PERSPECTIVES: The Benefit Receipt Patterns and Labor Market Experiences of Older Workers Who Were Denied Social Security Disability Insurance Benefits on the Basis of Work Capacity

Social Security Bulletin, Vol. 80 No. 2, 2020

This article examines the experiences of Social Security Disability Insurance (DI) applicants aged 51 or older who were initially denied benefits because the disability examiner determined that they could perform either their past work or other work. We use Health and Retirement Study survey data linked to administrative data on benefit application and receipt and earnings from the Social Security Administration. We find that few older DI applicants who were denied benefits on this basis resumed work at a substantial level following denial. More commonly, applicants denied at this stage continued to pursue benefits, often successfully. Nearly two-thirds of initial work capacity-related denials were ultimately allowed DI benefits after appealing the initial decision or reapplying, and our estimates suggest that many of the rest claimed Old-Age and Survivors Insurance benefits before they reached full retirement age.

The authors are with Mathematica. Jody Schimmel Hyde is a senior researcher, April Yanyuan Wu is a researcher, and Lakhpreet Gill is a systems analyst.

Acknowledgments: The authors thank Annie Doubleday, who worked at Mathematica at the time this work was completed, for her careful mapping of occupation categories in the HRS to O*NET codes. We also thank David Mann and David Stapleton of Mathematica's Center for Studying Disability Policy for insightful comments on the draft manuscript. A previous version of this article was published as Disability Research Consortium (DRC) Working Paper No. 2018-01 (

The research reported herein was performed pursuant to an SSA grant (no. 1-DRC12000001-01-00) and was funded as part of the Disability Research Consortium. The findings and conclusions presented in the Bulletin are those of the authors and do not necessarily represent the views of the Social Security Administration.


Selected Abbreviations
CPI Consumer Price Index
CYBF Cross-Year Benefits file
DDS Disability Determination Service
DI Disability Insurance
FRA full retirement age
HRS Health and Retirement Study
O*NET Occupational Information Network
OASI Old-Age and Survivors Insurance
OIG Office of the Inspector General
PRW past relevant work
RFC residual functional capacity
SEF Summary Earnings file
SGA substantial gainful activity
SSA Social Security Administration
SSI Supplemental Security Income

As workers approach retirement, they are more likely to experience a health condition that could limit their ability to remain employed. Among individuals aged 51 to 55 in 1992, one-quarter reported experiencing a health condition that, by age 62, had limited their ability to work at least once (Johnson, Mermin, and Murphy 2007). A worker becomes eligible for unreduced Old-Age and Survivors Insurance (OASI) benefits, commonly known as Social Security retirement benefits, on reaching his or her full retirement age (FRA: 65 to 67, depending on year of birth). If a new health condition significantly affects a worker's ability to remain in the labor force, and the worker has not yet reached FRA, he or she may be eligible for Social Security Disability Insurance (DI) benefits. Alternatively, a worker who has reached age 62 (but not FRA) may decide to claim actuarially reduced OASI benefits. Early (pre-FRA) OASI claiming reduces a monthly benefit by as much as 30 percent (if claimed at age 62 by a worker whose FRA is 67).1 When workers claim early OASI retirement benefits, the Social Security Administration (SSA) asks them whether they have experienced a health condition, injury, or illness in the past 14 months that left them unable to work. When applicants report any such experience, SSA considers their eligibility for DI benefits. Doing so is relatively costless for those who have already stopped working, and if DI benefits are awarded, the applicant will receive a higher monthly retirement benefit upon reaching FRA than he or she would receive with an early-claiming reduction. However, an award of disability benefits is far from certain.

This study examines the employment, earnings, and benefit-receipt outcomes of workers aged 51 or older whose applications for DI disabled-worker benefits are denied by the SSA Disability Determination Service (DDS) in the initial review levels.2 We find that almost half of the applicants aged 51 or older are initially denied benefits. In addition to documenting the employment, earnings, and benefits trajectory of these applicants, we compare outcomes by reason for denial.

SSA uses a five-step sequential evaluation process (detailed later) to determine whether an applicant meets the criteria for benefit award. In the first three steps, evaluators assess the applicant's insured status and the medical factors that affect the ability to continue or resume work. We focus on applications denied at the fourth and fifth steps, in which evaluators assess the applicant's work capacity.3 In most cases, the DDS assesses these applicants to determine whether, in light of their alleged medical impairments, their residual functional capacity (RFC) allows them to perform either past relevant work (PRW), at step 4, or other work, at step 5.4 In considering RFC at step 5, disability examiners also account for vocational factors—age, education, and work experience—using guidelines known as the “medical-vocational grids.”

Our study sheds light on the extent to which DI applicants denied for work-capacity reasons return to work, and compares their employment and earnings trajectories with those of applicants denied for other reasons or whose claims are ultimately allowed. Strictly speaking, the DI applications, and not the applicants themselves, are allowed or denied in the disability determinations. However, for practical reasons, we use “denied applicants” as shorthand to refer to our study sample, whose members applied for DI disabled-worker benefits, had their applications denied by the DDS at the initial review level, and were older than 50 and younger than FRA at the time of the initial denial.

We find that relatively few older denied applicants return to work, and highlight reasons why they might not. In particular, we examine the shares of initially denied applicants that appeal the denial or reapply for DI, and the shares that are ultimately awarded benefits. We also identify how many denied applicants claim OASI benefits after reaching the earliest age of eligibility (62) but prior to FRA—an option not available to younger workers, but a particularly salient one for older workers. We find some evidence that employment paths differ based on the sequential-evaluation step at which the initial denial occurred, and that the likelihood of working declines as time passes after the decision for all older denied applicants. Our findings also point to the possibility that applicants denied at step 5 who returned to work had slightly higher postdenial earnings than did those who were denied at other steps.

To conduct our analysis, we used Health and Retirement Study (HRS) data linked to SSA records on earnings and benefit application and receipt. Our linked sample consists of 805 HRS respondents who applied for DI at age 51 or older but younger than FRA from 1992 to 2012; of these, only 384 applicants were denied benefits, with 259 denied for work-capacity reasons. Although our sample is admittedly small, the richness of the longitudinal survey data in the HRS offers advantages that would not be possible using administrative data alone, including information about applicant demographics, socioeconomic status, and occupational characteristics. Using information from agency administrative records, we stratified DI applicants based on the outcome of and the basis for the initial decision. We measured the employment and earnings of groups of denied applicants using the HRS Respondent Cross-Year Summary Earnings file (hereafter, the Summary Earnings file, or SEF), which links survey results with earnings data from SSA's Master Earnings File. Finally, to better understand postdenial employment patterns, we used data from the linked SSA records on applicants' subsequent receipt of DI and OASI benefits.

Before describing our study, we outline the SSA disability determination process and review existing evidence on the earnings of denied applicants. We then describe our approach to identifying DI applicants using the linked HRS data and administrative records and examine the characteristics of denied applicants with different reasons for initial denial. We go on to consider applicants' benefit trajectories after denial, and follow that by tracing their employment and earnings trajectories both before application and after adjudication. We then conclude and discuss implications for policy.

The Disability Determination Process

An individual is eligible for DI benefits if he or she is unable to engage in substantial gainful activity (SGA) because of a medically determinable physical or mental impairment that has lasted or is expected to last for a continuous period of at least 12 months or is expected to result in death (SSA 2012b). Earnings that meet or exceed an annually adjusted monthly earnings threshold signify SGA. In 2020, SGA is indicated by $1,260 for nonblind individuals and $2,110 for blind individuals. SSA disability examiners follow the five-step sequential evaluation process outlined below to determine eligibility.

In step 1, the examiners determine whether the applicant's work and payroll-tax contribution history is sufficient to qualify him or her as insured. If so, disability examiners assess whether the applicant has an impairment that meets the medical eligibility criteria in step 2. If the medical evidence does not establish that the applicant's conditions are severe or will last for at least 12 months, the examiner denies the claim for medical reasons. In this article, we refer to these as “medical denials.”

If an impairment is deemed to be sufficiently severe in step 2, the examiner goes on to determine whether it satisfies the medical criteria for specific conditions contained in SSA's Listing of Impairments in step 3. The listing includes hundreds of conditions that result in a benefit award if sufficiently severe. If an applicant has more than one documented impairment, the examiner may find that the impairments in combination are equivalent to the medical criteria in a given listing. The adjudicator allows the application if the impairments are determined to meet or “equal” the listings.

If an application is not allowed at step 3, the examiner assesses the applicant's RFC (for details, see SSA 2018b) and compares it to the requirements of the applicant's PRW (see SSA 2011) at step 4.5 If the examiner deems that the applicant is able to perform PRW (which extends to any work performed within the last 15 years), the application is denied at step 4. Examiners compare RFC with PRW on a function-by-function basis; in other words, they compare the requirements of the past work, using information provided by the applicant or contained in SSA's Dictionary of Occupational Titles, with the applicant's RFC (for details, see SSA 2017b and SSA 2017c). Examiners consider the applicant's ability to perform PRW both as it was performed in his or her job (step 4a) or as it is generally performed in the national economy (step 4b). Assessing the ability to perform PRW does not account for the availability of such work in the current economy or for other economic conditions that might affect applicants' ability to find work (see SSA 2011).

The fifth step affects applicants whose RFCs are deemed incompatible with the performance of PRW. The cases of nearly half of insured DI applicants reach this final step (Wixon and Strand 2013; Mann, Stapleton, and de Richemond 2014). The examiner assesses ability to perform other work by comparing the applicant's RFC to the exertional requirements of work outlined in SSA's medical-vocational guidelines (see SSA 2015). Examiners deny benefits for applicants whose RFCs indicate that they can meet the requirements. As in step-4 decisions, step-5 medical determinations do not account for current economic conditions.

