a= CENTER for
RETIREMENT
RESEARCH

at BOSTON COLLEGE

WILL SURVIVORS OF THE FIRST YEAR OF THE COVID-19 PANDEMIC
HAVE LOWER MORTALITY?

Gal Wettstein, Nilufer Gok, Angi Chen, and Alicia H. Munnell

CRR WP 2022-10
August 2022

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

The authors are all with the Center for Retirement Research at Boston College. Gal Wettstein is
a senior research economist. Nilufer Gok is a senior research associate. Anqi Chen is a research
economist and the assistant director of savings research. Alicia H. Munnell is the director and
the Peter F. Drucker Professor of Management Sciences at Boston College's Carroll School of
Management. 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 would like to thank Stephen C. Goss and
Karen Glenn for helpful comments.

© 2022, Gal Wettstein, Nilufer Gok, Angi Chen, and Alicia H. Munnell. 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 University of Michigan and the National Bureau of Economic Research,
was established in 1998 through a grant from the 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 sponsors 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

The mortality burden of the COVID-19 pandemic was particularly heavy among older
adults, racial and ethnic minorities, and those with underlying health conditions. These groups
are known to have higher mortality rates than others even in the absence of COVID. Using data
from the 2019 American Community Survey, the 2018 Health and Retirement Study, and the
2020 National Vital Statistics System, this paper estimates how much lower the overall mortality
rate will be for those who lived through the acute phase of the early pandemic after accounting
for this selection effect of those who died from COVID. Such selection may have implications
for life insurance and annuity premiums, as well as assessments of the financial standing of
Social Security - if the selection is large enough to substantially alter projected survivor

mortality.

The paper found that:
e 10-year mortality rates, absent direct COVID deaths and long COVID, will likely be
lower in 2021 than anticipated in 2019.
e However, these differences are small, ranging from a decline of 0.4 percentage points for
people in their 60s to 1 percentage point for those in their 90s.
e The small difference is in spite of the fact that COVID mortality was, indeed, very
selective, with mortality declines exceeding half the maximum possible declines, holding

total COVID deaths constant, for every age group.

The policy implications of the findings are:
e That declines in mortality due to COVID selection likely will not impact overall population
mortality substantially enough to affect Social Security cost projections.
e Any impact of selection effects on Social Security costs will likely be swamped by ongoing

mortality increases directly attributable to acute and long COVID.
Introduction

The COVID-19 pandemic has claimed over a million American lives, with half of that
toll occurring in the year between March 2020 and March 2021. The burden was particularly
heavy among older adults, racial and ethnic minorities, and those with underlying health
conditions (Dyer et al. 2020; Alsan, Chandra, and Simon 2021; Ruhm 2021). These groups are
known to have higher mortality rates than others even in the absence of COVID, potentially
limiting the years of life lost to the pandemic.! What has not been as thoroughly explored is the
implication of this selection effect for the mortality rates of those who survived the pandemic.
This paper estimates how much lower the overall mortality rate will be for those who lived
through the acute phase of the early pandemic after taking into account the selection effect of
those who died from COVID. Such selection may have implications for life insurance and
annuity premiums, as well as assessments of the financial standing of Social Security - if the
selection is large enough to substantially alter projected survivor mortality relative to pre-
COVID expectations.

The analysis uses the American Community Survey (ACS) of 2019 and the Health and
Retirement Study (HRS) of 2018 to estimate the demographic and health distributions,
respectively, of the over-60 2019 population. The National Vital Statistics System (NVSS) data
from 2020 are then used to analyze deaths, by cause, in 2020 and, by extrapolation, early 2021.
These data are combined to arrive at an estimated distribution of the April 2021 population by
gender, race, ethnicity, and health status.

Based on this adjusted distribution, new life tables by gender are calculated, and
compared with the pre-COVID life tables. These estimates are made under the counterfactual
assumption that COVID will not directly impact mortality going forward either through new
infections or long COVID. This assumption is adopted since it provides the most conservative
estimate from the perspective of Social Security's finances.

The comparison of pre- and post-COVID life tables reveals that COVID mortality was,
indeed, very selective. However, despite the large death toll from COVID, the share of older
adults who died during the first year of the pandemic was not enough to substantially alter
mortality rates in the next few years. Moreover, COVID mortality was less selective among the

older population than among the younger cohorts.

 

1 For racial gradients in mortality, see Wettstein et al. (2021).
The rest of the paper proceeds as follows. The next section provides the background for
the analysis. The third and fourth sections describe the data and methodology. The fifth section
presents the results. The final section concludes that COVID victims were very concentrated in
otherwise high-mortality populations; however, the scale of COVID deaths was such that this

selection leads to only modest reductions in projected future mortality.

Background

The COVID pandemic was an enormous shock to all aspects of life, including mortality.
To date, over a million Americans (and many millions more around the world) have died of
COVID. This toll is evidenced by the many deaths clearly attributable to COVID, as well as
even higher excess mortality over the past two years.*

COVID deaths were not random. Some groups of people were more likely to be exposed
to COVID, and some groups, conditional on exposure, were more likely to suffer severe
consequences. In particular, Black and Hispanic individuals were more likely to come in contact
with COVID in the early months of the pandemic (Hooper, Napoles, and Pérez-Stable 2020;
Sarkar et al.2021). Meanwhile, for those contracting COVID, the disease was more dangerous to
those with certain preexisting health conditions and to older individuals (Imam et al. 2020;
Harrison et al. 2020). Specifically, the Centers for Disease Control (CDC) noted seven
categories of chronic health conditions that are associated with elevated risk from COVID:
cancer, cerebrovascular disease, diabetes, heart disease, kidney disease, lung disease, and
obesity.

With the exception of Hispanics, the groups most at risk from COVID were also more
likely to die within a given time period than other groups even in the absence of COVID.
Mortality was higher among Black than among white Americans through 2019 (Wettstein et al.
2021). Mortality rates also, of course, rise with age and are higher among individuals with a
large range of health problems, particularly those that lead to increased risk of COVID mortality
(Raghavan et al. 2019; Zucker et al. 2017; Soedamah-Huthu 2006; Badar et al. 2015; Lowe et al.
1998; Flegal et al. 2013; Mannino et al. 2003; Pedone 2010; Razmara et al. 2017; Ikeda et al.
2009; Florez-Perdomo et al. 2020; Ketchandji et al. 2009; Schonberg et al. 2011; Tu et al. 2018).

