Papers Recently Released by the NBER Retirement and Disability Research Center
Recently approved papers
NB22-01: Older Workers’ Employment and Social Security Spillovers through the Second Year of the COVID-19 Pandemic, Gopi Shah Goda, Emilie Jackson, Lauren Hersch Nicholas, and Sarah Stith
Abstract: The COVID-19 pandemic triggered a large and immediate drop in employment among US workers, along with major expansions of unemployment insurance and work from home. We use Current Population Survey and Social Security application data to study employment among older adults and their participation in disability and retirement insurance programs through the second year of the pandemic. We find ongoing improvements in employment outcomes among older workers in the labor force, along with sustained higher levels in the share no longer in the labor force during this period. Applications for Social Security disability benefits remain depressed, particularly for Supplemental Security Income. In models accounting for the expiration of expanded unemployment insurance, we find some evidence that the loss of these additional financial supports resulted in an increase in disability claiming. Social Security retirement benefit claiming is approximately 3 percent higher during the second year of the pandemic.
NB22-04: Healthcare Cost-Sharing and the Economic Security of Social Security Disability Insurance Beneficiaries: Medicaid Expansion and the SSDI-Medicare Population, Philip Armour, Zetianyu Wang, and Claire O’Hanlon
Abstract: Although a growing literature has focused on the impact of cash benefits of Social Security Disability Insurance (SSDI), beneficiaries also eventually gain coverage by Medicare—one of the few under-65 groups eligible for Medicare. This health care coverage has significant value given the limited health insurance alternatives available to the long-term disabled given their chronic health conditions; however, original Medicare carries with it substantial cost-sharing and premiums. Dual eligibility with Medicaid has been a source of supplemental coverage for the disabled population but, prior to 2014, only for those who could satisfy Medicaid’s asset and income tests. State-level Medicaid expansions, beginning in 2014, expanded a supplemental coverage option to a broader swath of the under-65 Medicare population—beneficiaries who resided in the states that expanded Medicaid and had income under 138 percent of the federal poverty line (FPL). In this article, we rely on difference-in-differences techniques to estimate the impact of Medicaid expansion on supplemental coverage for the SSDI-Medicare population, and the subsequent effects on beneficiaries’ physician visits and health care spending. We find a statistically robust average increase in Medicaid coverage of 4.0-5.5 percentage points among SSDI-Medicare beneficiaries (an over 10 percent increase relative to baseline Medicaid coverage rates). There is substantial heterogeneity in this increase, however, with disproportionate large increases in coverage among white beneficiaries, beneficiaries with self-care or ambulatory difficulties, rural beneficiaries, and married beneficiaries without children. Although impacts on most other outcomes were statistically insignificant, we found a significant reduction in any out-of-pocket medical spending, consistent with Medicaid’s limits on cost-sharing, as well as a decrease in premiums paid by rural beneficiaries. Our findings contribute to the growing literature on the effects of Medicaid expansion, specifically in the smaller body of work examining the effects of increasing coverage on the intensive margin; it also adds to the smaller literature on the structure of health insurance coverage of the disabled-Medicare population.
NB22-07: Mortality Differentials, the Racial/Ethnic Retirement Wealth Gap, and the Great Pandemic, Edward N. Wolff
Abstract: The story that unfolds is that the standard net worth gap between Black and white households was much the same in 2007 as in 1983, though it did lessen considerably for Hispanics. The Great Recession hit minority families much harder than white ones, pushing the ratio of mean net worth between Blacks and whites down from 0.19 in 2007 to 0.14 in 2010 and that between Hispanics and whites from 0.26 to 0.15. The racial wealth ratio remained stuck at 0.14 up through 2019 while the ethnic ratio did improve to 0.19.
When the definition of wealth is expanded to incorporate Social Security and defined benefit pension wealth, the racial and ethnic wealth gap was sharply reduced. In 2019, the ratio between Black and white households in mean augmented wealth was 0.27 and that between Hispanics and whites was 0.32. The ratio of median wealth was boosted from about zero in the two cases to 0.39 for the former and to a whopping 0.48 for the latter.
