Abstract
The Social Security Disability Insurance (SSDI) Employment Pilot in Wisconsin was one of four Social Security Administration authorized pilots to test a cash benefit offset feature for the SSDI program. Those allowed to use the offset only lost US$1 of their SSDI cash benefit for every US$2 earned when their monthly earnings reached the Substantial Gainful Activity (SGA) level after completing the Trial Work Period (TWP). Over the first two years following pilot enrollment, no statistically significant differences were observed in employment outcomes between the treatment and control groups. However, after these first two years, outcome trends diverged, ultimately leading to the treatment group exhibiting better outcomes. The differences between treatment and control participants were conditioned on whether participants completed their TWP by the end of 2008. Subsequently, there were statistically significant differences between outcome trends for the two groups of TWP completers. There were virtually no differences between the outcome trends for the groups with no TWP completers. These results are consistent with an interpretation that the cash benefit offset, given adequate time, can be an effective work incentive.
This article provides information about employment outcome trends for Social Security Disability Insurance (SSDI) beneficiaries participating in one of the Social Security Administration (SSA)–sponsored Benefit Offset Pilot Demonstration (BOPD) pilots for a 2-year period following that included in initial analyses. The crucial comparison is that between those in the treatment group actually eligible to use the benefit offset and those in the control group who would have been eligible had they been assigned to treatment. Outcomes for those eligible to use the offset feature are found to be considerably more positive than suggested by analyses of the earlier period.
The purpose of the SSDI Program is to provide income support for workers who, because of severe medically determinable impairments, are no longer able to perform substantial gainful work in the U.S. economy. Nonetheless, though establishing and maintaining SSDI eligibility requires demonstrating the incapacity to perform economically meaningful work, Congress, shortly after the program’s inception, started adding program features to encourage SSDI beneficiaries to return to work. The original goal was to encourage beneficiaries to leave benefit status. Although this goal remains foremost, since the 1990s, Congress and SSA have placed increasing emphasis on a second goal: encouraging those remaining indefinitely on SSDI to increase their employment outcomes.
This shift was particularly evident with the 1999 Ticket to Work and Work Incentives Improvement Act. One provision required that SSA test whether a benefit offset feature would encourage beneficiaries to increase their employment outcomes without entailing increases in program size and cost. Under SSDI program rules, having monthly earnings above the Substantial Gainful Activity (SGA) level, following completion of a Trial Work Period (TWP), results in the complete loss of the SSDI cash benefit. A benefit offset makes that reduction gradual. The required demonstration, the Benefit Offset National Demonstration (BOND), is currently in progress. Prior to BOND, SSA sponsored a smaller effort, the BOPD, to prepare for administering BOND and to get preliminary outcome information. The BOPD began in summer 2005 in four states. Wisconsin was one.
That project, the Wisconsin SSDI Employment Pilot (SSDI-EP), was implemented through a multi-organization collaborative housed at the state’s Department of Health Services. Participation was voluntary, but limited to adult beneficiaries who had cash benefits based on their own earnings histories and did not have concurrent Supplemental Security Income (SSI) benefits. Once individuals agreed to take part, they were randomly assigned to either the treatment or control group. Those assigned to the treatment group would have access to the benefit offset, but only following completion of their TWPs.
The TWP permits beneficiaries to test their ability to work at what SSA defines as economically meaningful levels for 9 (non-consecutive) months without losing any of their SSDI cash benefit. Expending a TWP month requires more than compensated employment. Earnings for the month must reach a considerable fraction (~70%) of SGA. In point of fact, TWP completion is more an indicator of relatively strong employment outcomes than a cause of them. There is, however, a substantial literature indicating that having relatively strong employment outcomes shortly after qualifying for a Social Security disability program is associated with positive outcomes later on (Burkhauser, Butler, Kim, & Weathers, 1999; Liu & Stapleton, 2011; Mitra & Brucker, 2004).
