.. CENTER for RETIREMENT RESEARCH at BOSTON COLLEGE WHAT FACTORS EXPLAIN THE DROP IN DISABILITY INSURANCE ROLLS FROM 2015 TO 2019? Siyan Liu and Laura D. Quinby CRR WP 2023-7 May 2023 Center for Retirement Research at Boston College Haley House 140 Commonwealth Avenue Chestnut Hill, MA 02467 Tel: 617-552-1762 Fax: 617-552-0191 https://crr.bc.edu Siyan Liu is a research economist at the Center for Retirement Research at Boston College (CRR). Laura D. Quinby is a senior research economist at the CRR. The research reported herein was pursuant to a grant from the U.S. Social Security Administration (SSA) funded as part of the Retirement and Disability Research Consortium. The findings and conclusions expressed are solely those of the authors and do not represent the views of SSA, any agency of the federal government, or Boston College. Neither the United States Government nor any agency thereof, nor any of their employees, makes any warranty, express or implied, or assumes any legal liability or responsibility for the accuracy, completeness, or usefulness of the contents of this report. Reference herein to any specific commercial product, process or service by trade name, trademark, manufacturer, or otherwise does not necessarily constitute or imply endorsement, recommendation or favoring by the United States Government or any agency thereof. The authors thank Lingfei Liu and Michael Wicklein for excellent research assistance. © 2023, Siyan Liu and Laura D. Quinby. All rights reserved. Short sections of text, not to exceed two paragraphs, may be quoted without explicit permission provided that full credit, including © notice, is given to the source. About the Center for Retirement Research The Center for Retirement Research at Boston College, part of a consortium that includes parallel centers at the National Bureau of Economic Research, the University of Michigan, and the University of Wisconsin-Madison, was established in 1998 through a grant from the U.S. Social Security Administration. The Center's mission is to produce first-class research and forge a strong link between the academic community and decision makers in the public and private sectors around an issue of critical importance to the nation's future. To achieve this mission, the Center conducts a wide variety of research projects, transmits new findings to a broad audience, trains new scholars, and broadens access to valuable data sources. Center for Retirement Research at Boston College Haley House 140 Commonwealth Avenue Chestnut Hill, MA 02467 phone: 617-552-1762 Fax: 617-552-0191 https://crr.bc.edu Affiliated Institutions: The Brookings Institution Mathematica - Center for Studying Disability Policy Syracuse University Urban Institute Abstract In 2015, the number of individuals receiving Social Security Disability Insurance (DI) benefits began to drop for the first time in two decades. This drop was caused by a wave of terminations, as beneficiaries aged into the Old-Age and Survivors Insurance (OASI) program, combined with a steep decline in the incidence rate (the number of new DI awards relative to the insured population). Yet, the forces driving down the incidence rate remain poorly understood. Prior studies suggest that three factors could have played a role: 1) demographic shifts due to population aging; 2) a strong economy following the Great Recession; and 3) policy changes at the Social Security Administration (SSA). Using data provided by the SSA's Office of Disability Programs, this study examines how each of these factors contributed to the drop in the incidence rate. The paper found that: e A strong economy accounted for about half of the drop in the incidence rate. e Policy changes - including a stricter process for awarding benefits on appeal - also accounted for a significant share. e Population aging put slight upward pressure on the incidence rate. e In terms of the total number on the disability rolls, the impact of aging on terminations far exceeded its impact on new awards. The policy implications are: e With the finances of DI now on a stronger trajectory, the time may have come to somewhat rebalance the goals of DI from encouraging labor force participation to protecting vulnerable people. o Congress may want to consider merging the DI and OASI trust funds. Introduction In 2015, the number of individuals receiving Social Security Disability Insurance (DI) benefits began to drop, reversing an upward trend that had persisted for two decades. Policymakers have wondered whether this downward trend reflects a permanent shift, especially since the reversal has had large positive implications for the program's finances.! Fundamentally, this recent drop in DI rolls is due to an acceleration of terminations, as beneficiaries age into the OASI program, combined with a steep decline in the incidence rate (the number of new DI awards relative to the insured population). Yet, the forces driving down the incidence rate remain poorly understood. Prior studies suggest that three broad factors could be playing a role. First, population aging may have reduced the number of DI applications as workers instead claimed their retirement benefits. Second, a strong economy following the Great Recession made DI less attractive to prospective applicants with some residual work capacity. And third, policy changes at the Social Security Administration (SSA) - notably, the closure of field offices and retraining of Administrative Law Judges (ALJs) in 2010 - increased the difficulty of applying for benefits and reduced the share of applicants who were accepted. The goal of this study is to determine what share of the drop in the incidence rate, from 2010 to 2019, is attributable to each of these three factors. Using data on applications and awards provided by the SSA's Office of Disability Programs, we find that a strong economy and policy change drove the recent drop. Although population aging is reducing the DI rolls through its impact on terminations, it is also putting slight upward pressure on the incidence rate. Currently, the number of terminations exceeds the number of new awards, so overall the DI rolls are declining. The rest of this paper proceeds as follows. The next section provides background on the design of the DI program. The third section motivates our analysis by highlighting trends in the number of beneficiaries over the past 30 years and reviewing what is known about these