Suffering, the Safety Net, and Disparities During COVID-19
Summary
A 2023 journal article, Suffering, the Safety Net, and Disparities During COVID-19, by Marianne P. Bitler, Hilary W. Hoynes and Diane Whitmore Schanzenbach, published in RSF: The Russell Sage Foundation Journal of the Social Sciences 9(3): 32–59. The article examines the economic shock of COVID-19 and the safety net response, focusing on differences across race and ethnicity, using Current Population Survey, Census Household Pulse and administrative SNAP data. Its abstract states that participation in SNAP increased more in counties with a larger employment shock, while the increase in total SNAP benefits was inversely related to that shock. The authors report that the seasonally adjusted unemployment rate rose to 14.7 percent in April from 3.5 percent in February. The article closes with its reference list.
Summary drafted by a model from the document's text below and checked by script against that text before publication. It is a navigation aid, not a reading of what the document proves. Where AI is used
Full text
Suffering, the Safety Net, and
Disparities During COVID-19
M a r i a n n e P. Bi t ler, Hil a ry W. Hoy n es, a n d
Di a n e W hi t mor e Sch a n zen bach
The economic and public health crisis caused by COVID-19 was devastating and disproportionately hurt
Blacks and Hispanics and some other groups. Unemployment rates and other measures of material hardship
were higher and increased more during the crisis among Blacks and Hispanics than among non-Hispanic
Whites. Congress authorized a historic policy response, incorporating both targeted and universal supports,
and expanding both the level and duration of benefits. This response yielded the remarkable result of an
estimated decline in the Supplemental Poverty Measure between 2019 and 2020. We study administrative
data to investigate the impact of the Supplemental Nutrition Assistance Program (SNAP) during the crisis.
We find that participation in SNAP increased more in counties that experienced a larger employment shock.
By contrast, the increase in total SNAP benefits was inversely related to the employment shock. The SNAP
benefit increases were less generous to Black and Hispanic SNAP participants than to White.
Keywords: COVID-19, SNAP, Child Tax Credit; unemployment insurance, Economic Impact payments,
racial-ethnic disparities, poverty, safety net
The COVID-19 crisis led to spiking unemploy- curred in addition to the large increases in
ment rates and unprecedented levels of food mortality and morbidity from COVID, which
hardship that fell disproportionately on low- also fell disproportionately on Blacks and His-
income families and among non-Hispanic panics (Hill and Artiga 2022).1 Food banks and
Black and Hispanic or Latino people. This oc- food pantries reported spikes in need. The re-
Marianne P. Bitler is professor of economics at the University of California, Davis. Hilary W. Hoynes is Haas
Distinguished Chair in Economic Disparities and professor of economics and public policy at the University of
California, Berkeley. Diane Whitmore Schanzenbach is Margaret Walker Alexander Professor of Human De-
velopment and Social Policy and director of the Institute of Policy Research at Northwestern University.
© 2023 Russell Sage Foundation. Bitler, Marianne P., Hilary W. Hoynes, and Diane Whitmore Schanzenbach.
2023. “Suffering, the Safety Net, and Disparities During COVID-19.” RSF: The Russell Sage Foundation Journal
of the Social Sciences 9(3): 32–59. DOI: 10.7758/RSF.2023.9.3.02. The authors thank Ted Carter, Nick Fleming,
Amelia Vasquez, and especially Raheem Chaudhry for excellent research assistance, and Dottie Rosenbaum,
Danny Schneider, Steve Raphael, and Sheldon Danziger for helpful comments. Direct correspondence to: Mar-
ianne Bitler, bitler@ucdavis.edu, Department of Economics, University of California, Davis, 1 Shields Avenue,
Davis, CA 95616.
Open Access Policy: RSF: The Russell Sage Foundation Journal of the Social Sciences is an open access journal.
This article is published under a Creative Commons Attribution-NonCommercial-NoDerivs 3.0 Unported Li-
cense.
1. Members of several other racial and ethnic groups, such as Native Americans and Alaska Natives, Hawaiians,
and Other Pacific Islanders, also suffered more than White and Asian Americans did. For example, after account-
ing for differences by age; adult persons who were Alaska Native or Native American or Hawaiian or Other
s u f f e r i n g , t h e s a f e t y n e t , a n d d i s p a r i t i e s d u r i n g c o v i d -1 9 33
sponse to this crisis from the formal and infor- Pulse decreases in response to relief payments,
mal safety net was robust (Bitler, Hoynes, and including economic impact payments (EIPs)
Schanzenbach 2020). Unemployment insur- and pandemic electronic benefit transfer (EBT)
ance (UI) participation soared as Congress ex- payments (Bauer et al. 2020). Detecting these
panded eligibility for the program, the length policy impacts would not be possible without
of time for which some UI benefits could be the frequent, real-time data. We then examine
received, and payment levels via a series of top- the change in annual food insecurity between
up payments. Participation in the Supplemen- 2019 and 2020, using reported estimates from
tal Nutrition Assistance Program (SNAP) and the CPS-FSS, the usual snapshot measure of an-
benefit levels increased. A series of relief pay- nual food insecurity. The annual food insecu-
ments provided cash to qualifying individuals. rity data show that whereas non-Hispanic
In this article, we examine the impact of the Whites (Whites) and White-headed families
economic shock and the safety net response to with children experienced a reduction in food
the COVID-19 crisis, focusing on differences insecurity from 2019 to 2020, non-Hispanic
across race and ethnicity. We also investigate Black (Black) and Hispanic families and Black
the relationship between safety net responses and Hispanic families with children faced sub-
and the alleviation of suffering; attempting to stantial increases in food insecurity from 2019
better understand the extent to which different to 2020, suggesting uneven impacts of both
groups experienced hardship at different levels, COVID and of the ability of the safety net to pro-
the extent to which the safety net responded vide protection against shocks. Both sources of
differently across groups, and who fell through food hardship data show large disparities be-
the many holes in the safety net. tween Whites and Black or Hispanic families
We start by analyzing the shock and levels both before and during the pandemic.
of hardship overall and by race and ethnicity, Next, we turn to a discussion of what we
using a combination of the monthly Current would have expected from the safety net based
Population Survey (CPS) to measure the eco- on previous downturns, and contrast that with
nomic shock, the CPS Annual Social and Eco- the COVID policy changes. The COVID re-
nomic Supplement to measure poverty, and the sponse marked an unprecedented expansion
December CPS Food Security Supplement in spending. We present changes in aggregate
(CPS-FSS) and the Census Household Pulse spending over time on UI, SNAP (EBT benefits
(Census Pulse) survey data to measure hard- for food for low-income persons), the Child Tax
ship. It is well known that even in strong labor Credit (CTC) (expanded during COVID to pro-
markets, levels of unemployment and hardship vide most families with children with tax re-
are higher for some racial and ethnic groups bates), and the EIPs (the relatively universal
than for others. We add to this by characteriz- stimulus payments offered to most families
ing the incidence of the COVID economic with low and moderate incomes); using
shock by race and ethnicity. We then turn to Monthly Treasury Statement data tracking fed-
examine the extraordinary safety net response, eral spending. We also discuss the policy re-
how it affected different groups, and who was sponses in these programs. We turn to survey
left out. We characterize suffering with data data from the CPS to investigate the incidence
from two sources. First, the Census Pulse pro- of economic hardship using the Supplemental
vided frequent, real-time data on economic Poverty Measure (SPM). We also document the
well-being that were not captured by our usual individual contributions—holding other fac-
data collection approaches (much of which be- tors constant—of each of our key safety net
came available for the COVID period only with programs to the reduction in SPM poverty ex-
a long lag, or only provides an annual snap- perienced in 2020. We find that the EIPs, the
shot). For example, food insufficiency in the Earned Income Tax Credit (EITC), and UI made
Hispanic Islanders had higher excess death rates due to COVID per hundred thousand in 2020 (relative to
normal rates from 2015 to before COVID) than Whites or Asian Americans (Zalla et al. 2022). These groups are
small in the general population and estimates of their characteristics in survey data are extremely noisy.
r sf: t he russell sage f ou n dat ion jou r na l of t he so ci a l sciences
34 t h e s o c i o e c o n o m i c i m p a c t s o f t h e c o v i d -1 9 p a n d e m i c
the largest contributions to the decline in pov- demic (Bitler, Hoynes, and Schanzenbach
erty for all groups except Black children, who 2020). We make several contributions to this
experienced a larger poverty reduction from literature. First, we update this earlier work
SNAP than from UI. However, the SPM mea- with a focus on the COVID crisis. Second, this
sures only annual poverty and has other limita- is the first article to examine the response of
tions, such as underreporting safety net bene- the social safety net to economic downturns
fits and nonrandom declines in response rates with a focus on examining the impacts across
during the pandemic (for a discussion of the race and ethnicity groups. Third, we focus on
challenges with CPS response rates during families with children, a group characterized
COVID, see Rothbaum and Bee 2021). by high poverty rates and economic vulnerabil-
We conclude with a detailed examination ity. Finally, this article is the first to use county-
of the responsiveness of SNAP participation level SNAP data to correlate changes in partici-
and benefit payments over the COVID crisis for pation and benefit payments with county
several reasons. First, various sources of ad- characteristics and the extent of the shock.
ministrative data on SNAP allow us more accu-
rately to measure the role of SNAP than that of Economic Suffering During
other programs that do not report such data. COVID-19, by Race and Ethnicity
Through 2019, we can describe SNAP receipt by Economic suffering was widespread and dispa-
characteristics such as race, ethnicity, and pres- rate during COVID-19. In this section we dem-
ence of children. Further, through January 2021 onstrate large differences across race and eth-
we can track—using administrative data—par- nicity in the labor-market shock, as well as in
ticipation and benefits received by county, al- material hardship as measured by food insecu-
lowing us to correlate county changes in SNAP rity and related measures.
with factors, including the magnitude of the
labor-market shock and health shock as well as The Labor-Market Shock
a variety of demographic and other character- COVID-19 hit the United States hard in March
istics. Second, SNAP is a relatively large pro- 2020 and President Trump declared a national
gram even in good times, so state-level imple- emergency on March 13. After reaching a busi-
mentation challenges in responding and ness cycle peak in February, the economy
adjusting to the crisis were likely less signifi- plunged as COVID spread, reaching a trough in
cant than for the UI program. Third, SNAP April (and representing the shortest peak to
makes a particularly interesting case study be- trough period since 1957, the start of the Na-
cause it was expanded during COVID to in- tional Bureau of Economic Research business
crease payment levels and to allow for some cycle dates). The seasonally adjusted unem-
temporary waiving of other rules about pro- ployment rate rose to 14.7 percent in April from
gram administration. We explore the extent to 3.5 percent in February and 4.4 percent in
which these policy expansions have offset the March. By October 2021, the unemployment
economic shock and their likely impact on dif- rate was back down to 4.6 percent, but still sta-
ferent demographic groups. Even though areas tistically elevated relative to February 2020.
that experienced a greater economic shock gen- Not only did traditional unemployment go
erally experienced larger increases in SNAP par- up to extraordinary levels, but also the number
ticipation levels, because of the unusual design of persons reporting they had a job but were
of the benefits expansions, they also saw not at work increased substantially. The Bureau
smaller increases in SNAP benefit payments. of Labor Statistics concludes that most of the
This article contributes to a large literature increase in reports of being employed but not
examining the response of the social safety net at work are miscategorized and should be
to economic cycles (Bitler and Hoynes 2010, counted as unemployed—a miscategorization
2016; Bitler, Hoynes, and Iselin 2020; Hardy, that occurred in part because of confusion in
Smeeding, and Ziliak 2018; Mueller, Rothstein, the early days of the pandemic on how workers
and von Wachter 2016; Ziliak 2015). In particu- who expected to experience only a temporary
lar, it builds on work early in the COVID pan- spell of joblessness due to pandemic shut-
r sf: t he russell sage f ou n dat ion jou r na l of t he so ci a l sciences
s u f f e r i n g , t h e s a f e t y n e t , a n d d i s p a r i t i e s d u r i n g c o v i d -1 9 35
downs would answer accurately labor-force sta- nificantly so. Second, figure 1 also shows the
tus questions. In addition, millions left the enormous shock to unemployment rates after
labor force as schools closed and care respon- COVID hit in March 2020, with Hispanic and
sibilities for children and others increased. Black adults experiencing the largest impacts.