 

2 World Health Organization (2022).
3 See, for example, Mulligan and Arnott (2022).
The implication of elevated COVID mortality among otherwise high-mortality groups is that
fewer members of those groups are likely to have survived into late 2021, and survivors of the
pandemic's first year are likely to have lower non-COVID mortality.

Such survivor selection might have implications for academic and practical forecasts of
mortality in the coming years. To be sure, the overwhelming impact of the pandemic has been to
increase mortality rates since 2019; however, a second order effect of selection may mitigate
mortality increases once the acute phase of the pandemic has passed. For researchers, life
insurers, and pension providers, looking beyond the direct effects of COVID on mortality to
these second-order effects is necessary for determining expected mortality trajectories, setting
life insurance and annuity premiums, and determining appropriate funding levels for life-
contigent policies. Furthermore, correctly attributing any mortality decline in future years to this
acceleration of deaths is necessary for understanding the rate of improvement in life expectancy
due to more permanent factors, such as medical innovation, in assessing longer-term trends.

The potential impact of survivor selection is particularly important for Social Security.
As one of the largest providers of life-contingent payments in the world, forecasting mortality
rates in the near- and long-term is essential to understanding the sufficiency of funding streams
for the program. In particular, since the Old-Age and Survivors Insurance (OASJD) program's
costs increase when mortality declines, the heavy death toll of the last few years has had a
beneficial impact on program finances. However, the program's actuaries are already looking
ahead to the effects on their life tables in the coming years. The acceleration of deaths of
otherwise high-mortality individuals due to the pandemic may require new life tables to be
estimated.*

Of course, if COVID-19 continues to account for many deaths in the next few years
despite widespread vaccination, future mortality will not decline as much, if at all. Similarly, if
survivors of COVID-19 have elevated mortality risk due to further health complications, like
"long COVID," that too would increase mortality.*> While both these effects would improve
OASI's finances, the analysis here assumes that they are negligible in order to provide a

conservative estimate of future improvement in mortality from the perspective of OASI.

 

4 The new life tables will be relevant for projections over the next decade or so. The decomposition of trends to
COVID-19 vs. more permanent changes can inform longer-term projections of mortality that rely on long-run trends
in mortality decline (U.S. Social Security Administration 2015).

5 For example, see Li et al. (2021).
A further limitation of the analysis lies in the availability of timely data. Two main
constraints are relevant here: data on deaths and data on the joint distribution of health conditions
and demographic characteristics in the population. The next section describes the data used in

the analysis and its limitations.

Data

Our analysis rests on three main data sources: the ACS, the HRS, and the NVSS. The
2019 ACS is used to estimate the demographic distribution of the population on the eve of the
pandemic - the share of those over 60 that fall into each category defined by the intersection of
gender, race, ethnicity, and 10-year age bins. Similarly, the 2018 wave of the HRS is used to
estimate the health distributions of the 2019 population - the share of each demographic group
that has each of the relevant chronic health conditions accounted for in the analysis.

In choosing which health conditions to include, the analysis is guided by two imperatives:
1) health conditions must be disproportionately affected by COVID; and 2) they must be
measurable in the HRS. To determine the former, the set of chronic conditions starts with those
cited by the CDC as of March 2021 (CDC 2021) as increasing the risk of severe illness from
COVID.® These conditions are then further restricted to those that are asked about in the HRS.

The following set of conditions meet these two criteria: cancer, cerebrovascular disease,
diabetes, heart disease, lung disease, and obesity.' Moreover, while comorbidities are common,
to maintain sufficient sample sizes of demographic-health conditions cells, individuals with
multiple conditions are assigned to the one with the highest risk of COVID mortality.

In the case of both the ACS and HRS, the years chosen reflect the final survey years
unambiguously prior to the pandemic. For the ACS, this choice does not seem to entail much
risk for error, since at best more timely data would have covered just the first two months of
2020 (assuming a pandemic start date of March 2020). However, in the case of the HRS, the
two-year gap between survey waves leads to some potential inaccuracy in the estimated disease

profile of the population if that profile changed substantially between 2018 and early 2020.

 

6 These are conditions that the CDC defines as "having a published meta-analysis or systematic review or
completing the CDC systematic review process. The meta-analysis or systematic review demonstrates a conclusive
increase in risk for at least one severe COVID-19 outcome."

7 Since March 2021, some conditions have been added to the CDC list that are not included in the analysis at this
stage. In particular, mental health conditions (such as mood disorders) and neurological conditions (such as
dementia) are not included in the analysis but could be incorporated in future work.
To examine whether the extrapolation of disease prevalence from 2018 to 2020 might
bias estimates, Figure 1 displays the time trends in these conditions since 1992 (the first wave of
the HRS). For ease of exposition, these trends are calculated for the full over-60 population,
rather than by demographic subgroups. The figure shows that changes in the prevalence of these
conditions have been fairly modest year-to-year, at least since 2006. Consequently, the analysis
proceeds under the assumption that the 2018 chronic condition distribution is a close
approximation of the early 2020 distribution.

Figures 2a and 2b display the estimated distribution of health conditions by gender and
age group on the eve of the pandemic.® The patterns in the figure illustrate some of the issues
present in the remainder of the analysis. First, the shorthand of "healthy" is used to describe
individuals who report having none of the enumerated chronic health conditions; however, such
individuals may well have other conditions not included in this analysis. They may even have
some of the conditions that are considered, but not be aware of them (e.g., undiagnosed
diabetes).

Second, the age trends and gender differences in the prevalence of the various conditions
demonstrate the importance of accounting for the joint distribution of health and demographic
groups. Unsurprisingly, the healthy group's size declines with age in both genders, but women
tend to be healthier in any given age category. Similarly, heart disease is more common among
men at all ages, while obesity prevalence is similar across genders. Finally, some of the trends
by age are likely due to survivors being selected to be healthier; however, some are an artifact of
the assignment of individuals to a single condition when they have comorbidities. For example,
diabetes does not likely decline with age so much as it is supplanted by more life-threatening
conditions, which tend to arise among those with diabetes.