Over time, from 1989 to 2019, the ratio of mean Social Security wealth between Blacks and whites climbed from 0.44 to 0.60 and that between Hispanics and whites from 0.48 to 0.76. In contrast, the ratio in mean augmented wealth between Blacks and whites was exactly the same in 2019 as in 1989. However, the ratio of median augmented wealth progressed from 0.24 to 0.39. The pattern is a little different for Hispanics. The ratio of both mean and median augmented wealth between Hispanics and whites advanced from 0.25 to 0.32 for the former and from 0.25 to 0.48 for the latter.
The COVID-19 Pandemic hit in 2020. Besides costing the United States one million plus lives, it lopped off over a quarter (26.7 percent) of Social Security wealth. Median Social Security wealth fell even more, 28.3 percent. The Pandemic effect was even stronger among young households – 29.4 percent and 30.6 percent, respectively. All told, mean augmented wealth dipped 8.1 percent among all households and 13.3 percent among young ones. The effect was even stronger on median values -- a 19.9 percent decline for the former and 27.4 percent for the latter.
The Pandemic reversed all of the absolute and relative gains made by minorities in terms of retirement and augmented wealth. It is first of note that the racial gap in life expectancy, which had declined from 6.2 years in 2006 to 4.8 years in 2019 for males and from 4.2 to 3.1 years for females, jumped to 7.0 years for the former and 4.5 years for the latter. The Hispanic-white gap went from 1.8 to 2.7 years in favor of Hispanics among males down to only 0.3 years in 2020 and from 2.3 to 3.1 years among females down to 2.2 years.
As a result, Black households saw their mean Social Security wealth fall by 29.1 percent. The reduction was particularly acute among young Black households, with a falloff of 34.2 percent. Hispanics were also hit very hard. Overall, mean Social Security wealth was down by 30.3 percent and 32.6 percent among younger households.
The Pandemic also enlarged the racial and ethnic gaps in retirement and augmented wealth. The Black-white ratio of mean Social Security wealth declined from 0.60 to 0.57 among all age groups and from 0.58 to 0.53 among the youngest group. The ratio of mean Social Security wealth between Hispanics and whites sank from 0.76 to 0.71 among all ages and from 0.83 to 0.78 for the youngest. Black households saw their mean augmented wealth down by 17.7 percent and median augmented wealth by 25.8 percent among all ages and 26.9 and 36.6 percent, respectively, among young Black households. Likewise, overall, mean augmented wealth was down by 17.2 percent and median augmented wealth by 28.0 percent among Hispanics and by 21.6 and 31.8 percent, respectively, among young Hispanics.
The Pandemic also widened the racial and ethnic gaps in augmented wealth. The Black-white ratio of mean values dropped from 0.27 before the Pandemic to 0.24 after it among all age groups and the ratio of median values from 0.39 to 0.35. Among the youngest group, the former fell from 0.31 to 0.26 and the latter from 0.42 to 0.35. The ratio of mean augmented wealth between Hispanics and whites tumbled from 0.32 to 0.28 among all ages and from 0.51 to 0.45 for the first age group. The ratio of median values showed an even steeper drop, from 0.48 to 0.42 among all ages and from 0.76 to 0.67 for the youngest.
A counterfactual experiment is also run for year 2019 in which white mortality rates are substituted for Black mortality rates and Social Security wealth recalculated. The substitution had a very modest effect on mean Social Security wealth calculated for all Black households, raising it by 10.4 percent. The impact was strongest among young households, 14.4 percent. It is calculated that 9.5 percent of the racial gap in mean Social Security wealth among all households was due to mortality rate differentials and 12.6 percent for the youngest group. The impact was stronger on median values. The substitution enlarges median Social Security wealth by 10.8 percent among all Black households and by 15.3 percent for the youngest age group. It also raises the Black/white ratio of median Social Security wealth by 0.12 among all households. Moreover, 17.6 percent of the racial gap in median Social Security wealth among all households is found to be due to mortality rate differentials and 25.3 percent for the youngest group. However, still the biggest explanatory factor is the residual which reflects difference in earnings history and coverage rates.