Following the TWP, earnings at or above SGA result in losing the entire cash benefit for that month and, potentially, continued SSDI eligibility. This is often referred to as the “cash cliff” and is effectively a 100% marginal tax rate on earnings over the SGA level (until earnings are equal to SGA plus the beneficiary’s full SSDI cash benefit). Under BOPD rules, those having access to the offset feature would lose only US$1 of their SSDI benefit for every US$2 of earnings above SGA. Although this represents a considerable loss relative to the situation during the TWP, offset users would still be much better off than TWP completers from the control group with equivalent earnings. Offset use was limited to the first seventy-two months after the TWP completion date, after which a beneficiary would again be subject to the cash cliff.
SSA required that each of the BOPD sites track participants’ employment outcomes over a 3-year period, including the year before project entry and 2 years thereafter. In Wisconsin, except for a higher rate of TWP completion in the treatment group, there were no significant differences in outcome trends related to random assignment (Delin, Hartman, Sell, & Brown-Reither, 2010). Both the treatment and control groups exhibited comparable increases in employment rates, mean earnings, and the proportion with quarterly earnings at least 3 times the monthly SGA level. Findings using pooled data from all four BOPD sites were also non-significant with one exception: Treatment group members were significantly more likely to increase their earnings beyond the SGA level (Weathers & Hemmeter, 2011). In addition, in at least two of the BOPD states, researchers identified subgroups which achieved greater than average gains in other employment outcomes (Chambless, Julnes, McCormick, & Reither, 2011; Tremblay, Smith, Porter, & Weathers, 2011).
Nevertheless, in assessing outcomes for the Wisconsin pilot, it is critical to remember that only a minority of either study group completed a TWP during the original analysis period. Less than 50% of the treatment group were ever eligible to utilize the offset provision and most of these achieved eligibility relatively late in the initial analysis period. Given this reality, we argue that a better strategy for assessing the offset would have been to compare the outcome trends of those who completed the TWP. In 2008, SSA made a decision that provided an opportunity to do this.
SSA decided to begin winding down the pilots 2 years before originally planned. Those in the treatment group who had completed their TWP by December 31, 2008 would be allowed to use the offset. Those in the treatment group who had not completed a TWP by this date were returned to regular program rules. Thus, as of January 1, 2009, the treatment group was permanently divided into two groups based on members’ prior behavior.
Thus, our focus, unlike prior analyses of BOPD outcomes, is on comparing outcomes between subgroups within the larger treatment and control groups based on TWP completion prior to 2009. Treatment group members who were returned to regular program rules are compared with those in the control group who had not completed a TWP. Treatment group members who completed their TWP are compared with TWP completers from the control group. We are especially interested in comparing the two completer groups as only they faced differing incentive structures going forward.
Comparing outcomes over a period subsequent to that examined in previous analyses speaks directly of the most frequently offered conjecture as to why study assignment had so little effect on most employment outcomes: Entering participants had not been given adequate time to complete their TWPs and, thus, if from the treatment group, to use the offset. Project directors and other stakeholders from all the BOPD sites observed that many beneficiaries needed to deal with factors that often substantially extended the time needed to complete a TWP. Some impediments were attitudinal, such as the fear of the negative consequences of work on keeping SSDI eligibility or valued non-cash benefits. Others reflected external circumstances, such as the time needed to complete education or training or awaiting improvements in personal health (Smith, Porter, Chambless, & Reiser, 2009).
Sample and Key Questions
We present data for SSDI-EP participants who had completed at least eight calendar quarters of participation following the calendar quarter of enrollment into the project. Although 468 individuals met this criterion, by December 31, 2008, some of these individuals had died or otherwise attrited from the project. In addition, we removed individuals who had already reached the 72nd month past their TWP completion date and thus, if in the treatment group, unable to further use the offset feature. This reduced sample size to 419.
For the purpose of analysis, we identified four groups within the sample:
Treatment group members who completed the TWP before the January 1, 2009 and would be eligible to use the offset for some period after that date (n = 114).
