The Decline in Earnings Prior to Application for Disability Insurance Benefits

Introduction

The Social Security Disability Insurance (DI) program provides benefits to individuals with disabilities and, in some cases, to their dependent family members. Individuals must accrue sufficient work histories and payroll tax contributions to be eligible for DI benefits. Some researchers, such as Autor and Duggan (2010), argue that the structure of the DI program supports a dependency on benefits that combines with other factors to increase program costs. However, other researchers have argued that many DI beneficiaries are simply unable to work (Bound 1989; Stapleton and others 2008).¹ Providing preapplication employment supports and related services to individuals at risk of applying for DI benefits might improve their economic well-being and lower the number of DI applications and awards.

To assist workers with disabilities effectively, researchers and policymakers might direct support services toward narrowly targeted subgroups (Wittenburg, Mann, and Thompkins 2013). In selecting groups to target for assistance, policymakers must consider factors such as the earnings history or the functional abilities of the prospective target populations. This article investigates one potential factor to consider when planning early intervention efforts: the decline in earnings in the period leading up to application for DI benefits. Although that factor has previously been observed (see von Wachter, Song, and Manchester 2011), this article focuses on the duration of the decline in earnings across disability types. The analyses in this article will serve as a first step in understanding the trends in earnings prior to application and will expand the exploration of how best to target intervention efforts.

Literature Review

Interventions to help maintain employment and earnings may be more effective if they occur prior to application for disability program benefits (Autor and Duggan 2010; Liebman and Smalligan 2013; Gimm, Hoffman, and Ireys 2014). Several initiatives have explored the effectiveness of early intervention. For example, the Demonstration to Maintain Independence and Employment (DMIE), administered by the Centers for Medicare & Medicaid Services in 2006–2009, tested the effectiveness of “wrap-around” services to prevent or delay employment loss and benefit receipt (Gimm and Weathers 2007; Whalen and others 2012). These services included employment-related supports, medical and behavioral services, life and work coaching, and person-centered case management (services varied by state).² In their evaluation of DMIE, Gimm, Hoffman, and Ireys (2014) found statistical evidence that early interventions significantly reduce the likelihood of federal disability-program benefit receipt after 12 months.

Similarly, Killackey, Jackson, and McGorry (2008) evaluated the effectiveness of Individual Placement and Support (IPS), an employment-support initiative in Melbourne, Australia that focused on enabling competitive employment for individuals with a mental illness.³ The authors found evidence suggesting that intervening at earlier stages of mental illness—specifically, at a first episode of psychosis—and incorporating IPS into the treatment regimen led to more jobs acquired and longer employment periods for their participants. Those receiving IPS treatment at the first stage of psychosis also worked more hours per week and relied less on welfare benefits than did the participants receiving the usual treatment regimen.

To date, most of the demonstrations administered by the Social Security Administration (SSA) have focused on employment efforts after award. Wittenburg, Mann, and Thompkins (2013) reviewed demonstrations conducted by SSA before and after Congress passed the Ticket to Work and Work Incentives Improvement Act of 1999. Although the authors concluded that none of the reviewed demonstrations were likely to reduce caseloads or benefit awards enough to reverse program growth, they found that demonstrations targeted narrowly to specific populations showed positive employment impacts. For example, after 24 months in the Mental Health Treatment Study, the targeted treatment group attained 61 percent employment, versus 40 percent for the control group (Frey and others 2011). In the Youth Transition Demonstration (YTD), the extent of the service provided determined the impact on the participants. The two YTD project sites that provided the most generous employment services reported the sharpest increases in employment rates (Fraker and others 2012). With studies suggesting that targeted earlier interventions may be effective, there has been strong interest in reaching potential DI applicants much earlier, before they apply for benefits. For example, SSA is implementing the Early Intervention Mental Health Demonstration (EIMHD) to enable people with disabilities to remain in the workforce. EIMHD will combine aspects of IPS, systematic medication management, and nurse-care coordination for those who allege a mental impairment.⁴

