and termination due to work

The number of months an SSI recipient (at VR exit) had his or her SSI payments suspended or terminated due to work during the first seven calendar years after program exit.

The number of months an SSDI beneficiary (at VR exit) had his or her SSDI benefits suspended or terminated due to work during the first seven calendar years after program exit.

Benefit receipt

Whether an individual not receiving SSI or SSDI benefits at program exit received any SSI payments during the first seven calendar years after program exit.

Whether an individual not receiving SSI or SSDI benefits at program exit received any SSDI benefits during the first seven calendar years after program exit.

Our data sources contained calendar year earnings information and monthly benefit receipt data. Consequently, our employment outcome measures capture the first six full calendar years after VR program exit (which excludes the year of program exit), whereas the benefit receipt outcome measures consider the first seven calendar years after program exit (including the months of the program exit year that were after program exit).

We defined employment as earning at least one quarter of coverage for Social Security benefits in a calendar year and SGA employment as earning greater than the annualized SGA amount (for non-blind beneficiaries) in a calendar year. Because the quarter of coverage threshold and SGA amount vary each calendar year according to the national average wage index, we determined whether an individual was employed or SGA employed in a given calendar year by comparing the applicant’s nominal earnings in that year to the year’s quarter of coverage threshold or SGA amount, respectively. The annualized SGA amount is a policy-relevant earnings threshold because SSI and SSDI program rules use the SGA amount to gauge whether someone is working at a substantive level. In contrast, the quarter of coverage earnings threshold captures whether an individual worked enough to earn a relatively modest amount—about $100 a month on average throughout the calendar year.

Turning to the benefit receipt measures, we examined benefit suspension and termination due to work because it indicates whether an SSI or SSDI beneficiary is earning enough to stop receiving benefits (even if just temporarily). The benefit receipt measures capture how many VR applicants who were not beneficiaries at VR exit eventually applied for and started receiving SSI or SSDI benefits. For the SSI and SSDI outcome regression analyses, we excluded those age 58 or older at closure because they would have transitioned to Old Age and Survivors Insurance before the end of the follow-up period.

3.3 Methods

To examine these six outcomes, we estimated two types of analytical models: logistic regression models and negative binomial regression models. The models estimated for this study do not capture causal relationships between VR service receipt and the outcomes of interest because they do not fully control for selection into VR services. Other unobserved factors such as ability or willingness to cooperate with an SVRA may strongly influence service receipt status as well as eventual benefit receipt and employment outcomes. Instead of measuring causal relationships, the regression model estimates provide information on the correlation between VR service receipt and outcomes after controlling for critical (observed) explanatory variables. The estimated correlations help us better understand how certain variables move together while keeping other influential variables constant, which allows us to generate informed hypotheses about the underlying causal mechanisms.

We used negative binomial regression models to investigate outcomes that can be measured as count variables (that is, random variables that take non-negative integer values). This includes both employment outcomes as well as the number of months an SSI or SSDI beneficiary (at program exit) had his or her SSI or SSDI benefits, respectively, suspended or terminated due to work. To help with the interpretability of the point estimates, we present the results as incidence rate ratios, which transform the parameter estimates by the exponential function to reveal by what factor the dependent variable is expected to increase if a certain independent variable increased by one unit, holding all other independent variables constant.

The negative binomial regression models had the following functional form: $\begin{matrix} Pr [{COUNT}_{i} = g_{i} | X_{i}, s_{e}, s_{w}] \\ = \frac{Γ (λ^{- 1} + g_{i})}{g_{i}! Γ (λ^{- 1})} {(\frac{λ^{- 1}}{y_{i} + λ^{- 1}})}^{λ^{- 1}} {(\frac{y_{i}}{y_{i} + λ^{- 1}})}^{g_{i}} \\ y_{i} = exp [ν + χ s_{e} + ψ s_{w} + X_{i} (Φ + K max {s_{e}, s_{w}})] \end{matrix}$

where COUNT is a count variable; Γ () is the gamma function; λ is an overdispersion parameter; X is a vector of demographic, disability status, and other explanatory variables; s_e is an indicator for whether someone has received VR services and closing with employment; and s_w is an indicator for whether someone received VR services and closed without employment.