Control group members who completed the TWP before January 1, 2009 and would have been eligible to use the offset had they been assigned to the treatment group (n = 81).
Treatment group members who did not complete a TWP before January 1, 2009 and were returned to regular SSDI program rules on that date (n = 118).
Control group members who did not complete a TWP before January 1, 2009 (n = 106).
While it is possible to compare outcome trends between any combination of the four groups, our primary focus is on comparisons that might answer the following questions.
Does continued access to the offset result in better outcome trends after December 2008, at least for the period for which data are available?
Does explicit denial of continued access to the offset result in worse outcome trends after December 2008, at least for the period for which data are available?
A final question of interest arises from the fact that, unlike the original assignment to the treatment and control groups, inclusion into one of these four groups also results from a specific action that each participant performed or failed to perform—completing a TWP. Thus, unlike the original study groups, inclusion in one of the four analytical groups is not random. Therefore, it is important to examine the extent to which any group’s characteristics differ from those of any group it is being compared with. Although this analysis will concentrate on characteristics at or prior to SSDI-EP entry, we will also look at some differences between what each group experienced following entry.
Data and Method
We use a deidentified data set prepared from SSDI-EP evaluation records. Employment outcome measures utilize data from Unemployment Insurance (UI) records from the Wisconsin Department of Workforce Development. Data about demographic and experiential characteristics are drawn from pilot records and administrative data provided by SSA.
Our main outcome measures are mean quarterly earnings, employment rates, and the percentage of cases with earnings at or above the SGA level. All monetary amounts are deflated into August 2005 dollars using the Consumer Price Index for Urban Consumers (CPI-U). The August 2005 baseline was chosen because it was the first month of participant enrollment.
Employment rates and attainment of SGA earnings are calculated from the quarterly earnings data. Having any reported earnings signifies employment during that calendar quarter. Unfortunately, the data do not permit identification of SGA earnings on a monthly basis. Our outcome measure imputes this based on having earnings of at least 3 times SGA in the relevant quarter. As the 2005 (non-blind) SGA level was US$830, the quarterly equivalent is set at US$2,490 in August 2005 constant dollars. In addition, not all employment or earnings are captured in the UI data. In particular, most earnings from self-employment are unreported.
We utilize employment outcome data over a 12 quarter period starting with the first calendar quarter of 2008. The last time period available is the fourth quarter of 2010. The analyses are conducted in calendar time relative to the fourth calendar quarter of 2008 as our emphasis is on describing employment outcome trends subsequent to that period. Data from the first three quarters of 2008 are included mainly for context and to assess whether the fourth quarter 2008 outcome value represents an outlier. The fourth quarter of 2008 is always designated as Q0 in the figures included in this article.
Analyses are mainly descriptive. We do not attempt to statistically control for any factor that might influence results. However, we have performed simple statistical tests on means (t tests, unequal variances) and distributions (chi-squares), assessing significance at the conventional .05 p value. Finally, the results cannot be generalized beyond the Wisconsin pilot. SSA has data that would allow it to perform a similar analysis for the BOPD as a whole, though it would have to utilize annual instead of quarterly earnings data.
Outcome Trends Prior to December 2008
Although no statistically significant differences were found between the trends in employment rates, mean earnings, or exceeding SGA between the SSDI-EP treatment and control groups during the initial 2 years following participants’ enrollment quarters, over the final quarters of this period there was increasing divergence between the two groups’ outcomes on all three measures. Differences were always to the treatment group’s benefit (Delin et al., 2010). This suggests to us that the original BOPD evaluation period was too short to capture the offset’s likely impact on employment outcomes.
In any case, we think prior BOPD evaluations did not include a critical comparison. From a policy perspective, the relevant intervention that should have been assessed was whether having access, as opposed to potential access, to the benefit offset feature would result in better outcomes. The natural experiment initiated at the end of 2008 allows us to disaggregate outcomes based on (a) whether a treatment group participant had access to use the offset feature and (b) whether, if Congress had amended the Social Security Act to include the tested offset, a control group participant would have met the condition (TWP completion) for having access. To perform this analysis, we now compare employment outcome trends across the four analytical groups described in the “Sample and Key Questions” section.