The goal of this exercise is to characterize the over-60 U.S. population after the first year
of the pandemic. Thus far, this population has been characterized in terms of its demographic
and health distribution as of 2019. To take the extra step of estimating the population
composition in 2021, the 2020 NVSS is used to analyze deaths, by demographic group and cause
of death, in 2020 and, by extrapolation, early 2021. The NVSS, accessed through the WONDER

 

8 The analysis uses an even finer partition of the population, with health conditions estimated by age, gender, race,
and ethnicity; see Figure 3. However, for the purposes of describing the pre-pandemic population, a more
aggregated partition with larger sample sizes allows for a more accurate discussion of broad trends.
database, records all deaths in the United States based on death certificates. Because deaths are
reported with a lag, the analysis relies solely on 2020 data to estimate deaths in each
demographic-health group through the March 2020-March 2021 period. The precise method for

this analysis is described in the next section.

Methods

The analysis has three conceptual parts: 1) to characterize the 2019 population of older
adults by their demographics and health conditions; 2) to adjust this estimated population for
COVID-19 and non-COVID-19 deaths in 2020 and early 2021, in proportion to the
demographics and health conditions identified, resulting in a predicted late 2021 population; and
3) to analyze the impact of selection in COVID deaths on life tables for the estimated late-2021

population relative to pre-pandemic expectations.

Characterizing the 2019 Population

The first part, estimating the distribution of the population over age 60 in 2019 into
demographic-health cells, is straightforward. Demographic groups, indexed by g, are defined by
the intersection of gender, race, ethnicity, and 10-year age bins starting at 60. The size of each
group is estimated from the 2019 ACS.

Next, the share of each group g suffering from each of the relevant chronic health
conditions, indexed by h, is estimated based on the HRS. The "healthy" group, with none of
these conditions, is denoted by h=/. The resulting partition of the population into cells defined
by demographics and health is shown in Figures 3a-3d, which are, themselves, a more detailed

breakdown of the simplified distribution displayed in Figure 2.

Adjusting the Population for Deaths in March 2020-March 2021

The base population from the first step is then adjusted to account for two types of
attrition: non-COVID and COVID mortality in the first year of the pandemic, between March
2020 and March 2021. Fundamentally, these data are from the NVSS. However, death records
in the NVSS list the demographics of the decedent, but not what health conditions they may have

had. Thus, assumptions must be made in assigning deaths to each demographic-health cell.
Specifically, three kinds of deaths need to be accounted for: deaths of causes related
directly to each of the COVID-relevant chronic health conditions; all other non-COVID deaths;
and deaths directly attributable to COVID. Each kind of death is handled differently. Particular

care is given to assigning COVID deaths, since this categorization is at the root of the analysis.

Assigning Non-COVID Deaths

First, deaths of causes directly linked to the chronic conditions of interest must be
allocated to the different demographic-health groups. The intuition here is that deaths
determined to be caused by, for example, heart disease, are more likely to occur among
individuals with known heart conditions than among the general population. Furthermore, the
pandemic seems to have exacerbated many causes of death that are not directly related to
COVID, and so accounting for specific changes in non-COVID causes of death during the
pandemic may be important in accurately forecasting who is alive after the pandemic's first year.

To account for this intuition, the analysis assumes that individuals in each cell who die
from any cause corresponding to the listed conditions come first out of the individuals with that
condition. For example, if among Black men ages 60-70 there were 100 cancer-related deaths
and 500 people with cancer, all cancer-related deaths would be assumed to come from that
group. If, instead, there were only 50 people with cancer, then all 50 would be assumed to have
died of cancer, while the remaining 50 would be assumed to have come from the other health-
condition groups, according to their relative baseline size.

In practice, this latter case of "excess" deaths spilling over into other categories is rare.
This outcome almost exclusively happens in either the 90-99 age groups or the "other" race
group due to small sample sizes of these groups in the HRS. Appendix Table Al documents
precisely which groups experience this situation.

All other causes of death (e.g., accidents) are treated much as chronic-condition-related
deaths that exceed their demographic-health group size. That is, within a demographic group,
such deaths are assigned to all the health groups in proportion to their size, implicitly assuming
an equal hazard of death for all health groups within a demographic cell. While this assumption

is no doubt incorrect, it is the most agnostic position to take.
Assigning COVID Deaths to Demographic-Health Cells

To determine which demographic-health group is assigned each COVID death in the
NVSS, the following procedure is adopted. First, for each chronic health condition, a literature
review was conducted to find the COVID mortality risk relative to a healthy individual. Thus,
for example, a person with diabetes has a 56-percent higher risk of death from COVID than a
healthy individual, according to Williamson et al. (2020).°

Table 1 lists the studies included in the literature review. The table lists, for each health
condition, the COVID mortality hazard ratios implied by the study for individuals with the
condition relative to healthy individuals. The table further details the country and age ranges of
the study sample and any additional notes and assumptions made to arrive at a single relative
hazard number.

The bolded studies are those that were ultimately chosen to determine the COVID
mortality risk associated with each condition. In making this choice, preference was given first
to studies in the United States. Barring that, studies in OECD countries were preferred. Finally,
in case multiple studies remained after the application of these criteria, the ones with a
population age range most closely approximating the over-60 population of interest were
selected.

Next, the number of COVID deaths in 2020 in each group g were acquired from the
NVSS. These deaths were then allocated between the health-conditions / within g based on h's
relative risk of dying from COVID-19, compared to the "healthy" population. This process
assumes the risk of getting COVID is independent of health status, but allows the risk to vary by
demographics. The procedure thus accounts for the documented elevated risk of COVID
exposure by race and ethnicity.

Specifically, the allocation of deaths to health conditions within a demographic group is

accomplished by solving the following system of equations:
H
a9=) dg
h

nd
di = tx, *di,h = 2,...,H,
ny

 

° Averaged over the different HbA 1c levels examined in that study.
where d4 is the number of COVID deaths in group g; af is the number of deaths within each
health category / within group g (the main variables to be solved for); nd is the share of group g
in the h health category; and x, is the relative mortality risk of the / category relative to h=/.
Overall, this procedure for assigning all deaths by their cause to each demographic-health
group yields an estimate of the number of people in each health status of each demographic
group who are still alive by the end of 2020. As data for 2021 may not yet be complete, the 2020
mortality rates are extrapolated to the first quarter of 2021. That is, each health-demographic
type is assumed to lose one quarter as many people in Q1 of 2021 as in all of 2020. The
resulting population as of April 1, 2021 is then the base population for the rest of the analysis.