We used logistic regressions to model the SSI and SSDI benefit receipt outcomes for non-beneficiaries at program exit. To help with the interpretability of the point estimates, we present the model estimation results as odds ratios, which transform the parameter estimates by the exponential function. Both logistic regression models had the following basic functional form: $\begin{matrix} Pr ({EVENT}_{i} = 1 | X_{i}, s_{e}, s_{w}) = \frac{1}{1 + e^{- f_{i}}} \\ f_{i} = α + γ s_{e} + δ s_{w} + X_{i} (M + B max {s_{e}, s_{w}}) + ɛ_{i} \end{matrix}$ (1)

where i is a VR case closure and EVENT is a Bernoulli random variable indicating whether the event occurred (a value of 1 indicates the event occurred).

For all regression models, VR service receipt status—regardless of whether service receipt ended with employment—was interacted with all other exogenous variables. We include these interactions because outcomes may vary in substantive ways for certain subgroups of those who receive services. By including these interaction terms, our model can capture this potential heterogeneity. To better capture how outcomes varied by service receipt status and case closure status, we include two stand-alone service receipt terms that reflect service receipt status and employment status at program exit. The exogenous variables included in vector X were indicators for gender, age category, race, educational attainment, primary impairment category, case closure year, and (for the employment analyses) SSI or SSDI benefit receipt status, as well as variables capturing state employment rate at closure, VR case length, and average SVRA funding level when the case was open.

Although this study provides useful information regarding how VR service receipt status was correlated with the employment and benefit receipt outcomes years after program exit, it does have a few limitations. First, the analysis does not estimate causal relationships between VR service receipt and the outcomes examined. VR applicants who receive services are likely different than those who do not, as are those who exit with and without employment after receiving services, across both characteristics we can and cannot observe in data. This selection issue is the central challenge to estimating the impact of VR services. Second, because the outcome period includes the recessionary period of 2007 through 2009, our results may understate employment and overstate benefit receipt outcomes relative to times of economic expansion (or less severe economic contraction).

4 Results

4.1 Employment

In this section, we examine two earnings-based employment outcomes—number of years earning at least one quarter of coverage for Social Security benefits in a calendar year (employment) and number of years earning more than the annualized SGA amount in a calendar year (SGA employment).

We estimated negative binomial models to examine how the number of years VR applicants spent with employment or SGA employment during the six full calendar years after VR program exit was correlated with various characteristics. The estimation results showed a strong link between VR closure status and employment when measured at the reference value for the explanatory variables. Relative to those who did not receive services, VR applicants who received services and closed without employment spent about one-quarter less years with employment. Also relative to those who did not receive services, VR applicants who received services and closed with employment spent about one-third more years with employment.

Employment outcomes for VR applicants who did not receive services varied across most of the characteristics we examined (Table 2). Women who did not receive services tended to be employed (that is, earn at least one quarter of coverage for Social Security benefits) for more years than men who did not receive services. Whites who did not receive services were employed more than those from all other racial groups who did not receive services except Asians. Among those who did not receive services, years of employment consistently decreased with age and increased with education level. Relative to VR applicants who did not receive services with medical/systemic impairments, the correlation between employment and impairment status varied, with those with learning disabilities having the most years with employment. Relative to non-beneficiaries who did not receive services, SSI, SSDI, and concurrent beneficiaries who did not receive services spent fewer years employed. With a range of 0.27 to 0.44, the employment incidence rate ratios for SSI, SSDI, and concurrent beneficiaries who did not receive services relative to non-beneficiaries who did not receive services were especially low. VR applicants who did not receive services in states with higher unemployment rates spent fewer years in employment.