Results
Mean Earnings
The first set of comparisons focus on quarterly UI earnings and are displayed in Figure 1. The most obvious difference is between the trends for the pair of groups defined by TWP completion by the end of 2008 and the pair that represents non-completers. Quarterly earnings for the two TWP completer groups are always much higher than for the non-completer groups. This result is expected given the level of earnings required to expend a TWP month.
Mean quarterly earnings by study group assignment and Trial Work Period (TWP) completion for the period of the first quarter of 2008 through the fourth of 2010.
Consequently, we turn to the more meaningful comparison, that between the two TWP completer groups. Figure 1 displays the increasing divergence in mean earnings between TWP completers from the treatment group and those from the control group. In the first two quarters of 2008, values were nearly identical. By Q0, the two groups began to diverge with the TWP completers from the treatment group always having higher quarterly earnings. As some TWP completers from treatment lose the ability to use the offset because more than 72 months has elapsed from their completion month, Figure 1 includes a secondary line describing mean earnings for only the TWP completers in the treatment group eligible to use the offset through the end of 2010. The outcomes are marginally higher, but the basic relationship to the trend line for peers from the control group remains essentially the same.
The trends in mean quarterly earnings from Q0 through Q5 are generally, if inconsistently, downward for both completer groups. Nonetheless, the decline in both percentage and absolute terms for TWP completers from the control group (35.4% and US$800) is greater than those for their peers from the treatment group (25.4% and US$630).
After Q5, earnings trends diverge to the clear advantage of TWP completers from the treatment group. Their earnings exhibit a rising trend, while those for control group completers appear more static. Over the Q0 through Q8 period, mean earnings for those in the treatment group who retained eligibility to use the offset declined, though only modestly (4.5% or US$111). By contrast, the decline for TWP completers from the control group over this period was considerably greater (28.0% or US$632). Moreover, the difference in mean earnings between TWP completers in the treatment group and TWP completers in the control group increased from US$219 in the fourth quarter of 2008 (Q0) to US$740 in the final quarter of 2010 (Q8). This finding is important, in that it suggests that those with access to the offset feature were better able to maintain their earnings gains over time, despite whatever downward pressures (e.g., a severe recession) were being commonly experienced by both these groups of beneficiaries. Nonetheless, there were no statistically significant differences between the mean earnings of these two groups in any of the calendar quarters included in the data series. Still, this “null” finding hardly closes the case that the SSDI-EP participants with continued access to the offset after December 2008 did not have substantially better outcomes than their control group peers.
SGA Earnings
The primary purpose of the offset is to remove the strong disincentive to earning above the SGA level following TWP completion. Figure 2 displays the proportions of the two TWP completer groups that had quarterly earnings that reached or exceeded 3 times the monthly SGA level. As the proportions of those with SGA equivalent earnings in the non-completer groups are small (never exceeding 5% in any quarter), quite stable across time, and exhibit no significant differences between those in treatment and control, we have chosen to show data only for the TWP completer groups.
Mean quarterly earnings equal to or greater than 3 times the Substantial Gainful Activity level for groups of Trial Work Period (TWP) completers for the period of the first quarter of 2008 through the fourth of 2010.
In every quarter, save Q3, the percentage with at least SGA equivalent earnings in the TWP completer component of the treatment group was higher than the percentage for the completers from the control group. By Q0 (the last quarter with any non-completers in either group) the TWP completers in treatment had 38.6% of its members earning at “SGA” compared with 25.9% of their peers in the control group. This is a nearly 13 percentage point difference.