NB22-10: How do Behavioral Approaches to Increase Savings Compare? Evidence from Multiple Interventions in the U.S. Army, Richard W. Patterson and William L. Skimmyhorn
Abstract: Information provision, choice simplification, social messaging, active-choice frameworks, and automatic enrollment all increase retirement savings. However, gauging the relative efficacy of these approaches is challenging because the supporting evidence spans widely different institutional settings, populations, and time periods. In this study, we leverage experimental and quasi-experimental variation in a constant setting, the U.S. military between 2016-2018, to examine the effects of nearly two dozen experiments for four leading policy options (i.e., information emails, action steps, target contribution rates, active choice, and automatic enrollment) designed to increase retirement savings. Consistent with previous literature, we find sizable effects of savings interventions on participation and cumulative contributions that increase with the intensity of the intervention. We then exploit cost data to complete the first cost-effectiveness analysis in the literature. Our analysis suggests that active choice programs are the most cost-effective method to generate new program participation and contributions for small, medium, and large firms, while automatic enrollment is more cost-effective for very large firms.
NB22-11: Do State Supplemental Nutrition Assistance Program Policies affect Older Adults and People with Disabilities? Donna K. Ginther and Michael Easterday
Abstract: This study examines the effect of state SNAP policies on access to SNAP benefits for the older adults and people with disabilities. Although SNAP is a federal program, states have considerable power to choose which policies to adopt, when to adopt them, and to what extent those policies cover their population. Previous research has focused on a single policy database and using policy indices to measure the impact of SNAP policies on caseloads. Using state policy variation from the SNAP Policy Database and the SNAP State Option Reports, this study uses two-way fixed effects and difference-in-differences models to understand the effects of both individual SNAP policies and the policy indices. Results indicate that SNAP policies that improve eligibility and reduce transaction costs increase participation among older adults and people with disabilities and restrictive policies reduce participation. The magnitudes of these coefficients are larger for the older adults and people with disabilities compared to the general population.
NB22-15: The Role of Stock-Flow Reasoning in Understanding the Social Security Trust Fund, Hal E. Hershfield, Stephen Spiller, Suzanne Shu, and Megan E. Weber
Abstract: The financial future of Social Security's trust funds is an important policy topic with significant implications for members of the public who pay taxes and expect to receive benefits in retirement. The funds were created to hold and invest surplus tax revenue not used to pay out benefits, but in recent years, Social Security has started to use this money to fulfill benefits obligations. The funds are projected to become depleted in 2035, at which point benefits payments will have to be reduced. In this research, we draw from the literature on stock-flow reasoning errors and inconsistencies to explore how communication about the trust funds impacts understanding of the situation. In Studies 1 and 2 we randomly assign participants to see information about the trust funds over time presented as a stock (i.e., balance) or in terms of flows (i.e., tax revenue and benefits payments), finding that those who see the stock presentation are significantly more likely to expect benefits to go away completely after depletion. In a third study, we show that explicitly prompting participants to reflect on the continuity of the inflows (via payroll taxes) significantly reduces this common misunderstanding even further. Applying the theoretical lens of stock-flow reasoning, results of this research highlight a key aspect of communications about the trust funds that may contribute to – or be used to remedy – the widespread misconception that benefits will cease when the funds are depleted.
NB21-13: Disability Heterogeneity in the Impact of the ACA’s Medicaid Expansions on Disability Employment, Ari Ne'eman and Nicole Maestas
Abstract: Objectives: To test for heterogeneous treatment effects in the impact of the Affordable Care Act’s (ACA) Medicaid expansion on the employment of people with disabilities.
Methods: Using difference-in-difference approaches, we estimate the impact of the ACA’s Medicaid expansion on employment outcomes for various subgroups of people with disabilities. Using the Current Population Survey (CPS) from June 2008 to December 2019, we segment the disabled population by disability type, disability recency and labor force attachment, leveraging the longitudinal aspect of the CPS.
Results: Among persons with higher labor force attachment, we find that Medicaid expansion reduced the employment rate of persons with new disabilities by a statistically significant -3.2%, while there was a precisely estimated null effect for persons with ongoing disabilities. Among those with lower labor attachment, we find suggestive evidence of offsetting treatment effects among persons with new versus ongoing disabilities. Medicaid expansion increased the employment rate of persons with ongoing disabilities by 10.5% but decreased the employment rate of persons with new disabilities by -9.2%. However, these latter estimates for persons with lower labor force attachment are imprecisely estimated, limiting the conclusions that can be drawn from them.