By law, SSA must consider the vocational factors—age, education, and work experience—in determining whether an applicant can engage in substantial work, although the law does not specify how those factors should be incorporated into the determination process. The grids show the type of work that can be done given the applicant's RFC and vocational factors. The types of work are grouped by exertional requirements, including sedentary, light, medium, heavy, or very heavy. Because the grids vary by age, with separate factors for ages 50–54, 55–59, and 60–64, they are particularly salient for older workers. Benefit awards are more likely with age because of the grids, even holding disability, education, and work experience constant. For example, a 54-year-old individual who could perform sedentary work would not receive an award based on the grid for 50- to 54-year-olds but could receive an allowance a year later when the examiner consults the grid for 55- to 59-year-olds, under which the individual is deemed unable to work. One final important factor in a step-5 determination is whether the applicant has transferable skills from PRW that enable him or her to meet the requirements of other occupations available in the RFC category. If the applicant is deemed to have transferable skills, that factor dominates age and education in reaching an award determination (Johns n.d.).

SSA's use of vocational factors and its assessment of ability to perform PRW have been criticized for many years. In reviewing the relationship between vocational factors and employment, Mann, Stapleton, and de Richemond (2014) did not find evidence in the literature that vocational factors alone can predict ability to perform work that one has not performed before—independent of other factors that might predict future work. For that reason and others, vocational factors have drawn policymakers' attention and reform proposals (for example, Warshawsky and Marchand 2015). SSA has also grappled with identifying the demands of work when comparing an applicant's RFC and the requirements of PRW. The agency continues to use the Dictionary of Occupational Titles, which is outdated and often does not reflect current job demands. More recently, SSA sponsored a new data set, the Occupational Information System (OIS), which is based on occupational information collected by the Bureau of Labor Statistics and is updated every 5 years.6

Prior Evidence About the Earnings of Denied Applicants

The earnings of denied DI applicants were the subject of a robust literature nearly three decades ago (for example, Bound 1989 and Parsons 1991). More recently, von Wachter, Song, and Manchester (2011) reexamined the topic, taking advantage of improvements in data and processing. However, those studies documented the earnings of denied applicants to benchmark what might have been expected for the earnings of allowed applicants in the absence of benefits. As such, the studies aggregated denied applicants and did not explore employment and earnings variations by reason for denial.

Two other recent studies focused on applicants denied because of their work capacity, which therefore relate closely to our analysis. However, the timeframe and applicant selection criteria used in those studies differed from ours. In one study (SSA 2017a), the agency's Office of the Inspector General (OIG) determined that less than half of applicants denied at steps 4 and 5 in 2013 returned to work. Moreover, average earnings after denial (among those with earnings) were less than $9,400 per year, or about 70 percent of that group's $13,640 preapplication average annual earnings. The OIG findings align with those of an earlier study (Strand and Trenkamp 2015), which focused on step-5 denials, although the studies analyzed different populations and observation periods.7 Strand and Trenkamp determined that employment declined by 22 percentage points (26 percent) from 2000 to 2008 among those who were initially denied in 2005. The authors also documented important differences in outcomes by preapplication earnings decile, but they found that postdenial earnings (both median and maximum) were persistently lower than preapplication earnings in nearly every decile, averaging around 77 percent for the sample overall.

Our work complements and extends the earlier studies along two important dimensions. First, we separate step-4 denials from step-5 denials and compare outcomes between those groups as well as with other groups of denied applicants. By contrast, the 2017 OIG study combined step-4 and step-5 denials, while Strand and Trenkamp considered only step-5 denials; in addition, neither study compared outcomes with those of other applicant groups. Second, we link administrative data to longitudinal survey data, allowing us to measure attributes such as health, household, and occupational characteristics—attributes that neither of the earlier studies measured because they are not contained in administrative data. This information provides a richer picture of denied applicants and how they differ from their allowed counterparts.

Although it has many advantages, the HRS limits our consideration of applicants to those aged 51 or older, whereas earlier studies considered applicants of all ages. As described earlier, vocational factors at step 5 are progressively less stringent for applicants as they age. As a result, the reasons we find for denials based on work capacity split differently from those found in other studies, reflecting the consideration of age in the vocational grid. For example, of denials based on work capacity in our study, more than two-thirds occurred at step 4 (shown later), in contrast with the OIG report, which found that only one-quarter of denials for work capacity occurred at step 4 and three-quarters occurred at step 5. Because the OIG report considered Supplemental Security Income (SSI) applicants as well as those for DI (and concurrent SSI and DI) benefits, its findings differ from ours, as postdenial employment for older DI applicants with substantial work histories will likely differ from those of SSI applicants of all ages without such experience. Given that our sample does not include younger applicants, our findings are not directly comparable with those of the earlier studies.

Data and Sample Selection

Our study capitalizes on a significant linkage of survey and administrative data. We use the HRS, a nationally representative longitudinal survey of Americans collected by the Institute for Social Research at the University of Michigan. The HRS interviews respondents aged 51 to 61 then conducts biennial follow-up interviews with them on a range of subjects. We use data collected in the 1992 through 2012 waves of the HRS. We link the HRS survey responses to administrative data from several SSA sources, with which we measure DI applications and benefit receipt and collect annual earnings data through 2012.8 From the Department of Labor's Occupational Information Network (O*NET), we also collect data on occupational characteristics.9 By linking the survey results with the data from various administrative sources, we are able to compile detailed information and avoid the reporting errors that are common with self-reported benefit receipt and earnings.

Our estimates are weighted to account for both the complex survey design and the respondents' consent to having their survey responses linked to administrative data. We use the weights provided by the HRS.10,11 The HRS consent weights account for nonrandom selection into the group that consented to the administrative data linkage. Even though more than three-quarters of the respondents in our observation period agreed to the linkage,12 evidence has shown that consenters differed from the full HRS sample in terms of age, race, sex, income, and education (see, for example, Gustman and Steinmeier 2001; Haider and Solon 2000; and Kapteyn and others 2006).

Identifying DI Applicants Using the Linked HRS-SSA Files

Our analysis focuses on 805 HRS respondents who applied for DI disabled-worker benefits (that is, benefits based on their own work history) at least once after age 50 and before FRA, following their first HRS interview.13 The DI applications were relatively evenly distributed from the first observation year (1992) to the last (2012). About one in four (26 percent) of these applicants concurrently applied for SSI. We identified these applicants using SSA's 831 file (named for the Disability Determination and Transmittal form SSA-831), which records information about all applications that receive a medical determination. We used the date of initial application recorded in that file, which we then aligned with the timing of the survey interviews in the linked HRS data to identify the date of first application that followed the first HRS interview.

Importantly, a denial recorded in the 831 file does not mean that the applicant was not ultimately awarded benefits, because the file does not include information about appeals to an administrative law judge (ALJ) or a federal court. Thus, we have information only on the outcomes adjudicated by the DDS. If an application was initially denied but was then reconsidered by the DDS at the request of the applicant, the result of the reconsideration is in the 831 file. In such cases, we used the reconsideration outcome—and the basis for that outcome—to classify the case. To track postdenial benefit trajectories, we developed an approach to identify subsequent ALJ benefit awards using the HRS Cross-Year Benefits file (CYBF), discussed later.

The fact that a respondent's first HRS interview is conducted no earlier than age 51 and as late as age 61 has two methodological implications for our analysis. First, the applications we observe do not necessarily reflect individuals' first applications for DI benefits; of the 805 applicants in our sample, approximately 10 percent had applied for DI at least once from 1988 (the earliest year for which SSA-831 data are available) to the time of their first HRS interview. First interviews in our sample occurred between 1992 (the first HRS year) and 2012 (the last year for which many SSA data are available, although the earnings data continue through 2013).14 Second, because the age at the first HRS interview ranges from 51 to 61, the first time we observe a DI applicant is left-censored. For instance, if a person participated in an HRS interview for the first time at age 56, but had previously applied for DI at age 52, we would only count applications that occurred at age 56 or older. Yet if another respondent had applied at age 52 but was first interviewed for the HRS at age 51, we would include that application in our analysis.

Categorizing DI Applicants Based on the Outcome of the Initial Determination

We group applicants based on the outcome of their initial decision as recorded in the 831 file. Chart 1 diagrams our sample based on the five-step sequential evaluation process described earlier. More than half of the older DI applicants in our sample were initially allowed benefits, a substantially higher allowance rate than that for DI applicants of all ages (just above 30 percent) found by Wixon and Strand (2013).15

Chart 1.
Distribution of DI applicants aged 51 or older but younger than FRA, by outcome of the initial determination (including DDS reconsideration), 1992–2012
Flowchart linked to data in narrative format.
SOURCES: Authors' calculations using HRS data linked to SSA's 831 file; Wixon and Strand (2013).
NOTE: Numbers are weighted to account for the complex HRS survey design and the varying probabilities of respondent consent to matching the HRS data to administrative data.
a. Consists primarily of applicants denied because the impairment was deemed not to be severe or likely to last 12 months or result in death. Also includes a small number of applicants (fewer than 10) who did not meet eligibility requirements before step 3 for other reasons, such as failing to follow prescribed treatment, submit to a consultative examination, or provide sufficient supporting evidence.