Estimating the Mortality Rates of the Forecasted 2021 Population

The previous steps yield a forecasted April 2021 population, with a best-guess estimate of
the distribution of that population into demographic-health cells. The final step is to estimate the
mortality of this forecasted population. The difficulty here is estimating the mortality rates of
different demographic-health types.

Rather than estimating the rates directly based on small sample sizes in the HRS
(particularly for minority groups at advanced ages with relatively rare health conditions), the
analysis once more relies on the medical literature. A second review of the literature was
undertaken to arrive at hazard rates for individuals with the chronic conditions of interest relative
to healthy individuals. The results of this review are in Table 2, which is laid out similarly to
Table 1.

With relative hazards in hand, these relative probabilities of death can then be applied to
the estimated population of each (g,h) type, adjusting the mortality rate for the full group g from
existing 2019 life tables by the hazard ratio for the 4 health condition. All that is required is the
baseline mortality probability of group g.

This baseline 2019 probability of death for each group g is estimated in a manner similar
to Wettstein et al. (2021). That is, using the NVSS for the numerator and the ACS for the
denominator, non-parametric death rates for each age are estimated. These rates are then fitted

with a Gompertz-Makeham function using nonlinear least squares. For the purpose of this
analysis, this function is fitted only to ages 60 and older.!° Using these fitted values, 10-year age
bin mortality rates are calculated.

One final hurdle must be overcome, which is that the mortality rates observed in the
population of each group g are composites of the mortality in each subgroup (g,h). The mortality
hazard rates from the medical literature give hazards relative to "healthy" individuals. So the
mortality of rates of individuals in group g who have none of the chronic conditions, that is,
groups (g, 1), must be found.

To do so, the overall mortality rate of g is made to conform to the rate resulting from
aggregating the rates of all the health conditions within g by solving for the mortality rate of
healthy group g-individuals. That is, the mortality of healthy individuals in g, m9, is solved for
in the equation: mg = ) W,2m¥, where m@ is the overall mortality rate of group g and W,? is
the share of g with condition h.

Finally, new forecasted life tables for men and women can be generated by aggregating
the estimated population of each (g,h) type up to the full population by gender and age group.
Here, the mortality of each gender-age type is the weighted average of mortality for all the (g,h)
types comprising the gender-age group, where the weights are given by the forecasted size of
each (g,h) type.

Two crucial assumptions are made at this point: first, that COVID deaths among older
adults are negligible after April 1, 2021. Second, that having had COVID and survived does not
substantially increase the rate of any of the listed health conditions or of all-cause mortality; that
is, "long COVID" has a negligible effect on mortality in the older population. These rosy
assumptions are clearly incorrect, but they provide a conservative estimate of the potential

improvement in mortality rates as a result of COVID eliminating the least healthy.

Results
The process outlined above yields gender-specific life tables for the U.S. population ages
60 and over. Table 3 shows these life tables for men and women in panel A and B, respectively.

To put the new estimates in context, Column 3 shows the 10-year mortality rate implied by

 

10 Restricting the ages targeted by the nonlinear least squares to over 60 yields 10-year mortality rates that are much
closer to those implied by published Social Security Administration life tables than using the full range of ages in
fitting the function.

10
published Social Security Administration (SSA) life tables from 2019. Column 4 shows the 10-
year mortality rate estimated through the Gompertz-Makeham approach, without adjusting for
any of the deaths observed in practice during 2020-2021, COVID or otherwise. Column 5 shows
the forecasted mortality rate accounting for the selection deaths that took place during the first
year of the pandemic.

Before turning to the main result regarding mortality in a post-COVID environment,
comparing the SSA life table to the pre-COVID life table estimated in this analysis provides a
baseline of what the methods here might have predicted in the absence of COVID. The two sets
of numbers are very close, although the estimates in this analysis are consistently slightly lower
than the SSA numbers. Thus, the remainder of the discussion will take the initial estimates from
the current analysis as the baseline, so as not to overestimate any reduction in mortality due to
COVID by choosing different methods for computing pre-COVID mortality.

The main result apparent in Table 3 is that mortality rates should be expected to be
modestly lower post-COVID than what had been expected before the pandemic. The differences
are not large, but they are particularly striking in the oldest age groups, where for both men and
women a 1-percentage point decline in 10-year mortality is expected due to the selection of
mortality during the pandemic's first year.

How large is the estimated reduction in mortality? To give it a proper scale, the
estimated decline in mortality can be compared with the maximum possible declines that might
have been observed based on overall mortality in 2020-2021. Column 6 of Table 3 shows the
predicted life table that would result if, instead of assigning 2020 COVID deaths to the
demographic-health groups from which they were most likely drawn based on relative hazard
rates, COVID deaths were instead mechanically deducted from the (g,h) groups with the highest
mortality rates overall. Thus, if COVID mortality were as selective as possible, given the total
number of actual COVID deaths, Column 6 shows what mortality rates would be by age.

Naturally, the mortality rates in Column 6 are always lower than Column 5, although for
younger ages these differences are not apparent after rounding. To focus in on these differences,
Figure 4 shows the change in 10-year mortality (in red), and the maximal potential change (in
gray). The figure shows that, as might be expected, absolute declines in mortality increase with

age, largely because mortality in general rises sharply with age. Furthermore, pandemic

11
mortality was very selective: mortality declines are more than half of the maximal possible
decline at all ages.!!

However, a more nuanced finding is that the reduction in mortality as a share of the
maximal possible reduction declines with age: for those in their 60s, around 80 percent of the
maximal mortality reduction is forecasted to actually take place. In contrast, for those in their
90s only about 50 percent (for men) and 60 percent (for women) of the possible decline is likely
to be realized. This pattern implies that COVID was selective in its victims of all ages - but
more selective among younger ages, where the most frail were much more likely to succumb
than at older ages, where COVID death was relatively likely to happen to all members of the
cohort.