trends. The fourth section introduces factors that have been proposed to explain the decline in the 11n 2015, when DI rolls were at their peak, the Social Security Trustees Report projected that the DI fund would deplete its reserves in 2016. In response, policymakers temporarily reallocated a portion of the Social Security payroll tax from the retirement program to the disability program. This infusion of revenue, combined with the falling DI rolls, greatly improved the DI program's financial position. The 2023 Trustees Report projected that the fund would never deplete its reserves over the 75-year horizon (U.S. Social Security Administration, 2015 and 2023a). incidence rate. The fifth section outlines the data and methodology for our analysis, while the sixth section displays the results. The final section concludes that demographic shifts, cyclical forces, and policy change have all played a meaningful role in the recent trajectory of the DI program. Background The DI program is intended to provide a basic level of income to people who cannot work due to disability or illness.? In practice, however, the design of the program reflects a tension between twin goals: on the one hand, to protect vulnerable people; and on the other, to encourage labor force participation. To be insured under the program typically requires at least 10 years of work history, with at least five of those years having occurred in the last 10 before disability onset.? Insured workers who experience a disability must undertake a lengthy application process in order to be considered for benefits. First, an SSA field office confirms the worker's insured status and checks that the worker has not engaged in Substantial Gainful Activity (SGA) during the past year. In 2023, SGA is defined as earning more than $1,470 per month.* Next, a medical examiner at a state-administered Disability Determination Services (DDS) office conducts a medical review. At this stage, the worker must meet several criteria. The disability must be expected to last for at least a year or result in death; and it must either be on SSA's "listing of impairments" (conditions that immediately qualify the worker for DI), or it must be severe enough to preclude the worker from performing any job in the national economy (with consideration for age, education, and work history, but without regard for geography).' Workers who are denied benefits have the option to appeal, but the appeals process depends on the applicant's state of residence. Those living in one of 10 "prototype" states proceed directly to a hearing before an Administrative Law Judge (ALJ).® Those living in "non- 2 For an excellent introduction to the DI program, see Maestas, Mullen, and Strand (2021). 3 Workers who experience a disability before age 31 can qualify with fewer years of work, and the Supplemental Security Income (SSI) program pays benefits to children who have a disability and live in a household with limited income or resources. 4 SGA increases to $2,460 for workers who are blind. 3 Workers who meet these criteria are awarded DI benefits starting on the sixth month after disability onset - with initial benefits enhanced to account retroactively for the application period - and are eligible for Medicare after a two-year waiting period. 6 The prototype states are: Alabama, Alaska, California (LA North and West only), Colorado, Louisiana, Michigan, Missouri, New Hampshire, New York, and Pennsylvania. prototype" states appeal for reconsideration to the same DDS office that made the initial decision. If they are denied a second time, they can obtain a hearing before an ALJ. If the ALJ still denies benefits, then the applicant can take their case to the Appeals Council, or even to federal court, but most do not. Consequently, the processing time for SSDI applications can be lengthy. Those whose initial applications are accepted typically wait less than five months for a decision (around one-third of applicants); but rejected applicants who appeal (around half of all applicants) often wait much longer." Once on the DI rolls, beneficiaries receive a monthly benefit equivalent to their Primary Insurance Amount under the Old-Age and Survivors Insurance (OASI) program, without an actuarial reduction for early claiming. Few beneficiaries ever return to work despite policies encouraging them to do so.® Instead, they leave the program upon death or at their Full Retirement Age (FRA) when they automatically transfer to the OASI program.' Trends in DI Rolls from 1990 to the Present From 1990 to 2015, the number of DI beneficiaries steadily increased due to three factors (see Figure 1). First, policy reforms in 1984 expanded the definition of disability and gave applicants and medical providers more influence over the decision process. Second, disability rates rise with age, and the baby boom population was aging into the more lenient eligibility criteria for benefits. Lastly, the secular rise in female labor force participation increased the fraction of women eligible for benefits, and they too aged into the more lenient eligibility criteria. At the same time, a strong labor market during much of this period put countervailing pressure on the number of new applications.!® Ultimately, the number of new DI awards each year always exceeded the number of beneficiaries leaving the program due to death or retirement (see Figure 2). Recent data suggest that the trajectory of the program has shifted. First, the early 2000s saw an acceleration of beneficiaries leaving the DI program, increasingly aging into the OASI program. Second, and more importantly, the number of new DI awards has been dropping 7 Autor et al. (2015); and Maestas, Mullen, and Strand (2021). & Maestas (2019). In 2019, less than one percent of DI beneficiaries left the program because they returned to work (Social Security Administration 2022). ? The amount of benefits received does not change at the Full Retirement Age, only the source of funding for those benefits. 101 jebman (2015) and Technical Panel on Assumptions and Methods (2015). continuously since 2010. Had the number of new awards continued to grow at its prior trajectory, then the DI rolls would not have gone down after 2015 even with more people leaving the program. This drop in new awards is not due to a contraction of the insured population, which actually grew by almost 3 percent from 2010 to 2019. Instead, it is due to a decline in the incidence rate, or the likelihood that eligible workers apply for and are awarded benefits (see Figure 3).11 What Factors Might Explain the Declining Incidence Rate? A number of prior studies have highlighted factors that could be contributing to the steep decline in the incidence rate.!