In addition to indicating whether an application was initially allowed or denied, the 831 file provides information on the step at which the examiner decided the case. We use that information to stratify applicants based on the scheme outlined in Wixon and Strand (2013). Among allowed applicants, about one-third were allowed benefits at step 3 because their impairment met or equaled medical criteria in the Listing of Impairments, and the other two-thirds were allowed at step 5 for reasons related to reduced work capacity. Among denied applicants, about one-third were denied at step 2 because their impairment was deemed not to be severe and the remaining two-thirds were denied for work-capacity reasons. Among the latter, denials for the ability to perform past work (step 4) were more than twice as common as denials for the ability to perform other work (step 5). That result is consistent with the sense that if there is cause to deny an application, examiners will aim to identify it at the first applicable step, and with the fact that some cases can be denied at step 4 but not at step 5.

Measuring Employment and Earnings Around the Initial DI Decision

We measure earnings and employment (indicated by having positive earnings) using the linked SEF, available with permission from the HRS. The file includes Master Earnings File information on annual earnings subject to OASI and DI payroll taxes, up to the taxable maximum in each year, as reported to the Internal Revenue Service.16 We consider earnings in the years before and after application using the application filing and decision dates. Virtually all HRS respondents who consent to the linkage to administrative data from SSA also have a record in the SEF, although not in every year (perhaps reflecting years with no reportable earnings). To align the administrative data on earnings with self-reported household income in the HRS, we follow the HRS convention of using the calendar year preceding the interview date. Using the Consumer Price Index (CPI), we adjust all income and earnings measures to 2012 dollars.

Tracing Preapplication and Postdenial Employment, Earnings, and Benefit Patterns

We open this section by outlining our methodology for tracing the pattern of benefit receipt for DI applicants in the period between the initial denial and FRA. We then describe our approach to measuring the employment and earnings trajectories of denied applicants in the 5 years before they applied for benefits and the 5 years after their initial denial.

We note that data enabling us to observe the benefits trajectory through FRA, and the employment and earnings of applicants for 5 years after denial, were not available for all applicants in our sample. There are two primary reasons for this right-censoring. First, some denied applicants died: 2 percent died within 5 years of the initial decision, and 4 percent died before they reached FRA. Second, some applicants' cases were not decided until shortly before our observation period ended. Applicants who were denied in or close to 2012 (the last year for which we obtained benefit data) had little time to reapply or appeal their decision and qualify for benefits within our observation period. Likewise, applicants who were denied within 5 years of 2013 (the last year for which we obtained earnings data) could not be observed for the full follow-up period. We discuss the magnitude and effects of censoring caused by partial data availability below.

Tracing Postdenial Benefit Patterns

To compile descriptive statistics on postdenial DI benefit trajectories, we link data from SSA's 831 file to information contained in the CYBF. The CYBF includes administrative data from SSA's Master Beneficiary Record and Payment History Update System, both of which provide monthly information on DI and OASI benefit receipt.

The CYBF link was necessary to determine whether applicants appealed an initial denial because the 831 file does not contain information on allowances and denials at higher levels of appeal. We assume that an applicant was granted benefits on appeal if the CYBF record indicates DI benefit receipt after an initial denial and we observe no later DI applications in the 831 file. We assume that an applicant reapplied and was awarded benefits if we observe DI benefit receipt in the CYBF and a subsequent DI application in the 831 file, although we do not know the adjudication level at which the case was allowed. For those whose records indicate a subsequent DI application but no benefit receipt, we assume that applicants reapplied and were denied, although they may have received an allowance after 2012, the last year in the CYBF. We categorize one final group: those who never applied again. This group comprises those with no subsequent application indicated in their 831 file and no DI benefits in the CYBF. It includes applicants who unsuccessfully appealed their initial denial as well as those who did not appeal; the data do not allow us to distinguish between the two groups.

One limitation of the CYBF data inhibits our ability to categorize postdenial DI benefit receipt correctly. The CYBF variable that indicates type of benefit (disability or retirement) has only accounted for conversions from the former to the latter—and enabled historical tracking of such changes—since May 2009. Before then, the prior status was overwritten when the status changed.17 To the extent that the absence of prior-status information in those records introduces errors in our results, it would lead us to underestimate the share of denied applicants who received DI benefits before reaching FRA. Specifically, we cannot determine whether DI applicants whose claims were initially denied, who became Social Security beneficiaries between age 62 and FRA, and who attained FRA before May 2009 received OASI benefits only or were first awarded DI benefits. Hence, an initially denied applicant who attained FRA before May 2009 and whom we have classified as entering the OASI beneficiary rolls before FRA may have actually entered the DI rolls first, albeit after age 62.

Tracing the Employment and Earnings of Denied Applicants Around the Application and Initial Denial

We estimate average employment rates and earnings for denied applicants from 5 years before application to 5 years after the initial denial was received. Right-censoring because of a lack of available data after the initial decision (because of death or a decision closing within 5 years of 2013, the last year of available earnings data) affects the share of records for which we can observe employment outcomes for the full period. The number of complete records for applicants through the end of the 5th year after denial is 15–25 percent lower than the number of denied applicants in our sample, with missing data most common among medical denials and least common among step-5 denials.

We did not find that censoring led to any systemic differences in average employment rates or earnings levels for denied applicants. We considered the effects of censoring by comparing mean earnings among all denied applicants in their postdenial years to those of denied applicants with an initial decision in 2008 or earlier; for the latter, we had 5 years of postdecision data available (excluding the few cases of applicants who died). Again, using unadjusted values, we did not find that the employment rate of denied applicants varied in any meaningful way when we imposed this restriction. To the extent that average earnings differed for a given group of denied applicants, the direction was not consistent and the magnitude was small relative to the overall standard deviation of earnings. To maintain our already small sample size, our findings are for all initially denied applicants.

Characteristics and Preapplication Occupational Attributes of Denied DI Applicants

The mean age at application across our entire sample was slightly younger than 58; ages ranged from 51 to 65. Most of the applicants we observe sought DI benefits before they were eligible to claim OASI benefits; only 12 percent had reached age 62 and were eligible for OASI when they applied for DI.

Table 1 shows that applicants denied because of work capacity were younger on average (57.4) than those denied for medical reasons (58.2). Consistent with the medical-vocational grids, work-capacity denials were concentrated in higher educational-attainment groups than medical denials; for example, 19 percent of work-capacity denials had less than a high school education, compared with 27 percent of medical denials. Relative to medical denials, work-capacity denials were less likely to be married (58 percent versus 75 percent) or Hispanic (8 percent versus 21 percent). They also had longer tenure in the job they held 2 years before application (10.3 years versus 6.5 years), more years with positive earnings at ages 22 to 50 (21 versus 20), higher average annual earnings in those years ($27,696 versus $19,638), and higher household income ($56,266 versus $45,652). Table 2 shows that work-capacity denials were also more likely to work in occupations requiring computer use (23 percent versus 15 percent), the ability to withstand stress (28 percent versus 20 percent) and to deal with unpleasant or angry people (14 percent versus 3 percent), and basic skills (59 percent versus 49 percent). However, they were less likely to work in jobs requiring social skills (21 percent versus 44 percent) or “system” skills (6 percent versus 17 percent).18