Mamun and others (2011) examined employment rates for beneficiaries who had been on the disability rolls for at least 1 calendar year. The authors found little variation in 2007 employment rates across disability types (from 9.7 percent to 12.9 percent) except for beneficiaries with intellectual disability, who had an employment rate of 15.5 percent. Using linear probability model regressions, the authors also found that beneficiaries with intellectual disability were 2.7 percentage points more likely than the reference group (beneficiaries with nonmusculoskeletal physical disorders or with missing records) to be employed, all else being equal. Mann, Mamun, and Hemmeter (2015) extended this analysis by comparing employment and earnings outcomes across 25 categories of disabilities using logistic regression models. The authors' analyses included probabilities of employment and earnings exceeding the substantial gainful activity (SGA) level.⁵ They found that DI beneficiaries with hearing impairments, intellectual disability, visual impairments, human immunodeficiency virus/acquired immune deficiency syndrome (HIV/AIDS), and cancers were more likely to be employed than were beneficiaries with primary impairments of respiratory diseases, which was their reference group. The authors also reported that DI beneficiaries with primary impairments of anxiety disorders, schizoaffective disorders, endocrine/nutritional/metabolic diseases, back disorders, and affective disorders were less likely to be employed than were DI beneficiaries in the reference group. Studies by Bound (1989) and von Wachter, Song, and Manchester (2011) showed that DI applicants typically experienced decreases in their earnings in the period immediately before application, especially among those whose claims were rejected. Both studies analyzed postapplication employment and earnings trends and found that a portion of rejected applicants were able to find and maintain employment.

Burkhauser, Butler, and Weathers (2001/2002) looked at the influence of policy variables on the timing of DI application. Using data from the Health and Retirement Study, the authors found that certain variables significantly predict the time to application. The authors reported that the median time elapsed from the onset of a work-limiting condition to DI application among working-age individuals is 7 years for men and 8 years for women. They further found that individuals in states with high allowance rates are disproportionately more likely to apply within 1 year of the onset of a work-limiting condition. Other factors affecting the timing of DI application include the size of the benefit (a 20 percent increase in DI benefits reduces the time to application by 1.2 years); state allowance rates (a 20 percent increase in the allowance rate reduces the time to application by 0.88 years for men); and employer accommodations for work-limiting conditions (universal accommodation increases the time to application by 4.36 years for men and 3.76 years for women).

In this article, I analyze DI applicants by type of disability, sex, age group, and claim outcome to see how long individuals experience an earnings decline prior to filing an application. The article seeks to add to the current literature on employment experiences and variations across disability types and other characteristics. It differs from the studies mentioned above in multiple ways. First, Mamun and others (2011) and Mann, Mamun, and Hemmeter (2015) focused on employment after award, while this study examines earnings and employment prior to award. Second, in addition to examining differences across broader diagnosis groups (as in Mann, Mamun, and Hemmeter 2015), this study examines differences across more specific disability categories. Third, although Bound (1989) and von Wachter, Song, and Manchester (2011) mention employment trends prior to award, those studies do not distinguish between disability types, as this article does. Lastly, this study also analyzes variations by sex and age group.

The period of earnings decline prior to DI application is a potential measure of how much time SSA has to offer intervention services to reduce the rate of DI application. Overall, I find differences in the earnings decline periods among the general impairment categories. However, differences in the decline periods are more pronounced when the disability categories are disaggregated at specific levels. The preapplication earnings decline period also differs by sex. Across age groups, the earnings decline period increases with successively older applicant cohorts. I present separate results for DI applicants whose claims were allowed and those whose claims were denied to highlight differences in the earnings-decline patterns for these groups. For denied claims, the type of impairment is based on the disability the applicant alleged, which was not verified by a disability examiner.

