Papers Recently Released by the NBER Retirement and Disability Research Center 2021


Recently approved papers

NB20-08: Induced Entry in Disability Insurance: Evidence from Canada, Giacomin Favre, Andreas Haller & Stefan Staubli
Abstract: Disability insurance (DI) programs are large and experienced substantial growth over the last decades. This paper studies the induced entry effect of two important DI program parameters: (i) DI benefit generosity and (ii) financial work incentives for DI recipients. Using two Canadian DI reforms and administrative tax records from 20 percent of Canadians, we find that more generous DI benefits induce substantial entry into DI. Specifically, a 1,000 CAD increase in benefits increases average DI take up by 0.126 percentage points, implying a DI take up elasticity with respect to benefits of 0.67. On the other hand, we find that the introduction of an earnings disregard has a quantitatively small effect on labor supply and entry into DI. Specifically, a 1,000 CAD increase in the earnings disregard leads to induced entry of 0.002 percentage points.

NB20-09: Misperceptions of the Social Security Earnings Test and the Actuarial Adjustment: Implications for Labor Force Participation and Earnings, Alexander M. Gelber, Damon Jones, Ithai Lurie & Daniel W. Sacks
Abstract: Budget set kinks are much studied in economics, including in the context of “bunching” estimators that assume individuals react to the true marginal tax rate. We document that individuals disproportionately “left-bunch” below rather than above kinks in the context of the Social Security Earnings Test where incentives from its actuarial adjustments should instead push many rational agents to bunch above kink. We show that the left bunching cannot be explained through standard, rational reactions to the incentives. We demonstrate that this represents the first empirical evidence consistent with “spotlighting,” wherein individuals misperceive the local marginal tax rates as applying throughout the tax schedule and therefore treat the kink as a notch. In the context of the Earnings Test, this misperception provides an explanation for why literature has found large earnings responses despite the fact that the Earnings Test typically creates weak incentives for rational agents to adjust earnings. More generally, if individuals perceive kinks as notches, then we suggest that elasticities estimated from bunching at kinks where this misperception may be at play may be significantly over-estimated.

NB20-14: The Evolution of U.S. Firms’ Retirement Plan Offerings: Evidence from a New Panel Data Set, Antoine Arnoud, Taha Choukhmane, Jorge Colmenares, Cormac O'Dea & Aneesha Parvathaneni
Abstract: This paper documents, using a newly-constructed data set, the evolution of the characteristics of employer-sponsored DC schemes. The features we focus on are their match schedules, vesting schedules, and the extent of ‘auto-features’ (i.e. presence of auto-enrollment, the level of any default contribution, and presence and details of auto-escalation). The data we construct is formed by hand-coding the details in narrative plan descriptions attached to plan filings. Our data covers approximately 5,000 plans, covering up to 37 million participants annually, for the period 2003-2017. We document that matching schedules, when they are offered, have become more generous over time. However, the proportion of firms offering a match fell sharply during the Great Recession and the proportion offering one did not recover to its pre-financial crisis level for almost a decade. Vesting schedules for DC plans have remained essentially unchanged since 2003, while the proportion of plans with auto-enrollment has increased dramatically over the same period. We find that the vast majority of plans that offer auto-enrollment have a default rate that is substantially lower than the level that would fully exploit the match offered by the employers.

NB20-19: The Impact of COVID-19 on Older Workers' Employment and Social Security Spillovers, Gopi Shah Goda, Emilie Jackson, Lauren H. Nicholas & Sarah See. Stith
Abstract: The COVID-19 pandemic and associated mitigation strategies exacted a large economic toll on large portions of the United States population. For older and disabled workers, the effects could be more persistent and fiscally costly than the impacts experienced by young, healthy workers due to the spillovers onto Social Security. We use Current Population Survey, Social Security administrative data on applications for retirement and disability benefits, and Google Trends data to assess the impact of COVID-19 on older adults age 50-70. We find that employment for this group dropped substantially more than would have been predicted prior to the pandemic: employment for 50-61 year-olds was 5.7 pp (8.3 percent) lower, while employment for 62-70- year-olds was 3.9 pp (10.7 percent) lower. For people aged 50-61, unemployment and labor force exits due to reasons other than disability and retirement represented 63 and 30 percent of the employment decline, respectively. For those aged 62-70, the two largest components of the reduction were unemployment (50 percent) and retirement-driven labor force exits (30 percent). We find evidence of declines in reporting a labor force exit due to disability (4-5 percent), applications for disability insurance (15 percent), and Google search intensity for disability (7 percent). Retirement benefit claiming remains largely unchanged overall, though we find evidence that applicants substituted towards filing for benefits via the internet. We explore potential mechanisms and find evidence for both supply- and demand-side explanations.

NB21-01: Children Receiving OASDI and SSI by State and County, 1970-2019: Description and Fifty Years of Data, Timothy J. Moore
Abstract: This paper describes the digitization and analysis of county-level data from Social Security Administration publications reporting children receiving Social Security benefits and Supplemental Security Income. Data sets are created that begin in the 1970s and go through until 2019, which are merged with population data to create rates of receipt in the population. The variables are described and then used to analyze patterns in receipt across counties and over time. Other county-level data are also merged with these data to examine what is correlated with the number of beneficiaries and recipients at the county level. The data will be made freely available for use in examining a wide variety of questions around social insurance.