Conclusion

COVID took a heavy toll in the United States. The populations that bore the brunt of
mortality from COVID were not random; instead, minorities were more likely to be infected and,
conditional on infection, older adults andthose with certain chronic health conditions were more
likely to suffer severe illness and death.

A consequence of COVID mortality selection is that those who lived through the
pandemic are a slightly different population than those who entered the pandemic. The
differences lie in both demographic and health condition distributions. Survivors of the first year
of the pandemic are therefore less likely to be members of some of these high-mortality groups.
This selection might have resulted in a lower mortality rate for the next few years than would
have prevailed if the pandemic had not taken place.

This analysis shows that while, directionally, the selection effect is likely to reduce
mortality in the near future, the magnitude of the impact of selection on near-term mortality rates
is modest. Mortality is anticipated to decline by around one percentage point among those ages
90-99, and less at younger ages.

In addition, the assumptions made in the analysis were conservative from the perspective
of OASI's finances. In particular, the analysis assumed that COVID will not directly cause
higher mortality going forward and that long COVID will not be associated with elevated

 

1! Note that, had COVID mortality been completely random, no decline in mortality relative to pre-COVID forecasts
would be expected.

12
mortality risk in the future. Both these assumptions are clearly optimistic. In conjunction with
the small impacts that selection among survivors of the first year of the pandemic might have on

mortality, we are likely to continue observing above-trend mortality in the next few years.

13
References

Alsan, M., A. Chandra, and K. Simon. 2021. "The Great Unequalizer: Initial Health Effects of
COVID-19 in the United States." The Journal of Economic Perspectives 35(3): 25-46.

Aveyard, P. et al. 2021. "Association between Pre-Existing Respiratory Disease and Its
Treatment, and Severe COVID-19: A Population Cohort Study." The Lancet Respiratory
Medicine 9(8): 909-923.

Badar, A. A. et al. 2015. "Clinical Characteristics and Outcomes of Patients with Coronary
Artery Disease and Angina: Analysis of the Irbesartan in Patients with Heart Failure and
Preserved Systolic Function Trial: Analysis of the Irbesartan in Patients with Heart
Failure and Preserved Systolic Function Trial." Circulation: Heart Failure 8(4): 717-724.

Centers for Disease Control and Prevention. 2021. "Evidence Use to Update the List of
Underlying Medical Conditions That Increase a Person's Risk of Severe Illness from
COVID-19." Atlanta, GA. Available at: https://www.cdc.gov/coronavirus/2019-
ncov/science/science-briefs/underlying-evidence-table.html

Centers for Disease Control and Prevention, National Center for Health Statistics. 2021.
Underlying Cause of Death 1999-2020 on CDC WONDER Online Database. Atlanta,
GA.

Dyer, Owen. 2020. "COVID-19: Black People and Other Minorities Are Hardest Hit in US."
BMJ 369: m1483.

ElGohary, G. M. et al. 2020. "The Risk and Prognosis of COVID-19 Infection in Cancer
Patients: A Systematic Review and Meta-Analysis." Hematology/Oncology and Stem
Cell Therapy.

Flegal, K. M., B. K. Kit, H. Orpana, and B. I. Graubard. 2013. "Association of All-Cause
Mortality with Overweight and Obesity Using Standard Body Mass Index Categories: A
Systematic Review and Meta-Analysis." JAMA 309(1): 71.

Florez-Perdomo, William Andrés, Sergio Andrés Serrato- Vargas, Pilar Bosque-Varela, Luis
Rafael Moscote-Salazar, Andrei F. Joaquim, Amit Agrawal, Alvaro Ricardo Soto-Angel,
and Leidy Tatiana Tovar-Montenegro. 2020. "Relationship between the History of
Cerebrovascular Disease and Mortality in COVID-19 Patients: A Systematic Review and
Meta-Analysis." Clinical Neurology and Neurosurgery 197: 106183.

Gu, T. et al. 2020. "History of Coronary Heart Disease Increased the Mortality Rate of Patients
with COVID-19: A Nested Case-Control Study." BMJ Open 10(9): e038976.

Hajifathalian, K. et al. 2020. "Obesity Is Associated with Worse Outcomes in COVID-19:
Analysis of Early Data from New York City." Obesity (Silver Spring) 28(9): 1606-1612.

14
Harrison, Stephanie L. et al. 2020. "Comorbidities Associated with Mortality in 31,461 Adults
with COVID-19 in the United States: A Federated Electronic Medical Record Analysis."
PLoS Medicine 17(9): ¢1003321.

Hooper, Monica Webb, Anna Maria Napoles, and Eliseo J. Pérez-Stable. 2020. "COVID-19 and
Racial/Ethnic Disparities." JAMA 323(24): 2466-2467.

Ikeda, A. et al. 2009. "Living Arrangement and Coronary Heart Disease: The JPHC Study."
Heart (British Cardiac Society) 95(7): 577-583.

Imam, Zaid. 2020. "Older Age and Comorbidity Are Independent Mortality Predictors in a Large
Cohort of 1,305 COVID-19 Patients in Michigan, United States." Journal of Internal
Medicine 288(4): 469-476.

Ketchandji, Melanie, Yong-Fang Kuo, Vahakn B. Shahinian, and James S. Goodwin. 2009.
"Cause of Death in Older Men After the Diagnosis of Prostate Cancer: Death in Older
Men with Prostate Cancer." Journal of the American Geriatrics Society 57(1): 24-30.

Khoury, E. et al. 2022. "Differences in Outcomes and Factors Associated with Mortality among
Patients with SARS-Cov-2 Infection and Cancer Compared with Those without Cancer:
A Systematic Review and Meta-Analysis." JAMA 5(5): e2210880.

Li, Jie et al. 2021. "Epidemiology of COVID-19: A Systematic Review and Meta-Analysis of
Clinical Characteristics, Risk Factors, and Outcomes." Journal of Medical Virology
93(3): 1449-1458.

Lowe, L. P., P. Greenland, K. J. Ruth, A. R. Dyer, R. Stamler, and J. Stamler. 1998. "Impact of
Major Cardiovascular Disease Risk Factors, Particularly in Combination, on 22-Year
Mortality in Women and Men." Archives of Internal Medicine 158(18): 2007-2014.

Mangone, L. et al. 2021. "Cumulative COVID-19 Incidence, Mortality and Prognosis in Cancer
Survivors: A Population-Based Study in Reggio Emilia, Northern Italy." International
Journal of Cancer 149(4): 820-826.