Data and Methodology

This analysis uses data from SSA's 2014 Disability Research File (DRF). Created annually, the DRF is a longitudinal file containing data on all disability claims filed with SSA in the previous 10 years; the 2014 DRF thus covers DI and Supplemental Security Income (SSI) applications filed from 2004 through 2013. The DRF combines data from several agency administrative data files, including the Numerical Identification System (Numident) file, for beneficiary/client information; the Master Beneficiary Record, for information on Old-Age, Survivors, and Disability Insurance benefits; the Supplemental Security Record, for information on SSI payments; the 831 File, for information on disability; and the Summary Earnings Record, for earnings histories. This analysis uses DRF records on DI-only or concurrent DI and SSI claims.

The data used in this analysis have some inherent limitations. First, the study examines only the impairment types identified as the primary diagnosis; secondary diagnoses, or any other subsequent diagnoses, are not considered. Second, the primary diagnosis is not always what the applicant alleges; it is rather what SSA and the Disability Determination Service (DDS) consider the applicant's primary disability to be. An applicant may allege more than one disability but be determined by DDS and SSA to be eligible on the basis of only one of the diagnoses. If DDS recognizes more than one disability, the order in which the examiner lists the disabilities determines the diagnosis group to which the individual belongs.⁶ Lastly, because my sample consists of applications filed 2004–2013, different economic conditions may have affected earnings trends across applicant cohorts.

I analyze disability data at two tiers of disaggregation: by general diagnosis and by specific diagnosis. SSA identifies a specific diagnosis in its records with a 4-digit impairment code. To provide comparisons at the broader tier, I categorize these 4-digit impairment codes into 23 groups of general diagnoses (see Appendix Table A-1 for the general-diagnosis classification scheme).

Before conducting my analysis, I adjusted certain values for selected variables. I recoded the earnings of applicants when they were younger than 18 to “missing” to avoid lowering their average earnings based on a child's level of work. I relabeled the specific diagnoses of “autistic disorders” and “schizophrenic and other psychotic disorders” as “autistic disorders and other pervasive developmental disorders” and “schizophrenic, paranoid, and other functional psychotic disorders,” respectively, to match the terminology in published SSA tables. Lastly, I adjusted all earnings to real 2014 dollars using the Consumer Price Index for Urban Wage Earners and Clerical Workers (CPI-W).

Because this analysis aims to facilitate discussions on targeting applicant groups to help them maintain employment and earnings, I include only DI applicants aged 26–55. The lower bound of age 26 includes individuals who joined the labor force, reached their maximum earnings level, and experienced earnings decline at relatively young ages. Setting the upper bound at age 55 (rather than an older age) focuses the analysis on a population with greater labor force attachment and likelihood of return to work.⁷ I exclude from the sample 20,646 applicants whose claims remained pending as of year-end 2013. After applying all restrictions to the data, my sample consists of 8,767,497 individuals.⁸

Personal Characteristics

Table 1 presents selected characteristics of the sample (overall and by claim outcome) with allowance rates. A little less than half (49.0 percent) of claims are allowed. About 70 percent of applicants are aged 40–55. Among all applicants, the two most common diagnoses are diseases of the musculoskeletal system and connective tissue (31.7 percent) and mood disorders (14.0 percent). Autistic disorders, developmental disorders, childhood and adolescent disorders not elsewhere classified, intellectual disability, congenital anomalies, diseases of the blood and blood-forming organs, and diseases of the skin and subcutaneous tissue are among the least prevalent primary general diagnoses, each accounting for less than 1 percent of the sample. The distribution by general diagnosis in my sample is broadly similar to that of SSA statistical publications (for example, SSA 2015, Table 40).

Allowance rates are low for the younger age groups: 31.1 percent for applicants aged 26–29 and 36.8 percent for those aged 30–39. The allowance rate for those aged 50–55 is 64.1 percent, more than double the allowance rate of the youngest applicant cohort. Applicants with a primary diagnosis of neoplasms, intellectual disability, and genitourinary system diseases (such as kidney failure) have the highest allowance rates, at 83.8 percent, 82.6 percent, and 79.1 percent, respectively. Apart from those with an “unknown” diagnosis, applicants with a primary diagnosis of childhood and adolescent disorders not elsewhere classified; developmental disorders; and endocrine, nutritional, and metabolic diseases have the lowest allowance rates, at 20.3 percent, 21.4 percent, and 37.5 percent, respectively.