Table 1. Demographic, employment, and income characteristics of older DI applicants denied at the initial level, by reason for denial, 1992–2012
Characteristic Denied for medical reasons (step 2) Denied for work capacity
Total (step 4 or 5) Able to perform past work (step 4) Able to perform other work (step 5)
Statistic Standard error Statistic a Standard error Statistic Standard error Statistic b Standard error
Number (weighted) 780,144 1,661,495 1,130,297 531,198
Average age 58.2 0.3 57.4** 0.2 58.2 0.2 55.7*** 0.4
Women 58.1 4.5 53.6 3.2 59.7 3.7 40.7*** 5.9
Married 74.7 4.0 58.2*** 3.2 54.3 3.8 66.1* 5.7
White 68.6 4.2 73.6 2.8 69.1 3.4 83.0** 4.5
Black 18.0 3.5 20.2 2.5 22.5 3.1 15.1 4.3
Hispanic 20.5 3.7 7.5*** 1.7 10.4 2.3 1.2** 1.3
Percentage with—
Less than high school diploma 27.1 4.0 18.8* 2.5 17.9 2.9 20.7 4.9
High school diploma 39.2 4.4 39.6 3.1 47.7 3.7 22.5*** 5.0
More than high school diploma 33.7 4.3 41.6 3.1 34.4 3.5 56.8*** 6.0
Employment status and work experience
Employed 62.5 4.6 65.8 3.1 66.2 3.7 64.9 6.0
Employed full time 71.8 5.4 73.5 3.5 74.8 4.1 70.6 7.0
Self-employed 13.2 4.1 12.3 2.6 10.3 2.8 16.9 5.8
Working for a firm with—
More than 500 employees 11.8 4.2 18.5 3.5 25.9 4.8 4.6*** 3.4
Fewer than 100 employees 67.8 6.0 63.1 4.4 61.9 5.3 65.4 7.8
Average number of years—
In current job c 6.5 0.9 10.3*** 0.9 10.8 1.1 9.3 1.8
With positive earnings at ages 22–50 19.8 0.5 21.2* 0.4 21.2 0.4 21.3 0.6
Earnings and income ($)
Hourly wage 14.0 0.8 16.4 1.0 15.1 1.1 19.3** 1.9
Annual earnings (self-reported) 18,025 2,512 20,019 1,519 21,033 1,800 17,760 2,835
Total household income 45,652 3,879 56,266* 3,453 55,134 4,094 58,789 6,472
Annual earnings at ages 22–50 19,638 1,066 27,696*** 1,083 26,317 1,199 30,577* 2,258
SOURCE: Authors' calculations using HRS data linked to selected administrative data files from SSA.
NOTES: Sample consists of applicants aged 51 or older but younger than FRA who were denied at the initial level (including DDS reconsiderations).
Earnings and income values are adjusted to 2012 dollars using CPI.
All results are weighted to account for the complex HRS survey design and the varying probabilities of respondent consent to matching the HRS data to administrative data.
Chi-square tests compared the distribution of multinomial outcomes across groups; t-tests compared binomial outcomes.
* = statistically significant at the 0.10 level; ** = statistically significant at the 0.05 level; *** = statistically significant at the 0.01 level.
a. Indicators of statistical significance are relative to medical (step-2) denials.
b. Indicators of statistical significance are relative to step-4 denials.
c. Job held 2 years before DI application.
Table 2. Attributes of the preapplication job held by older DI applicants denied at the initial level, by reason for denial, 1992–2012 (in percent)
Characteristic Denied for medical reasons (step 2) Denied for work capacity
Total (step 4 or 5) Able to perform past work (step 4) Able to perform other work (step 5)
Statistic Standard error Statistic a Standard error Statistic Standard error Statistic b Standard error
Applicants with O*NET data (weighted)
Number 581,770 1,252,512 890,475 362,037
Percentage of total applicants 13.1 28.1 20.0 8.1
O*NET occupational requirement
General physical demands
Any 38.5 5.8 43.1 4.0 36.1 4.5 58.2 7.6
High-level 3.8 2.3 2.9 1.3 3.2 1.7 2.2 2.2
Flexibility or dexterity
Any 32.4 5.6 28.6 3.6 23.6 4.0 39.3 7.5
High-level 4.3 2.4 12.8 2.7 9.2 2.7 20.6** 6.2
Vision 10.4 3.6 15.2 2.9 17.3 3.6 10.6 4.8
Cognitive ability demands
Any 46.6 6.0 48.1 4.0 49.8 4.7 44.4 7.7
High-level 30.4 5.5 18.0 3.1 15.1 3.4 25.2 6.7
Computer use 15.4 4.3 23.4** 3.4 25.7 4.1 18.5** 6.0
Interpersonal skills 35.4 5.7 26.8 3.6 37.5 4.6 3.9*** 3.0
Stress tolerance
Any 19.9 4.8 27.8** 3.6 32.3 4.4 18.2 6.0
High-level 6.6 3.0 8.9 2.3 10.2 2.9 5.9 3.7
O*NET occupational skill
Ability to—
Deal with unpleasant or angry people 3.2 2.1 14.3*** 2.8 10.8 2.9 21.8* 6.4
Use and update relevant knowledge 8.3 3.3 8.9 2.3 4.3 1.9 18.6*** 6.0
Handle difficult working conditions 13.2 4.0 23.0 3.4 17.0 3.6 35.9** 7.4
Skill type
Basic 48.7 6.0 58.9* 4.0 61.5 4.6 53.2 7.7
Complex problem solving 2.7 1.9 4.3 1.6 0.3 0.5 12.8*** 5.2
Resource management 21.8 4.9 14.6 2.8 12.6 3.1 19.0 6.1
Social 43.9 5.9 20.6** 3.3 22.8 4.0 15.8 5.6
System 17.1 4.5 5.7** 1.9 2.4 1.5 12.8* 5.2
Technical 13.2 4.1 17.3 3.0 11.0 3.0 30.8** 7.1
SOURCE: Authors' calculations using HRS data linked to selected administrative data files from SSA.
NOTES: Sample consists of applicants aged 51 or older but younger than FRA who were denied at the initial level (including DDS reconsiderations).
Earnings and income values are adjusted to 2012 dollars using CPI.
All results are weighted to account for the complex HRS survey design and the varying probabilities of respondent consent to matching the HRS data to administrative data.
Chi-square tests compared the distribution of multinomial outcomes across groups; t-tests compared binomial outcomes.
* = statistically significant at the 0.10 level; ** = statistically significant at the 0.05 level; *** = statistically significant at the 0.01 level.
a. Indicators of statistical significance are relative to medical (step-2) denials.
b. Indicators of statistical significance are relative to step-4 denials.

Table 1 shows that applicants denied at step 4 were older than those denied at step 5 (on average, 58.2 versus 55.7).19 Relative to applicants whose initial determination was a denial at step 5, those who were denied at step 4 were more likely to be women (60 percent versus 41 percent), to be Hispanic (10 percent versus 1 percent), or to have a high school education (48 percent versus 23 percent). They were less likely to be married (54 percent versus 66 percent), to be white (69 percent versus 83 percent), or to have any postsecondary education (34 percent versus 57 percent). Further, step-4 denials were more likely than step-5 denials to work in large firms (26 percent versus 5 percent) but with a lower hourly wage rate ($15 versus $19) and lower average annual earnings at ages 22–50 ($26,317 versus $30,577).

Highlighting some intriguing differences in job attributes, Table 2 shows that step-4 denials were more likely than step-5 denials to work in an occupation requiring computer use (26 percent versus 19 percent) or interpersonal skills (38 percent versus 4 percent). However, they were less likely to have jobs that require high-level flexibility or dexterity (9 percent versus 21 percent) or the abilities to deal with unpleasant or angry people (11 percent versus 22 percent), to continually update their skills and knowledge (4 percent versus 19 percent), and to handle difficult working conditions (17 percent versus 36 percent). Interestingly, step-4 denials were also less likely to have jobs that demand complex problem-solving skills (0 percent versus 13 percent), system skills (2 percent versus 13 percent), or technical skills (11 percent versus 31 percent).

How Postdenial Appeal, Reapplication, and Allowance Patterns Vary by Reason for Initial Denial

In the first of the two subsections that follow, we report our findings on the likelihood that DI applicants continue to pursue benefits after an initial denial. In the second subsection, we discuss how the availability of OASI benefits as early as age 62 may affect DI claiming behavior after initial denial.

Applicants Who Appeal or Reapply for DI Benefits After Initial Denial

After the DDS denies an application at the initial or reconsideration level, applicants have several options for continuing to seek DI benefits (as well as the option not to continue). Chart 2 diagrams the options and shows how the older applicants who constitute our sample responded. The applicant must first decide whether to appeal the initial denial. For applicants who do not appeal, or whose appeal of the initial denial is likewise denied, the second decision is whether to reapply for benefits—perhaps several years after the initial denial. Overall, 56 percent of denied applicants in our sample ultimately received an award, of whom the majority had appealed the first observed denial. Some denied applicants (28 percent) did apply again, either after an appeal was denied or without first appealing.

Chart 2.
DI appeals and reapplications following initial denials for older applicants, 1992–2012
Flowchart linked to data in narrative format.
SOURCES: Authors' calculations using HRS data linked to SSA's 831 file.
NOTES: Sample consists of applicants aged 51 or older but younger than FRA who were denied at the initial level (including DDS reconsiderations).
To meet the requirements for using restricted HRS-SSA linked data, we have rounded the percentages to avoid disclosing potentially identifiable information without diminishing the qualitative findings.
Percentages do not sum to 100.0 because of rounding.

More than one-third of applicants with an initial-level denial (39 percent) never applied again, although that figure likely overstates the actual proportion for two reasons. First, the 831-file data prevent us from excluding two groups: individuals who appealed the initial decision but were denied; and applicants who appealed, were awarded DI benefits, then converted to OASI benefits on reaching FRA prior to May 2009. Second, as discussed earlier, some members of our sample may have reapplied after the period for which data were available. About 26 percent (weighted) had initial decisions within 3 years of the end of the period with administrative data available; within a longer follow-up period, we would expect more appeals or reapplications. Nonetheless, we have no reason to believe that the censoring would dramatically affect the differences in reapplication patterns across groups.

The overall pattern of appeals and reapplications among older applicants with initial-level denials reflects the important divergence in the paths of those denied for medical reasons and those denied for work-capacity reasons (Table 3). In particular, 67 percent of work-capacity denials were allowed after appeal or reapplication, more than twice the 31 percent rate at which those denied for medical reasons were ultimately allowed. Applicants with work-capacity denials have higher eventual DI allowance rates than applicants with medical denials have because a much higher share of the former successfully appeal the initial decision. Compared with the 20 percent of medical denials who appealed the initial decision and subsequently received benefits, 43 percent of those denied at step 4 and 32 percent of those denied at step 5 filed an appeal and received an allowance.

Table 3. Postdenial appeals and reapplication for older DI applicants denied at the initial level, by reason for denial, 1992–2012 (in percent)
Characteristic All denials Denied for medical reasons (step 2) Denied for work capacity
Total (step 4 or 5) Able to perform past work (step 4) Able to perform other work (step 5)
Statistic Standard error Statistic Standard error Statistic a Standard error Statistic Standard error Statistic b Standard error
Number of applicants (weighted) 2,441,638 780,144 1,661,494 1,130,296 531,198
Percentage of applicants who—
Appealed and were allowed 33.1 2.4 19.8 3.6 39.4*** 3.1 42.9 3.7 32.0*** 5.6
Reapplied and were—
Allowed 22.6 2.2 11.2 2.9 28.0*** 2.8 26.5 3.3 31.0*** 5.6
Denied 5.7 1.2 8.1 2.5 4.5*** 1.3 4.5 1.6 4.5*** 2.5
Never reapplied c 38.7 2.5 60.9 4.4 28.3*** 2.9 26.1 3.3 33.0*** 5.7
SOURCE: Authors' calculations using HRS data linked to selected administrative data files from SSA.
NOTES: Sample consists of applicants aged 51 or older but younger than FRA who were denied at the initial level (including DDS reconsiderations).
All results are weighted to account for the complex HRS survey design and the varying probabilities of respondent consent to matching the HRS data to administrative data.
To meet the requirements for using restricted HRS-SSA linked data, we have rounded the percentages to avoid disclosing potentially identifiable information without diminishing the qualitative findings.
Chi-square tests compared the distribution of multinomial outcomes across groups.
*** = statistically significant at the 0.01 level.
a. Indicators of statistical significance are relative to medical (step-2) denials.
b. Indicators of statistical significance are relative to step-4 denials.
c. Includes applicants who appealed an initial denial and did not receive an allowance. The administrative data do not allow us to distinguish that group from initially denied applicants who neither appealed nor reapplied.