These unprecedented labor-market fluctua- Third, by the end of August 2021, Black adult
tions mask large differences across race and unemployment rates remain the most elevated
ethnicity. Even in strong labor markets, differ- (relative to White or Asian adults), followed by
ences across groups are clear, with Black and Hispanic adult rates.
Hispanic Americans experiencing higher un- Many analysts conclude that this unemploy-
employment rates—and often nearly twice as ment rate was understated in the early months
high—than White Americans. We use monthly of the COVID-19 pandemic (see, for example,
CPS micro data to compare unemployment Aaronson 2021). The share of workers reporting
outcomes for race-ethnicity groups over time to be “employed, but not at work” increased
(Flood et al. 2021). We compare four groups in- dramatically, and many of these workers were
cluding those reporting they are Hispanic and likely affected by closures of their place of work
of any race, and those who are non-Hispanic due to COVID-19 and would have been more
and reporting that they are only White, Black, appropriately classified as unemployed (Bureau
or Asian. Based on the February 2020 CPS, 62 of Labor Statistics 2020). Furthermore, early in
percent of the population reported being the pandemic, when stay-at-home orders were
White, 15 percent reported being Hispanic, 12 in place, the unemployed were less likely to
percent reported being Black, and 6 percent re- search for a new job than is typical for a host of
ported being Asian.2 reasons, which led to a spike in the share of
We begin by showing trends in unemploy- people reporting being not in the labor force
ment levels by race-ethnicity. Figure 1 shows (Bureau of Labor Statistics 2020). Further, some
the seasonally unadjusted monthly unemploy- adults may have left the labor force to care for
ment rate among adults ages eighteen through family members. All three of these data issues
sixty-four for every month from January 2019 mean that the measured unemployment rate
through August 2021, by race-ethnicity. We understates the true experience of COVID-
show these rates for White, Black, Hispanic, inflicted labor-market shocks. Thus we also
and Asian adults. The x-axis denotes calendar look at an alternative measure of the shock—
time, and the y-axis is the unemployment rate changes over time in the share of adults who
(in percentage points) for each group. The are unemployed, not in the labor force, or have
filled-in markers for Black, Hispanic, and Asian a job and are not at work during the survey
adults indicate the estimate is statistically sig- week, where we difference out the shares rela-
nificantly different from the value for Whites tive to the same calendar month during the
(for that month). Several facts are notable. twelve months before March 2020. In particu-
First, Black and Hispanic adults have persis- lar, for each race-ethnicity group we estimate a
tently higher unemployment rates than White regression model with indicator variables for
adults, even in the booming labor market lead- each month in the COVID period (March 2020
ing up to the COVID crisis. In March 2020, through August 2021) along with indicator vari-
White adults experienced an unemployment ables for each calendar month. We adjust the
rate of 2.8 percent, versus 5.0 percent for Black standard errors for clustering at the state level.
adults and 4.7 percent for Hispanic adults. Figure 2 shows these estimated monthly shocks
Asian adults tended to have lower unemploy- for each race-ethnicity group. As with figure 1,
ment rates than White adults in the months solid (hollow) symbols for Black, Hispanic, and
leading up to COVID, but not statistically sig- Asian adults indicate that the coefficient is (is
2. Because of the small shares of the population, our analysis excludes those reporting non-Hispanic American
Indian, Alaska Native, Hawaiian Native, or Pacific Islander (1 percent) and those reporting non-Hispanic multiple
race (2 percent). We omit them and those who refused or did not know or did not answer (1.8 percent) from the
graphs, but include them in all the regressions and comparisons.
r sf: t he russell sage f ou n dat ion jou r na l of t he so ci a l sciences
36 t h e s o c i o e c o n o m i c i m p a c t s o f t h e c o v i d -1 9 p a n d e m i c
Figure 1. Unemployment Rate, by Race and Ethnicity
14
12
10
Unemployment rate, percent
8
6
4
2
0
Ja n
Fe ua
br r y
u a 20
M r y 19
a r 20
c 1
Ap h 2 9
ri 019
M l 20
a 1
Ju y 20 9
ne 19
Ju 20
Se Aug ly 2 19
pt u 01
em s t 9
O be 201
c
N to r 2 9
ov b 01
D em r 2 9e
ec b 01
em e r 9
Ja be 01
n r 9
Fe ua 20
br r y 19
u a 20 2
M r y 20
a r 20
c 2
Ap h 2 0
r i 020
M l 20
a 2
Ju y 20 0
ne 20
Ju 20
Se Aug ly 2 20
p t u 02
em s t 0
O b e 202
N cto r 2 0
o v b 02
D em r 2 0
e c b 02 e
em e r 0
Ja be 02
nu r 2 0
Fe a 0 2
br r y 20
u a 20
M r y 21
a r 20
c 2
Ap h 2 1
ri 021
M l 20
a 2
Ju y 20 1
ne 21
J 2
Au uly 021
g u 20
st 21
2021
White Black Hispanic Asian
Source: Authors’ calculations from Current Population Survey, as compiled by IPUMS (Flood et al.
2021).
Note: Data for adults ages eighteen through sixty-four. Solid (hollow) symbols for Black adults, His-
panic adults, and Asian adults indicate that the coefficient is (is not) statistically significantly different
from the unemployment rate among White adults in the same month. Calculations use sample weights
and cluster the standard errors at the state level. Groups are mutually exclusive (with, for example,
Black being short for non-Hispanic Black).
not) statistically significantly different from the percentage points among White adults, 15.9
unemployment rate among White adults in the percentage points among Black adults, 17.4 per-
same month. centage points among Hispanic adults, and 14.1
Figure 2 shows that this broader shock hit percentage points among Asian adults. A year
Hispanic adults and Black adults even harder later, in April 2021, the increase among White
than White adults, who already experienced an adults had fallen to 2.5 percentage points, ver-
enormous shock. In April 2020, the increase in sus 5.5 and 4.7 percentage points among Black
the sum of those unemployed plus those re- and Hispanic adults, respectively. Asian adults
porting being not in the labor force plus those generally returned to values no different from
reporting having a job and not at work was 12.7 White adults by August 2020.3
3. American Indians, Alaska Natives, Hawaiian Natives, and Pacific Islanders generally had higher levels of un-
employment pre-COVID and had had increases in unemployment (relative to pre-crisis monthly averages) that
were statistically indistinguishable from Whites, and Multiple Race adults had higher levels pre-COVID and
higher increases than Whites (not shown on graph).
r sf: t he russell sage f ou n dat ion jou r na l of t he so ci a l sciences
s u f f e r i n g , t h e s a f e t y n e t , a n d d i s p a r i t i e s d u r i n g c o v i d -1 9 37
Figure 2. Broader Labor Market Shock, by Race and Ethnicity
20
15
White Black Hispanic Asian
Percentage point change
10
5
0
ay
2020l2 Ju ne 20
20 20
20
02 Se Au Ju ly
g u 20
st 20
0
pt
O em be r220
0220
–5
N ov ct ob er 0
ch ri D ecJa
em
em be
be r
r2
20
20
02
20
20
ar Ap M Fe br
M
nuua
ar
ar
ryy220020
1
21
M ApM
ch
ri
ay
20
l202
20
21
1
Ju
June
ly 20 21
21
Au g u 21
st 20
20 21
Source: Authors’ calculations from Current Population Survey, as compiled by IPUMS (Flood et al.
2021).
Note: Labor-market shock calculated as the change in the rates of “unemployment, not in labor force,
and employed but not at work” relative to the same month in the year prior to March 2020. Data for
adults ages eighteen through sixty-four. For White adults, solid (hollow) symbols indicate the change in
the unemployment measure is statistically (not) different from zero. For the other groups, solid (hollow)
symbols indicate that the coefficient is (is not) statistically significantly different from the unemploy-
ment measure among White adults in the same month. Calculations use sample weights and cluster
the standard errors at the state level.
Measures of Material Hardship (U.S. Census Bureau 2020c). Nonetheless, the
In the early days of the pandemic, food banks data—especially the food hardship data—
reported dramatic surges in need for emer- have been shown to be sensitive to changes in
gency relief. Within weeks, survey data be- economic conditions and receipt of relief pay-
came available to track food hardship over ments. For example, Lauren Bauer and her
the course of the pandemic. One of the most colleagues (2020) show that reported food
important sources of real-time data on eco- hardship declines among low-income fami-
nomic hardship is the Census Bureau’s ex lies in the weeks after pandemic EBT pay-
perimental Household Pulse Survey, which ments for missed school meals are received
released new data first every week then sub- across states.
sequently every two weeks during the course The share of adult respondents with chil-
of the pandemic. To be sure, the data are im- dren, by race and ethnicity, and adult respon-
perfect, characterized by low response rates dents, by race and ethnicity, who answered that
(not atypical for online surveys) and imper- they sometimes or often did not have enough
fect sample designs and, in some cases, can- to eat during the prior week from April 2020
not be directly compared with other sources through October 2021 are presented in the on-
r sf: t he russell sage f ou n dat ion jou r na l of t he so ci a l sciences
38 t h e s o c i o e c o n o m i c i m p a c t s o f t h e c o v i d -1 9 p a n d e m i c
Figure 3. Annual Food Insecurity, by Race, Ethnicity, and Presence of Children, 2019–2020
30
2019 2020 27.3
25
23.0
21.7 21.8
Percent with food insecurity
20 19.1
17.2 17.0
15.6
15 14.8
13.6
10.5 10.5 10.7
10 9.7
7.9
7.1
5
0
Overall White Black Hispanic Overall White Black Hispanic
All households Households with children
Source: Coleman-Jensen et al. 2020, 2021.
line appendix (see figures 1a and 1b).4 Despite slight or no confidence in their ability to pay
period-to-period variation, the share generally their next housing payment. Between 8 and 9
climbed during the fall of 2020 and fell—some- percent of Black and Hispanic respondents re-
times sharply when relief payments were ported that they received food from a food pan-
paid—starting in January 2021. Rates of food try in the prior week, relative to around 2 per-
hardship are generally twice as high among cent of White and Asian respondents.