> These factors can be grouped into three broad categories: Demographic Shifts As discussed previously, the retirement of the baby boomers is accelerating the rate at which DI beneficiaries leave the program, leading some to speculate that it might also be affecting the number of new applicants. In actuality, population aging is also putting pressure on the incidence rate, but in an upward direction. Specifically, the average age of the population targeted by DI is still rising, as mid- and late-boomers have not yet reached full retirement. '3 Moreover, application rates increase with age (see Figure 5).!* Consequently, the average age of DI applicants has increased since 2009 (see Figure 6).1> An older applicant pool also implies a higher allowance rate for benefits, since applicants above age 50 are subject to more lenient eligibility criteria. Ultimately, population aging is affecting the DI rolls through two channels that currently offset each other: more recipients leaving the program and upward pressure on the incidence rate. The question for this study is which channel dominated from 2010 to 2019? 1 The insured population has climbed steadily since the 1990s, so any recent reverse in the incidence rate is attributable to the number of new DI awards. 12 See U.S. Social Security Administration (2019) for a summary of these potential explanations. 13 See Appendix Figure Al. We define the eligible population as 18 to 64 year-olds who are not yet receiving Social Security benefits. 14 We calculate the age gradient in DI applications using data on the number of applications by age provided by the SSA's Office of Disability Policy, and eligible population counts by age from the Current Population Survey. For simplicity, Figure 5 shows the age gradient in DI applications in 2019; Appendix Table A2 calculates the gradient separately for each year between 2001 and 2019. Although the baseline rate of DI application varies over the business cycle, the difference between age groups is reassuringly constant over time. 15 Part of the increase is also due to the business cycle, since younger, healthier workers stay in the labor force when unemployment is low. Some have also pointed to the changing industry composition of the workforce as an explanation for the declining incidence rate. The shift from manufacturing to service industries may mean that fewer workers end up with health conditions that qualify for DI. Figure 7 uses the Current Population Survey to plot the share of workers in physically intensive industries from 1990 to 2019.16 The physically intensive share declined during the 1990s and early 2000s - from 35 to 27 percent - but leveled out in 2010 and has actually increased slightly again.!" So, while the decline in manufacturing might have played a significant role in earlier years, it is not relevant during the period that we study. Business Cycle Effects The most obvious explanation for the decline in the incidence rate is the business cycle: since many workers with disabilities retain some work capacity, the long and costly DI application process became less attractive when unemployment rates dropped following the Great Recession.!® Indeed, Figure 4 highlights the counter-cyclical nature of the DI program, with applications and unemployment tracking each other closely.!® As noted earlier, prior research has long documented a relationship between cyclical unemployment and DI applications.?' Most studies use state-level data on DI applications and exploit local variation in the business cycle from the 1980s through the Great Recession. They all find that DI 16 Physically intensive industries include agriculture, forestry, fisheries, mining, manufacturing, construction, transportation, communications, and other public utilities. 17 See Appendix Figure A2. As expected, the decrease is driven by the well-documented decline in the manufacturing sector; however, manufacturing employment has stabilized since 2010. Charles, Hurst, and Schwartz (2019) show similar trends in manufacturing employment and provide an overview of potential factors driving the changes. 18 A strong economy also shifts the applicant pool toward those with more severe health conditions, raising the allowance rate (Cutler, Meara, and Richards-Shubik 2012; Liebman 2015; Maestas 2019; and Maestas, Mullen, and Strand 2021). 1% Throughout, we use SSA fiscal years as our unit of time and measure applications for DI with the number of claims that receive an initial determination, either favorable or unfavorable, in that fiscal year. In recent years, the number of claims that are dismissed on technical grounds due to fraudulent or incomplete files has increased substantially (authors' calculations from the Annual Statistical Report on the Social Security Disability Insurance Program 2020; and Foote, Grosz, and Rennane 2018). Additionally, to account for processing time, determinations made in the first three months of a fiscal year are assumed to represent applications from the prior fiscal year. Similarly, the unemployment rate is measured on a fiscal-year basis. 20 Stapleton et al. (1988); Cutler, Meara, and Richards-Shubik (2012); Liebman (2015); Maestas, Mullen and Strand (2015); and Maestas, Mullen, and Strand (2021). applications are positively associated with the unemployment rate, although the estimates vary due to the different periods studied.?! Policy Change During the period considered, the SSA made two notable changes to the DI program that could have reduced both the number of applications and the approval rate.?> Most noticeable to the public, budgetary pressures forced the SSA to close about seven percent of its field offices between 2001 and 2013 (see Figure 8). Although these closures had already begun gradually before 2010, the rate of closure picked up significantly between 2010 and 2013. Since field offices are an important source of in-person assistance and program information, the closures increased the cost of applying and resulted in an 11-percent drop in DI applications in affected zip codes, and an even larger drop in benefit awards (Deshpande and Li 2019). Moreover, in 2010 the SSA undertook a comprehensive retraining of ALJs to improve consistency in their decision making and reduce appellate approval rates. This retraining did not affect initial allowance rates, which stayed largely flat between 2010 and 2019, but corresponds with a steep decline in the final allowance rate (see Figure 9). The exact effect of ALJ retraining on the incidence rate has not yet been determined in the literature.?