Employment was also correlated in various ways with the interaction of VR service receipt with the other explanatory variables. Women who received services spent slightly fewer years in employment than men who received services. Relative to those ages 25–34 who received services, those ages 14–18 and 19–24 who received services typically spent more years with employment, whereas members of all other age groups who received services typically spent fewer years with employment. Those who did not complete high school and received services spent fewer years in employment than high school graduates who received services. Similar to the result for those who did not receive services, years with employment varied by impairment status for those who received services. SSI, SSDI, and concurrent beneficiaries who received services spent fewer years with employment than non-beneficiaries who received services.

Turning to the SGA employment model estimates, the primary VR service receipt estimates provide further evidence that, relative to those who did not receive services, years with SGA employment was quite different for those who did and did not close with employment (Table 2). Similar to the employment result, when measured at the reference value for the explanatory variables, VR applicants who received services but did not close with employment spent about one-third fewer years with SGA employment than did those who did not receive services. Those who received services and closed with employment spent about 48% more time with SGA employment than did those who did not receive services. Years of SGA employment were lower among women (relative to men) both for those who did receive services and those who did not. The other estimates for the employment and SGA employment models vary somewhat across race, age, and impairment status.

4.2 Benefit suspension or termination due to work

The negative binomial regression-adjusted results in Table 3 reveal how the number of months in benefit suspension or termination during the first seven calendar years after program exit was correlated with various factors, all else constant. When measured at the reference value for the explanatory variables, SSI and SSDI beneficiaries who received services and closed with employment had their benefits suspended or terminated more often than those who did not receive services. SSDI beneficiaries who applied, received services, but did not close with employment spent less time in benefit suspension or termination than those who applied but did not receive any services.

The amount of time SSI and SSDI beneficiaries who applied for but did not receive VR services eventually spent in benefit suspension or termination varied with certain characteristics. Asian, Black, and unknown race SSDI beneficiaries and Black SSI recipients who did not receive services spent more time in benefit suspension or termination than Whites who did not receive services. Relative to other working-age SSI and SSDI beneficiaries who did not receive services, beneficiaries ages 19–34 were most likely to have their benefits suspended or terminated due to work. Relative to high school graduates who did not receive services, SSI and SSDI beneficiaries with at least some postsecondary education who did not receive services spent more time in benefit suspension or termination than high school dropouts. Time in benefit suspension or termination varied by primary impairment status for those who did not receive services, with beneficiaries with developmental (SSDI only), mental, neurological, substance abuse, unknown (SSDI only), and trauma (SSI only) impairments having fewer months than those with medical/systemic impairments. SSI and SSDI beneficiaries with multiple VR closures from 2004 to 2006 who did not receive services spent more time in benefit suspension or termination relative to beneficiaries with a single VR closure during that period who did not receive services. SSI recipients in states with higher unemployment rates who did not receive services spent less time in benefit suspension or termination.

Benefit suspension and termination patterns for SSI and SSDI beneficiaries also varied among those who received services. For example, relative to high school graduates who received services, SSI and SSDI beneficiaries with at least some postsecondary education spent more months in benefit suspension or termination after receiving VR services. Relative to White SSDI beneficiaries who received services, SSDI beneficiaries from other racial categories who received services had more suspension or termination months. SSDI beneficiaries ages 19–24 were the only age group of those who received services to spend more time in benefit suspension or termination than SSDI beneficiaries ages 25–34 who received services.

4.3 Benefit receipt

Table 4 contains odds ratios from the logistic regression revealing how eventual SSI or SSDI benefit receipt among VR applicants during the first seven years after program exit was correlated with VR service receipt and other factors. Among those who were non-beneficiaries at program exit, VR service receipt was correlated with SSDI receipt within six full calendar years of program exit in some interesting ways when measured at the reference value for the explanatory variables. Most notably, non-beneficiaries who received services and closed with employment were 19% more likely (relative to those who did not receive services) to eventually apply for and receive SSDI benefits. This result is somewhat counterintuitive because employment at program exit is typically linked conceptually with benefit non-receipt after program exit, though among non-beneficiaries who receive services, employment for a limited period might allow some to acquire enough quarters of coverage to become administratively eligible for SSDI. In addition, non-beneficiaries who received services and closed without employment were about 7% more likely to eventually become SSDI beneficiaries than non-beneficiaries who did not receive services. However, for SSI, we observe the opposite, more intuitive pattern—non-beneficiaries who received services and closed with employment were relatively less likely to receive SSI benefits, whereas non-beneficiaries who received services and did not close with employment were relatively more likely to enroll in SSI.