Black and Hispanic families with children as Annual food insecurity data have been col-
they are among White and Asian families with lected in the December Current Population
children. Food hardship rates among those Survey for nearly twenty years and provide a
with children are uniformly higher than for the consistently measured annual snapshot of food
overall population. Rates among Blacks and hardship. Rates by race and ethnicity, and by
Hispanics in the general population are gener- presence of children, in 2019 and 2020 are pre-
ally two to three times those among Whites and sented in figure 3. The overall household food
Asians. insecurity rate was unchanged across the two
Similar patterns across race and ethnicity years, but the average masks heterogeneous ex-
are found in the Census Household Pulse data periences across groups. Black and Hispanic
in other financial hardship domains. Relative persons experienced higher food insecurity in
to White and Asian respondents, Black and 2020 relative to 2019; Whites experienced a de-
Hispanic respondents are substantially more cline. Among households with children, the
likely to report that it was somewhat or very dif- same pattern holds but the magnitudes of the
ficult to pay for their usual household expenses, increases among Black and Hispanic families
and a higher share reported that they had only is larger.5
4. See the online appendix (https://www.rsfjournal.org/content/9/3/32/tab-supplemental).
5. Jonathan Rothbaum and Adam Bee (2021) suggest disruptions to some CPS response rates, with those ex-
pected to have lower incomes having lower response rates.
r sf: t he russell sage f ou n dat ion jou r na l of t he so ci a l sciences
s u f f e r i n g , t h e s a f e t y n e t , a n d d i s p a r i t i e s d u r i n g c o v i d -1 9 39
S a fe t y N e t a n d S o c i a l I n s ur a n c e nomic recessions and expansions and the ex-
R e s p o n s e to t h e Pa n d e m i c tent to which it does. In particular, the litera-
This section presents existing evidence on how ture examines the extent to which safety net
our safety net responds in economic recessions programs are countercyclical (spending and
and how the response has changed over time participation rise during recessions and fall
before discussing the relief bills implemented during expansions) thereby providing needed
during the COVID crisis. assistance during economic downturns or pro-
cyclical (the opposite pattern). For example,
Programs and Evidence from Marianne Bitler and Hilary Hoynes (2016) use
Prior Recessions data through 2012 to explore how per capita
The United States has many programs that help real spending on various safety net programs
low-income families smooth their consump- responds to changes in local labor-market con-
tion in economic downturns and avoid hunger, ditions measured by state-year unemployment
poverty, or other negative outcomes. These in- rates. They find that UI, including the fully
clude social insurance programs—with the federally funded extensions and emergency
most relevant such program being unemploy- programs as well as the usual state and state-
ment insurance. These social insurance pro- federal program, is the most countercyclical
grams are universal (not income targeted), are program, although SNAP also has a strong
paid for using payroll taxes while working, and countercyclical response. Bitler, Hoynes, and
are triggered by an event, such as losing one’s John Iselin (2020) extend that work and docu-
job through no fault of one’s own for UI. Addi- ment the countercyclical nature of a host of
tionally, means-tested safety net programs such programs pre-COVID using data through 2019.
as SNAP, a program for low-income, low-asset They find that since 2007, UI has shown a par-
individuals and families, provide benefits in ticularly robust countercyclical response, with
the form of grocery vouchers, which are deliv- a 1 percentage point increase in the unemploy-
ered by EBT card. It also includes tax credits ment rate leading to an 18 percent increase in
such as the EITC or the CTC, which provide re- UI spending. SNAP has a significant economi-
fundable (or partially refundable) tax credits to cally meaningful but weaker response, with a 1
eligible families with earned income as well as percentage point increase in the unemploy-
cash benefits through Temporary Assistance ment rate leading to a 7 percent increase in
for Needy Families (TANF). SNAP spending. Interestingly, neither the work-
In response to the massive economic shock conditioned tax credits (EITC) nor cash welfare
and increase in material hardship associated for families with children (TANF) provide any
with the COVID crisis, the pre-COVID U.S. countercyclical response to economic down-
safety net, under then-current law, would have turns, as might be expected given their eligibil-
provided some protection. In addition, in se- ity rules and timing of EITC payout and the fact
vere downturns, Congress often enhances the that TANF spending has been fixed in nominal
generosity of existing programs. For example, terms since 1996.
Congress can authorize emergency unemploy- To put these responses into context, we
ment compensation (which tends to be fully highlight how the U.S. social safety net has
federally funded), such as the program provid- changed over time. In many cases, the pro-
ing greatly extended duration for UI benefits grams have been redesigned in recent decades
during the Great Recession. During the Great in ways that have made it less responsive to
Recession, Congress also temporarily raised economic downturns. In the years following
maximum SNAP benefits. Congress has also the Great Recession, many states reduced the
authorized relatively universal tax credits or re- generosity of their UI programs with the stated
bates, such as the Recovery Rebates in response goal of reducing taxes for firms. In 2019, UI
to the Great Recession, which provided credits replacement rates—measured as the share of
of $600 for individuals or $1,200 for joint filers. pre-unemployment earnings replaced by UI—
Research documents whether the social averaged 45 percent, and many states had re-
safety net expands and contracts with eco- placement rates below 40 percent including
r sf: t he russell sage f ou n dat ion jou r na l of t he so ci a l sciences
40 t h e s o c i o e c o n o m i c i m p a c t s o f t h e c o v i d -1 9 p a n d e m i c
Arkansas (31 percent), Arizona (37 percent), In- gest the EITC is pro-cyclical: spending per filer
diana (37 percent), and Louisiana (34 percent).6 rises in economic expansions. Further, the
Elira Kuka and Bryan Stuart (2021) document EITC is paid out in a lump sum tax refund in
systematically lower UI replacement rates February or March in the year after the earn-
among Black than among White workers. Fur- ings which qualify individuals are accrued, and
ther, UI coverage is not complete and excludes thus unlikely to be responsive to current need.
unauthorized immigrants, those with inconsis- Thus, despite its important role in reducing
tent work histories, new labor-market entrants, poverty, the EITC is poorly suited to insure con-
and the self-employed.7 More generally, the so- sumption against job loss. More generally, the
cial safety net has shifted toward being more authors show that the move from the previous
work conditioned, using earnings subsidies to out-of-work safety net (higher participation in
increase incomes among workers with children Aid to Families with Dependent Children and
but offering relatively little out-of-work assis- limited tax credits for working) to the current
tance to those not elderly or disabled (Hoynes in-work safety net (the EITC providing substan-
and Schanzenbach 2018). These changes were tive tax credits for workers) led to a reduced
ushered in through the 1996 federal welfare re- overall cyclical response from the means-tested
form law; expansions to the EITC; and, for safety net.
some populations (notably able-bodied adults In sum, the literature shows that before
without dependents), work requirements for COVID-19, the safety net was providing uneven
SNAP.8 The result is a social safety net with a and incomplete protection during economic
strong emphasis on promoting and rewarding downturns. The EITC is not designed to pro-
work—a system that may be adequate during vide insurance against job loss and TANF no
times of low unemployment but provides too longer responds to aggregate economic need
little insurance against job loss and economic and benefits are extended to few households.
shocks. While UI is strongly countercyclical overall, its
The EITC provides an important example of coverage is incomplete. SNAP expands during
why these work-conditioned programs may not economic downturns, but SNAP benefits are
provide much protection. The EITC is the larg- more modest than UI, and because SNAP pro-
est antipoverty program for children in the vides vouchers for food, benefits are only par-
United States, but eligibility requires earned tially fungible and cannot be used for many
income. Bitler, Hoynes, and Kuka (2017) ana- other needs.
lyze Internal Revenue Service data on EITC pay-
ments and find no relationship between local COVID-19 Pandemic Recession
unemployment rates and EITC spending. In To date, five federal laws responded directly to
fact, for single filers with children (the largest the COVID-19 economic crisis. These include
group of recipients), the point estimates sug- the Families First Coronavirus Response Act
6. For data on replacement rates, see U.S. Department of Labor 2004.
7. Bitler, Hoynes, and Diane Schanzenbach (2020) use a UI calculator (Ganong, Noel, and Vavra 2020) and the
2019 CPS-ASEC and document that 4 percent of workers (14 percent of workers in poverty) would be ineligible
for UI if they lost their jobs because they were likely unauthorized, 4 percent (7 percent of those in poverty) would
be ineligible because they are self-employed, and 5 percent (17 percent of those in poverty) would be ineligible
because of insufficient earnings. The latter two groups were covered by the PUA program but the unauthorized
were left out of the UI expansions and are ineligible for SNAP. They are also ineligible for the economic impact
payments and their citizen and authorized family members were excluded from the first EIP.
8. In addition, policy changes during the end of the Trump administration risked further reducing the protective
effects of SNAP by imposing stricter work requirements and discouraging participation among immigrants and
families with mixed immigration status with proposals to include SNAP in public charge rules about immigrants
attempting to convert their immigration status. Many of these policies have been rescinded by the Biden ad-
ministration.
r sf: t he russell sage f ou n dat ion jou r na l of t he so ci a l sciences
s u f f e r i n g , t h e s a f e t y n e t , a n d d i s p a r i t i e s d u r i n g c o v i d -1 9 41
(FFCRA), enacted March 18, 2020; the Corona- increases—a maximum monthly benefit of
virus Aid, Relief, and Economic Security about $170 per person is reduced by 30 cents
(CARES) Act, enacted March 27, 2020; the Con- for each additional dollar in income. The FF-
tinuing Appropriations Act 2021 and Other CRA authorized states to increase benefits for
Extensions Act, enacted October 1, 2020; the all SNAP participants to the maximum benefit,
Consolidated Appropriations Act 2021, enacted a provision known as the Emergency Allot-
December 27, 2020; and the American Rescue ment (EA), while state and federal health emer-
Plan Act of 2021 (ARPA), enacted March 11, gency declarations are in place. Notably, this
2021).9 As of August 31, 2021, these laws are re- expansion provides an increase in benefits to
ported to have resulted in $3.4 trillion in SNAP recipients who were not already receiving
spending.10 In this article, we focus on a subset the maximum; these are the more “advan-
of safety net programs for which the response taged” of the SNAP population and include
to the COVID crisis was substantial and some those with earned income and those with other
administrative data are available. Four ele- income support (such as the elderly receiving
ments of this response are particularly impor- Social Security). Therefore, even though SNAP’s
tant for lower-income families: expansions to general structure is progressive (the highest
SNAP, expansions to UI, the EIPs, and the re- benefits accrue to the lowest-income groups),
fundable monthly CTC payments. As we show, the first COVID-era expansion of SNAP was re-
these four policies account for almost $1.4 tril- gressive, at least within the SNAP population.
lion in new spending from April 2020 through Subsequent expansions to SNAP during COVID
December 2021 and were the main sources were not regressive, however. The Consoli-
of direct payments to households during dated Appropriations Act (December 2020) in-
COVID-19.11 Wherever possible, we examine creased maximum benefit amounts for all re-
data on participation and benefits by race and cipients by 15 percent from January through
ethnicity. September 2021. Later, the previously enacted
SNAP is structured to respond quickly to in- EA payments were revised upward to require
creased need because the program is an enti- that all recipients received a monthly benefit
tlement (not subject to annual funding limits), increase of at least $95, giving the lowest-
benefits are fully federally funded, and house- income families who previously received no EA
holds that newly become eligible due to unem- payments a boost in benefit levels. In addition,
ployment or other loss of income can apply states were temporarily allowed to extend eli-
and generally receive benefits with thirty days gibility periods for currently participating
(Hoynes and Schanzenbach 2019). During the households for six months (under normal cir-
pandemic, Congress made temporary changes cumstances, recipients are required to reapply
that increased both participation and (for for benefits every six to twelve months), allow-
many participants) benefit levels. Like those of ing offices already stretched by health-related
most income support programs, SNAP benefits office closures and the need to socially dis-
are typically reduced as a household’s income tance to concentrate on screening new appli-
9. This section draws on Randy Aussenberg and Kara Billings (2021), Julia Whittaker and Katelin Isaacs (2021),
and Margot Crandall-Hollick (2021).
10. Of the agencies whose programs we focus on, the Treasury had disbursed $1.4 trillion in new spending tied
to the recovery by this point, the Department of Agriculture distributed $81 billion, and the Department of Labor
distributed $650 billion. A large amount of SNAP and other Department of Agriculture and UI spending auto-
matically increases in bad times, and much of this additional Treasury spending is the tax credits (USASpending
2021).