* In addition to depressing the allowance rate directly, a tightening of allowance rates could also discourage new applications if marginal candidates believe that their cases have a low likelihood of success.?* Ultimately, the goal of this study is to determine the relative importance of these three factors - population aging, business cycle, and policy change - in explaining the recent drop in 21 Appendix Table A1 compares the estimates from these papers. Additionally, Maestas, Mullen, and Strand (2021) use restricted individual-level data to establish a relationship between the state unemployment rate and DI awards during the Great Recession. 22 Non-DI policy changes may also affect DI applications. Some argue that the Affordable Care Act (ACA) reduced reliance on DI by expanding access to health insurance. However, ACA Medicaid expansions occurred after the drop in the incidence rate, and evidence of their impact is ambiguous at best (see Anand et al. 2019; Maestas, Mullen, and Strand 2014; and Schmitt, Shore-Sheppard, and Watson 2020). Similar arguments can be made regarding recent expansions of Temporary Disability Insurance programs at the state level (Autor et al. 2013; Autor, Duggan, and Gruber 2014). 23 Suggestive evidence indicates a 10-percentage-point decrease in the appellate allowance rate (Ray and Lubbers 2015; U.S. Social Security Administration 2019). 24 The estimated elasticity of DI applications with respect to the approval rate ranges between 0.2 and 0.6. See Bound and Burkhauser (1999) for a review of this literature. More recently, Lahiri, Song, and Wixon (2008) find little effect of state-level variation in allowance rates on applications; other studies focus on the elasticity of applications with respect to benefit levels (Low and Pistaferri 2015). the incidence rate. The next section describes the data and methodology used to perform this decomposition. Data and Methodology Conceptually, we decompose the drop in the incidence rate by taking the level change in each factor of interest (such as the unemployment rate or number of field offices) and multiplying that change by the marginal impact of each factor on awards. Practically, we implement this approach in four stages. The first stage estimates the impact of population aging. We begin by calculating age- specific incidence rates in 2010 using administrative data provided by the SSA's Office of Disability Programs.?'> Then we re-weight the age-specific rates by the share of the eligible population in each age group in subsequent years. This exercise yields the counterfactual incidence rate if all the factors, except aging, had remained at their 2010 levels.? The second stage estimates the additional impact of the business cycle in two steps. The first step uses regression analysis to estimate how the unemployment rate affects DI applications. The SSA's Office of Disability Programs provided administrative data on applications, by state and year, between 1990 and 2019.27 We combine these administrative records with eligible population counts and unemployment rates - by state and year - from the 1990-2019 CPS.?® The CPS also provides demographic control variables, although these turn out not to be very important in the estimation. With these data, we run an OLS regression at the state-year level: 25 We calculate incidence rates for the following eight age groups: 18-29, 30-39, 4044, 45-49, 50-54, 55-59, 60-61, and 62-64. Population age shares are also calculated for each of the eight age groups in the eligible population, excluding those who are younger than 18 or older than 64. 26 The counterfactual incidence rate is equivalent to the age-adjusted incidence rate in U.S. Social Security Administration (2023b), using the 2010 age group distribution as the base for adjustments. 27 Note that we cannot use the same regression methodology to directly estimate the impact on awards because many initial rejections go through two levels of appeal, and the outcome is not tracked at the state level in the data provided by SSA's ODP. 28 The eligible population is defined here as 18 to 64 year-olds not yet receiving Social Security benefits,. We choose this denominator, rather than the insured population as measured by the SSA, for two reasons: 1) to maintain consistency with the control variables calculated in the CPS; and 2) to address the unavailability of state-level insured population data. Our eligible population estimates are higher than the insured population, since we do not exclude those with an insufficient work history. Reassuringly, the post-2010 national trends of these two measures look similar (see Appendix Figure A3). Rt,s = a+ fllUt,s + BZXt,s +cs+ &5 ¢y Where R, is the DI application rate in year ¢ and state s and U, ¢ is the local unemployment rate.?' The regression also includes state fixed effects (c) and controls for other characteristics (X¢,s), such as the age composition of the eligible population in the state, industry mix, education level, female labor force participation (LFP), and race. The hope is that 8, captures the causal relationship between DI applications and unemployment. Of course, this assumption will not hold if other omitted variables also changed during the analysis period, or if the effect of unemployment varies over time. Our approach to both of these issues is to run many versions of the regression, testing the sensitivity of our coefficient to the time period examined. The second step uses the regression results to estimate how the decline in unemployment translated to a drop in the incidence rate. We multiply 8, by the level change in unemployment nationally between 2010 and year ¢: AApplicationratey, = 1 (Ur - Uzp10) 2) The resulting drop in applications is then multiplied by an allowance rate to estimate the impact of falling unemployment on the final incidence rate. Alncidence ratey, = AApplication ratey . » Allowance rate, 3) A key decision at this stage is what allowance rate to choose. We do not want to take the observed allowance rate between 2011 and 2019 because this reflects the impact of ALJ retraining. For simplicity, we use the observed allowance rate in 2010 (57.5 percent) with an upward adjustment to reflect the impact of population aging since, as discussed previously, older applicants are subject to more lenient eligibility criteria.3° 29 Maestas, Mullen, and Strand (2021) run a very similar regression with lagged as well as contemporaneous unemployment. Since they find that most of the association is with contemporaneous unemployment, we adopt this specification for parsimony. 