Analysis

To qualify for DI benefits, an applicant must be unable to work enough to earn at the SGA level, which in 2015 was $13,080 for nonblind individuals. Chart 1 shows that among individuals who applied for DI benefits during 2004–2013, the percentages who had any earnings or had earnings exceeding the SGA threshold in a given year dropped significantly in the period leading up to the year of application.⁹ The red lines of Chart 1 represent the applicants whose DI claims were eventually allowed, and the blue lines represent denied claims. In my sample of individuals who applied during the period 2004–2013, the percentage whose claims were allowed and who earned more than the SGA level during the 3 years prior to application dropped from 68 percent to 63 percent, 53 percent, and finally 22 percent in the year of application.

On average, earnings at the time of DI application are drastically lower than the maximum earnings within the 10 previous years, regardless of diagnosis group. Chart 2 displays the stark differences between an individual's maximum earnings (in the 10-year period leading up to application) and his or her earnings in the year before application by primary general diagnosis. By definition, applicants must earn less than the SGA level to be eligible for DI benefits; therefore, some may choose to reduce their earnings around the time they apply. Average earnings in the year before application accurately measures the decline in earnings because it reflects a year in which applicants do not have to reduce earnings in order to meet the eligibility threshold. In Chart 2, for each diagnosis, the red bar shows the earnings differences for the applicants whose DI claims were allowed. The largest percentage decline in earnings (not labeled in Chart 2) occurred for those who had schizophrenic and other psychotic disorders (65 percent), childhood and adolescent disorders not elsewhere classified (62 percent), mood disorders (57 percent), and autistic disorders (56 percent). On average, applicants with schizophrenic and other psychotic disorders experienced a drop in average earnings from $29,664 at the maximum to $10,398 in the year before application. The smallest percentage drop in earnings was for applicants diagnosed with neoplasms (39 percent); that group experienced a decline in average earnings from about $48,088 at maximum to $29,326 in the year before application.

The blue bars in Chart 2 illustrate the earnings differences for DI applicants whose claims were denied. For many diagnoses, results are similar for both claim outcomes. On average, applicants whose claims were denied and who alleged schizophrenic and other psychotic disorders experienced an earnings drop from $23,532 at maximum to $7,256 in the year before application. Those who alleged childhood and adolescent disorders not elsewhere classified experienced a drop from $24,547 to $8,018, and those alleging other mental impairments had their earnings drop from $27,719 to $9,362. The earnings declines for those three diagnoses were 69 percent, 67 percent, and 66 percent, respectively. Applicants with claims denied who alleged neoplasms experienced the smallest proportional decline in earnings at 47 percent.

Decline Periods

Determining the diagnosis groups that experienced the quickest earnings declines is complicated because of the difficulty of properly identifying the beginning of the decline for those whose earnings did not fall monotonically. For this analysis, I measure the period that begins with an individual's year of maximum earnings (among the 10 years prior to DI application) and ends with the year of application. The period is measured in full years only; I do not account for the month of maximum earnings or application. Hereafter, I refer to that span as the “decline period.”

Knowing the differences in the average decline period between diagnosis types can inform researchers and policymakers who are assessing the selection of target groups and the timing of intervention efforts. For example, the data may reveal that certain groups, on average, experience a quicker earnings decline; therefore, intervening prior to application may not be as feasible for that particular group as it might be for others.