11. Other spending through nutrition programs included pandemic EBT (replacement payments for school meals
while schools were closed), enhanced WIC benefits, directly provided school meals, and other meals. Eviction
moratoria and housing spending also likely helped a host of families. Further, many of these safety net programs
reduced or suspended recertification requirements temporarily, likely increasing participation.
r sf: t he russell sage f ou n dat ion jou r na l of t he so ci a l sciences
42 t h e s o c i o e c o n o m i c i m p a c t s o f t h e c o v i d -1 9 p a n d e m i c
cants. This temporary policy increased SNAP employed and gig-economy workers and other
participation by reducing the flows out of the workers who were previously excluded from
program during the pandemic.12 eligibility on the basis of low earnings or insuf-
Congressional policy responses also in- ficient work history.14 Initial UI claims surged,
cluded expansive changes to the joint state- rising from 221,000 for the week of March 14,
federal UI program. The Federal Pandemic Un- 2020, to 5.9 million the week of March 28, 2020,
employment Compensation (FPUC) program and maxing out at 6.1 million the week of April
increased weekly benefits by $600 for weeks of 4, 2020. Ongoing claims went up nearly seven-
unemployment through the end of July 2020. fold, before returning to pre-pandemic levels
This was followed by the Lost Wages Assistance by December 2021.
program, which allowed participating states to The largest and most universal of the relief
increase benefits by $300 to $400 per week for efforts came through direct payments to fami-
up to six additional weeks, for unemployment lies. The EIP included in the CARES Act pro-
spells from the expiration of FPUC through vided $1,200 per adult ($2,400 for a married cou-
early fall (weeks of unemployment ending Sep- ple filing jointly) and $500 per dependent
tember 5, 2020). After a period with no benefit under age seventeen. This was structured as a
top-ups, benefits were again increased by $300 fully refundable tax credit, phased out begin-
per week for spells from December 26, 2020, ning at annual incomes of $150,000 for married
through early September 2021.13 All of these couples, $112,000 for head of household filers,
benefit increases were federally funded. The and $75,000 for single filers. Treasury provided
length of eligibility for UI was also extended, automatic payments for all who filed federal
including an initial thirteen-week extension of taxes in tax years 2018 or 2019 as well as to those
fully federally funded benefits (Pandemic receiving payments through Social Security or
Emergency Unemployment Compensation) Veteran’s Affairs programs.15 The initial pay-
that was eventually expanded to provide up to ments were made to those with direct deposit
thirty-nine additional weeks through early Sep- information during the week of April 17, 2020,
tember 2021 for those exhausting other bene- and paper checks followed more slowly after
fits. Overall, through October 1, 2022, total that. However, nonmilitary families that in-
spending on UI from the federal government cluded any immigrant adult without a Social
from the CARES Act and subsequent laws to- Security number were ineligible, thus exclud-
taled $674 billion above and beyond the regular ing many citizen children and spouses.
UI program spending (U.S. Department of La- A second round of direct payments went out
bor 2022). as part of Consolidated Appropriations Act of
Additionally, important expansions were 2021 (enacted December 27, 2020, payments
made to the eligibility criteria for UI. The Pan- starting in January 2021). This was a smaller
demic Unemployment Assistance (PUA) pro- payment of $600 for each eligible individual
gram expanded UI eligibility to the self- and $1,200 for joint married filers, and an ad-
12. A revised Thrifty Food Plan, on which SNAP benefits are based, was announced in the summer of 2021 and
took effect on October 1. This increased regular SNAP benefits by about 27 percent relative to basic benefits
without pandemic-related increases. Because the 15 percent pandemic increase ended at the same time, net
benefits went up by a smaller amount.
13. States had to opt in to participate in the UI expansions and twenty-six ended some of these other programs
before they expired in September 2021, citing concerns about work disincentives. Additionally, the Mixed Earner
Unemployment Compensation program provided $100 additional per week for unemployed workers with self-
employment and wage and salary income not getting UI for weeks of unemployment from December 27, 2020,
to early September 2021.
14. The federal government also funded the waiting week for UI so that benefits would get out more quickly and
most states suspended search requirements for obtaining UI during the health crisis through May 2020.
15. Some of the Social Security Administration groups had to submit forms to receive dependent payments.
r sf: t he russell sage f ou n dat ion jou r na l of t he so ci a l sciences
s u f f e r i n g , t h e s a f e t y n e t , a n d d i s p a r i t i e s d u r i n g c o v i d -1 9 43
ditional $600 per qualifying child under age new spending came from the EIPs—the bene-
seventeen. It also phased out for higher income fits least targeted to those who experienced a
individuals. A third round went out starting in direct economic shock or who have low levels
March 2021 as part of the ARPA; phase-outs of income. More than 25 percent came from UI,
were similar but payments were higher, $1,400 9 percent came from SNAP, and 7 percent came
per individual or dependent (and including all from the monthly CTC payments that started
dependents, not just those under seventeen). in July 2021.
In addition, the ARPA included a consider- As shown in figure 4, variation in new
able expansion of the CTC for tax year 2021. The monthly spending is driven by the timing of the
National Academies (Duncan and Le Menestre EIPs, and most new spending occurred in April
2019) and other researchers (Shaefer et al. 2018; 2020 and January and March 2021. Unemploy-
Bitler, Hines, and Page 2018) have laid out evi- ment insurance payments are generally smooth
dence about the benefits of a child allowance across months, averaging $23 billion per month
in reducing poverty, and the CTC expansion from April 2020 through March 2021 but in-
was modeled after these proposals. The maxi- creasing and decreasing somewhat in relation
mum CTC was expanded from $2,000 to $3,000 to the availability of federal top-up payments.
per year per child ($3,600 for children ages five From April through August 2021, UI payments
and younger) and payments were made fully averaged $13 billion per month, declining fur-
refundable so children in households with no ther in the months that followed with the expi-
or low earnings were eligible for the full bene- ration of COVID-era policies. SNAP payments,
fit. ARPA also changed the timing of payments the program most targeted to the low-income
so half of the annual credit would be issued population, grew over this period: spending in-
monthly starting in July (2021), and the rest creases were driven by an increase in participa-
would come when filing 2021 taxes in early tion levels in the first months of COVID, then
2022. Eligibility was also extended to seventeen- by subsequent increases in benefits levels. The
year-olds (who are usually ineligible). Real-time refundable monthly CTC payments were rela-
analysis has shown that these expansions sub- tively stable across July to December 2021, and
stantially reduced child poverty and child food in magnitude were about three times the new
insufficiency (Parolin, Curran, et al. 2021; Paro- monthly spending on SNAP and 80 percent of
lin, Ananat, et al. 2021). the monthly average new UI spending.
Figure 4 displays the timing and magnitude
of new spending on these programs, reported T h e Pa n d e m i c , t h e S o c i a l
monthly between April 2020 and December S a fe t y N e t, a n d P ov e r t y
2021. The information is drawn from Monthly The Annual Social and Economic Supplement
Treasury Statements from the Department of (ASEC) to the Current Population Survey is ad-
the Treasury, which provide information on ministered to most households in March every
monthly receipts and outlays of the federal gov- year and is an annual survey that collects labor
ernment (U.S. Department of the Treasury market, income, and program participation in-
2022). For SNAP and UI, we measure the change formation for individuals for the previous cal-
in spending relative to the programs’ February endar year; as well as demographic information
2020 levels, which were $4.9 billion and $2.8 bil- from the time of the survey.
lion, respectively.16 Spending on the EIPs is re- We begin by examining poverty rates by race
ported directly, as are payments of the CTC that and ethnicity for calendar years 2019 and 2020.17
exceed tax liabilities (the refundable portion of We measure poverty using the SPM, which is
the CTC). Cumulatively, throughout these available from the Census Bureau beginning in
twenty-one months, nearly 60 percent of the 2009 and is released alongside the official pov-
16. Payments to SNAP participants of the pandemic EBT benefits to replace missed school meals are also in-
cluded in the Monthly Treasury Statements. Pandemic EBT payments to SNAP nonparticipants are not included.
17. The CPS faced challenges with interviewing in COVID. Rothbaum and Bee (2021) document nonresponse
issues in the 2020 ASEC used for measuring 2019 poverty. Their adjusted 2020 measure adjusting for lagged
r sf: t he russell sage f ou n dat ion jou r na l of t he so ci a l sciences
44 t h e s o c i o e c o n o m i c i m p a c t s o f t h e c o v i d -1 9 p a n d e m i c
Figure 4. New Monthly Spending in Economic Impact Payments, Unemployment Insurance, SNAP, and
the Child Tax Credit
400
350
300
250
Billions of dollars
200
150
100
50
0
Ju l2
ay 0
ne 0 02
20 2
Ju 20
Au y 20
S e g u 20
pt s t
em 20 l 20
N O er 2
ct
o v erob
em 20b 0 20
20
20
D ec er
em 20b
b e 20
ri Ja r
–50
Fe nu 202
br a ry 0
ua 021 2
Ap M M ry
ar 021
ch
Ap 021
ri
M 021
ay l22
2
2
Au Ju 021
ne
Ju 21
ly 20
20
Se g u 21
N pt s t
em 20
O er 2
ct b
ob 021 21
D o v er
em 20
ec er
em 20b
be 21
r2 21
02 1
EIP UI SNAP CTC
Source: Authors’ tabulations of Monthly Treasury Statements, February 2020 through December 2021
(U.S. Department of the Treasury 2022).
Note: We difference monthly expenditures relative to their February 2020 level to net out new pay-
ments.
erty measure.18 A person is in poverty if their come taxes including the tax credits—EITC,
family’s SPM resources are below their SPM CTC—and the EIPs). The Census Bureau’s SPM
threshold. SPM resources include all cash in- Thresholds are the average between the 30th
come (earnings, pensions, cash transfers, So- and 36th percentiles of the distribution of con-
cial Security Administration payments for dis- sumer expenditures on food, clothing, shelter,
ability, retirement and supplemental security and utilities, plus an additional 20 percent to
income) plus the cash value of in-kind transfers account for additional necessary expenditures.
(SNAP, the National School Lunch Program, Additionally, the thresholds are adjusted to re-
housing subsidies, energy assistance, WIC [the flect family size, owner versus renter status,
Special Supplemental Nutrition Program for and geographic variation in housing costs (for
Women, Infants, and Children]) minus deduc- more detail on the SPM, see Fox and Burns
tions (medical out-of-pocket expenditures, 2021).
child support paid, work expenses, childcare) Figure 5 presents the share in poverty for all
and taxes (payroll taxes, federal and state in- persons (left) and for children (right), by race
administrative and historical responses suggests the nonrespondents were lower-income individuals (pre-
pandemic), and thus that official poverty might have been underestimated.