30 Specifically, we combine our estimates from the first stage with data on DI applications by age to find the allowance rate by age in 2010. We then re-weight the age-specific allowance rates by the share of the population in each age group in subsequent years. This counterfactual allowance rate may still be an underestimate since individuals who apply for DI during a tight labor market tend to be in worse health and are approved for benefits at a higher rate (Coe and Rutledge 2013; Cutler, Meara, and Richards-Shubik 2012; Maestas, Mullen, and Strand 2015; and Maestas, Mullen, and Strand 2021). The third stage of the analysis estimates the decline in the incidence rate due to SSA policy change. To begin, we focus on field offices because Deshpande and Li (2019) have already established the marginal impact of closures. Specifically, the authors find that each closure reduced the number of DI awards by 15.5 percent in the zip codes directly affected. Hence, the percentage decrease in the incidence rate due to field office closures can be modeled as:3! % Alncidence ratefieiq offices,t = ¥ 5010(0.155 * number of closures, * share of population af fected by each closure,) ") Multiplying this percentage change in the incidence rate by the observed level in 2010 yields the percentage-point decrease in the incidence rate due to field office closures. Data on the number of field office closures is drawn from Figure 8; we use summary statistics from Deshpande and Li (2019) to estimate the share of the total population affected by each closure.3? The final policy to be accounted for is ALJ retraining. The challenge here is that we lack convincing empirical evidence on the impact of this policy on final awards.3* Consequently, we assume that any residual difference between the actual observed incidence rate and the counterfactual incidence rate - accounting for the first three factors already modeled - is the effect of ALJ retraining. We implement this approach by estimating a counterfactual DI incidence rate if population aging, the unemployment rate, and the number of field offices had remained at their 2010 levels: Incidence ratecoynterfactuar,e = Incidence ratezgo + Y3-1AIncidence rate,; (5) 3! This approach slightly underestimates the impact of field office closures since Deshpande and Li (2019) also find spillover effects in areas serviced by neighboring field offices, due to longer wait times and administrative congestion. However, it also overestimates the impact if discouraged applicants eventually apply at another field office. 32 Specifically, Deshpande and Li (2019) provide the number of directly affected, indirectly affected, and unaffected zip codes in their analysis, as well as the average population of each type of zip code in 2000. By drawing on these statistics, we implicitly assume that the share of the population affected by each new closure does not change over time. 33 Simply comparing the allowance rate before-and-after 2010 might not give the correct answer because, as mentioned above, the allowance rate also fluctuates with the business cycle. And we cannot use regression analysis to control for the business cycle, as we did earlier for applications, because appeals are determined on a national basis, rather than state-by-state. Where x denotes each of the three factors listed above and Alncidence rate, represents the change in the incidence rate due to each factor. The impact of ALJ retraining can then be estimated as a residual: Alncidence ratey,;, = Incidence rate,pgerveq,r - Incidence ratecoynierfactual 6) While this approach has the advantage of simplicity, it will overstate the importance of ALIJs if other factors not considered in this study are also driving down the incidence rate. Results Before turning to the main result, Table 1 presents regression results for the relationship between the unemployment rate and DI application rate. Column (1) shows results for the full time period covered by our data (1990-2019), while columns (2) through (5) consider only expansionary periods (1992-2000; 2002-2007; 2010-2019; and all expansion years). Overall, we find that a one percentage-point increase in the unemployment rate is associated with a 0.033 percentage-point increase in the DI application rate.3* Changing the time period hardly changes this result, and our point estimate is consistent with prior literature.' Figure 10 presents the main result: how much of the drop in the incidence rate is attributable to population aging, the business cycle, field office closures, and ALJ retraining. Specifically, the figure plots counterfactual incidence rates from 2010 to 2019, holding the factors at their 2010 levels and then changing each factor sequentially. For instance, the black line shows that the incidence rate would have risen by 0.02 percentage points due to population aging, between 2010 and 2019, if all the other factors had stayed constant. The red line then adds the impact of the business cycle, which decreased the incidence rate by 0.14 percentage points. The tan line incorporates field office closures, decreasing the counterfactual rate by another 0.01 percentage points.>® Lastly, we assume that the -0.13 percentage-point difference 34 The mean DI application rate was 0.84 percent during this period. 35 We also test the robustness of our estimates with an alternative specification used in prior studies: regressing the log unemployment rate on the log of the DI application rate. We find that the DI application rate increased by 0.3 percent for every 1 percent increase in the unemployment rate (see Appendix Table A2). Appendix Table A3 benchmarks our results against prior literature. 36 The effect of field office closures is small because they only affected about three percent of the population during our analysis period. 