Mannino, D. M., A. Buist, T. Petty, P. Enright, and S. Redd. 2003. "Lung Function and Mortality
in the United States: Data from the First National Health and Nutrition Examination
Survey Follow Up Study." Thorax 58(5): 388-393.

Mulligan, C. and R. Arnott. 2022. "Non-Covid Excess Deaths, 2020-21: Collateral Damage of
Policy Choices?" Working Paper 30104. Cambridge, MA: National Bureau of Economic
Research.

Nakeshbandi, M. et al. 2020. "The Impact of Obesity on COVID-19 Complications: A
Retrospective Cohort Study." International Journal of Obesity 44(9): 1832-1837.

15
Pedone, Claudio, Simone Scarlata, Claudio Sorino, Francesco Forastiere, Vincenzo
Bellia, and Raffaele Antonelli Incalzi. 2010. "Does Mild COPD Affect Prognosis in the
Elderly?" BMC Pulmonary Medicine 10(1): 35.

Pranata, Raymond, Ian Huang, Michael Anthonius Lim, Eka Julianta Wahjoepramono, Julius
July. 2020. "Impact of Cerebrovascular and Cardiovascular Diseases on Mortality and
Severity of COVID-19-Systematic Review, Meta-Analysis, and Meta-

Regression." Journal of Stroke and Cerebrovascular Diseases 29(8): 104949.

Rabbani, G. et al. 2021. "Pre-existing COPD Is Associated with an Increased Risk of Mortality
and Severity in COVID-19: A Rapid Systematic Review and Meta-Analysis." Expert
Review of Respiratory Medicine 15(5): 705-716.

Razmara, A. et al. 2017. "Depression Is Associated with a Higher Risk of Death among Stroke
Survivors." Journal of Stroke and Cerebrovascular Diseases 26(12): 2870-2879.

Ruhm, C. J. 2021. "Excess Deaths in the United States During the First Year of COVID-
19." Working Paper 29503. Cambridge, MA: National Bureau of Economic Research.

Sarkar, S. et al. 2021. "Health Disparity and COVID-19 - A Retrospective Analysis." Health
Science Reports 4(3): p. e345.

Schonberg, M. A. et al. 2011. "Causes of Death and Relative Survival of Older Women After a
Breast Cancer Diagnosis." Journal of Clinical Oncology 29(12): 1570-1577.

Soedamah-Muthu, S. S. et al. 2006. "All-Cause Mortality Rates in Patients with Type 1 Diabetes
Mellitus Compared with a Non-Diabetic Population from the UK General Practice
Research Database, 1992-1999," Diabetologia 49(4): 660-666.

Tu, H. et al. 2018 "Cancer Risk Associated with Chronic Diseases and Disease Markers:
Prospective Cohort Study." BMJ (Clinical Research Ed.) 360: k134.

U.S. Social Security Administration. 2015. "Technical Panel on Assumptions and Methods."
Washington, DC.

U.S. Social Security Administration. 2020. "Memorandum: Updated Baseline for Actuarial
Status of the OASI and DI Trust Funds, Reflecting Pandemic and Recession Effects."
Washington, DC.

Wettstein, Gal, Alicia H. Munnell, Wenliang Hou, and Nilufer Gok. 2021. "The Value of
Annuities." Working Paper 2021-5. Chestnut Hill, MA: Center for Retirement Research
at Boston College.

Williamson, E. J. et al. 2020. "Factors Associated with COVID-19-Related Death using
OpenSAFELY." Nature 584(7821): 430-436.

16
World Health Organization. 2022. "WHO Coronavirus (COVID-19) Dashboard." Geneva,
Switzerland. Available at: https://COVID19.who. int/

Zeng, J. et al. 2021. "The Association between BMI and metabolically Unhealthy Status with
COVID-19 Mortality: Based on 3,019 Inpatients from Wuhan, China," Nutrition,
Metabolism, and Cardiovascular Diseases: NUCD 31(11): 3219-3226.

Zhu, L. et al. 2020. "Association of Blood Glucose Control and Outcomes in Patients with
COVID-19 and Pre-Existing Type 2 Diabetes." Cell Metabolism 31(6): 1068-1077.e3.

Zucker, Inbar, Tamy Shohat, Rachel Dankner, and Gabriel Chodick. 2017. "New Onset Diabetes
in Adulthood Is Associated with a Substantial Risk for Mortality at All Ages: A
Population Based Historical Cohort Study with a Decade-Long Follow-

Up." Cardiovascular Diabetology 16: 105.

17
81

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Oey ICY SOUSpIOUT G/< pue YS> ATPI] 6C°1 sour) (1Z0Z) Te 39 'elon'y] 'suosueyy
SISATeUe 8}9UI WOL Sorpnys vsn , p19 '
Ire Jo Wu ow Woy pofeay P60!  nedg 'eu CEE 190k) (0Z0Z) 'Te 49 "epeyD "Areyoold
PL-SE -_--S@LNUNOD BZ 69'I Ja00%,) (ZZ0Z) Te 19 "eurarg "Amouy
(JvaisojojpMayuoU
pup pvr1sojojpmay sapnjout) +81 yn OTT sdIUDD (0207) 10 12 "fq 'uosmmyiY
SUN] SISOUSDIP 4040 pasvsaap
+0r VIN 8€Z sree (0Z0Z) Te 19 'puomdey 'ereuRIg
JBTNOSBAOIGSIOT)

+81 ya 917 UY (0707) 10 12 "fy 'uosumyy
sisA[eue-e}OUW ¢ WY E1'€ +81 SN Mn "euryD 61  asvasiq sunT (1Z0Z) 'Te 19 "WepoH 'Tueqqey
DUYISD PU" Td OD 4240 pasviaap +07 ya cr i Me aspasig sun] (IZ0@) 'Jv Ja nog 'pavdaap
uowom Joy 77] "ATUO syusHed pO¢ , (0707)
'AQ WOK MON { ISNY OAT +81 vs vt "8°40 39 'poueyoyy '"puequsoyeN

APIA 6¢ "siusned . AjISo ( ) Te 10 "Yoary "uereujeysile
OL, 'AID MOK MAN "Fsny oanEpoy +81 vsn SIT Is9qQ  (0Z0Z) 'Te 19 "Youry "uerepjeytfey