18. The official poverty measure is of limited use to understand hardship because it is based only on cash pretax
income, thus not inclusive of SNAP, EITC, CTC, or EIPs.
r sf: t he russell sage f ou n dat ion jou r na l of t he so ci a l sciences
s u f f e r i n g , t h e s a f e t y n e t , a n d d i s p a r i t i e s d u r i n g c o v i d -1 9 45
Figure 5. Supplemental Poverty Measure 2019, All Persons and Children
25%
All persons Children
20.1%
20% 19.5%
18.8%
18.2%
15%
12.4%
11.7% 11.5%
10% 9.6%
8.2%
7.0%
5%
0%
An H H H H An H H H H
yr e,
N
e,
N
e,
N ispa yr e,
N
e,
N
e,
N ispa
ace on on on ni ace on on on ni
al al al c al al al c
te ac k n te ack n
hi ia hi ia
W Bl As W Bl As
Source: Authors’ tabulations using 2020 and 2021 Annual Social and Economic Supplement to the
Current Population Survey (U.S. Census Bureau 2020b; U.S. Census Bureau 2021a).
and ethnicity, in the pre-pandemic baseline of fell by 5.5 points. Across all groups, these rep-
calendar year 2019. Overall, in 2019, 11.7 percent resent significant declines on the order of 20 to
of all persons, and 12.4 percent of children, 25 percent of the pre-pandemic level (the low-
were poor in the United States. The disparities est percentage decline was 12 percent for Black
across race and ethnicity are striking. For ex- children).
ample, 19.5 percent of Black children and 20.1 Clearly, a decline in poverty in the midst of
percent of Hispanic children are poor, versus an economic crisis is not a typical finding. Al-
7.0 percent of White children and 9.6 percent though UI and SNAP are strong automatic sta-
of Asian children. bilizers (Bitler and Hoynes 2010, 2016; Bitler,
Despite the dramatic increases in unem- Hoynes, and Iselin 2020), poverty has consis-
ployment, between 2019 and 2020–2021, annual tently increased during recessions in the
poverty rates across all groups declined (see fig- United States (Bitler and Hoynes 2010, 2015;
ure 6); for a partial caveat related to differential Bitler, Hoynes, and Kuka 2017). The 2019 to
nonresponse by income groups, see note 17. 2020 decline in poverty is a direct result of the
The overall poverty rate fell by 2.6 percentage dramatic pandemic policy response. Figure 7
points (from 11.7 to 9.1 percent) for all persons presents the effect of individual policies on
and by 2.7 percentage points for children. De- SPM poverty rates in 2020 for all persons (panel
clines in poverty rates were experienced across A) and all children (panel B). To make these
all race and ethnic groups. For example, the calculations, we zero out a given tax or trans-
share of Black children in poverty fell by 2.4 fer program and recalculate the poverty rate
percentage points, and for Hispanic children it assuming no change in behavior. We also in-
r sf: t he russell sage f ou n dat ion jou r na l of t he so ci a l sciences
46 t h e s o c i o e c o n o m i c i m p a c t s o f t h e c o v i d -1 9 p a n d e m i c
Figure 6. Reduction in SPM Between 2019 and 2020 (Percentage Points)
25
All Persons Children
20
2.4
5.5
Percent in SPM poverty
3.6 4.8
15
2.7
10 2.6 2.9
1.7 3.6
1.4
5
0
An yr H An yr
te al on
ace ispa
te al on e,
ace
Bl ac
e, N H nic
Bl ac k al on
N H
k al on As ia n e, N H
As ia e, N H al on e,
hi n al on hi H N H
e, N ispa ni
W H W c
2020 Decline from 2019 to 2020
Source: Authors’ calculations based on 2020 and 2021 Annual Social and Economic Supplement to the
CPS (U.S. Census Bureau 2020b and U.S. Census Bureau 2021a).
clude the antipoverty effects for 2019 as a pre- 2019, we note several findings. First, the effects
pandemic baseline comparison. of the EITC-CTC are smaller in 2020 than in
For all persons, Social Security leads to the 2019 (consistent with Bitler, Hoynes, and Kuka
largest poverty reduction at a staggering 8.1 2017), illustrating that the EITC is pro-cyclical
percentage points (the same poverty reduction (decreases during recessions) for single-parent
for both 2019 and 2020). Focusing on 2020, we families. (Importantly, the ARPA-expanded
see the EIPs reduced poverty by 3.6 percentage CTC did not take place until July 2021 and thus
points, followed by UI at 1.7 percentage points, is not reflected in these calculations.) Second,
the combined effect of the EITC and the CTC in 2019 UI played a very small role in poverty
at 1.6 percentage points and the combination reduction whereas in 2020 it was the third larg-
of SNAP and school lunch at 1 percentage point. est antipoverty program for all persons and for
Among children, the largest poverty reduction children.19 This highlights the significance of
resulted from the EIPs at 4.5 percentage points, the COVID-era UI expansions, particularly the
followed by the combined impact of the EITC benefit top-ups. These calculations make it
and CTC at 3.8 percentage points, UI at 2.0 per- very clear that without the increase in pan-
centage points and SNAP and school lunch at demic aid, poverty rates in 2020 would have in-
1.8 percentage points. Social Security plays a creased dramatically. Of course, these are static
smaller role in poverty reduction among chil- comparisons, limited because they simply
dren. Comparing these poverty reductions to compare poverty calculated with and without
19. The CPS greatly understated receipt of UI, suggesting an even larger role of UI during COVID (Larrimore,
Mortensen, and Splinter 2022).
r sf: t he russell sage f ou n dat ion jou r na l of t he so ci a l sciences
s u f f e r i n g , t h e s a f e t y n e t , a n d d i s p a r i t i e s d u r i n g c o v i d -1 9 47
Figure 7. Effect of Individual Elements of Social Safety Net on 2020 Versus 2019, Percentage Point
Change in SPM Rate
A. All persons (2020 base = 9.1 percent, 2019 base = 11.7 percent)
–8.1
Social Security
–8.1
0.0
Economic impact payments
–3.6
–0.2
Unemployment insurance
–1.7
–2.3
2019 EITC/CTC
–1.6
2020
–1.1
SNAP/school lunch
–1.0
–0.9
SSI
–0.8
–0.8
Housing subsidies
–0.7
–0.1
TANF/GA
–0.1
–9 –8 –7 –6 –5 –4 –3 –2 –1 0
Percentage Points
B. All children (2020 base = 9.7 percent, 2019 base = 12.5 percent)
0.0
Economic impact payments
–4.5
–5.5
EITC/CTC
–3.8
–0.2
Unemployment insurance
–2.0
–2.3
SNAP/school lunch
2019 –1.8
2020 –2.0
Social Security
–1.5
–1.0
Housing subsidies
–1.1
–0.7
SSI
–0.5
–0.2
TANF/GA
–0.3
–6 –5 –4 –3 –2 –1 0
Percentage Points
Source: Authors’ calculations based on Fox 2020; Fox and Burns 2021.
Note: We suppress very small changes in poverty rates for LIHEAP, worker’s compensation, and WIC.
r sf: t he russell sage f ou n dat ion jou r na l of t he so ci a l sciences
48 t h e s o c i o e c o n o m i c i m p a c t s o f t h e c o v i d -1 9 p a n d e m i c
Figure 8. Effect of Individual Elements of Social Safety Net on the 2019 and 2020 Child SPM Rates, by
Race and Ethnicity
A. White children B. Black children
0.0 0.0
EIP EIP
–2.8 –6.5
–2.9 –8.6
EITC/CTC EITC/CTC
–2.1 –5.6
–0.1 –0.3
UI UI
–1.2 –3.0
–0.8 –3.1
SNAP SNAP
–0.8 –3.5
–10 –9 –8 –7 –6 –5 –4 –3 –2 –1 0 –10 –9 –8 –7 –6 –5 –4 –3 –2 –1 0
Percentage Points Percentage Points
C. Asian children D. Hispanic children
0.0 0.0
EIP EIP
–3.5 –6.8
–3.7 –9.3
EITC/CTC EITC/CTC
–2.6 –6.4
–0.5
UI UI
–3.6 –2.4
–0.4 –1.7
SNAP SNAP
–0.3 –1.8
–10 –9 –8 –7 –6 –5 –4 –3 –2 –1 0 –10 –9 –8 –7 –6 –5 –4 –3 –2 –1 0
Percentage Points Percentage Points
2019 2020
Source: Authors’ calculations based on 2020 and 2021 Annual Social and Economic Supplement to the
CPS (U.S. Census Bureau 2020b; U.S. Census Bureau 2021a).
various income components but do not include SNAP at 3.5 percentage points, and UI at 3 per-
any behavioral responses were the programs to centage points (we have suppressed the other
be removed.20 safety net policies for clarity). EIPs had the larg-
The program-driven reductions in poverty est impact on poverty for White, Black, and
are experienced across all groups. Figure 8 Hispanic children; UI had the largest impact
shows the effects of individual policies on child for Asian children. The effects of SNAP and UI
poverty, separately for White, Black, Asian, and are lower for Hispanic than for Black children
Hispanic children. We include estimates for despite their similar baseline poverty rates.
2019 and 2020, as before, to highlight the ef- This is likely a result of incomplete eligibility
fects of the COVID era policies. Focusing on the or lower take-up of these programs among fam-
data for 2020, among Black children, the EIPs ilies with unauthorized members.21 Overall,
reduced poverty by 6.5 percentage points, fol- these results show that universal policies, such
lowed by EITC-CTC at 5.6 percentage points, as the uniform $300 to $600 UI top-up and the
20. Zachary Parolin, Meghan Curran, and colleagues (2022) present an approach to calculating a monthly SPM
and use it to explore well-being through the beginning of COVID. Parolin, Elizabeth Ananat, and colleagues
(2021) and Parolin, Sophie Collyer, and colleagues (2021) explore the effects of the CTC.
21. The EIPs and EITC-CTC are not measured directly in the ASEC and are imputed by the census. This may
generate somewhat higher antipoverty effects than are realized if true participation is not 100 percent.
r sf: t he russell sage f ou n dat ion jou r na l of t he so ci a l sciences
s u f f e r i n g , t h e s a f e t y n e t , a n d d i s p a r i t i e s d u r i n g c o v i d -1 9 49
relatively universal EIPs, can reduce disparities shock. Participation increased by 11 percent in
across groups. April 2020 and through December 2021 re-
mained elevated by an average of 13 percent
A C lo s e r Lo o k at SNAP ’ s relative to the February 2020 level.22
R e s p o n s e to COV ID -1 9 In addition to the increase in SNAP partici-
The ASEC is useful given the ability to measure pation, total SNAP expenditures increased due
family resources and to identify race and ethnic to legislated increases in benefit payments, as
groups, but has disadvantages in regard to sur- described in more detail above. First, all par-
vey measures of poverty and well-being. One ticipants were awarded the maximum SNAP
concern is the well-documented misreporting benefit through the EA payments starting in
(typically underreporting) of various programs April 2020. Next, there was an across-the-board
(such as Meyer, Mok, and Sullivan 2015) as well 15 percent increase in maximum benefits in
as earnings (Bollinger et al. 2019). The Census January 2021. This was followed by action that
Pulse data were extremely valuable by provid- set minimum EA payments equal to $95 per
ing real-time information about hardship but month, awarding these additional EA pay-
are not comparable to pre-COVID measures. ments for the first time to those who had previ-
Further, all survey measures may have differen- ously been receiving maximum SNAP benefits,
tial nonresponse (Rothbaum and Bee 2021). rolled out in April and May 2021. Each of these
This leads us to examine administrative data, increases can be seen clearly in the time series
which do not suffer from this underreporting of total benefits, which peaked at a 130 percent
issue and do not require individuals to report increase relative to February 2020 spending.23
program use. Ideally, we would look at all Some states opted to terminate their EA pay-
sources of administrative data, but this is not ments in the summer months in 2021, reducing
possible. In the case of SNAP, we use adminis- benefit payments in those months. The 15 per-
trative data on county-level participation and cent increase in maximum benefits ended in
benefits, as well as case-level data that includes October 2021, coinciding with the increase in
information on benefits, income sources, and maximum benefits resulting from a recalibra-
demographic characteristics such as race- tion of the Thrifty Food Plan (Food and Nutri-
ethnicity; some data extend to January 2021. tion Service 2021c).
SNAP is also an important case study because We rely primarily on two sources of SNAP
of its central role in the social safety net and administrative data. The first source is the
the many policy changes made in response to quality control (QC) data, which for a sample
COVID. of cases have detailed administrative informa-
Thus we more closely investigate SNAP’s re- tion on benefits, resources, and household
sponse to the COVID-19 crisis, paying special composition (including race-ethnicity). At the
attention to impacts across racial and ethnic time of this analysis, the QC data are available
groups (overall and among children) and by only during the pre-COVID era through 2019.
geographic area. As shown in figure 9, admin- The second source is the Department of Agri-
istrative data from the Department of Agricul- culture’s Bi-Annual State Project Area and
ture shows that SNAP participation increased County Level Participation and Issuance data,
sharply after COVID’s onset, likely a function reported for January and July of each year and
of both increased need and the policy change available through January 2021 (Food and Nu-
that temporarily allowed states to automati- trition Service 2021a). This source provides
cally recertify existing SNAP cases so their ad- county-level data from most states but a few
ministrators could concentrate on serving states report only state-level data. Together, we
those made newly eligible due to the economic use these data sources to both see how spend-
22. Average monthly participation in SNAP in calendar year 2019 was thirty-five million persons, and benefits
spending was $4.57 billion.