10 between the tan line and the actual observed incidence rate reflects the effect of ALJ retraining. Ultimately, the business cycle and a lower benefit allowance rate emerge as the two most important factors driving down the incidence rate in recent years. 37 One final question is whether the increase in new DI awards due to population aging exceeds the number of age-related terminations. In 2019, we estimate that aging increased the number of new awards by around 30,000. Meanwhile, an additional 186,000 DI beneficiaries transitioned to the OASI program that year, relative to 2010. Consequently, population aging is currently reducing the DI rolls overall. The magnitude of this downward pressure - around 156,000 fewer beneficiaries in 2019 - is still slightly less than the impact of the business cycle or ALJ retraining (around 200,000 fewer beneficiaries in 2019 due to each factor). Conclusion Since 2015, the DI rolls have steadily dropped due to an acceleration of terminations as beneficiaries age into the OASI program, and a steep decline in the incidence rate. Since this dramatic reversal has helped improve the program's finances, a key question facing policymakers is whether the decline is likely to persist? This study decomposes the recent drop in the incidence rate into the portions attributable to population aging, the business cycle, and policy changes at the SSA. We find that a strong economy following the Great Recession accounts for about half of the drop in the incidence rate, and policy change accounts for the rest. In particular, the retraining of Administrative Law Judges, which reduced the benefit allowance rate, seems to have had a significant impact. Although population aging is reducing the DI rolls through its impact on terminations, it is also putting slight upward pressure on the incidence rate. Currently, the impact on terminations outweighs the impact on new awards, so the overall effect on the DI rolls is negative. 37 Our result is consistent with Technical Panel on Assumption and Methods (2019)'s finding that the improving economy and decreasing allowance rates both contribute to the decline in incidence rate. Although the exact numbers are somewhat sensitive to the underlying modelling assumptions, the conclusion holds for a reasonable range of parameters. For example, we re-estimated the counterfactuals assuming a stronger relationship between unemployment and DI applications (specifically we use the estimate from Table 1 for expansion years); as well as the weaker relationship found in Maestas, Mullen, and Strand (2021). We also tested the sensitivity of the results to different allowance rates. These sensitivity tests are available from the authors upon request. 11 These results suggest that demographic shifts, cyclical forces, and policy change have all played a meaningful role in the recent trajectory of the DI program. Of course, strong employment has always reduced DI incidence. The difference between the current period (2010- 2019) and the earlier one (1990-2009) is that policy changes are reinforcing the cyclical effect today whereas in the earlier period population aging, the rise in female labor force participation, and the expanded definition of disability were often counteracting the employment effect. As always, these results are subject to caveats. Our results would overstate the importance of ALJ retraining if other factors not considered here are also driving down the incidence rate. And, more importantly, we end our analysis before the COVID-19 pandemic. Although the incidence rate continued to decline during the pandemic, economic conditions, population health, and the policy environment also changed markedly. Most notably, SSA closed all its field offices for a period of two years, coinciding with a sharp drop in DI applications. Hence, the future trajectory of the DI program will depend on the longer-run implications of the pandemic as well as the other factors explored in this study. 38 Recessionary periods in the early 1990s, early 2000s, and 2008-2009 were notable exceptions, with high unemployment driving up the incidence rate. 12 References Anand, Priyanka. 2019. "The Impact of Affordable Care Act Medicaid Expansions on Applications to Federal Disability Programs." Forum for Health Economics and Policy 21(2): 1-20. Bound, John and Richard V. Burkhauser. 1999. "Economic Analysis of Transfer Programs Targeted on People with Disabilities." Handbook of Labor Economics 3: 3417-3528. Charles, Kerwin Kofi, Erik Hurst, and Mariel Schwartz. 2019. "The Transformation of Manufacturing and the Decline in U.S. Employment." NBER Macroeconomics Annual 2018 33(1): 307-372. Coe, Norma B. and Matthew R. Rutledge. 2013. "How Does the Composition of Disability Insurance Applicants Change across Business Cycles?" Working Paper 2013-5. Chestnut Hill, MA: Center for Retirement Research at Boston College. Cutler, David M., Ellen Meara, and Seth Richards-Shubik. 2012. "Unemployment and Disability: Evidence from the Great Recession." Working Paper No. NB 12-12. Cambridge, MA: National Bureau of Economic Research. Deshpande, Menasi and Yue Li. 2019. "Who is Screened Out? Application Costs and the Targeting of Disability Programs." American Economic Journal: Economic Policy 11(4): 213-248. Foote, Andrew, Michel Grosz, and Stephanie Rennane. 2018. "The Effect of Lower Transaction Costs on Social Security Disability Insurance Application Rates and Participation." Journal of Policy Analysis and Management 38(1): 99-123. Liebman, Jeffrey B. 2015. "Understanding the Increase in Disability Insurance Benefit Receipt in the United States." Journal of Economic Perspectives 29(2): 123-50. Low, H. and L. Pistaferri. 2015. "Disability Insurance and the Dynamics of the Incentive Insurance Trade-off." American Economic Review 105(10): 2986-3029. Maestas, Nicole. 2019. "Identifying Work Capacity and Promoting Work: A Strategy for Modernizing the SSDI Program." The Annals of the American Academy of Political and Social Science 686(1): 93-120. Maestas, Nicole, Kathleen J. Mullen, and Alexander Strand. 2015. "Disability Insurance and the Great Recession." American Economic Review: Papers and Proceedings 105(5): 177- 182. Maestas, Nicole, Kathleen Mullen, and Alexander Strand. 2015. "Disability Insurance and Health Insurance Reform: Evidence from Massachusetts." American Economic Review 104(5): 329-335. 13 Maestas, Nicole, Kathleen Mullen, and Alexander Strand. 