We observed several patterns in the benefit receipt of non-beneficiaries who did not receive services. Relative to men who did not receive services, female VR applicants who did not receive services were somewhat less likely to eventually receive SSI or SSDI benefits. Relative to those in other racial groups who did not receive services, Whites who did not receive services were more likely to eventually receive SSDI benefits but just as or less likely to receive SSI payments. Benefit receipt was more likely as people who did not receive services aged, with those ages 45–54 being more likely to receive SSI or SSDI benefits relative to those ages 25–34. The correlation between eventual benefit receipt and educational attainment varied by program, with high school dropouts who did not receive services being less likely to receive SSDI benefits but more likely to receive SSI payments and those with at least some postsecondary education who did not receive services being less likely to receive SSI payments. The probability of eventual benefit receipt also varied by impairment category among those who did not receive services, with those having learning, trauma, substance abuse or unknown impairments being the least likely to eventually receive benefits. Non-beneficiaries with cases closed in 2006 who did not receive services were less likely to eventually receive benefits than those with cases closed in 2004 who did not receive services. Among those who did not receive services, non-beneficiaries at program exit living in states with higher unemployment rates were less likely to eventually receive SSDI benefits. Higher SVRA funding was positively correlated with benefit receipt but negatively correlated with VR case length for those who did not receive services.

The eventual benefit receipt outcomes of non-beneficiaries who received services were also correlated with several factors. Women who received services were less likely than men who received services to eventually receive SSI or SSDI benefits. Blacks, Asians, Native Americans, and Pacific Islanders who were non-beneficiaries and received services were less likely to eventually receive SSDI benefits (but not necessarily SSI payments) than Whites who received services. Interestingly, older non-beneficiaries who received services were relatively more likely to eventually receive SSI or SSDI benefits than non-beneficiaries ages 25–34 who received services. Being a high school dropout and receiving services was positively correlated with eventual SSI and SSDI benefit receipt relative to being a high school graduate who received services. In contrast, non-beneficiaries who received services and had at least some postsecondary education were relatively less likely to eventually receive either SSI or SSDI benefits. SSI and SSDI benefit receipt were more likely for non-beneficiaries who received services from SVRAs with higher funding levels.

5 Discussion

The analysis results reveal that outcomes varied substantively by employment status at program exit as well as by VR service receipt status. Perhaps unsurprisingly, applicants who received VR services and closed with employment spent more months in SSI or SSDI benefit suspension or termination and had more years with employment than those who did not receive services. However, applicants who received services but did not close with employment had relatively fewer months of benefit suspension or termination and poorer employment outcomes than those who did not receive VR services, as well as those who received services and closed with employment.

For those who received VR services, employment status at closure was correlated with how difficult it was for those VR applicants to work for years into the future, even after receiving personalized services and supports designed to help them overcome their barriers to employment. Applicants who received services and closed with employment had barriers that they were able to overcome, whereas the opposite was true for those who received services but closed without employment. In effect, VR service receipt and case closure status act together as an ex post sorting mechanism, identifying those who are more or less likely to work after receiving assistance.

VR applicants who did not receive services are a more heterogeneous group than those who received services. Our data are limited regarding the employment barriers of VR applicants who did not receive services, so we cannot separate them into groups based on the employment barrier they experienced. Some were found not eligible for VR services. Others were eligible but may have found employment and discontinued seeking VR services, whereas others might have had a health condition that worsened. For others, either the wait to receive such services was too long or the fit between the applicant’s goals and the agency’s services was not a good one. Those who did not receive VR services (as a group) likely have longer-term outcomes that are superior to those who closed without employment but inferior to those who closed with employment.