ES"T :(sre0A ; ap ia "Burp

eul Iso e Jo 'Sui 'suo
Co>)MH 68°¢ :(pI0 wok CO=<) HH +81 Ne) v7 AyIsoqO (1ZOZ) 'Te 39 "Burp 'Suoz
Sasspjo Ajisago Jao paspsady +81 ya bOT 4j1saqQ (07207) 10 32 "fA 'uosupiyyi

- eur . oseesid 'Te 19 'UBIT 'n
v6-bZ myo ZV syeopy Areu01075 (0Z0Z) 'Te 19 "UeI], "ND
+81 ya soLIT een (0207) 1018 "fy 'uosmmypLY

jAvay] 21u0dy) "
69°09 05 +02 OorXayAl 6r'l soyoqeic] (1202) "0 BOsLIQ HosT00,
SI [p'] 'sdnoss ode sesh QT Aq YH . . '

SL-81 Bury) 6r'l soyoqrid (0Z0Z) 'Te 39 "enyry "nyZ
8249] ITVQH 4240 paspsaay +81 ya 9ST sajaquid (0202) 10 1a "fy 'uosumiyiy
SOJON osuel os V Ayvuno7) YH Uoylpuos wyeoH Apis

 

soupy pavzoy AYDIAOW CIAQD uo aaniosayyy] pajaajas "| s[qe
61

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

yes] Josue.) Jo ysry +81 uemiel = S| your') (S10Z) 'Te 19 'Sueyeny 'ny,
4adUuD) IspaIq ' '
'fo sans sano paSnacay +29 ysQ Itt sa2UDD Isvasg ([[QZ) 10 1a "y van 'Bsaquoyos
daouvo ajojsoid ' '
fo sans sano paSnusay 48-99 VSQ SET ADDU] BIDISOLT (6007) 19 Ia 'alunjayy 'Upuvynay
O8-0€ Ajeyy 29 eur) QL 7 dSBOSIC] Je[NOSeAOIGSIOZ (OZOZ) 'Te 19 OWOpIag-zd10[ 4
SOYO.NS YIM USUIOM . . cee
10} pi" 3 USUI 10F €8'T 69-0b ueder §=610'% = SYONS oFWayos] ano (6002) 'Te 38 'Tv "epax]
bL-S9 ysa LST ayous (L107) 1» 3a 'NV 'vapuxny
+69 Ay 97 I ddOd prea (O10Z) 'orpne[D 'ouopeg
fo &ptssags 1940 poSpaaay bL-SZ VSQ §ST asvasig sun] (£007) 19 12 "Wd 'ouluunyy
ay) suIpnyjaul
sistjpun-vjayy os-r¢ 8 on 'june - , eT 4918290 (E107) 10 ta "W autdayjwy yosa1 J
st CAD HHA woMI0M 10) r9-Ob VSN PET esvosiq senoseacrpreg (3661) 'Te 30 "Wuky 'omoT]
nuisuy/qV2 +09 VSQ == €T asnasig Auajay Aanuos0g (S107) 1 1a "y ayy "anpog
Jopuaes pue 'dnois ode Ag +-SE MN 8'€-9'T sojoqeiq] [ odAT, (9002) 'S*S "nyN}-Yyewepsog
+S€ eis] =. SET sajoqeiq (LOZ) 'Te 19 "Tequy "JoyonZ
$9 asp uvayy VSQ 6771 sajaqvid (6107) '10 Ja 'uvavypus 'uvanysoy
SON dssURlosy Aqyunos YH UOTIPUOS TTeoH Apis

 

AQIJDIAOPY ASNDZ ]]P UO aANJDABIT'T PajIajag "7 S|qeL
Table 3. 10-year Mortality Rates, SSA 2019 Life Tables versus Post-COVID Adjustments

 

 

 

Gender Age group SSA Initial rate | Adjusted nossible effec t
M 60-69 14.7% 14.6% 14.6% 14.6%

M 70-79 29.6 29.2 29.1 29.0

M 80-89 63.1 61.9 61.5 61.2

M 90-99 94.3 94.9 93.8 92.7

F 60-69 9.2 9.2 9.2 9.2

F 70-79 21.7 20.8 20.7 20.7

F 80-89 53.1 51.4 51.0 50.9

F 90-99 90.1 91.4 90.3 89.6

 

Sources: SSA (2019); and authors' calculations.

Figure 1. Distribution of Health Conditions for 60+ in HRS, 1992-2018

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

60%
- - Healthy
\ Cancer
N Stroke
N . Lung disease
5 Seek Diabetes
40% ~~. Heart disease
~AL Obesity
20% --$-- -
7
0% -

 

1992 1996 2000 2004 2008 2012 2016

Source: Authors' calculations from University of Michigan, Health and Retirement Study (HRS) (1992-2018).

20
Figure 2a. Distribution of Health Conditions by Age Group for Men

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

100% 4 7
75% EY '
50% +
25% Th 9% 6%
34%
23% 18% 22%
0% : : !
60-69 70-79 80-89 90-99

 

| CO Healthy 0 Cancer & Stroke @ Lung @ Diabetes @ Heart @ Obesity,

 

Source: HRS (2018).

Figure 2b. Distribution of Health Conditions by Age Group for Women

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

100% - "7
ao ws a
75% T Vie °
50% -
6%
25% ~~ = ~
0
0)
39% 27% 28% 27%
0% 1
60-69 70-79 80-89 90-99

 

| 0 Healthy 0 Cancer @ Stroke S@ Lung & Diabetes @ Heart ™ Obesity

 

Source: HRS (2018).