2021. "The Effect of Economic Conditions on the Disability Insurance Program: Evidence from the Great Recession." Journal of Public Economics 199: 1-22. Ray, Gerald K. and Jeffery S. Lubbers. 2015. "A Government Success Story: How Data Analysis by the Social Security Appeals Council (with a Push from the Administrative Conference of the United States) Is Transforming Social Security Disability Adjudication." The George Washington Law Review 83(4/5). 1575-1608. Schmitt, Lucie, Lara D. Shore-Sheppard, and Tara Watson. 2020. "The Impact of the ACA Medicaid Expansion on Disability Program Application." American Journal of Health Economics 6(4): 444-476. Social Security Administration, Office of Retirement and Disability Policy. 2019. "Trends in Social Security Disability Insurance." Briefing Paper No. 2019-01. Available at: https://www.ssa.gov/policy/docs/briefing-papers/bp2019-01.pdf Social Security Administration. 2022. "Annual Statistical Supplement to the Social Security Bulletin, 2022." Washington, DC. Social Security Administration, Office of Disability Programs. 2023. "Unpublished Data on Historical SSDI Applications and Applications by Age." Washington, DC. Stapleton, David, Kevin Coleman, Kimberly Dietrich, and Gina Livermore. 1988. "Empirical Analyses of DI and SSI Application and Award Growth." In Growth in Disability Benefits: Explanations and Policy Implications, edited by Kalman Rupp and David Stapleton, 31-92. Kalamazoo, Michigan: W.E. Upjohn Institute for Employment Research. Technical Panel on Assumptions and Methods. 2015. "Report to the Social Security Advisory Board." Washington, DC. Technical Panel on Assumptions and Methods. 2019. "Report to the Social Security Advisory Board." Washington, DC. U.S. Census Bureau. Current Population Survey, 1990-2019. Washington, DC. U.S. Social Security Administration. 2015. Annual Report of the Board of Trustees of the Federal Old-Age and Survivors Insurance and Federal Disability Insurance Trust Funds. Washington, DC. U.S. Social Security Administration, Office of Retirement and Disability Policy. 2019. "Trends in Social Security Disability Insurance." Briefing Paper No. 2019-01. Washington, DC. Available at: https://www.ssa.gov/policy/docs/briefing-papers/bp2019-01.pdf 14 U.S. Social Security Administration. 2022a. Annual Statistical Report on the Social Security Disability Insurance Program. Washington, DC. U.S. Social Security Administration. 2022b. Annual Statistical Supplement to the Social Security Bulletin, 2022. Washington, DC. U.S. Social Security Administration. 2023a. Annual Report of the Board of Trustees of the Federal Old-Age and Survivors Insurance and Federal Disability Insurance Trust Funds. Washington, DC. U.S. Social Security Administration, Office of the Chief Actuary. 2023b. "The Long-Range Disability Assumptions for the 2023 Trustees Report." Washington, DC. Available at: https://www.ssa.gov/OACT/TR/2023/2023 Long-Range Disability Assumptions.pdf 15 91 (020Z-1007) doang uonvindo 1ua.riny) o) pue sweidord AIIqesiq JO 91FO S.VSS o3 £q popraoid eyep woy $ejewunsd SIONY 294108 *$1JOUSq AILNosg [BI90S SUIAIONI JOU {9 PUB §T JO SoF. o) USOMIS] 950U} SOPN[OUT JOJRUITOUSD YL, '0007 ur woryerndod 21e3s o) Aq PAIYSIOM oI8 puR BIqUINIO)) JO IS Y opnhour suoissorSay 'S0'0>d xx '10°0>d xxx "SISOYIUIRd UI SIOLD pIepue)s ISNqOY SION [s IS IS IS IS §97%)S JO IoquuniN 8Y°0 L90 ¥Co L90 9%'0 pazenbs-y 971 01¢ 90¢ 6SY 0€S'T SuoneAIdsSqO SOX SOX SOX SOX SOX 7 AeIS SOX SO X SO X SO X SO X XIW Ansnpuy SOX Sa X SaX SaX SOX (s19j0r1q JeL-0A1)) uonisoduros 98y SOX SO X SOX SO X SOX dAT srewag SOA SO X SOA SO X SaA Jyoerq 9019 SOX SOX SOX SOX SOX 00139p Vg © UM JUo0Idd (900°0) (210°0) (L00°0) (800°0) (000 #%2L£0°0 *x6C0°0 #xx££0°0 #xx9C0°0 *%xx££0°0 ojer yudwAordwoun (0661 Surpnjout) s1eak 610Z-010C L00T-TOOT 000T-T661 610T-0661 aje1 uoneorddy uorsuedxs [y spouad uorsuedxy 0 © @) © @ ! 6102-0661 210y uoyvonyddy [T puv 2wy jusuidojduiauy) ay usamiag drysuoyviay ayy 40f S|nsay uo1ssaL32y °1 dqe, Figure 1. Number of DI Beneficiaries, 1990-2019 12 Millions N / 0 1990 1994 1998 2002 2006 2010 2014 2018 Source: U.S. Social Security Administration (2022a). Figure 2. Number of New Awards and Terminated Beneficiaries, 1990-2019 L.5 - New awards Terminated beneficiaries ) \ Millions o W N\ 0.0 | | T T | | | 1990 1994 1998 2002 2006 2010 2014 2018 Source: U.S. Social Security Administration (2022a). 17 Figure 3. DI Incidence Rate, 1990-2019 0.8% 0.6% /\ 0.4% 4 0.2% 0.0% 1990 1994 1998 2002 2006 2010 2014 2018 Sources: Author's calculations from U.S. Social Security Administration (2022a, 2022b). 18 Figure 4. DI Application Rate and Unemployment Rate, 1990-2019 1.5% 1.2% 0.9% 0.6% 0.3% 0.0% -- Application rate (left y-axis) Unemployment rate (right y-axis) 1990 1994 1998 2002 2006 2010 2014 2018 15% 12% 9% 6% 3% 0% Notes: The application rate is total fiscal year determinations divided by the insured population. The unemployment rate is measured as the average during the fiscal year. Sources: Authors' calculations from administrative data provided by the SSA's Office of Disability Programs and the Current Population Survey (CPS) (1990-2019). 19 Figure 5. DI Application Rate by Age Group, 2019 1.8% 797 1.5% 1.2% . 0.6% 0 3% 0.3% 0.0% 18-24 25-44 4549 50-54 55-59 60-61 62-64 Age groups Sources: Authors' calculations from administrative data provided by the SSA's Office of Disability Programs and the CPS (2019). 20 Figure 6. Average Age of DI Applicants, 2001-2019 48 C- NN 47 " o~ 45 Age 44 43 2001 2003 2005 2007 2009 2011 2013 2015 2017 2019 Note: The data are not available prior to 2001. Sources: Authors' calculations from administrative data provided by the SSA's Office of Disability Programs and the CPS (2001-2019). 21 Figure 7. Share of Workers in Physically Intensive Industries, 1990-2019 40% 35% N\ 30% - 25% T T T T T 1990 1994 1998 2002 2006 2010 2014 2018 Notes: Physically intensive industries include agriculture, forestry, fisheries, mining, manufacturing, construction, transportation, communications, and other public utilities. Source: Authors' calculations from the CPS (1990-2019). 22 Figure 8. Number of SSA Field Offices, 2001-2019 1,600 1,400 - 1,200 1,000 800 T T T T T T T T T 2001 2003 2005 2007 2009 2011 2013 2015 2017 2019 Notes: Field offices include level 1 and 2 offices, as well as resident stations. The data are not available prior to 2001. Source: U.S. Social Security Administration (2022a). 