23. Unlike the monthly Treasury statement data in figure 4, which include P-EBT payments to SNAP participants
in SNAP spending, Department of Agriculture benefits data in figure 9 include only SNAP benefits.
r sf: t he russell sage f ou n dat ion jou r na l of t he so ci a l sciences
50 t h e s o c i o e c o n o m i c i m p a c t s o f t h e c o v i d -1 9 p a n d e m i c
Figure 9. Percentage Increase in SNAP Participation and Spending Relative to February 2020,
February 2020–July 2021
Increase in SNAP Spending, Participation During COVID
140%
120%
Percent increase in SNAP participants, spending
100%
80% Benefits
60%
40%
20% Participation
0%
ar y
a r 202
ch 0
A p 202
ri 0
M 02 l2
ay 0
Ju 202
ne 02
Ju 020
Au ly 2
S e g u 02
pt s t 0
em 2 b 02
N
D O er 0
ct o
o v b er 20
em 2
e c b e r 20
em 2 200
Ja er 0b 02
nu 20
Fe ar 20
br y 2
ua 02
r 1
M y2
Febr M
ar 02
ch 1
Ap 202
ri 1
M 02 l2
ay 1
Ju 202
ne 1
Ju 0212
u Au ly 2
Se g u 02
N
D
pt s t 1
em 2
O er 1
ct o b 02
ov ber 21
em 2
ec ber 21
em 2 200
be 021
r2 021
Source: Authors’ calculations from USDA, Food and Nutrition Service, SNAP National Level Monthly
Data (Food and Nutrition Service 2021b).
ing and participation vary with the characteris- in the nationally representative QC data and in
tics of areas and simulate who obtained more the CPS ASEC. For both calculations, the de-
generous increases to their average benefit by nominator is based on population counts by
race/ethnicity. race-ethnicity in the CPS-A SEC, limited to
We start by investigating the change in SNAP those with incomes below 150 percent of the
participation during COVID-19. Participation federal poverty line to proxy the number of per-
can grow from two sources: those who are eli- sons likely eligible for SNAP.
gible but not participating in SNAP can enroll, Using the QC data as the numerator, the
and more people can become eligible to par- SNAP participation rate overall is 62 percent.
ticipate due to income losses. Table 1 shows Variation in participation across racial and eth-
baseline participation rates (where the denom- nic groups is wide: averages are 77 percent
inator is households with incomes below 150 among Blacks, 52 percent among Whites, and
percent of poverty), averaged across calendar 35 percent among Hispanics. Estimated par-
years 2017 through 2019, by race-ethnicity and ticipation rates are substantially lower across
presence of children, to demonstrate variation the board when the CPS is used for the numer-
in room to grow through increased participa- ator, as expected given the known underreport-
tion rates. We take two approaches to calculat- ing of SNAP participation in the data. We pres-
ing the numerator in this rate: calculating the ent this to highlight the drawbacks to relying
number of SNAP participants by race-ethnicity on CPS survey data (as we did earlier) and the
r sf: t he russell sage f ou n dat ion jou r na l of t he so ci a l sciences
s u f f e r i n g , t h e s a f e t y n e t , a n d d i s p a r i t i e s d u r i n g c o v i d -1 9 51
Table 1. SNAP Participation Rates, by Group (2017–2019)
Overall White Black Hispanic
(1) (2) (3) (4)
Panel A. All participants
SNAP admin data 61.6 51.6 76.5 35.3
CPS survey data 31.9 27.3 44.7 30.6
Panel B. Participants in
families with children
SNAP admin data 77.6 76.7 93.6 38.2
CPS survey data 39.2 37.4 53.5 32.9
Source: Authors’ calculations based on 2017–2019 CPS-ASEC (U.S. Census Bureau 2018, 2019, and
2020b) and SNAP Quality Control data (Mathematica Policy Research, 2018, 2019, and 2020).
Note: Participation is calculated relative to a denominator of population counts in households with in-
comes below 150 percent of the poverty threshold calculated from the CPS-ASEC. The first row in
each pair calculates the numerator (SNAP participation) from SNAP administrative data, and the sec-
ond row in each pair calculates it from the CPS-ASEC.
desirability of using administrative data when pendix figure 2), with a weak but positive rela-
possible to understand program spending. tionship between baseline participation rates
Panel B repeats the exercise for families with among eligibles and percentage growth during
children. Participation rates are higher across COVID.
the board among those with children: an esti- We also measure the relationship between
mated 78 percent overall and nearly 94 percent the magnitude of the COVID economic shock
among Black families with children. All else and changes in SNAP participation, using
equal, then, the opportunity for participation county-level data, as shown in figure 10. The
to grow was higher among childless families, x-axis shows the percentage change in number
and among Whites and Hispanics relative to of people employed from quarter 1 to quarter
Blacks. 2, 2020 and the y-a xis shows the percentage
We next explore determinants of the magni- change in SNAP participation from January
tude of participation changes from January 2020 to January 2021. As we would expect, we
2020 to January 2021. We first test the correla- find counties that experienced a larger employ-
tion between the state-level increase in SNAP ment loss also had a larger increase in SNAP
participation and state-level participation rates participation.
among eligible persons from 2017 (the most re- We next analyze changes in total SNAP ben-
cent available) calculated by Mathematica Pol- efit payments, exploring by how much, when,
icy Research (Cunnyngham 2020). The hypoth- and for whom benefits increased. Because data
esized relationship could go in either direction. were at the time of our writing available only
We may expect the increase in participation to through January 2021, we observed only the pe-
be larger in states that previously had lower riod for which the original EA payments were
participation rates among eligibles, given more in place—a policy that paid everyone the max-
room to grow. On the other hand, if high par- imum benefit but provided no additional ben-
ticipation rates in part reflect an efficient and efits to those who had already been receiving
inclusive state administrative system, we may the maximum benefit. But we can model the
expect participation to increase more in these likely impacts of the series of payment changes
states as they are better equipped to process using participant characteristics from the 2017–
applications among those newly eligible due to 19 SNAP QC data. The first two rows of table 2
the economic shock. We find evidence consis- show the average benefit amount (as a share of
tent with the latter hypothesis (see online ap- the maximum benefit) and the share of house-
r sf: t he russell sage f ou n dat ion jou r na l of t he so ci a l sciences
52 t h e s o c i o e c o n o m i c i m p a c t s o f t h e c o v i d -1 9 p a n d e m i c
Figure 10. Percent Change in SNAP Participation (January 2020–January 2021) Versus Number
Employed (2020, Quarter 1 to 2020, Quarter 2)
60%
β = –0.5125***, R 2 = 0.015
40%
Percent change in participation
Poplulation
10k
50k
20% 100k
500k
1m
10m
0%
–20%
–40% –20% 0% 20%
Percent change in number employed (from Q1 to Q2 2020)
Source: Authors’ calculations based on USDA’s Bi-Annual State Project Area and County Level Partici-
pation and Issuance data and the U.S. Bureau of Labor Statistics Quarterly Census of Employment and
Wages (U.S. BLS 2021).
Note: The figure indicates the ordinary least squares estimate of the effect of the county employment
shock on SNAP participation (β) and the fit of that regression (R2), and *** indicates that beta is statis-
tically significant at the 1 percent level.
holds receiving the maximum benefit (who did were already receiving the maximum SNAP
not receive a payment increase under the orig- benefit at baseline, and therefore would not
inal EA policy).24 Under the regular SNAP ben- have received any benefit increase under the
efits schedule, benefits are awarded as the dif- original EA policy implemented in March
ference between the maximum benefit and 30 2020.25 Black recipients were more likely to be
percent of a household’s net income after a se- receiving the maximum benefit at baseline,
ries of deductions including a portion of earn- meaning that more of this population would
ings and some expenses such as dependent not have received a benefit increase under the
care and excess shelter cost (Center on Budget original EA policy. Within each racial-ethnic
and Policy Priorities 2022). Those with zero net group, households with children were less
income receive the maximum SNAP benefit. likely to be receiving the maximum benefit at
Understanding benefits receipt at baseline clar- baseline.
ifies who received extra resources, and how The original EA policy increased benefits by
many, during the COVID policy changes to SNAP. 44 percent overall and by 39 percent for those
As shown in table 2, 32 percent of house- with children. We project that White recipients
holds and 29 percent of those with children received larger percentage increases than
24. Maximum benefits depend on family size, and in 2019 were $192 per month for a household of 1, increasing
by approximately $142 per month for each additional household member (Center on Budget and Policy Priorities
2019).
25. Some states implemented EAs in April 2020.
r sf: t he russell sage f ou n dat ion jou r na l of t he so ci a l sciences
s u f f e r i n g , t h e s a f e t y n e t , a n d d i s p a r i t i e s d u r i n g c o v i d -1 9 53
Table 2. SNAP Benefits and COVID-Era Increases, by Race and Ethnicity and Presence of Children
Overall White Black Hispanic
(1) (2) (3) (4)
Panel A. All participants
Baseline receiving max benefit 31.6 31.0 34.3 31.6
Baseline benefits, maximum 69.3 66.9 71.0 73.1
Benefit increase, EA only 44.4 49.4 40.8 36.7
Increase, all policy changes 87.5 94.5 84.0 78.1
Panel B. Participants in families with children
Baseline receiving maximum benefit 28.9 28.3 31.4 27.8
Baseline benefits, maximum 72.0 70.4 73.2 73.8
Benefit increase, EA only 39.0 42.0 36.6 35.6
Increase, all policy changes 71.7 74.8 69.2 69.2
Source: Authors’ calculations based on SNAP quality control data (Mathematica Policy Research, 2018,
2019, and 2020) corresponding to years 2017 to 2019.
Note: All figures in percentages. The first row in each panel is average pre-pandemic benefits as a
share of the maximum benefit. The second row calculates the share of participants receiving the maxi-
mum benefit. The third row predicts the benefit increase from the original Emergency Assistance pol-
icy change enacted in March–April 2020 that moved all participants to the maximum benefit. The final
row predicts benefit increases from additional COVID-era policy changes (original EA, 15 percent max-
imum benefit increase, and new EA requiring a $95 monthly minimum enacted in March–April 2020,
January 2021, and March–April 2021, respectively).