21

 

 
Figure 3a. Distribution of Health Conditions by Age Group for Whites

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

100% Men 100% omen
0 7 oO iy
. PY Zi ae 00, ney
v (0) , v
75% se aa 75% -Hay uh a"
50% - 50%
7%
° 7% 11%
9%
25% 8% Ta 6%, 25% , °
34% , 38% 30% 28% 0
23% 17% 20% 25%
0% T T T 0% T T T
60-69 70-79 80-89 90-99 60-69 70-79 80-89 90-99
OHealthy OCancer G@Stroke WMLung OHealthy GCancer G@Stroke MLung
HDiabetes @ Heart @ Obesity & Diabetes @ Heart # Obesity
Source: HRS (2018).
Figure 3b. Distribution of Health Conditions by Age Group for Blacks
Men Women
100% ; 100% ;
BY, 12% aA) ur ae
ry, 1, vLy, RY
75% 75% yf
50% +- 50% 6%
11%
9 0
25% 5% we 25% [846 49%
30% 40% = °
° 23% 25% 22% 16% 20%
0% T T T 0% T T
60-69 70-79 80-89 90-99 60-69 70-79 80-89 90-99
OHealthy OCancer O®Stroke WMLung OHealthy OCancer GStroke WMLung
G Diabetes @ Heart @ Obesity HB Diabetes @ Heart @ Obesity

 

 

 

Source: HRS (2018).

22

 

 

 
Figure 3c. Distribution of Health Conditions by Age Group for Hispanics

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

  

   

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

100% wen yg 100% Women oF
ee aA ae o Uh 1%
iy ' UZ
: LA wy,
75% 75% :
50% 50%
9% 6% 6%
25% VA 6% 25% 6
31% ° 0 32%
0% T T T 0% T T T
60-69 70-79 80-89 90-99 60-69 70-79 80-89 90-99
Healthy OCancer GStroke MHLung OHealthy OCancer GStroke MLung
@ Diabetes ™ Heart @ Obesity @ Diabetes @ Heart _™ Obesity
Source: HRS (2018).
Figure 3d. Distribution of Health Conditions by Age Group for the Other Race Group
100% Men 100% 7 (Omen
Key % , Koy
why; W : y
75% 75% 25%
50% 50%
0 0
0% 1% ~ 0%
0 0 0%
25% 43% 25% 39% 38%
31% 24% 26%
0% 1 0% 1
60-69 70-79 80-89 90-99 60-69 70-79 80-89 90-99
OHealthy OCancer Stroke MLung OHealthy OCancer Stroke WLung
@ Diabetes @ Heart @ Obesity @ Diabetes ™ Heart @ Obesity

 

 

 

Source: HRS (2018).

23

 

 

 
Figure 4. The Change in 10-year Mortality, and the Maximal Potential Change

 

 

 

 

 

 

 

 

 

 

 

 

0% 7 T _-e
= i !
I
I
I
a I
~ I
3% |
Q I
o I
op) !
g I
Sc I
8 I
5 -2% i
Q4 oo
I
@ Estimated change
I
© Maximal potential change |!
-3% '

SP SP KF SF NM F
sr wr YF YT F YT LY

ewe m» m Ew Rue Pua ENS mm
@ Ww Ww Ww 3 Psa @ oO ©

SS oF So

Source: Authors' calculations.

24
Appendix

Appendix Table Al.

 

Age group Demographic group Cancer Stroke Lung Diabetes Heart Obesity

 

60-69 White male - - - - - -
60-69 White female - - - - - -
60-69 Black male - - - - - -
60-69 Black female - - - - - -
60-69 Hispanic male - - - - - -
60-69 Hispanic female - - - - - -
60-69 Other male xX - - - - -
60-69 Other female - - - - - -
70-79 White male - - - - - -
70-79 White female - - - - - -
70-79 Black male - - - - - -
70-79 Black female - - - - - -
70-79 Hispanic male - - - - - -
70-79 Hispanic female - - - - - -
70-79 Other male - - - - - -
70-79 Other female - - - - - -
80-89 White male - - - - - -
80-89 White female - - - - - -
80-89 Black male - - xX - - -
80-89 Black female - - - - - -
80-89 Hispanic male - - - - - -
80-89 Hispanic female -
80-89 Other male xX -
X

80-89 Other female - X - - -
90-99 White male - - x - - -
90-99 White female - - - - - -
90-99 Black male - - x - x ><
90-99 Black female - - Ȣ - - -
90-99 Hispanic male - - Xx - - -
90-99 Hispanic female x - X - - -
90-99 Other male xX xX xX - - -
90-99 Other female xX - - xX - -

 

Source: Authors' calculations.

25
RECENT WORKING PAPERS FROM THE
CENTER FOR RETIREMENT RESEARCH AT BOSTON COLLEGE

Understanding the Increased Financial Hardship Experienced by Older Adults with
Disabilities during the COVID-19 Pandemic
Zachary A. Morris, August 2022

A Framework for Evaluating the Adequacy of Disability Benefit Programs and Its
Application to U.S. Social Security Disability
Zachary A. Morris, August 2022

Comparative Regression Discontinuity and Regression Discontinuity as Alternatives to
RCT for Estimating Average Treatment Effects
Duncan Chaplin, Charles Tilley, Denise Hoffman, and John T. Jones, August 2022

Work Overpayments Among New Social Security Disability Insurance Beneficiaries
Denise Hoffman, Monica Farid, Serge Lukashanets, Michael T. Anderson, and John T. Jones,
July 2022

What Is the Relationship Between Deprivation and Child SSI Participation?
Michael Levere, David Wittenburg, and Jeffrey Hemmeter, May 2022

What Share of Noncovered Public Employees Will Earn Benefits that Fall Short of Social
Security?

Jean-Pierre Aubry, Siyan Liu, Alicia H. Munnell, Laura D. Quinby, and Glenn Springstead,
April 2022

Employer Concentration and Labor Force Participation
Angi Chen, Laura D. Quinby, and Gal Wettstein, March 2022

Will the Jobs of the Future Support an Older Workforce?
Robert L. Siliciano and Gal Wettstein, March 2022

Employment Outcomes for Social Security Disability Insurance Applicants Who Use
Opioids
April Yanyuan Wu, Denise Hoffman, Paul O'Leary, and Dara Lee Luca, February 2022

Would 401(k) Participants Use a Social Security "Bridge" Option?
Alicia H. Munnell and Gal Wettstein, December 2021

The Alignment Between Self-Reported and Administrative Measures of Application to and
Receipt of Federal Disability Benefits in the Health and Retirement Study
Jody Schimmel Hyde and Amal Harrati, December 2021

All working papers are available on the Center for Retirement Research website
(https://crr.bc.edu) and can be requested by e-mail (crr@bc.edu) or phone (617-552-1762).

26