Conclusions: Existing literature on the disability employment effects of Medicaid expansion is mixed in part due to different study designs picking up different effects on distinct groups of people with disabilities. We show that accounting for disability heterogeneity allows for more precise estimates of policy impacts for some populations while providing suggestive evidence of countervailing treatment effects for others.
NB21-14: Final Grant Report: Cash vs. Food? How Does Food Stamp Eligibility A ect Food Stamp Enrollment and Food and Health Outcomes of SSI Recipients? Marianne Bitler, Amelia A. Hawkins, Lucie Schmidt, and Hilary Seligman
Abstract: Supplemental Security Income (SSI) has been a source of cash assistance for low-income disabled, blind, and elderly persons since it was created in 1974. One choice states had at the start was whether to make SSI recipients eligible for Food Stamps (a means-tested program to provide vouchers for food) or to instead provide additional cash in the SSI grant. Five states initially included the "cash-out" option in 1974, but between 1974 and 1992, four transitioned to grant Food Stamps eligibility to SSI recipients. In 2019, the final hold-out state of California also ended the cash-out program, making SSI recipients in the state eligible for Food Stamps for the first time. In this report, we present findings about the effects of both the beginning of cash out (did Food Stamp participation and food security change for of recipients of the new SSI program differently for those in cash-out states versus non-cash-out states?) and the end of cash out (how did Food Stamp participation and health outcomes change for SSI recipients when cash out ended in their states, making them eligible for Food Stamps?). We use data from a number of sources, including survey data on beneficiaries of SSI and its precursor programs immediately before and after the implementation of SSI, Food Stamp quality control administrative data, hospital-discharge data, and data on program use from the state of California. We find several intriguing results. First, when SSI was initially implemented in 1974, a large number of recipients in cash-out states lost Food Stamp eligibility, and we show that they experienced a significant increase in food insecurity as as result. Second, using data around the 1992 end of cash out in Wisconsin, we show that Food Stamp use went up, and we show suggestive evidence that hospitalizations for food-related diagnoses went down among the low-income elderly population that was likely to be eligible for SSI. Finally, our preliminary analysis of California suggests that county offces that used more different types of outreach and used more accommodations saw a greater increase in SNAP applications among SSI recipients after cash out ended in California.
NB21-15: The Impact of Paid Family Leave on Families with Health Shocks, Courtney Coile, Maya Rossin-Slater, and Amanda Su
Abstract: This paper analyzes the impact of paid family leave (PFL) policies in California, New Jersey, and New York on the labor market and mental health outcomes of individuals whose spouses or children experience health shocks. We use data from the 1996-2019 restricted-use version of the Medical Expenditure Panel Survey (MEPS), which provides state of residence and the precise timing of hospitalizations and surgeries, our health shock measures. We use difference-in-difference and event-study models to compare the differences in post-healthshock labor market and mental health outcomes between spouses and parents before and after PFL implementation relative to analogous differences in states with no change in PFL access. We find that PFL access leads to a 7.0 percentage point decline in the likelihood that the (healthy) wives of individuals with medical conditions or limitations who experience a hospitalization or surgery report “leaving a job to care for home or family” in the post-healthshock rounds. Impacts of PFL access on women’s mental health outcomes and on men whose spouses have health shocks are more mixed, and we find no effects on parents of children with health shocks. Lastly, we show that improvements in job continuity are concentrated among caregivers with 12 or fewer years of education, suggesting that government-provided PFL might reduce disparities in leave access.