General diagnosis groups. Tables 2A and 2B present the mean decline periods for each primary general diagnosis for applicants with allowed claims and denied claims, respectively. The general-diagnosis rankings for shortest to longest mean decline period are also given. The mean times from maximum earnings to application range from 4.65 to 5.69 years for allowed applicants (Table 2A) and from 5.11 to 5.74 years for denied applicants (Table 2B). Among the allowed applicants, those with intellectual disability experienced an average decline period of 4.65 years, the shortest among the general diagnosis groups (Table 2A). Autistic disorders, congenital anomalies, diseases of the blood and blood-forming organs, and developmental disorders followed, with decline periods ranging from 4.71 to 5.04 years. The longest mean decline periods were experienced by applicants with diseases of the musculoskeletal system and connective tissue (5.52 years); diseases of the circulatory system (5.55 years); diseases of the respiratory system (5.60 years); endocrine, nutritional, and metabolic diseases (5.64 years); and diseases of the digestive system (5.69 years).

On average, applicants with denied claims experienced slightly longer decline periods than did applicants with allowed claims. However, the primary general diagnoses of congenital anomalies, developmental disorders, intellectual disability, and autistic disorders were among the five groups with the shortest mean decline periods (Table 2B), as was the case with allowed claims. The mean decline periods for those groups ranged from 5.11 to 5.27 years. Likewise, the diagnosis groups with the longest mean decline periods among denied-claim applicants were similar to those for allowed-claim applicants.

Specific diagnosis groups. A better understanding of the preapplication earnings trends of DI applicants is provided by examining the decline periods among specific diagnoses. Although the shortest and longest mean decline periods among the general disability groups differ by only about 1 year, the differences are greater between specific diagnoses within each general diagnosis group. There are 239 distinct specific diagnoses listed among the applicants in my full sample; 238 of them are listed among the allowed-claim applicants and 228 are listed among the denied-claim applicants.¹⁰ Table 3 presents the mean decline periods for each of the 10 specific diagnoses with the shortest decline periods among the full sample, the allowed-claim subsample, and the denied-claim subsample. (A table presenting results for all specific diagnoses is available on request from the author: Jackson.Costa@ssa.gov.)

Among all DI applicants, those with the specific diagnoses of organic mental disorders, chromosomal anomalies, and intellectual disability experienced the three shortest decline periods, ranging from 3.65 to 3.96 years. Of the 2,393 allowed-claim applicants who had a general diagnosis of childhood and adolescent disorders not elsewhere classified, only 14 did not have a more specific level of disability identified with a 4-digit impairment code. That group of 14 had the shortest average earnings decline period in the subgroup.¹¹ Allowed-claim applicants with specific diagnoses of chromosomal anomalies, organic mental disorders, intellectual disability, neoplasms, and congenital anomalies ranked next, with decline periods ranging from 3.76 to 4.03 years. Allowed-claim applicants with specific diagnoses of chronic liver disease and cirrhosis, gout, and alcohol addiction disorders¹² had the longest average earnings declines (about 6 years; not shown). The differences between the shortest and longest earnings decline periods are greater among the specific diagnoses than they are at the general-diagnosis level—a finding that could be useful when considering groups to target for interventions or support services.

Denied-claim applicants with the following specific diagnoses experienced the quickest average earnings declines: malignant neoplasms of the thymus, heart, or mediastinum; hereditary hemolytic anemias; malignant neoplasms of other parts of the nervous system; disorders of the metabolism (cystic fibrosis); and developmental and emotional disorders of newborn and younger infants. Those groups experienced average earnings decline periods ranging from 4.41 to 4.77 years. Denied-claim applicants experiencing the longest decline periods (ranging from 6.2 to 6.8 years) were those with substance addiction disorders, malignant neoplasms of the liver and intrahepatic bile ducts, Epstein-Barr hepatitis, anterior horn cell disease, and asbestosis (not shown). Denied-claim applicants experienced longer average decline periods than those experienced by allowed-claim applicants.

The population of some specific-diagnosis groups is small enough that the aggregate positive impacts of an early intervention program may not significantly exceed the resource requirements of recruiting and serving participants. For that reason, Table 4 presents the mean decline periods for each of the 10 most populous specific diagnosis groups among all, allowed-claim, and denied-claim applicants.