NB21-02: The Impact of Biopharmaceutical Innovation on Disability, Social Security Recipiency, and Use of Medical Care of U.S. Community Residents, 1998-2015, Frank R. Lichtenberg
Abstract: 88% of privately-funded U.S. funding for biomedical research came from pharmaceutical and biotechnology firms. This study analyzes the overall impact that biopharmaceutical innovation had on disability, Social Security recipiency, and the use of medical services of U.S. community residents during the period 1998-2015. Most of the data come from the Medical Expenditure Panel Survey. The effect of biopharmaceutical innovation is identified by differences across over 200 medical conditions in the growth in the lagged number of drug classes ever approved. The FDA believes that 70% of first-in-class drugs offer a “significant improvement” compared with products already on the market. The estimates indicate that the probability of disability, Social Security recipiency, and medical care utilization is inversely related to the number of drug classes previously approved. The length of the estimated lag is generally 6-9 years, which is not surprising, due to the gradual diffusion of new drug classes. The effect of biopharmaceutical innovation related to a medical condition on the overall health of a person with that condition depends on the number of (other) medical conditions a person has: the smaller the number of conditions, the larger the effect. Previous innovation is estimated to have reduced: the number of people who were completely unable to work at a job, do housework, or go to school in 2015 by 4.5%; the number of people with cognitive limitations by 3.2%; the number of people receiving SSI in 2015 by 247 thousand (3.1%); and the number of people receiving Social Security by 984 thousand (2.0%). Previous innovation is also estimated to have caused reductions in home health visits (9.2%), inpatient events (5.7%), missed school days (5.1%), and outpatient events (4.1%). We estimate the value in 2015 of some of the reductions in disability, Social Security recipiency, and use of medical care attributable to previous biopharmaceutical innovation. This value ($115 billion) is fairly close to 2015 expenditure on drug classes that were first approved by the FDA during 1989-2006 ($127 billion). However, for a number of reasons, the costs are likely to be lower, and the benefits are likely to be larger, than these figures.

NB21-03: How Does the Income Effect Vary with Skill Level for Workers with Disabilities? Evidence from Workers’ Compensation, Kathleen J. Mullen & Stephanie L. Rennane
Abstract: Understanding how skilled vs. unskilled workers respond to disability benefits could help SSA determine how skill level should factor into disability determinations. Individuals with transferable skills could be better able to adapt to new occupations after the onset of a health condition that prevents them from returning to their previous occupations, which could make them more or less responsive to the amount of disability benefits. However, given that that benefits are determined by one's prior earnings using the same formula for all disabled workers, it is impossible to test whether return-to-work outcomes for workers with and without transferable skills differ in their sensitivity to the amount of disability benefits. This paper takes advantage of a policy change to permanent partial disability (PPD) benefits in Oregon workers’ compensation to identify an income effect in response to disability benefits, and to examine how this income effect varies with a worker’s skill level. In practice, the policy change resulted in nearly every individual receiving a different PPD benefit after 2005 compared to what she would have gotten prior to 2005. We find that unskilled workers are more responsive to benefit levels than skilled workers for a range of return-to-work outcomes including employment, earnings and weekly hours. For example, a $1,000 increase in PPD benefits results in a 0.35% decline in the probability of return-to-work for unskilled workers, compared to a 0.15% decline in the probability of return-to-work for skilled workers. While these estimates seem small it is important to keep in mind that a change in PPD benefits represents a one-time change in nonlabor income. Indeed, if we rescale estimates from the prior literature to represent a one-time change in benefit levels then our estimates are in line with, if not larger than, other recent estimates of income effects for this population.

NB21-04: Communicating the Implications of How Long to Work and When to Claim Social Security Benefits, Megan E. Weber, Stephen Spiller, Suzanne Shu & Hal E. Hershfield
Abstract: The decision of when to begin claiming SSA retirement benefits is part of an interconnected set of choices including how to draw down individual retirement savings and when to exit the workforce. Because the decision to begin claiming is often irrevocable, for retirees who begin claiming but then continue working or decide to return to work, an understanding of the applicable rules is extremely important. Retirees who claim before full retirement age and earn more than a specified (relatively low) income threshold receive reduced benefits prior to full retirement age (FRA) and increased benefits after FRA. Prior attempts to enhance understanding of the policy that results in these tradeoffs, termed the Retirement Earnings Test (RET), have not generally been successful. In four studies, we aim to increase understanding via visualizations. Study 1 assesses how well prospective retirees understand the impact on SSA benefits of continuing to work. Studies 2A, 2B, and 3 are randomized experiments that test whether alternative graphical ways to present these tradeoffs enhance understanding or affect labor decisions, using situations analogous to those of the RET (2A, 2B) or the RET itself (3). We find that a visualization that more clearly displays the tradeoff is able to improve understanding and application of RET policies. Prior research suggests that improvements in understanding of the social security system can generate large welfare gains; we would argue that improving understanding of RET can have similarly large effects. These tests of alternative visual presentations of the tradeoffs, informed by psychological research on how individuals understand income flows, can provide insight into how to improve RET communication.