Denied Applicants Who Claim OASI Before FRA

In addition to seeking DI benefits through appeal or reapplication, some denied applicants aged 62 or older may claim OASI benefits. For those who claim OASI before FRA, pursuing a DI application is relatively costless; SSA considers the DI eligibility for OASI claimants who report a health condition or impairment that might be significant enough to meet the DI criteria. Actuarially reduced OASI benefits claimed before FRA are an important component of the safety net for older adults with work-limiting health conditions (Leonesio, Vaughan, and Wixon 2003; Bound and Waidmann 2010; Schimmel and Stapleton 2012; Wu and Schimmel Hyde 2018). Even though OASI beneficiaries who are younger than FRA can earn more than the SGA level and retain some benefits,20 we hypothesize that denied DI applicants who ultimately claim retired-worker benefits before FRA are unlikely to reenter the workforce. Moreover, although pre-FRA OASI benefits offer a strong safety-net function and do not have the SGA restrictions that DI benefits do, applicants consider DI benefits to be financially preferable. Because OASI benefits claimed before FRA are actuarially reduced, a monthly OASI benefit claimed at age 62 will be up to 30 percent lower21 than a monthly DI benefit at the same age, and the actuarial reduction will persist until death.

We find that a majority of applicants who were initially denied DI benefits go on to receive DI or OASI benefits before they reach FRA. Table 3 shows that approximately one-third of applicants denied for medical reasons and two-thirds of those denied for work-capacity reasons were observed to receive DI benefits by FRA. Among the cases for whom we did not observe a DI award, more than 70 percent began to receive either OASI or DI benefits at age 62 to FRA, with little difference across the groups of denied applicants (not shown).22 Among those who began to receive benefits from age 62 to FRA, three-quarters received their first payment at age 62, suggesting that OASI claimed at the first possible age accounts for most of the benefits received among this group.

Most of the applicants in our sample were younger than the earliest OASI eligibility age when they applied for DI benefits, perhaps indicating that DI was their best option for income support following disability onset. Nonetheless, the ability to claim OASI at age 62 is relevant to individuals who apply for DI at ages nearing 62, as the initial decision takes months and appeals may take many months or years. Comparing step-4 and step-5 denials reveals that 68 percent of applicants denied at step 4 were aged from 57 to 61 when they applied, compared with 45 percent of applicants denied at step 5 (not shown). Our earlier findings indicated that many of these denied applicants appealed that decision, but for others, the ability to claim OASI benefits at age 62 may in part have driven a decision not to appeal.

A minority of applicants (12 percent) in our sample were aged 62 or older when they applied for DI benefits (not shown). Members of this group may have begun to receive OASI benefits while awaiting their DI decision, or they may have filed a claim for OASI benefits but SSA advised them that they might be eligible for DI benefits because of their limitations. There were slight, although not statistically significant, differences between denied and allowed DI applicants in the shares who were older than 62 at application.

The Employment and Earnings Patterns of Denied Applicants Around the Time of the Application Decision

We now turn to the employment and earnings trajectories of denied applicants in the 5 years before application and 5 years following the initial decision. In Charts 3 and 4, we indicate the application year with t and the decision year with T; for many applicants, those events occur within the same year, and for them, earnings in t and T would thus be the same.

Five years before application, the employment rate among applicants who will be denied for work-capacity reasons is significantly higher than that of applicants who will be denied for medical reasons, although the employment rates of these groups converge by the year of application (Chart 3). In − 5, 85 percent of step-4 denials and 95 percent of step-5 denials were working, compared with 62 percent of those denied for insufficiently severe impairments at step 2 (that is, for medical reasons). As the application date approaches, the employment rates of all three groups drop, reaching 47 percent of step-2 denials, 48 percent of step-4 denials, and 57 percent of step-5 denials in the calendar year of application. The preapplication employment decline is steeper among those denied at steps 4 and 5 than for those denied at step 2. The preapplication employment rates we find for our sample resemble those in SSA (2017a) and Strand and Trenkamp (2015), despite differences in study-population age and timing before application. We omit confidence intervals to preserve the visual clarity of Chart 3; however, we note that the standard error on most of these estimates is relatively large (see Appendix Table A-1).23 We are unable to reject the hypothesis that the likelihood of any earnings is the same in each year for applicants denied at steps 4 and 5. We are able to reject the hypothesis that the likelihood of any earnings is the same for medical and work-capacity denials, but only in the fifth and fourth ysears before denial; after that point, the likelihood of earnings is not statistically different across groups.

Chart 3.
Employment rates (percentages of initially denied older applicants with positive earnings) in the calendar years before DI application and after denial, by reason for denial, 1992–2012
Line chart linked to data in table format.
SOURCE: Authors' calculations using HRS data linked to selected administrative data files from SSA.
NOTES: t = year of application; T = year of initial denial. For many (but not all) applicants, t = T.
Sample consists of applicants aged 51 or older but younger than FRA who were denied at the initial level (including DDS reconsiderations).
See Appendix Table A-1 for tabulations, standard errors, and analogous results for various alternative combinations of covariates.

Although Chart 3 shows that the likelihood of working declines across all groups as the application date nears, the drop for those who will be denied for work-capacity reasons is more precipitous than that for applicants who will be denied at step 2. This pattern is most clearly seen when comparing employment rates 1 year before application (− 1) and 1 year after decision (T + 1). Although the small (and declining) sample size for the postdecision years diminishes the precision of the estimates, applicants denied for medical reasons appear more likely to have earnings in the first few postdenial years than do those denied for work-capacity reasons. By the fifth calendar year after application, when the average age of a denied applicant in our study is from 61 to 63, 14–20 percent of sample members have positive earnings. This finding stands in contrast to Strand and Trenkamp's assessment: Observing younger denied applicants, they found that the likelihood of employment rose for the first 3 years after denial, before tapering off to a level that was similar to each group's employment rate a year before application.

Among older denied applicants who are working in a given year, we find that average earnings for those with step-2 and step-4 denials were much lower after denial than before application (Chart 4). Although applicants with step-5 denials had higher average earnings than other denied applicants, their postdenial earnings were about 15 percent lower than their preapplication averages. Given the small sample size and the small percentage of applicants who work after a denial, the standard errors for the earnings values are high and the results need to be interpreted with caution.24 For example, applicants who were denied at step 5 and had observed earnings in T + 5 numbered around a dozen.

Chart 4.
Mean earnings (among initially denied older applicants with positive earnings) in the calendar years before DI application and after denial, by reason for denial, 1992–2012
Line chart linked to data in table format.
SOURCE: Authors' calculations using HRS data linked to selected administrative data files from SSA.
NOTES: t = year of application; T = year of initial denial. For many (but not all) applicants, t = T.
Sample consists of applicants aged 51 or older but younger than FRA who were denied at the initial level (including DDS reconsiderations).
See Appendix Table A-1 for tabulations, standard errors, and analogous results for various alternative combinations of covariates.

The pool of individuals who return to work after DI denial may be subject to nonrandom self-selection; for example, those who continue to work after denial may have had above-average earnings prior to application and experienced a greater-than-average decline in earnings. The opposite could also be true. To assess this, we considered average (unadjusted) earnings in the years before application among those who worked in at least 1 of the 5 years after denial, and compared those values with average (unadjusted) preapplication earnings of the full group (not shown). No clear pattern emerged to suggest the direction of any selection into the group who returned to work. Applicants who worked after a step-4 denial generally had higher-than-average earnings in the years prior to application. Applicants who worked after denial at step 5 had lower-than-average earnings 3 to 5 years before application, but higher-than-average earnings in the application year and the 2 preceding years. Because we did not explore the hours worked among these applicants, we do not know whether those findings indicate individuals employed in higher-paying occupations or individuals working higher-than-average weekly hours before they applied.

The unadjusted values shown in Charts 3 and 4 do not account for demand-side or supply-side factors that might have affected the employment and earnings trajectories around the time of application. As shown in Tables 1 and 2, there were important differences that could explain the likelihood of returning to work after DI application, and we considered variants of our results that accounted for these factors. For example, applicants denied at step 4 were older on average than those denied at step 5, which could affect the likelihood of finding postdenial work. Results accounting for these observable differences across the groups are available in Appendix Table A-1. We present several variants of our results to show that they are relatively robust to specification changes. In particular, we highlight the changes resulting from the mechanism we use to account for group differences, and for more limited and fuller sets of controls.

In general, the pattern of regression-adjusted results is largely consistent with the pattern we describe based on the unadjusted values shown in Appendix Table A-1—employment rates decline for all denied-applicant subgroups and do not substantially recover in the 5 years after denial. Similar to the unadjusted mean employment rates, we cannot reject the hypothesis that the groups of denied applicants have the same employment rate in the regression-adjusted model in most instances. Average earnings continue to be volatile in a way that limits our ability to draw strong conclusions, and results vary across specifications. Further, if any individual characteristics are correlated with the reason for denial, holding characteristics constant across groups overcontrols for the factors that explain postdenial outcomes. Because the unadjusted results are more intuitive, they are our preferred specification.