This difference would grow over the succeeding 8 calendar quarters reaching over 18 percentage points. Over this period, the trends for the proportions of those with SGA equivalent earnings generally mirror those observed in the mean earnings data. The proportion of TWP completers from treatment declines from Q0 through Q5, reaching a minimum of 28.9%, that is, a reduction of almost 10 percentage points. However, this decline is less steep than that experienced by completers from the control group. Following Q5, the proportion with SGA equivalent earnings rises. By Q8, the proportion of those with SGA equivalent earnings has returned to 36.8%, less than 2 percentage points (4.7%) below the Q0 reference value.
By contrast, the proportion of TWP completers in the control group that posted SGA equivalent earnings declined steadily through most of the period following the final quarter of 2008. By Q5, the proportion with SGA equivalent earnings had decreased from 25.9% to 16.0%. Although this decrease is just marginally greater in percentage point terms than that for the treatment group completers, it is substantially more in percentage terms (38.2% versus 25.1% for treatment group completers). By the end of 2010, the proportion of TWP completers in the control group had rebounded to 18.5%. This still represents a 28.6% decline relative to Q0.
Moreover, the within-quarter differences between the two TWP completer groups are statistically significant in almost all of the calendar quarters in the data series. This finding lends support to the expectation that the offset would incent earnings over SGA for those who, upon entering their Extended Period of Eligibility (i.e., the 36 months following TWP completion) would have otherwise had a strong incentive to reduce earnings below SGA.
Employment Rates
The aggregated earnings measures we use are quite sensitive to changes in employment rates. Figure 3 displays quarterly UI employment rates over the 12 quarter analysis period. Compared with the SSDI population as a whole, the employment rates for all four of the groups are relatively high. Even the 2 non-completer groups typically had UI employment rates of 20% or higher. Indeed, what is most notable about the non-completer groups is the stability and similarity of their UI employment rates over the full period. In fact, the employment rates for these two groups are almost identical after the TWP non-completers from treatment are returned to regular SSDI rules after Q0.
Quarterly employment rates by study group assignment and Trial Work Period (TWP) completion for the period of the first quarter of 2008 through the fourth quarter of 2010.
Although UI employment rates for both of the TWP completers groups are always far higher than that for the non-completer groups, the trends are substantially different. Both TWP completer groups display generally downward trends over the entire data period. Using the fourth quarter of 2008 (Q0) as a reference point, the employment rate for TWP completers from the SSDI-EP treatment group falls from 77.2% to 64.9% by the fourth quarter of 2010. This is a decrease of over 12 percentage points (15.9%). TWP completers from the control group start with a lower Q0 employment rate (64.2%) and then experience a somewhat greater decrease in employment rates over the next 8 quarters. By the fourth quarter of 2010, their employment rate is 49.4%. This is a reduction of nearly 15 percentage points (23.1%) relative to Q0. The difference in employment rates for TWP completers from the treatment group and the rates for their peers from the control group increases from roughly 9 percentage points at Q0 to about 15½ percentage points at Q8.
This result is reasonably consistent with the trends seen for the other two outcome variables. Are the differences between the two TWP completer groups statistically significant? The answer is yes for most calendar quarters, particularly those after the Q0 reference (p < .05). Overall, the data are supportive of a conclusion that continued access to the offset feature tends to sustain continued work effort among those who have completed their TWP. However, though we think it improbable that these results reflect any differential effect of poor labor market conditions on the TWP completer groups, they may be a result of more than offset availability. For example, it is possible that because of SSA’s need to more closely monitor the earnings of those who retained access to the offset that TWP completers from the original treatment group retained greater access to services such as work incentive benefits counseling.
Interactions Between Mean Earnings and Employment Rates
Figure 4 displays mean earnings for only those participants in each of the four analytical groups who had UI earnings in any particular quarter. Thus, the number and identity of cases used to calculate means changes from quarter to quarter. The purpose of the data displayed in Figure 4 is to “control” for the effect of employment rates on mean earnings.
Mean quarterly earnings for those employed by study group assignment and Trial Work Period (TWP) completion for the period of the first quarter of 2008 through the fourth of 2010.