NB21-22: Disability, Earnings, Income and Consumption, Bruce D. Meyer and Wallace K.C. Mok
Abstract: Using longitudinal data for the period 1968-2005 for a sample of male household heads, we determine the prevalence of disability during the working years and examine how the extent of disability affects a range of outcomes, including earnings, income, and consumption. We have seven main findings. First, disability rates are high. We divide the disabled along two dimensions based on the persistence and severity of their work-limiting condition. We estimate that a person reaching age 56 has a 53 percent chance of having been disabled at least once during his working years, and a 19 percent chance that he has begun a chronic and severe disability. Second, the economic consequences of disability are frequently profound. Ten years after disability onset, a person with a chronic and severe disability on average experiences a 68 percent decline in earnings, a 32 percent decline in after-tax income, a 22 percent decline in food and housing consumption and a 21 percent decline in food consumption. Third, the various economic consequences differ sharply across disability groups. The outcome declines for those with a chronic and severe disability are often more than twice as large as those for the average disabled. Fourth, our findings show the partial and incomplete roles that individual savings, family support and social insurance play in reducing the consumption drop that follows disability. Only about half of this most disabled group reports receiving Social Security Disability Insurance or Supplemental Security Income. Fifth, we find a noticeable fall in earnings and income prior to the onset of a reported disability. Consumption also falls somewhat, suggesting that future disability is partially but incompletely predictable in the short run. Sixth, time use and detailed consumption data further indicate that disability is associated with a decline in wellbeing. Seventh, the quantities we have estimated, combined with elasticities from the literature, allow us to examine the optimality of current compensation for the disabled. We find that the current compensation for our most disabled group appears to be lower than is optimal.
NB20-07: Long-Run Effects of Incentivizing Work After Childbirth, Elira Kuka and Na'ama Shenhav
Abstract: This paper identifies the impact of increasing post-childbirth work incentives on mothers’ long-run careers. We exploit variation in work incentives across mothers based on the timing of a first birth and eligibility for the 1993 expansion of the Earned Income Tax Credit. Ten to nineteen years after a first birth, single mothers who were exposed to the expansion immediately after birth (“early”), rather than 3–6 years later (“late”), have 0.62 more years of work experience and 4.2% higher earnings conditional on working. We show that higher earnings are primarily explained by improved wages due to greater work experience.
NB19-16: The Long-Term Effects of Workplace Injury on Labor Market Outcomes: Evidence from California, Michael Dworsky and David Powell
Abstract: Although workplace injury is more common than job displacement, we have limited evidence about the long-term impacts of workplace injury on earnings, employment or labor force exit due to disability or retirement. We link workers' compensation claims data for California workers injured in 2005 to earnings records spanning 2003-2019, providing a long panel for analyzing the long-term effects of workplace injuries. Our difference-in-differences research design compares injured workers who were paid benefits for lost work time (temporary disability) or permanent impairment (permanent disability) to "medical-only" workers with minor injuries. We estimate large reductions in employment and earnings due to workplace injuries. Difference-in-differences estimates controlling for worker characteristics and earnings losses averaged $920 per quarter over 14 years post-injury (or $51,000 without discounting). Event-study estimates show that earnings losses as a percentage of counterfactual earnings do shrink over time (from 19.6% over years 1-4 post-injury to 10.9% over years 10-14 post-injury), yet the presence of a 10.9% earnings reduction more than 10 years after injury suggests that average impacts on labor market outcomes are highly persistent. We also estimate hazard models that examine whether the rate of labor force exit responds to incentives created by Social Security's disability and retirement programs, exploiting age-specific thresholds in program eligibility. We find no evidence of increased labor force exit for injured workers at the Early or Normal Retirement Age, but we do find suggestive evidence that more favorable disability evaluation rules for workers aged 55 and over are associated with increased labor force exit among injured workers.
NB19-17: Mortality Impacts of Disability Insurance Payments: Context and Implications, Alexander M. Gelber and Timothy J. Moore
Abstract: Previous research has found that Social Security Disability Insurance (DI) income reduces beneficiaries’ mortality (Gelber, Moore, Pei, and Strand 2022). A key contribution in this paper is to provide context for these results by further describing the economic and demographic circumstances of DI recipients relative to non-recipients, using restricted-use “gold standard” Survey of Income and Program Participation data linked to Social Security Administration records as well as data from the Consumer Expenditure Survey. We find that DI recipients show important signs of economic disadvantage, particularly in the lower-income groups where DI income has the largest mortality effects. In more preliminary work, we suggest that incorporating the mortality benefits to these lower-income groups can have implications for the analysis of optimal DI benefits, though the implications vary significantly across assumptions and no clear takeaway is possible regarding optimal DI payments.