23 Figure 9. Initial and Final Allowance Rates, 1992-2019 70% = Final Initial 60% N\ 4 50% j 40% 30% T T T T T T T T T 1992 1995 1998 2001 2004 2007 2010 2013 2016 2019 Note: The data are not available prior to 1992. Source: U.S. Social Security Administration (2022a). 24 Figure 10. Counterfactual DI Incidence Rates, 2010-2019 0.008 - 0.006 \ 0.004 -Effect of population aging 0.002 -+ --Effects of aging and the business cycle --Effects of aging, the business cycle, and field offices Observed incidence rate 0.000 T T T T T T T T 1 2001 2003 2005 2007 2009 2011 2013 2015 2017 2019 Note: The counterfactual rates hold all other factors constant at 2010 levels. Sources: Authors' estimates from data provided by SSA's Office of Disability Programs; the CPS (1990-2019); and Deshpande and Li (2019). 25 Appendix Table Al. Summary of Previous Studies on the Relationship Between the Unemployment Rate and DI Application Rate Study Elasticity Time period Time series frequency Stapleton et al. (1988) DI Only 0.28%* 1980-1993 Annual DI/SSI Concurrent 0.26** 1980-1993 Cutler, Meara, and Richards-Shubik (2012) 0.31*** 2001-2011 Quarterly Maestas, Mullen, and Strand (2015) 0.19%*x* 1992-2012 Monthly Maestas, Mullen, and Strand (2021) 0.20*** 2006-2012 Monthly This paper 0.31%*x* 1990-2019 Fiscal Year 0.33*** Expansion years Note: Significance level is *** p<0.01, ** p<0.05. Source: Maestas, Mullen, and Strand (2021) and authors' estimates from data provided by SSA's Office of Disability Programs and the CPS (1990-2019). 26 Table A2. Age Differences in DI Application Rates, 2001-2019 25-44 Difference from the 25-44 group Year application rate 45-49 50-54 55-59 60-61 62-64 2001 0.54% 0.35% 0.65% 1.08% 1.76% 1.69% 2002 0.61 0.39 0.70 1.04 1.78 1.61 2003 0.67 0.41 0.72 1.07 1.73 1.66 2004 0.69 0.43 0.73 1.05 1.74 1.65 2005 0.69 0.43 0.71 1.01 1.70 1.57 2006 0.65 0.42 0.70 0.96 1.70 1.52 2007 0.63 0.44 0.69 0.95 1.46 1.47 2008 0.65 0.43 0.67 0.92 1.41 1.45 2009 0.69 0.44 0.67 0.91 1.39 1.15 2010 0.79 0.46 0.78 0.97 1.43 1.23 2011 0.82 0.52 0.89 1.13 1.55 1.30 2012 0.77 0.51 0.86 1.10 1.54 1.26 2013 0.71 0.47 0.84 1.06 1.42 1.13 2014 0.66 0.45 0.89 1.13 1.40 1.13 2015 0.62 0.41 0.80 1.05 1.34 1.07 2016 0.57 0.36 0.76 1.04 1.34 1.04 2017 0.52 0.33 0.72 0.96 1.27 0.98 2018 0.47 0.29 0.69 0.94 1.20 1.00 2019 0.48 0.28 0.64 0.95 1.19 0.94 Average 0.41% 0.74% 1.02% 1.49% 1.31% Note: The data are not available prior to 2001. Sources: Authors' calculations from administrative data provided by the SSA's Office of Disability Programs and the CPS (2001-2019). 27 8¢ (6102-0661) SAD o pue swerdord AIIqesI(T JO IO S, VSS Y Aq papiaoid ejep wox sajeun)sa SIOYMNY 5224108 "sygeuaq AJLmoag [e100g SUIAISDaI J0U 9 pue 8] Jo safe o) USSMIDq 9SOT) S PSULJSP ST JOJBUIIIOUIP oyl '000z w uonerndod s1e1s o) Aq PAIYSIOM QIR pUR BIQUINIOY) JO 101NSKT oY) 9pN[our SUOISSAZY '10'0>d 444 SOSoIuaIed U SI0II0 prepue)s 1snqoy $ION IS IS IS IS IS S3JE}S JO JSqUINN 050 1L°0 LTO L9°0 6v°0 parenbs-y 97¢T 01¢ 90¢ 65 0€S°T SUONEAIOSGO SOX SOX o X SOX SOA qq 91018 SOX SOX SOX S9X SOX X1 Ansnpuj SOX SO SO SO SO (s1ojorIq 18L-2A1)) uonisodwoos a8y S X S9X Sax SOX Sax dAT1 srews g SOX SOX o X SOX SOX Jyoe[d 1ud10d Sax SO S SaX SaX 90139p Vg © YyNM U019 (€0°0) (900°0) (+0'0) (90°0) (zo'0) +4%€€0 44600 x%xCC0 +4xLT0 +4%1€°0 aje1 juowKo[dwaun 307 (0661 Surpnjout) sreak 6102-010C LO0T-CT00C 000T-C661 610C-0661 oyex uoneordde So] uorsuedxd [Ty spouad uorsuedxyq M © @) © @ ! 610Z-0661 21y uoypanddy [(q 307 puv a1y juswidojduou) 307 ay) usamzag diysuoyp]ay ayj L0f S|nsay uoissaL3ay "¢y d[qel Figure Al. Average Age of the Population Eligible for DI, 2001-2019 41 40 - 38 T T T T T T 2001 2004 2007 2010 2013 2016 2019 Note: We define the eligible population as 18 to 64 year-olds who are not yet receiving Social Security benefits. Source: Authors' calculations from the CPS (2001-2019). Figure A2. Percentage of Workers in Physically Intensive Industries by Industry Type, 1990- 2019 25% Manufacturing Mining and construction 20% Public utilities \- = Agriculture, forestry, and fisheries 10% 5% O% T T T T T T T 1990 1994 1998 2002 2006 2010 2014 2018 Source: Authors' calculations from the CPS (1990-2019). 29 Figure A3. DI Eligible Population and Insured Population, 1990-2019 200 -Eligible population (CPS) Insured population / 180 160 / 120 Millions = S 100 1990 1994 1998 2002 2006 2010 2014 2018 Notes: The eligible population from the Current Population Survey (CPS) is defined as those ages 18 to 64 who are not currently receiving OASDI benefits. Sources: Authors' calculations from the CPS (1990-2019) and U.S. Social Security Administration (2022a). 30 RECENT WORKING PAPERS FROM THE CENTER FOR RETIREMENT RESEARCH AT BOSTON COLLEGE Why Do Late Boomers Have So Little Wealth and How Will Early Gen-Xers Fare? Angi Chen, Alicia H. Munnell, and Laura D. Quinby, May 2023 The Importance of Schools in Driving Children's Applications for Disability Benefits Michael Levere, Jeffrey Hemmeter, and David Wittenburg, March 2023 Does the Drop in Child SSI Applications and Awards During COVID Vary by Locality? Michael Levere, Jeffrey Hemmeter, and David Wittenburg, March 2023 What Factors Drive Racial Disparities in Housing Wealth Accumulation? Siyan Liu and Laura D. Quinby, March 2023 What Matters for Annuity Demand: Objective Life Expectancy or Subjective Survival Pessimism? Karolos Arapakis and Gal Wettstein, January 2023 How Will COVID Affect the Completed Fertility Rate? Angi Chen, Nilufer Gok, and Alicia H. Munnell, January 2023 County-Level Drivers of Disability Benefit Claims in Times of COVID-19 R. Vincent Pohl and David R. Mann, December 2022 Health, Disability, and the Evolving Nature of Work Barbara A. Butrica and Stipica Mudrazija, December 2022 Racial Disparities in COVID-19 Experiences among Older Adults with Disabling Conditions Marisa Shenk, Amal Harrati, Bernadette Hicks, and Ana Quifiones, December 2022 What Is the Risk to OASI Benefits from Unpaid Student Loans? Gal Wettstein and Siyan Liu, November 2022 How Does Local Cost-Of-Living Affect Retirement for Low and Moderate Earners? Laura D. Quinby and Gal Wettstein, November 2022 Did the Stimulus Checks Improve Household Balance Sheets? Andrew G. Biggs, Anqi Chen, and Alicia H. Munnell, November 2022 All working papers are available on the Center for Retirement Research website (https://crr.be.edu) and can be requested by e-mail (crr@bc.edu) or phone (617-552-1762). 31