Although the absolute differences in earnings between the TWP completer and non-completer groups are reduced, TWP completion status remains a far greater determinant of results than the original assignment to treatment or control. For example, there are no statistically significant differences in the means within the pair of TWP completer groups or the pair of TWP non-completer groups. On an absolute basis, the quarterly mean earnings of control group completers with UI reported employment are generally more than twice those of non-completers from the control group. This ratio is typically in the 3:1 range for the 2 comparable components of the treatment group.
Outcome trends after Q0 are best characterized as mildly upward, save for TWP completers from the control group. Relative to Q0, the Q8 mean earnings of those employed from the TWP non-completers from treatment and control are, respectively, US$397 (37.3%) and US$328 (23.4%) greater. Starting from a much higher base, mean earnings for employed individuals from the TWP completer component of the treatment group grew only 13.6% but with a higher absolute increase of US$436. The decline in quarterly earnings for control group completers over the Q0–Q8 period was US$224 (6.3%). Unfortunately, as UI data lack information about hours worked or hourly wages, we cannot assess what these factors contributed to these generally positive trends.
It is also notable, that quarterly mean earnings for employed members of both TWP completer groups always remain well above the SGA equivalent level (US$2,490). While perhaps not surprising for the group retaining access to the offset, we had not anticipated this result for the TWP completers subject to the “cash cliff.” Unfortunately, we do not have data allowing us to identify how many of these beneficiaries had incremental earnings beyond SGA that were greater than their SSDI payments. Nevertheless, we would not be surprised to find that a considerable proportion of employed TWP completers from the control group meet this criterion. Indeed, though we can imagine reasons why a beneficiary with incremental earnings less than her or his cash benefit might forgo the entire SSDI payment (e.g., placing greater value on performing work than on the income generated), given beneficiaries’ generally low incomes we would expect such behavior to be rare.
Outcomes for Participants Returned to Regular Rules in January 2009
Most treatment group members who were returned to regular SSDI program rules after December 2008 had very modest employment outcomes. Still, we wondered whether there might be individuals in this group making progress toward TWP completion who, in reaction to SSA’s change to pilot rules, decided to substantially reduce their work activity. The answer appears to be no.
We compared the earnings trends of TWP non-completers from both study groups who had positive UI earnings in the fourth quarter of 2008 (n = 30 from control, n = 31 from treatment). After Q0, the mean earnings trends for these subgroups are nearly identical, albeit declining from Q0 levels. There is nothing here to suggest that non-completers from treatment, after being returned to regular SSDI rules, behaved differently than non-completers from the control group.
Figure 5 displays employment rates for these two groups. The 100% employment rate at Q0 is an artifact of how these groups are defined. Once again, employment rate trends for TWP non-completers from the treatment and control groups are very similar. For both groups, there is a steep decline in employment rates immediately following Q0, with some divergence in trend (to the slight advantage of the treatment component) from Q3 onward. Again, there is nothing to suggest that this subgroup of treatment group members, when returned to regular program rules, reduced their work effort relative to employed non-completers from the control group.
Quarterly employment rates for those employed in the fourth quarter of 2008 but who had not yet completed a Trial Work Period (TWP) completion for the period of the first quarter of 2008 through the fourth of 2010.
There is positive news here. The employment rates for both groups of employed TWP non-completers held at 60% or higher 2 years beyond the Q0 reference point, values well above those expected for SSDI beneficiaries. It appears that, even among those with little current interest in or ability to work at the SGA level, employment at any earnings level is associated with continuing attachment to work.
Comparison of Analytical Group Characteristics
In our descriptive comparisons of employment outcome trends, we did not statistically control for differences in participant characteristics across the analytical groups. Nonetheless, claims about the importance of TWP completion and of the subsequent availability of the offset in motivating better employment outcomes would be strengthened if the following conditions are met: (a) the four groups were demographically and experientially similar at project entry and (b) to the extent that there were differences, the TWP completer groups would be more similar to each other than to either of the groups that did have any completers. It would also be pertinent to identify differences in participant experience attributable to BOPD implementation, especially any between the two groups of TWP completers.