Black or Hispanic recipients did, a direct result providing larger increases for those who were
of their lower baseline SNAP benefits (relative already better off.
to the maximum benefit, shown in row 2).26 Given this background, figure 11 shows the
Subsequently, benefits were increased across relationship between the county-level employ-
the board in January 2021 and the EA payments ment shock (change in number employed from
were reformed so that all households received Q1 to Q2 2020) and county-level change in SNAP
a minimum of a $95 payment starting (depend- benefits (from January 2020 to January 2021,
ing on the state of residence) in April or May and only includes the EA expansion). Recall
2021. Together, these policy changes boosted that the measure of SNAP benefits includes ad-
benefits relative to their pre-pandemic levels ditional resources from both increased partici-
by 88 percent overall and 72 percent for house- pation and the EA benefits, and we previously
holds with children, and for the first time pro- demonstrated in figure 10 that participation
vided additional resources to those previously increased more in counties with larger employ-
receiving the maximum benefit allotment. Cu- ment shocks. The relationship between the em-
mulatively, under all of the policy changes, ployment shock and change in SNAP benefits
White participants still saw a larger percentage is the inverse of what is expected—that is,
increase in their benefits than Black and His- counties that experienced a smaller drop in em-
panic participants did, in part because a larger ployment received larger increases in SNAP
share of Black and Hispanic participants were benefits. In other words, the policy-induced
already receiving the maximum allotment be- benefit increases were more generous to coun-
fore the EA. Overall, the SNAP policy changes ties less affected by the economic shock. Unfor-
were regressive within the SNAP population, tunately, at the time of this writing the SNAP
26. As a check, we compare our predictions of SNAP benefit increases based on pre-pandemic SNAP caseload
characteristics with actual benefit increases from January 2020 to January 2021 at the state level in appendix
figure 3. We predict benefit increases from EAs only and do not model increases due to higher enrollment. We
find that the actual benefit increase is positively correlated with our prediction.
r sf: t he russell sage f ou n dat ion jou r na l of t he so ci a l sciences
54 t h e s o c i o e c o n o m i c i m p a c t s o f t h e c o v i d -1 9 p a n d e m i c
Figure 11. Percent Change in SNAP Benefits (January 2020–January 2021) Versus Number Employed
(2020, Quarter 1 to 2020, Quarter 2)
150%
Poplulation
Change in SNAP benefits
10k
100% 50k
100k
500k
1m
50% 10m
β = 0.5144***, R 2 = 0.0021
0%
–40% –20% 0% 20%
Percent change in number employed (from Q1 to Q2 2020)
Source: Authors’ calculations based on USDA (Food and Nutrition Service 2021a) and the U.S. Bureau
of Labor Statistics Quarterly Census of Employment and Wages (U.S. BLS 2021).
Note: The figure indicates the ordinary least squares estimate of the effect of the county employment
shock on the SNAP benefits (β) and the fit of that regression (R2), and *** indicates that β is statistically
significant at the 1 percent level.
data do not yet extend to the later and more creases in SNAP participation. The share of
progressive SNAP increases. households with broadband, which may be a
We estimate population-weighted bivariate proxy for the ability to sign up online for SNAP
correlations to further explore how SNAP par- during COVID, is positively correlated with the
ticipation and benefit changes from January increase in SNAP participation. Places with
2020 to January 2021 are related to county char- higher population density also had more SNAP
acteristics. Figure 12 reports point estimates participation growth. Places with more COVID
and 95 percent confidence intervals from these deaths per capita experienced less SNAP par-
bivariate regressions. As shown in figure 10, the ticipation growth.
top row indicates that counties that experi- The increase in county-level SNAP benefits
enced larger declines in employment had is often less strongly correlated with character-
larger increases in SNAP participation. Coun- istics than the increase in participation, likely
ties with a higher share of the population iden- because of the policy changes that made the
tifying as Black also saw larger increases in par- program more generous to those among the
ticipation. The Hispanic population share is SNAP population who were somewhat better
weakly negatively related to increases in SNAP off. As shown in figure 11, counties that experi-
participation, as is the share of households enced larger declines in employment had
with children. More advantaged counties— smaller increases in SNAP benefits. Although
measured as higher median household in- the change in SNAP benefits is positively re-
comes or lower poverty rates—saw larger in- lated to the Black share of the population, it is
r sf: t he russell sage f ou n dat ion jou r na l of t he so ci a l sciences
s u f f e r i n g , t h e s a f e t y n e t , a n d d i s p a r i t i e s d u r i n g c o v i d -1 9 55
Figure 12. Correlations Between County Characteristics and Percent Changes in SNAP Outcomes
% Decline in employment
% Black, NH
% Hispanic
% Black, NH or Hispanic
% Non-citizens
% Households with children
% Households with broadband
Median household income
Poverty rate
Population density
COVID deaths per capita (2020)
–0.4 –0.2 0.0 0.2 0.4
Correlation
% Change in SNAP benefits % Change in SNAP participation
Source: Authors’ calculations based on USDA (Food and Nutrition Service 2021a), Bureau of the
Census American Community Survey 2015–2019 5-Year Data Release (U.S. Census Bureau 2020a),
Census Bureau 2020 Census Population Density (U.S. Census Bureau 2021b), and 2020 Covid
Deaths from USAFacts (USAFacts 2022).
negatively related to the Hispanic share, the enced smaller employment shocks. Future
combined Black and Hispanic share, and the work can extend this analysis through the full
share of noncitizens. Counties with higher me- COVID policy response period.
dian incomes, lower poverty rates, and lower
COVID death rates saw larger SNAP benefit in- Su m m a ry a n d C o n c lu s i o n s
creases. The economic and public health crisis caused
In summary, the response from SNAP—in by COVID-19 was devastating and dispropor-
terms of participation and monthly pay- tionately hurt Blacks and Hispanics. We show
ments—was sizable. While data are not yet that unemployment rates were higher and in-
available to know whether participation in- creased more during the crisis among Blacks
creased disproportionately across racial and and Hispanics than among Whites. Other
ethnic groups, we find that participation in- measures of material hardship, including lack
creased more in counties with a higher share of access to adequate food, being behind on
of the population that is Black but is unrelated housing payments, and use of food banks,
to the population’s Hispanic share. We also were two to three times as prevalent among
find that because the design of the EA payment Blacks and Hispanics as among Whites and
increases was more generous to those who were Asians.
already better off (among a disadvantaged Without policy intervention, the U.S. safety
SNAP population), these increases provided net is not well designed for an economic down-
less assistance to places with larger shares of turn, let alone a crisis of this magnitude. The
Blacks and Hispanics and larger shares of chil- replacement rates and duration of state unem-
dren. Further, counties that received larger in- ployment insurance benefits are on the de-
creases in SNAP benefits during COVID experi- cline, and our means-tested social safety net
r sf: t he russell sage f ou n dat ion jou r na l of t he so ci a l sciences
56 t h e s o c i o e c o n o m i c i m p a c t s o f t h e c o v i d -1 9 p a n d e m i c
has grown increasingly conditional on work. cause the less-targeted SNAP benefit increases
The result is less insurance against job loss. were more generous to SNAP participants who
Congress authorized a historic policy response, were already better off. Simulating the benefits
incorporating both targeted and universal sup- increases from pre-COVID administrative data,
ports and expanding the reach, duration, and we predict that Black and Hispanic SNAP par-
level of benefits. This response yielded the un- ticipants received a smaller percentage in-
usual outcome of a decline in the poverty rate crease in their benefits than White partici-
between 2019 and 2020 (measured using the pants, as families with children across the
Supplemental Poverty Measure) amid an his- board did. Overall, this suggests that the target-
toric recession. ing in SNAP may not have been ideal.
This article also examines changes in these
poverty rates across groups as well as the R e fe r e n c e s
poverty-alleviating impacts of the array of so- Aaronson, Stephanie. 2021. “What Does the Unem-
cial safety net benefits. We find that in 2020 the ployment Rate Measure?” Up Front (Brookings
near-universal economic impact payments re- Institution blog), February 18. Accessed Novem-
duced overall poverty by 3.6 percentage points ber 2, 2022. https://www.brookings.edu/blog/up
and the children’s poverty rate by 4.5 percent- -front/2021/02/18/what-does-the-unemploy
age points. The EIPs reduced poverty among ment-rate-measure/.
children by more than any other targeted pro- Aussenberg, Randy, and Kara Billings. 2021. “USDA
gram for Whites, Blacks, and Hispanics; for Nutrition Assistance Programs: Response to the
Asians, UI had a slightly larger impact. This COVID-19 Pandemic.” CRS Report no. R46681.
suggests that universal programs can reduce Washington, D.C.: Congressional Research Ser-
disparities between groups. The increases in vice. Accessed November 2, 2022. https://crsrep
unemployment insurance protected millions orts.congress.gov/product/pdf/R/R46681/4.
of families from falling into poverty. Bauer, Lauren, Abigail Pitts, Krista Ruffini, and Diane
We augment the findings, based on survey Whitmore Schanzenbach. 2020. “The Effect of
data, with detailed administrative data on Pandemic EBT on Measures of Food Hardship.”
SNAP participation and benefit payments. The Hamilton Project, Economic Analysis. Wash-
SNAP is of particular interest for several rea- ington, D.C.: Brookings Institution. Accessed No-
sons. First, it is the only program that is quite vember 2, 2022. https://www.brookings.edu/wp
large when times are good; UI is small outside -content/uploads/2020/07/P-EBT_LO_7.30.pdf.
recessions and the other programs did not exist Bitler, Marianne P., Annie Laurie Hines, and Mari-
in the form they took during COVID. Second, anne Page. 2018. “Cash for Kids.” RSF: The Rus-
SNAP had both more targeted and less targeted sell Sage Foundation Journal of the Social Sci-
expansions during COVID. The more targeted ences 4(2): 43–73. DOI: https://doi.org/10.7758
expansion resulted from suspending temporar- /RSF.2018.4.2.03.
ily rules that require participants to recertify Bitler, Marianne P., and Hilary W. Hoynes. 2010. “The
for the program regularly—a feature of the pro- State of the Safety Net in the Post-Welfare Re-
gram shown to decrease participation (Ho- form Era.” Brookings Papers on Economic Activity
monoff and Somerville 2021; Gray 2019). The No. 2 (Fall): 71–127. Accessed November 2, 2022.
less targeted expansion involved paying every- https://www.brookings.edu/bpea-articles/the
one the maximum benefit for much of the pan- -state-of-the-social-safety-net-in-the-post
demic, which increased benefits for the partic- -welfare-reform-era-with-comments-and
ipants who were relatively better off but did not -discussion/.
change them for the worst off. We find that par- ———. 2015. “Heterogeneity in the Impact of Eco-
ticipation in SNAP increased more in counties nomic Cycles and the Great Recession: Effects
that experienced a larger employment shock, Within and Across the Income Distribution. ”
consistent with the standard countercyclical American Economic Review 105(5): 154–60.
role of SNAP. By contrast, the increase in total ———. 2016. “The More Things Change, the More
SNAP benefits received was inversely related to They Stay the Same? The Safety Net and Pov-
the employment shock. This likely occurred be- erty in the Great Recession.” Journal of Labor
r sf: t he russell sage f ou n dat ion jou r na l of t he so ci a l sciences
s u f f e r i n g , t h e s a f e t y n e t , a n d d i s p a r i t i e s d u r i n g c o v i d -1 9 57
Economics 34(S1): 403–44. DOI: https://doi.org https://crsreports.congress.gov/product/pdf/IN
/10.1086/683096. /IN11613/10.
Bitler, Marianne P., Hilary W. Hoynes, and John Ise- Cunnyngham, Karen. 2020. “Reaching Those in
lin. 2020. “The Cyclicality of Cash and Near Need: Estimates of State Supplemental Nutrition
Cash Transfer Programs.” National Tax Journal Assistance Program Participation Rates in 2017.”
73(3): 759–70. Washington: U.S. Department of Agriculture,
Bitler, Marianne P., Hilary W. Hoynes, and Elira Kuka. Food and Nutrition Service.
2017. “Do In-Work Tax Credits Serve as a Safety Duncan, Greg, and Suzanne Le Menestre, eds.. 2019.
Net?” Journal of Human Resources 36(2): 358– A Roadmap to Reducing Child Poverty. Washing-
89. ton, D.C.: National Academies Press. DOI: https://
Bitler, Marianne P., Hilary W. Hoynes, and Diane doi.org/10.17226/25246.