Disorders of the back and affective disorders were the two most prevalent specific diagnoses, regardless of claim outcome. Of the 10 most common specific diagnoses among allowed-claim applicants, those with the shortest mean earnings decline periods (ranging from 5.06 to 5.34 years) were anxiety-related disorders, chronic renal failure, organic mental disorders, affective disorders, and late effects of cerebrovascular disease; those with the longest were chronic pulmonary insufficiency and other and unspecified arthropathies (5.70 years). Of the 10 most common specific diagnoses among denied-claim applicants, those with the shortest mean earnings decline periods (ranging from 5.30 to 5.53 years) were fractures of lower limbs; disorders of muscle, ligament, and fascia; anxiety-related disorders; and affective disorders. Those with the longest were osteoarthrosis and allied disorders (5.70 years), diabetes mellitus (5.71 years), and essential hypertension (5.80 years).

Other characteristics. Differences by sex and age provide a fuller understanding of the trends in earnings prior to DI application. Table 5 presents the decline periods for men and women and among four age groups.

Women tended to have shorter average decline periods than did men, but the difference was less than a year. Among allowed-claim applicants, men experienced a mean decline period of 5.54 years and women experienced a mean decline period of 5.19 years. The difference among denied-claim applicants was smaller, with mean periods of 5.70 years for men and 5.40 years for women.

Allowed-claim applicants aged 26–29 at application experienced decline periods averaging 3.72 years, whereas denied-claim applicants in that age group experienced decline periods averaging 4.14 years. The average decline period increased for each successively older age group. Allowed-claim applicants aged 50–55 at the time of application experienced a mean decline period of 5.68 years. Denied-claim applicants of the same age experienced a mean decline period of 6.01 years. The large difference in mean decline periods between the oldest and youngest age groups is to be expected because older applicants have longer exposure to the labor market and thus have higher earnings levels from which a subsequent decline can more sharply differ.

Conclusion

In this article, I have examined the average duration of the period from the year of maximum earnings to the year of application for DI benefits by primary disability diagnosis, restricting the analysis to the 10 years prior to application. Although all applicants experience a significant decline in earnings prior to their application, the speed and severity of the declines differ across general diagnosis groups, and more so across specific diagnosis groups. Decline periods also differ by sex and age group. In my sample, the oldest age group (50–55) experienced a mean earnings decline period almost 2 years longer than that of the youngest age group (26–29).

With further research and analysis, the results of this study—more specifically, the trends in earnings prior to application—could assist in the planning of future early intervention services. Typically, denied-claim applicants experience a decline period slightly longer than that of allowed-claim applicants, and applicants with physical impairments experience longer earnings declines than do applicants with mental impairments. The contrast in decline periods among the different diagnosis groups can inform targeting strategies. For example, providers could design shorter and quicker support services for individuals known to experience shorter decline periods or extended services for those known to experience longer decline periods. Knowledge of the relative length of earnings decline periods could also enhance SSA's Quick Disability Determination and Compassionate Allowance processes, with which the agency identifies and quickly provides benefits to claimants whose medical conditions are particularly severe and demonstrably meet SSA disability standards. Because the number of individuals in each of the diagnosis groups is also an important research consideration, I have presented the mean decline periods for each of the 10 most common specific diagnoses. Among those groups, applicants experienced relatively similar decline periods (between 5 and 6 years) regardless of claim outcome.

Supplemented with further analysis, this study could inform the creation and design of intervention services. Additional research could explore how information on decline periods can be applied to targeting and directing early intervention initiatives. Further research could also expand the study period beyond this article's 10-year earnings histories for 2004–2013 applicants. Economic conditions experienced by individuals who applied in 2004 differed from those experienced by individuals who applied in 2013. For example, the Great Recession (2007–2009) likely affected the earnings of individuals applying in 2007 or later, but would not have had an impact for those who applied earlier. For that reason, it may be worth exploring the earnings decline by application year. In addition, future research could analyze the variance in the decline periods within given diagnosis groups; those with wide variances may require further analysis before decline-period information is used in planning intervention services.

Appendix A