We examined whether the distribution of a set of participant characteristics differed significantly across the four analytical groups, using chi-square when a variable was categorical and ANOVA when interval. Characteristics in place prior to or at entry into the SSDI-EP included mean age at entry, gender, race, educational attainment, SSA impairment category (musculoskeletal, neurological, mental retardation, mental non-retardation, and all others), whether the participant had been employed for some period subsequent to entering the SSDI program, the SSDI Primary Insurance Amount (an indicator of benefit size), and mean quarterly UI earnings in the year prior to entering the pilot. We also looked at a limited number of variables that captured some aspect of participant experience of the pilot: whether a participant received benefits counseling services, how much of that service was received, and when during the 15-month project enrollment period the participant entered the SSDI-EP.
The critical finding for our purposes is that there were no statistically significant or even near significant differences between the two TWP completer groups on any pre-enrollment or post-enrollment variable we examined (p always > .05). However, there were some statistically significant differences (p < .05) in the distributions of participant characteristics across the 4 groups. Although of less analytical importance than the differences (or lack thereof) between the two TWP completer groups, these data provide insight into both pre- and post-enrollment differences associated with having completed a TWP by the end of 2009.
We found statistically significant differences for only two of the pre-enrollment variables, having employment for some period between establishing SSDI eligibility and pilot entry (p < .001) and mean quarterly earnings in the year prior to entering the offset pilot (p < .001). However, these results reflect differences between the completer and non-completer groups, rather than within-pair differences. The percentage of TWP non-completers having UI employment after establishing SSDI eligibility was 66.9% for those assigned to treatment and 69.8% for those assigned to control. The mean quarterly earnings in the year prior to study entry for these two groups were, respectively, US$223 and US$211. By contrast, the percentage of TWP completers having employment subsequent to their SSDI eligibility date was 89.5% in the treatment group and 93.8% in the control group. Mean quarterly earnings in the year prior to study entry was US$1,439 for this component of the treatment group and US$1,368 for their peers in control.
These results support the conclusion that the differences among these four groups are far more closely related to TWP completion than to assignment to either the treatment or control group. Indeed, the fact that the analytical groups with stronger post entitlement employment outcomes also had stronger outcomes in the post-2008 period is consistent with prior research about return to work for those participating in a Social Security disability program (Burkhauser et al., 1999; Liu & Stapleton, 2011; Mitra & Brucker, 2004).
There were also statistically significant differences in the distributions for variables capturing some aspect of participant experience in the pilot. Once again, it was the pairs of groups who completed a TWP by the end of 2008 that differed most markedly from the groups of non-completers. On average, participants in the non-completer groups were less likely to get work incentive benefits counseling through the SSDI-EP, received less when they did, and tended to have enrolled later in the project enrollment period. The first two findings appear to reflect the greater utility of benefits counseling services for those actively seeking work or wanting to mitigate the negative impacts of higher earnings and income. The last finding is chiefly a consequence of how SSDI-EP enrollment processes were implemented (Delin et al., 2010).
Discussion of Results
This article has examined selected employment outcomes for four groups of SSDI-EP participants. These groups are defined by two factors: (a) original assignment to the SSDI-EP treatment and control groups and (b) TWP completion by December 31, 2008. Our principal findings are as follows:
Those who had completed a TWP had far better outcomes after December 2008 than those who had not, irrespective of continued access to the offset.
Of those who had completed a TWP, those who because of their original assignment to the SSDI-EP treatment group had access to the offset exhibited better outcome trends after the end of 2008 than TWP completers from the control group.
Differences in outcome trends between the two TWP completer subgroups can be characterized as statistically significant for employment rates and, especially, the proportion with earnings over the SGA level. However, the differences in mean earnings did not reach statistical significance at the .05 level.
These differences tended to increase over time reaching their maxima at the end of the data period.