Conclusions and Policy Implications

We find that few older DI applicants who were initially denied benefits returned to work and that those who did return to work generally appeared to earn less on average than they had in the years before they applied. This general pattern is consistent with the findings of Strand and Trenkamp (2015), who focused on a younger population that had many more potential working years after a DI denial than did the denied applicants in our sample. As might be expected, we conclude that the likelihood of returning to work is lower among older denied applicants than among younger ones. We also find that older denied applicants exit the labor force earlier than their nonapplicant age-group peers.

Relatively few older workers go back to work following a denial, regardless of the reason for denial. For step-5 denials, disability examiners deemed the applicants to be able to perform other work given their age, education, and work experience, and our descriptive analysis documents characteristics that are consistent with those criteria—they are younger, more educated, and working in what appear to be more skilled jobs than are applicants denied at step 4. We find evidence suggesting that applicants denied at step 5 who returned to work were slightly more likely to earn amounts that were similar to their preapplication earnings than were other applicant groups. Yet, small sample sizes lead us to interpret those results with caution.

After a denial, low employment among applicants appears to be related to appeal of the denial, DI reapplication, or filing an OASI claim. Among initially denied DI applicants aged 51 or older, virtually all who continued to seek benefits—whether DI or OASI—were receiving them by the time they reached FRA. Nearly two-thirds of the DI applicants in our sample who were initially denied for work-capacity reasons ultimately received a DI award after either appealing the initial decision or reapplying. That award rate is twice the rate for those who appealed or reapplied following a denial for medical reasons. Applicants who continue to pursue DI benefits have an incentive not to work at the SGA level, which would make them ineligible. Among denied applicants who do not receive DI benefits after a subsequent appeal or reapplication, more than two-thirds claim OASI benefits between the earliest age of eligibility (62) and their FRA. Although the work disincentives of OASI beneficiaries in this age range are less substantial than are those of DI beneficiaries, few of them work.

Thus, among the many older DI applicants who are initially denied benefits, few go back to work and most become Social Security disability or retirement program beneficiaries before attaining FRA. Two-thirds of DI applicants with initial work-capacity denials ultimately received DI benefits before FRA, and more than two-thirds of initially denied DI applicants who did not receive a subsequent DI allowance opted to receive actuarially reduced OASI benefits before reaching FRA. Many DI benefit allowances follow a lengthy appeal or reapplication. Because a DI award is predicated on not engaging in SGA, denied applicants awaiting a decision on an appeal are unlikely to work at significant levels and could experience financial hardship during a long wait. Over time, the size of SSA's backlog of appealed cases has varied, and one would expect an older denied applicant to be likelier to return to work when the backlog is shorter than when it is longer. This is because older workers who waited 2 to 3 years during a period of large backlogs for their appeal to be decided would have only a year or 2 after that decision and before FRA in which to reconsider work.

Two SSA demonstration projects test the feasibility of early interventions to support individuals with potential work-limiting medical conditions either after impairment onset (Retaining Employment and Talent After Injury/Illness Network, or RETAIN) or after a DI application is denied (Supported Employment Demonstration, or SED). Because the SED focuses on younger applicants (aged 18–50) with behavioral health issues, it is not directly relevant to the population we analyze. Yet the results of that demonstration may offer important insights about whether postdenial targeting of applicants is early enough to prevent labor force exit. RETAIN, which SSA is implementing jointly with the Department of Labor, will target workers after work-limitation onset but earlier in their trajectory toward benefit receipt. For older workers who may already be nearing labor force exit, this approach seems more likely to maintain potential connections to the labor force. However, the RETAIN demonstration is in its early stages and results are still many years away.

Appendix A

Table A-1 presents regression model estimates of the conditional adjusted probabilities (in the case of employment) or means (in the case of earnings), based on the Stata statistical software package's margins command. Results are shown for each of four combinations of covariates, the first of which excludes all covariates (corresponding with the values plotted in Charts 3 and 4). For the second combination, the results are conceptually similar to those in a typical regression, where the group's own covariates are used to generate predicted probabilities. For the third and fourth combinations, we show the conditional adjusted value for each group assuming that it had the same average characteristics as the group of applicants denied at step 4. Stated another way, the conditional adjusted value for each group identifies the outcome if the average applicant who was denied at step 4 had been denied at step 2, 4, or 5, accounting for the covariates described in the narrative. The selection of group 4's characteristics was arbitrary, though the intent of holding constant at one group's characteristics was to minimize the effect of differences in characteristics across groups.