Differences in benefit receipt results across programs could be partially attributable to differences in SSI and SSDI program eligibility rules. Relative to applicants who did not receive services, the eventual SSI payment receipt of VR service recipients typically varied depending on employment status at program exit. However, for SSDI benefit receipt, we find that VR service receipt was positively correlated with being more likely to eventually receive benefits regardless of employment status at closure. Although the medical eligibility criteria for both programs is the same, SSI and SSDI eligibility criteria vary substantively, with SSI providing income support to people with limited income and assets and SSDI requiring a sufficient work history (or in special cases people qualify based on a parent or spouse’s work history). Consequently, for those who do not yet qualify for SSDI benefits, working actually helps establish future SSDI eligibility.

Findings from this study have several implications for VR applicants, policymakers, VR counselors, VR administrators, and other stakeholders. The correlation between employment status at closure and subsequent outcomes over seven years provides an opportunity to target further assistance to former VR clients. Because VR service recipients who close without employment are less likely to be employed in the future, spend fewer months in benefit suspension or termination, and are more likely to eventually receive SSI payments, this group may include strong candidates for other, non-employment based services and supports designed to help people with disabilities who are unable to be economically independent. Conversely, VR applicants who receive services and close with employment – who are relatively most likely to be employed in the future, spend time in benefit suspension or termination, and avoid receiving SSI payments in future years – may benefit from additional employment services and supports that help sustain their workforce participation.

Another takeaway from our results involves the potential value of service receipt in consideration of state- and agency-level variables. As expected, employment was lower when the state unemployment rate at exit was relatively higher, but receipt of VR services mitigated that relationship. These findings suggest a possible benefit of VR service receipt via increased resources or in response to poor economic environments, although these findings are not consistent across our outcomes.

Conflict of interest

The authors have no conflict of interest to report.

Footnotes

Because we analyze the entire VR population during the analysis period, we technically do not need to use statistical inference to understand the proximity of our parameter estimates to the population parameters—our estimates are the population parameters. Nevertheless, we follow convention, reporting standard errors and significance levels.

Acknowledgments

The authors appreciate the assistance of Xiao Barry, Svetlana Bronnikov, and Miaomiao Shen for programming support; Jody Schimmel Hyde, Purvi Sevak, Debra Brucker, and SSA reviewers for helpful comments on the analysis; and Connie Qian and Cara Stepanczuk for other support. Funding for this study was provided by the Research and Training Center on Disability Statistics and Demographics at the University of New Hampshire, which is funded by the National Institute for Disability, Independent Living, and Rehabilitation Research, in the Administration for Community Living, at the U.S. Department of Health and Human Services (DHHS) (Grant No: 90RT5022-02-00). The contents do not necessarily represent the policy of DHHS and you should not assume endorsement by the federal government (EDGAR, 75.620 (b)).

References

Capella

M. E.

(2002). Inequities in the VR system: Do they still exist?Rehabilitation Counseling Bulletin, 45(3), 143–153.

Chan

J. Y.

, Wang

C.-C.

, Ditchman

, Kim

J. H.

, Pete

, Chan

, & Dreis

(2014). State unemployment rates and vocational rehabilitation outcomes: A multilevel analysis. Rehabilitation Counseling Bulletin, 57(4), 209–218.

Dean

, Dolan

, & Schmidt

(1999). Evaluating the vocational rehabilitation program using longitudinal data: Evidence for a quasiexperimental research design. Evaluation Review, 23(2), 162–189.

Dean

, Pepper

J. V.

, Schmidt

R. M.

, & Stern

(2014). State vocational rehabilitation programs and federal disability insurance: An analysis of Virginia’s vocational rehabilitation program. IZA Journal of Labor Policy, 3(7). doi:10.1186/2193-9004-3-7

Hayward

B. J.

, & Schmidt-Davis

(2003). Longitudinal study of the vocational rehabilitation services program. Final report 1: How consumer characteristics affect access to, receipt of, and outcomes of VR services. (Report submitted to Rehabilitation Services Administration, U.S. Department of Education.) Research Triangle Park, NC: RTI International.