There were no statistically significant differences between the two TWP completer groups in the distributions of any of the pre- or post-enrollment demographic or experiential characteristics we looked at.
Taken as a whole, the evidence from this natural experiment is highly supportive of a conclusion that the offset feature ultimately led to better employment outcomes for TWP completers, though it required a longer period than observed in the original SSDI-EP evaluation to generate those effects. We find it impressive that this result occurred during a severe recession which has been followed by a slow recovery that has had exceedingly modest job growth.
Conclusion and Policy Implications
Given our results, in conjunction with those from other studies, should Congress add an offset feature to the SSDI program? Any answer must consider at least two impact domains: the effects on beneficiaries and the effects on the SSDI program. We think that from the perspective of SSDI beneficiaries, the answer is clearly yes, particularly from the standpoint of achieving stronger employment outcomes from the roughly 40% of beneficiaries who appear to be “work oriented” (Livermore, 2011). Saying this, it is necessary to be mindful that beneficiaries’ willingness to use an offset and thus the level of its positive effects on employment outcomes no doubt depends on “optimizing” multiple factors. Some may involve the design of the offset itself; for example, the impact on mean earnings might be greater under some alternative combination of offset’s threshold level and slope. The offset tested in the BOPD involved a fairly strong (50%) inframarginal disincentive to earnings at its threshold level, that is, SGA (Gokhale, 2013). Similarly, an offset might induce stronger outcomes, especially above the SGA level, if beneficiaries had less fear that such outcomes would lead to termination of their SSDI eligibility. Although policy features such as expedited reinstatement or Medicaid Buy-ins may help mitigate some beneficiaries’ concerns, achieving the full potential of an offset provision may require changes to the Social Security disability definition and adjudication process to assess disability based on functional limitations rather than the specific ability to perform work at the SGA level. Finally, based on participants’ experiences in the four pilots, effective implementation of an offset feature is likely to be an important factor in encouraging beneficiaries to use it. In particular, inaccurate check adjustments and prolonged work initiated Continuing Disability Reviews may have a discouraging influence on the level of employment outcomes (Delin et al., 2010; Smith et al., 2009).
However, in addition to the potential gains, economic and social, for beneficiaries and the larger society, an offset feature may incur additional costs for SSA and, more generally, the public purse that need to be weighed. One finding from the BOPD was that the increased employment outcomes also led to at least a short-term increase in benefit payments (Weathers & Hemmeter, 2011). It is possible that other factors, for example, the ability to utilize the offset over longer periods, access to appropriate support services, inclusion of the concurrent SSDI/SSI population, or changing the features of the offset itself to ameliorate inframarginal disincentives, would result in a more positive “cost–benefit” assessment. Our data cannot directly support that type of analysis. It is our understanding that the results of the BOND will speak of all but the last of these possibilities.
In addition, we suggest deepening research efforts aimed at understanding why some beneficiaries make substantially greater use of available work incentives and supports than others who appear to share similar characteristics. After all, every participant assigned to the SSDI-EP treatment group had the legal right to complete a TWP and presumably interest in doing so as a necessary condition for being able to use the offset. Even with SSA’s decision to shorten the BOPD, every one of these individuals had a minimum of 26 months in which to complete their TWP. Yet, slightly less than half did so. It is possible that a small contribution to more fully understanding the dynamics of low utilization can be made by analyzing attitudinal and experiential data contained in the data set used for this article. However, we think it far more important to make sure that research designs for future tests of work incentives and other return to work innovations include concerted efforts to learn why only some beneficiaries make use of such opportunities and to then use that knowledge to inform policy making and implementation.
Footnotes
Acknowledgements
The authors thank the staff of Pathways to Independence and of the Office of Independence and Employment, Wisconsin Department of Health Services for their cooperation and support.
Authors’ Note
The descriptions and interpretations in this article are solely those of the authors. A data set may be obtained from the corresponding author, subject to completion of a data use agreement.
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 work was supported through Social Security Administration Contract No. SS00-05-60008.