Table A-1. Employment and earnings of older DI applicants from 5 years before application to 5 years after initial decision, by initial outcome: Estimates using alternative combinations of covariates, 1992–2012
Year No covariates a Using each group's average characteristics, full set of covariates Using average characteristics of step 4 denials—
Percent with earnings b Standard error Mean of nonzero earnings ($) Standard error Percent with earnings b Standard error Mean of nonzero earnings ($) Standard error With full set of covariates Excluding covariates related to medical-vocational disability determinations
Percent with earnings b Standard error Mean of nonzero earnings ($) Standard error Percent with earnings b Standard error Mean of nonzero earnings ($) Standard error
  Allowed for medical reasons
5 years before application 94.1 2.5 41,231 3,373 89.6 5.1 40,413 3,414 93.4 5.8 27,209 5,969 87.9 7.8 30,224 6,727
4 years before application 93.7 2.6 41,038 3,574 89.6 5.1 38,960 3,303 93.4 5.8 26,231 5,775 87.2 8.0 30,003 6,732
3 years before application 89.1 3.7 39,677 3,495 94.5 3.9 38,294 3,411 96.6 3.5 25,783 5,822 81.9 10.4 27,646 6,189
2 years before application 84.1 4.2 39,949 2,972 88.0 5.4 35,942 3,558 92.3 6.7 24,199 5,662 73.6 12.5 27,408 6,258
1 year before application 83.1 4.3 34,727 2,681 86.4 5.8 28,243 3,786 91.1 7.5 19,016 4,854 71.5 13.0 24,144 5,570
Application year 72.7 5.0 22,162 2,904 73.0 6.5 19,623 3,378 81.2 12.6 13,212 3,496 57.9 14.7 15,485 4,041
Initial decision year 57.3 5.6 19,042 3,351 48.4 8.1 15,770 2,990 59.7 18.9 10,618 2,937 39.6 14.1 14,463 4,007
1 year after initial decision 17.6 4.2 13,405 4,330 16.1 5.9 15,246 6,480 23.7 14.9 10,265 4,813 11.4 6.3 8,696 3,354
2 years after initial decision 13.5 3.6 15,553 4,863 14.0 5.8 24,183 4,442 20.8 14.1 16,282 4,686 8.1 4.8 11,490 4,141
3 years after initial decision 14.4 3.9 14,355 5,446 17.1 6.9 15,462 7,651 25.0 16.0 10,410 5,614 8.7 5.1 10,077 4,498
4 years after initial decision 12.7 3.7 4,848 1,442 5.7 4.5 13,456 3,034 9.1 9.2 9,060 2,407 7.3 4.5 3,032 1,256
5 years after initial decision 8.4 3.2 5,928 2,055 8.6 5.7 12,047 2,515 13.4 11.9 8,111 1,945 5.5 3.6 3,708 1,702
  Allowed for medical-vocational reasons
5 years before application 96.9 1.5 38,524 1,988 95.0 2.7 32,631 2,229 97.3 2.5 23,729 4,413 93.5 4.6 26,392 5,786
4 years before application 96.7 1.5 37,927 2,062 95.4 2.6 33,147 2,422 97.6 2.3 24,104 4,540 93.1 4.7 26,076 5,727
3 years before application 93.7 2.0 35,477 2,144 95.2 2.3 29,663 2,678 97.4 2.3 21,571 3,967 87.4 7.2 24,885 5,551
2 years before application 88.7 2.6 33,227 2,088 90.3 3.4 28,748 2,804 94.6 4.4 20,905 3,780 78.6 10.2 23,581 5,213
1 year before application 83.3 3.0 28,890 2,146 85.4 3.8 24,997 3,560 91.5 6.5 18,178 3,602 72.6 11.7 19,968 4,556
Application year 67.4 3.6 17,145 2,181 58.1 5.2 22,446 5,887 70.3 15.9 16,323 4,232 52.3 14.1 12,655 3,263
Initial decision year 51.8 3.9 13,014 2,286 39.9 5.1 17,605 7,325 52.8 18.5 12,802 5,119 36.9 13.0 9,418 2,768
1 year after initial decision 18.8 3.2 10,140 3,860 15.5 3.9 16,503 6,021 24.0 14.0 12,001 4,464 11.8 6.1 5,473 3,291
2 years after initial decision 12.6 2.7 5,074 1,107 12.3 3.8 5,035 1,624 19.5 12.5 3,662 1,332 7.9 4.4 2,944 1,104
3 years after initial decision 12.7 2.8 5,355 1,304 13.0 4.1 4,305 1,403 20.4 13.0 3,131 1,179 8.0 4.5 3,260 1,418
4 years after initial decision 10.9 2.7 6,201 1,610 14.2 4.7 6,384 1,538 22.2 13.9 4,643 1,381 6.7 3.9 4,542 2,089
5 years after initial decision 9.3 2.3 10,432 2,193 9.2 3.8 14,240 5,321 14.9 10.7 10,356 4,151 5.9 3.4 6,842 2,490
  Denied for medical reasons (step 2)
5 years before application 61.9 5.9 27,946 3,133 79.3 6.8 29,794 4,613 83.8 6.7 28,630 5,184 68.6 6.0 29,141 3,160
4 years before application 60.1 5.9 27,137 3,003 74.8 6.8 30,714 4,859 79.7 7.0 29,513 5,533 66.6 5.9 29,099 3,165
3 years before application 61.0 5.9 24,611 3,073 81.3 6.2 29,775 4,360 85.6 6.1 28,611 5,118 67.6 6.0 27,725 3,080
2 years before application 57.9 5.9 23,551 3,094 80.5 6.3 26,285 4,300 84.9 6.3 25,257 4,948 64.0 6.1 25,758 3,230
1 year before application 54.1 6.0 21,892 3,435 74.3 7.2 24,421 5,355 79.2 7.8 23,467 5,829 61.7 6.3 23,459 3,869
Application year 46.8 6.0 12,219 3,083 59.7 8.5 9,444 2,044 64.8 9.6 9,075 2,130 53.5 6.7 13,935 4,449
Initial decision year 41.2 6.0 7,737 957 53.9 8.5 7,904 1,241 58.8 9.5 7,595 1,278 46.6 6.9 7,467 1,032
1 year after initial decision 31.4 5.9 10,828 2,593 40.8 9.3 13,826 3,551 45.0 10.6 13,285 3,476 37.5 7.1 11,176 2,653
2 years after initial decision 28.0 6.0 14,934 3,134 36.4 8.8 18,777 2,766 40.3 9.8 18,044 3,084 34.1 6.8 15,535 2,884
3 years after initial decision 30.2 6.5 12,250 3,623 33.2 9.3 18,409 3,584 36.9 10.2 17,690 3,786 32.1 7.1 14,425 3,922
4 years after initial decision 24.2 6.0 17,601 4,526 26.8 8.9 18,891 3,882 30.0 9.4 18,153 4,237 28.9 6.5 17,351 4,066
5 years after initial decision 13.5 4.4 9,731 2,146 10.2 5.8 9,407 1,697 11.6 6.3 9,039 1,911 17.0 5.7 8,385 1,728
  Denied for capacity to perform past work (step 4)
5 years before application 85.3 3.6 28,642 2,454 82.5 6.7 24,286 4,696 82.5 6.7 24,286 4,696 82.9 4.0 30,303 2,927
4 years before application 83.5 3.7 25,796 2,301 84.7 6.4 26,783 3,318 84.7 6.4 26,783 3,318 80.9 4.1 27,845 2,901
3 years before application 78.6 4.1 24,806 2,505 85.6 5.9 25,734 4,031 85.6 5.9 25,734 4,031 80.0 4.1 26,742 3,091
2 years before application 70.4 4.6 24,331 2,497 82.3 6.3 22,594 3,902 82.3 6.3 22,594 3,902 74.0 4.5 24,213 2,844
1 year before application 69.6 4.5 20,469 2,656 75.6 6.2 17,465 5,609 75.6 6.2 17,465 5,609 69.8 4.4 20,903 3,336
Application year 48.4 5.0 13,694 3,004 36.8 7.4 10,353 2,901 36.8 7.4 10,353 2,901 43.5 5.1 16,729 5,105
Initial decision year 34.8 4.7 11,892 3,839 31.3 7.3 6,768 1,516 31.3 7.3 6,768 1,516 33.1 4.9 14,671 6,291
1 year after initial decision 18.1 3.8 16,140 8,152 13.4 5.8 11,421 3,307 13.4 5.8 11,421 3,307 20.4 4.3 18,060 10,118
2 years after initial decision 17.2 3.9 15,744 9,264 15.2 6.4 10,056 3,036 15.2 6.4 10,056 3,036 16.4 4.3 25,852 12,222
3 years after initial decision 16.3 4.1 14,844 8,043 15.0 6.7 5,788 4,123 15.0 6.7 5,788 4,123 15.8 4.5 20,750 9,116
4 years after initial decision 15.6 3.8 6,093 1,705 19.3 7.4 5,811 3,636 19.3 7.4 5,811 3,636 15.4 4.1 5,333 1,375
5 years after initial decision 17.0 4.1 8,399 2,001 12.8 5.0 14,324 4,644 12.8 5.0 14,324 4,644 16.9 4.5 7,490 1,800
  Denied for capacity to perform other work (step 5)
5 years before application 95.2 2.5 30,061 3,721 96.9 2.3 22,353 3,938 97.3 2.1 20,766 4,199 94.1 3.0 21,676 2,890
4 years before application 92.8 3.9 30,400 4,182 97.4 2.2 29,625 6,916 97.7 1.9 27,523 7,045 91.1 4.3 20,900 4,444
3 years before application 83.1 6.2 30,668 5,063 91.5 8.3 33,036 7,268 92.3 7.7 30,691 7,312 79.5 6.9 21,896 5,379
2 years before application 76.1 6.9 27,738 4,898 99.1 0.9 15,312 3,967 99.2 0.8 14,225 3,996 71.4 7.8 18,399 4,183
1 year before application 64.5 7.7 22,470 4,526 88.5 7.5 10,861 3,629 89.6 7.1 10,090 3,491 62.5 8.0 16,095 4,285
Application year 57.0 7.8 11,931 2,999 65.1 11.8 7,420 4,019 67.3 11.9 6,894 3,719 57.1 7.9 9,930 2,887
Initial decision year 41.0 7.6 13,873 4,691 57.6 13.3 4,908 3,408 60.0 13.3 4,559 3,103 44.6 8.0 12,548 4,677
1 year after initial decision 18.9 5.7 24,607 7,639 11.4 7.6 40,333 16,469 12.4 7.8 37,470 14,247 21.6 6.1 25,834 7,539
2 years after initial decision 18.1 5.8 24,780 5,681 16.9 8.1 27,699 18,273 18.4 8.2 25,733 16,096 20.4 6.0 21,870 4,912
3 years after initial decision 19.1 6.0 25,653 6,783 17.5 8.3 30,936 18,111 19.0 8.5 28,741 15,840 20.9 6.2 19,849 6,291
4 years after initial decision 17.0 5.8 25,380 6,405 17.8 8.4 32,108 18,633 19.4 8.5 29,829 16,252 18.6 6.0 17,511 6,123
5 years after initial decision 19.6 6.5 20,985 5,854 17.8 8.4 31,015 18,977 19.4 8.5 28,813 16,616 22.4 7.0 15,567 4,613
SOURCE: Authors' calculations using HRS data linked to selected administrative data files from SSA.
NOTES: Values were developed in a generalized linear model regression with year, group, and year-by-group interactions and sampling weights applied.
Sample consists of applicants aged 51 or older but younger than FRA who were denied at the initial level (including DDS reconsiderations).
All results are weighted to account for the complex HRS survey design and the varying probabilities of respondent consent to matching the HRS data to administrative data.
SEF data are top-coded to the Social Security taxable maximum amount.
Earnings are adjusted to 2012 dollars using CPI.
The full set of covariates includes demand-side factors (the national employment rate in the specified year and the decision year); factors correlated with the decision to work (sex, race, Hispanic origin, marital status, self-reported health status, and self-reported functional limitations 2 years before application); and factors more closely related to the disability determination process, including age, education, and employment-related measures (firm size and job tenure in job held 2 years before application, and working-age years with positive earnings).
a. Values for denied applicants are plotted in Charts 3 and 4.
b. Limited to applicants with SEF data. Sample size and mean nonzero earnings decline in postdecision years because of right-censoring.


1 Technically, the 30 percent maximum actuarial reduction will take effect in 2022, when the first of the workers whose FRA is 67 (those born in 1960 or later) turn 62. In 2020, the maximum reduction for claiming at age 62 is 28.3 percent.

2 We use “initial denial” to refer to a DDS decision in either its first review or a first-level appeal known as a reconsideration. We exclude all subsequent appeals, which are heard by non-DDS entities (an administrative law judge, the Appeals Council, or a federal court).

3 For simplicity, we use “work capacity” throughout the article to refer to step-4 and step-5 determinations, although that term is not SSA's official nomenclature.

4 RFC is not considered if the individual has a severe impairment(s), has no past relevant work, is aged 55 or older, and has no more than a limited education. In such a case, the individual is simply deemed disabled (SSA 2018a).

5 Since July 2012, SSA has allowed examiners to bypass step 4 for applicants whose past job requirements are difficult to determine from submitted information. For applicants found unable to adjust to another job in the economy in step 5, the examiners “return to the fourth step to develop the claimant's work history and make a finding about whether the claimant can perform his or her past relevant work” (SSA 2012a). This approach does not affect applicants during our period of study but may explain some of the differences in findings between our study and those of SSA's Office of the Inspector General, described later.

6 More information about the OIS is available at /occupational_info_systems.html.

7 The OIG report considered DI and Supplemental Security Income applicants aged 18–88 who were denied at steps 4 and 5 in 2013. Strand and Trenkamp focused on DI disabled-worker benefit applicants aged 18–61 who were initially denied at step 5 in 2005. Our study is limited to individuals aged 51 or older but younger than FRA who filed DI applications during 1992–2012.

8 These linkages required permission from the HRS to access its restricted data, which was granted after receiving study approval from an independent Institutional Review Board under contract with Mathematica at the time of the study.