Honeycutt

, Thompkins

, Bardos

, & Stern

(2016). Youth with disabilities at the crossroads: The intersection of vocational rehabilitation and disability benefits for youth with disabilities. Advance online publication. Rehabilitation Counseling Bulletin. doi:10.1177/0034355215621897

Institute for Community Inclusion. (2010), A selected look at the literature base on vocational rehabilitation and implications for future research. Boston, MA: University of Massachusetts Boston.

Martin

, & Stapleton

(2016). What have we learned using merged administrative data from SSA and RSA?Manuscript submitted for publication.

Mwachofi

A. K.

, Broyles

, & Khaliq

(2009). Factors affecting vocational rehabilitation intervention outcomes: The case for minorities with disabilities. Journal of Disability Policy Studies, 20(3), 170–177.

10.

Nazarov

Z. E.

(2013). Factors associated with wage differentials between Caucasian and African-American vocational rehabilitation clientele. Journal of Vocational Rehabilitation, 39, 183–191.

11.

Nord

, Hewitt

, & Nye-Lengerman

(2013). Economic and programmatic indicators explain vocational rehabilitation employment rates of persons with cognitive disabilities. Journal of Vocational Rehabilitation, 38, 149–156.

12.

Northwest Economic Research Center. (2013). Vocational rehabilitation: Return on investment in Oregon. Retrieved from https://www.pdx.edu/nerc/sites/www.pdx.edu.nerc/files/Voc%20Rehab%20Final%20Report_0.pdf

13.

O’Neill

, Mamun

A. A.

, Potamites

, Chan

, & Cordoso

E.D.

(2015). Return to work of disability insurance beneficiaries who do and do not access state vocational rehabilitation agency services. Journal of Disability Policy Studies, 26(2), 111–123.

14.

Rehabilitation Services Administration. (2016a). FY 2014 state vocational rehabilitation performance. Retrieved from http://www2.ed.gov/programs/rsabvrs/resources/fy2014-state-vocrehab-performance.pdf

15.

Rehabilitation Services Administration. (2016b). Wall chart of state vocational rehabilitation performance. Retrieved from http://www2.ed.gov/programs/rsabvrs/resources/fy2014-state-voc-rehab-performance.pdf

16.

Rigles

, Ipsen

, Arnold

, & Seekins

(2011). Experiences of rural vocational rehabilitation clients who leave the system prematurely: A qualitative exploration. Rehabilitation Counseling Bulletin, 54(3), 164–174.

17.

Schimmel Hyde

, Honeycutt

, & Stapleton

(2014). The relationship between timely delivery of vocational rehabilitation services and subsequent federal disability benefit application and receipt. IZA Journal of Labor Policy, 3(15). doi:10.1186/2193-9004-3-15

18.

Stapleton

D. C.

, & Martin

F. H.

(2012). Vocational rehabilitation on the road to social security disability: Longitudinal statistics from matched administrative data. (Report submitted to the Michigan Retirement Research Center.) Washington, DC: Mathematica Policy Research.

19.

Tremblay

, Smith

, Xie

, & Drake

R. E.

(2006). Effect of benefits counseling services on employment outcomes for people with psychiatric disabilities. Psychiatric Services, 57(6), 816–821.

20.

U.S. Government Accountability Office. (2005). Vocational rehabilitation: Better measures and monitoring could improve the performance of the VR program(GAO-05-865). Washington, DC: Author.

21.

U.S. Government Accountability Office. (2007). Vocational rehabilitation: Improved information and practices may enhance state agency earnings outcomes for SSA beneficiaries(GAO- 07-521). Washington, DC: Author.

22.

Wilhelm

, & Robinson

J. L.

(2012). The economic impact of Utah’s vocational rehabilitation program. Journal of Disability Policy Studies, 24(3), 148–157.