NB21-10: The Long-Term Externalities of Short-Term Disability Insurance, Michael Stepner
Abstract: This paper shows that employer-provided short-term disability insurance (STDI) increases long-term disability insurance (LTDI) take-up and imposes a negative fiscal externality on the government budget. Expanding private STDI has been touted as a way to lower public LTDI costs by giving employers a financial incentive to provide workplace accommodations. But private STDI can also raise public LTDI costs, since STDI generates moral hazard by providing benefits during the waiting period for LTDI. Using variation in private STDI coverage caused by Canadian firms ending their plans, I find that the moral hazard effect dominates and private STDI raises two-year flows onto LTDI by 0.07 percentage points (33%). Extrapolating to Canada’s entire population, private STDI generated 18,300 LTDI recipients and CA$230 million dollars (5%) of public LTDI spending in 2015. The efficient Pigouvian tax on Canadian private STDI that internalizes this externality is approximately CA$35 per insured worker.

NB21-16: Projecting Trends in Undocumented and Legal Immigrant Populations in the United States, Ryan A. Bhandari, Benjamin Feigenberg & Darren Lubotsky
Abstract: We use administrative data on over 9 million Matr´ıcula (identification) cards issued by the Mexican government between 2008 and 2017 to Mexican-born individuals living the United States to improve estimates of the undocumented foreign-born population. These cards are held by those who do not have legal status in the United States and therefore do not have other forms of valid identification. The key contribution of our work is to use this data to produce estimates of the undocumented population from Mexico and from other countries, carefully laying out the relevant issues, assumptions, and sources of uncertainty. The ability to use the Matr´ıcula data to inform estimates of the undocumented population is particularly important because of the general lack of direct data on this group. Our preferred estimates indicate that there were on average 8.3-8.7 million undocumented Mexican individuals in the United States per year between 2008 and 2012 and 7.5-8.2 million between 2013 and 2017; both estimates are somewhat higher than the well-known estimates produced by the Pew Center. Our estimates of the undocumented immigrant population from other Latin American and Caribbean countries are more closely aligned with those from the Pew Center. Finally, we conclude that Matr´ıcula data is unlikely to be useful in estimating the undocumented population from outside of the Latin America and Caribbean region.

NB21-17: The Impact Of Health And Education On Labor Force Participation In Aging Societies – Projections For The United States And Germany From A Dynamic Microsimulation, René Böheim, Thomas Horvath, Thomas Leoni & Martin Spielauer
Abstract: Using a highly stylized dynamic microsimulation model, we project the labor force of the United States up to the year 2060 and contrast these projections with projections for Germany to assess differential effects on outcomes The projections are consistent with the U S Census Bureau’s and Eurostat’s demographic projections. Our modeling approach allows to show and quantify how policy changes the future size of the labor force, which we assess with a series of what-if scenarios. Both the US and Germany are expected to undergo demographic aging, but their demographic fundamentals differ starkly. This has strong implications for their labor force developments. According to our microsimulation, the US labor force will, despite population aging, increase by 16.2 percent in the age groups 15 to 74 (corresponding to 25.2 million workers) between 2020 and 2060, while Germany will experience a decline by 10.7 percent (4.4 million workers). In these baseline projections, improvements in the education structure will add about two million persons to the US labor force and about half a million persons to the German labor force by 2060. In the what-if scenarios, we examine the implications of improvements in the educational structure of the population and of policies which address the health impediments for labor force participation. Of the educational scenarios that we evaluate, increasing the number of persons who achieve more than lower education has the strongest positive impact on labor force participation, relative to the number of additional years of schooling implied by the various scenarios. Shifting people from intermediate to higher education levels also increases labor force participation in higher age groups, however, this is partially offset by lock in effects at younger ages. Our projections highlight that improvements in the labor market integration of people with health limitations provide a particularly promising avenue to increase labor force participation rates and thus help to address the challenges posed by demographic aging. If the health gap in participation rates in the United States were similar to that currently observed in Sweden, the labor force in 2060 would be larger by about 14.9 million persons.

NB21-19: Beyond Health: Non-Health Risk and the Value of Disability Insurance, Manasi Deshpande & Lee Lockwood
Abstract: The public debate over disability insurance has centered on concerns about individuals without severe health conditions receiving benefits. We go beyond health risk alone to quantify the overall insurance value of U.S. disability programs, including value from insuring non-health risk. We find that disability recipients, especially those with less-severe health conditions, are much more likely to have experienced a wide variety of non-health shocks than non-recipients. Selection into disability receipt on the basis of non-health shocks is so strong among individuals with less-severe health conditions that by many measures less-severe recipients are worse off than more-severe recipients. As a result, under baseline assumptions, benefits to less-severe recipients have an annual surplus value (insurance benefit less efficiency cost) over cost-equivalent tax cuts of $7,700 per recipient, about three-fourths that of benefits to more-severe recipients ($9,900). Insurance against non-health risk accounts for about one-half of the value of U.S. disability programs.