9 For more information on O*NET, see Schimmel Hyde, Wu, and Gill (2018, Appendix B).

10 Using the weights provided in the survey (HRS 2017), we estimate that our sample represents approximately 5.5 million individuals who ever applied for DI disabled-worker benefits after age 50 and before FRA from 1992 through 2012. Because the HRS consent weights were developed as cross-sectional weights at three points as new cohorts of HRS respondents entered the sample (1992, 1998, and 2004), we took the first weight available for each person, which effectively produces a sum of three cross sections over the period that are difficult to align with the flow of older applicants during our study period. Our estimated weighted sample size is significantly smaller than the 13.3 million DI applications filed between 1992 and 2012 by individuals older than 50 (calculated by SSA's Office of the Chief Actuary). SSA counts repeated applications from the same individual and we do not; this accounts for some of the difference in sample sizes. As we will show, repeated filings by older applicants may be common. Further, we restrict our sample to individuals who apply after the first HRS interview, which excludes applicants who filed before that interview (which can occur as late as age 56).

11 The substance of our findings did not change between the weighted and unweighted versions of our analysis.

12 Through the 2004 wave of the HRS, respondents who consented to the data linkage gave access to their records only through the year in which consent was granted. Since 2006, the HRS has asked respondents for prospective permission to link their records for 30 years. The HRS has obtained prospective permission from most respondents who previously offered permission as well as from the majority of new sample members, with the exception of those in early cohorts who were not reinterviewed after 2004.

13 These respondents include 324 members of the 1931–1941 birth cohort (first interviewed by the HRS in 1992), 226 members of the 1942–1947 cohort (first interviewed in 1998), and 255 members of the 1948–1953 cohort (first interviewed in 2004). For respondents in the two oldest cohorts, we are able to observe all DI applications prior to the respondent's FRA (65 or 66, depending on birth year). For members of the 1948–1953 cohort, we are able to observe DI applications only through 2012, when those sample members were aged 59 to 64. For this latter cohort, we do not have complete information on OASI claiming, as none had reached FRA by the end of our observation period, and many had not reached age 62, the earliest age of OASI eligibility.

14 We identified 242 respondents who applied for DI at least once before their first HRS interview but did not subsequently reapply; we omitted those respondents from our analysis.

15 Wixon and Strand excluded reconsideration determinations in their statistics. If their statistics had included reconsiderations, the difference between our allowance rates and theirs would be smaller, but the rate for our sample of older applicants would still be higher.

16 More than 95 percent of our sample members had at least 1 year of earnings data in the SEF. We assumed that an individual did not have taxable earnings in any year with no SEF data. We estimate employment rates (the shares of individuals with earnings) as percentages of the sample members with at least 1 year of earnings data rather than percentages of the full sample.

17 This issue is not exclusive to the CYBF; it applies to the Type of Claim field in the Master Beneficiary Record from which the CYBF is derived.

18 System skills are defined by the Department of Labor's O*NET database as capacities used to understand, monitor, and improve sociotechnical systems. These include judgment and decision making, systems analysis, and systems evaluation.

19 This finding is consistent with the medical-vocational grid's consideration of age, in that the likelihood of allowance, which can occur at step 5 but not at step 4, increases for older applicants.

20 If an individual collects OASI benefits before FRA, his or her benefits are reduced by $1 for every $2 earned in excess of an annual limit (Song and Manchester 2007). In 2020, that limit is $18,240 annually. Any such benefit reductions result in actuarially fair increases in benefits paid to the beneficiary after attaining the FRA.

21 See note 1.

22 As we have discussed, the CYBF data do not allow us to identify whether those who receive an OASI or DI benefit after age 62 ultimately were awarded DI benefits or claimed OASI at an actuarially reduced amount. As an additional reason to interpret these statistics with caution, the data for 15 percent of our sample did not cover all years through age 62 and, for an additional 5 percent, the data extended beyond age 62 but not through FRA. With those cases excluded from consideration, virtually all of those whom we observe as able to claim OASI or DI benefits did so before FRA.

23 For a version of this chart that includes the confidence intervals, see Schimmel Hyde, Wu, and Gill (2018, Figure V.1).

24 For a version of this chart that includes the confidence intervals, see Schimmel Hyde, Wu, and Gill (2018, Figure V.2).


Bound, John. 1989. “The Health and Earnings of Rejected Disability Insurance Applicants.” American Economic Review 79(3): 482–503.

Bound, John, and Timothy Waidmann. 2010. “The Social Security Early Retirement Benefit as Safety Net.” MRRC Working Paper No. 2010-240. Ann Arbor, MI: University of Michigan Retirement Research Center.

Gustman, Alan L., and Thomas L. Steinmeier. 2001. “How Effective Is Redistribution Under the Social Security Benefit Formula?” Journal of Public Economics 82(1): 1–28.

Haider, Steven, and Gary Solon. 2000. “Nonrandom Selection in the HRS Social Security Earnings Sample.” Report No. DRU-2254-NIA. Labor and Population Program Working Paper No. 00-01. Santa Monica, CA: RAND Corporation.

[HRS] Health and Retirement Study. 2017. “Health and Retirement Study Restricted Data Social Security Weights: Data Description and Usage.” Ann Arbor, MI: University of Michigan Institute for Social Research.

Johns, Tom. n.d.SSA's Sequential Evaluation Process for Assessing Disability.”

Johnson, Richard W., Gordon B. T. Mermin, and Dan Murphy. 2007. “The Impact of Late-Career Health and Employment Shocks on Social Security and Other Wealth.” CRR Working Paper No. 2007-26. Chestnut Hill, MA: Center for Retirement Research at Boston College.

Kapteyn, Arie, Pierre-Carl Michaud, James P. Smith, and Arthur van Soest. 2006. “Effects of Attrition and Non-Response in the Health and Retirement Study.” Labor and Population Working Paper No. WR-407. Santa Monica, CA:RAND Corporation.

Leonesio, Michael V., Denton R. Vaughan, and Bernard Wixon. 2003. “Increasing the Early Retirement Age Under Social Security: Health, Work, and Financial Resources.” Health and Income Security for an Aging Workforce Brief No. 7. Washington, DC: National Academy of Social Insurance.

Mann, David R., David C. Stapleton, and Jeanette de Richemond. 2014. “Vocational Factors in the Social Security Disability Determination Process: A Literature Review.” DRC Working Paper No. 2014-07. Washington, DC: Mathematica Policy Research.

Parsons, Donald O. 1991. “The Health and Earnings of Rejected Disability Insurance Applicants: Comment.” American Economic Review 81(5): 1419–1426.

Schimmel, Jody, and David C. Stapleton. 2012. “The Financial Repercussions of New Work Limiting Health Conditions for Older Workers.” Inquiry 49(2) 141–163.

Schimmel Hyde, Jody, April Yanyuan Wu, and Lakhpreet Gill. 2018. “The Benefit Receipt Patterns and Labor Market Experiences of Older Workers Who Were Denied SSDI on the Basis of Work Capacity.” DRC Working Paper No. 2018-01. Washington, DC: Mathematica Policy Research.

Social Security Administration. 2011. “Program Operations Manual System (POMS) DI 25005.001: Determination of Capacity for Past Relevant Work (PRW)—Basics of Step 4 of the Sequential Evaluation Process.”

———. 2012a. “Expedited Vocational Assessment Under the Sequential Evaluation Process: Notice of Proposed Rulemaking (NPRM).” Federal Register 76(177): 56357–56362.

———. 2012b. “Program Operations Manual System (POMS) DI 00115.015: Definitions of Disability.”

———. 2015. “Program Operations Manual System (POMS) DI 25025.005: Using the Medical-Vocational Guidelines.”

———. 2017a. Disability Applications Denied Because of Claimants' Ability to Work. Congressional Response Report No. A-01-17-50222. Baltimore, MD: SSA, Office of the Inspector General.

———. 2017b. “Program Operations Manual System (POMS) DI 25005.025: Past Relevant Work (PRW) as Generally Performed in the National Economy.”

———. 2017c. “Program Operations Manual System (POMS) DI 25005.020: Past Relevant Work (PRW) as the Claimant Performed It.”

———. 2018a. “Program Operations Manual System (POMS) DI 24510.006: Assessing Residual Functional Capacity (RFC) in Initial Claims (SSR 96-8p).”

———. 2018b. “Program Operations Manual System (POMS) DI 24510.005: General Guidelines for Residual Functional Capacity (RFC) Assessment.”

Song, Jae, and Joyce Manchester. 2007. “Have People Delayed Claiming Retirement Benefits? Responses to Changes in Social Security Rules.” Social Security Bulletin 67(2): 1–23.

SSA. See Social Security Administration.

Strand, Alexander, and Brad Trenkamp. 2015. “When Impairments Cause a Change in Occupation.” Social Security Bulletin 75(4): 1–13.

von Wachter, Till, Jae Song, and Joyce Manchester. 2011. “Trends in Employment and Earnings of Allowed and Rejected Applicants to the Social Security Disability Insurance Program.” American Economic Review 101(7): 3308–3329.

Warshawsky, Mark J., and Ross A. Marchand. 2015. “Modernizing the SSDI Eligibility Criteria: A Reform Proposal That Eliminates the Outdated Medical-Vocational Grid.” Mercatus Working Paper. Arlington, VA: George Mason University.

Wixon, Bernard, and Alexander Strand. 2013. “Identifying SSA's Sequential Disability Determination Steps Using Administrative Data.” Research and Statistics Note No. 2013-01. Washington, DCSSA.

Wu, April Yanyuan, and Jody Schimmel Hyde. 2018. “The Postretirement Well-Being of Workers with Disabilities.” Journal of Disability Policy Studies 30(1): 46–55.