Whitmore Schanzenbach. 2020. “The Social Flood, Sarah, Miriam King, Renae Rodgers, Steven
Safety Net in the Wake of COVID-19.” Brookings Ruggles, J. Robert Warren, and Michael West-
Papers on Economic Activity No. 2: 119–58. Ac- berry. 2021. Integrated Public Use Microdata Se-
cessed November 2, 2022. https://www. ries, Current Population Survey: Version 9.0
brookings.edu/wp-content/uploads/2020/06 [dataset]. Minneapolis, Minn.: IPUMS. https://doi
/SU20_S2_Bitler-et-al_-final-paper.pdf. .org/10.18128/D030.V9.0.
Bollinger, Christopher, Barry Hirsch, Charles Ho- Food and Nutrition Service. 2021a. “Bi-Annual (Jan-
kayem, and James Ziliak. 2019. “Trouble in the uary and July) State Project Area/County Level
Talks: What We Know about Earnings Non- Participation and Issuance Data.” Washington:
Response Thirty Years after Lillard, Smith, and U.S. Department of Agriculture. [dataset]. Ac-
Welch.” Journal of Political Economy 127(5): cessed November 2, 2022. https://www.fns.usda
2143–85. .gov/pd/supplemental-nutrition-assistance
Center on Budget and Policy Priorities. 2019. “Policy -program-snap.
Basics: The Supplemental Nutrition Assistance ———. 2021b. “SNAP National Level Monthly Data.”
Program.” Updated June 25, 2019. Washington, Washington: U.S. Department of Agriculture. Ac-
D.C.: Center on Budget and Policy Priorities. Ac- cessed November 2, 2022. https://www.fns.usda
cessed November 2, 2022. https://www.cbpp.org .gov/pd/supplemental-nutrition-assistance
/research/food-assistance/the-supplemental -program-snap.
-nutrition-assistance-program-snap. ———. 2021c. Thrifty Food Plan, 2021. Report no.
———. 2022. “A Quick Guide to SNAP Eligibility and FNS-916. Washington: U.S. Department of Agri-
Benefits.” Updated January 6, 2022. Washington, culture. Accessed November 2, 2022. https://
D.C.: Center on Budget and Policy Priorities. Ac- FNS.usda.gov/TFP.
cessed November 2, 2022. https://www.cbpp.org Fox, Liana 2020. “The Supplemental Poverty Mea-
/research/food-assistance/a-quick-guide-to sure: 2019.” Current Population Report P60-272,
-snap-eligibility-and-benefits. Washington: U.S. Census Bureau. Accessed Feb-
Coleman-Jensen, Alisha, Matthew P. Rabbitt, Chris- ruary 18, 2023. https://www.census.gov/content
tian A. Gregory, and Anita Singh. 2020. “House- /dam/Census/library/publications/2020/demo
hold Food Security in the United States in /p60-272.pdf.
2019.” Report no. ERR-275. Washington: U.S. Fox, Liana, and Kalee Burns. 2021. “The Supplemen-
Department of Agriculture, Economic Research tal Poverty Measure: 2020.” Current Population
Service. Report No. P60-27. Washington: U.S. Census Bu-
———. 2021. “Household Food Security in the United reau. Accessed November 2, 2022. https://www
States in 2020.” Report no. ERR-298. Washing- .census.gov/library/publications/2021/demo/p60
ton: U.S. Department of Agriculture, Economic -275.html.
Research Service. Ganong, Peter, Pascal Noel, and Joseph Vavra. 2020.
Crandall-Hollick, Margot L. 2021. “The Child Tax “US Unemployment Insurance Replacement
Credit: Temporary Expansion for 2021 Under the Rates during the Pandemic.” NBER working pa-
American Rescue Plan Act of 2021.” CRS Insight per no. w27216. Cambridge, Mass.: National Bu-
#IN11613. Washington, D.C.: Congressional Re- reau of Economic Research. Accessed November
search Service. Accessed November 2, 2022. 2, 2022. https://www.nber.org/papers/w27216.
r sf: t he russell sage f ou n dat ion jou r na l of t he so ci a l sciences
58 t h e s o c i o e c o n o m i c i m p a c t s o f t h e c o v i d -1 9 p a n d e m i c
Gray, Colin. 2019. “Leaving Benefits on the Table: 2015. “Household Surveys in Crisis.” Journal of
Evidence from SNAP.” Journal of Public Econom- Economic Perspectives 29(4): 199–226.
ics 179 (November): 104054. DOI: https://doi.org Mueller, Andreas, Jesse Rothstein, and Till von
/10.1016/j.jpubeco.2019.104054. Wachter. 2016. “Unemployment Insurance and
Hardy, Bradley, Tim Smeeding, and James Ziliak. Disability Insurance in the Great Recession.”
2018. “The Changing Safety Net for Low-Income Journal of Labor Economics 34(S1): S445–75.
Parents and Their Children: Structural or Cyclical Parolin, Zachary, Elizabeth Ananat, Sophie Collyer,
Changes in Income Support Policy.” Demography Megan Curran, and Christopher Wimer. 2021.
55 (February): 189–221. “The Initial Effects of the Expanded Child Tax
Hill, Latoya, and Samantha Artiga. 2022. “COVID-19 Credit on Material Hardship.” Brief no. 20413.
Cases and Deaths by Race/Ethnicity: Current New York: Columbia University, Center on Pov-
Data and Changes over Time.” KFF Issue Brief, erty and Social Policy.
August 22. San Francisco: Kaiser Family Founda- Parolin, Zachary, Sophie Collyer, Megan Curran, and
tion. Accessed November 2, 2022. https://www Christopher Wimer. 2021. “Monthly Poverty Rates
.kff.org/coronavirus-covid-19/issue-brief/covid-19 Among Children After the Expansion of the Child
-cases-and-deaths-by-race-ethnicity-current Tax Credit.” Brief no. 20412. New York: Columbia
-data-and-changes-over-time/. University, Center on Poverty and Social Policy.
Homonoff, Tatiana, and Jason Somerville, 2021. “Pro- Parolin, Zachary, Megan Curran, Jordan Matsudaira,
gram Recertification Costs: Evidence from Jane Waldfogel, and Christopher Wimer. 2022.
SNAP.” American Economic Journal: Economic “Estimating Monthly Poverty Rates in the US.”
Policy 13(4): 271–98. Discussion paper. New York: Columbia Univer-
Hoynes, Hilary W., and Diane Whitmore Schanzen- sity, Center on Poverty and Social Policy.
bach 2018. “Safety Net Investments in Children.” Rothbaum, Jonathan, and Adam Bee. 2021. “Corona-
Brookings Papers on Economic Activity No. 1 virus Infects Surveys Too: Survey Nonresponse
(Spring): 89–150. Bias and the Coronavirus Pandemic.” Working
———. 2019. “Strengthening SNAP as an Automatic Paper no. SEHSD WP2020-10. Washington: U.S.
Stabilizer.” In Recession Ready: Fiscal Policies to Census Bureau. Accessed November 2, 2022.
Stabilize the American Economy, edited by https://www.census.gov/library/working-papers
Heather Boushey, Ryan Nunn, and Jay Sham- /2020/demo/SEHSD-WP2020-10.html.
baugh. Washington, D.C.: Brookings Institution. Shaefer, H. Luke, Sophie Collyer, Greg Duncan,
Kuka, Elira, and Bryan A. Stuart. 2021. “Racial In- Kathryn Edin, Irwin Garfinkel, David Harris, Tim-
equality in Unemployment Insurance Receipt and othy Smeeding, Jane Waldfogel, Christopher
Take-Up.” NBER working paper no. w29595. Wimer, and Hirokazu Yoshikawa. 2018. “A Univer-
Cambridge, Mass.: National Bureau of Economic sal Child Allowance: A Plan to Reduce Poverty
Research. and Income Instability Among Children in the
Larrimore, Jeff, Jacob Mortenson, and David Splinter. United States.” RSF: The Russell Sage Founda-
2022. “Unemployment Insurance in Survey and tion Journal of the Social Sciences. 4(2): 22–42.
Administrative Data.” FEDS Notes. Washington, DOI: https://doi.org/10.7758/RSF.2018.4.2.02.
D.C.: Board of Governors of the Federal Reserve USAFacts. 2022. “Deaths from COVID_19.” Down-
System. DOI: https://doi.org/10.17016/2380-7172 loadable data file. Accessed January 3, 2023.
.3135. https://usafacts.org/visualizations/coronavirus
Mathematica Policy Research. 2018. FY 2017 SNAP -covid-19-spread-map.
QC database. Accessed January 3, 2023. https:// USASpending.gov. 2021. “The Federal Response to
snapqcdata.net/datafiles. COVID-19.” Accessed November 8, 2021. https://
———. 2019. FY 2018 SNAP QC database. Accessed www.usaspending.gov/disaster/covid-19
January 3, 2023. https://snapqcdata.net ?publicLaw=all.
/datafiles. U.S. Bureau of Labor Statistics (BLS). 2020. “Fre-
———. 2020. FY 2019 SNAP QC database. Accessed quently Asked Questions: The Impact of the
January 3, 2023. https://snapqcdata.net Coronavirus (COVID-19) Pandemic on the Em-
/datafiles. ployment Situation for April 2020.” Accessed
Meyer, Bruce, Wallace Mok, and James Sullivan. October 8, 2022. https://www.bls.gov/covid19
r sf: t he russell sage f ou n dat ion jou r na l of t he so ci a l sciences
s u f f e r i n g , t h e s a f e t y n e t , a n d d i s p a r i t i e s d u r i n g c o v i d -1 9 59
/employment-situation-covid19-faq-april-2020 Bureau of Labor Statistics. Washington: U.S. De-
.htm. partment of Commerce. Accessed January 3,
———. 2021. “Quarterly Census of Employment and 2023. https://www.census.gov/data/datasets
Wages.” QCEW Open Data Access. Accessed /time-series/demo/cps/cps-asec.2021.html#list
January 3, 2023. https://www.bls.gov/cew -tab-OGPJT8V40Y9IK01AGX.
/additional-resources/open-data/home.htm. ———. 2021b. “Population Density, from 2020 Cen-
U.S. Census Bureau. 2018. “Current Population Sur- sus Demographic Data Map Viewer.” Washing-
vey, 2017 Annual Social and Economic ton: U.S. Department of Commerce. Accessed
(ASEC).” Supplement conducted by the Bureau January 3, 2023. https://www.census.gov/library
of the Census for the Bureau of Labor Statistics. /visualizations/2021/geo/demographicmapviewer
Washington: U.S. Department of Commerce. Ac- .html.
cessed January 3, 2023. https://www.census.gov U.S. Department of Labor. 2004. “UI Replacement
/data/datasets/time-series/demo/cps/cps-asec Rates Report.” Created March 29, 2004. Updates
.2017.html#list-tab-QZJHN1DP8SLZ903PMP. July 7, 2022. Accessed November 1, 2022.
———. 2019. “Current Population Survey, 2018 An- https://oui.doleta.gov/unemploy/ui_replacement
nual Social and Economic (ASEC).” Supplement _rates.asp.
conducted by the Bureau of the Census for the ———. 2022. “Families First Coronavirus Response
Bureau of Labor Statistics. Washington: U.S. De- Act and Coronavirus Aid, Relief, and Economic
partment of Commerce. Accessed January 3, Security (CARES) Act Funding to States through
2023. https://www.census.gov/data/datasets October 8, 2022.” Accessed October 7, 2022.
/time-series/demo/cps/cps-asec.2018.html#list https://oui.doleta.gov/unemploy/docs/cares_act_
-tab-LOWBNPSTB2TGIHZ7BR. funding_state.html.
———. 2020a. “American Community Survey 2015– U.S. Department of the Treasury. Bureau of the Fis-
2019 5-Year Data Release.” Washington: U.S. De- cal Service. 2022. “Monthly Treasury Statement,”
partment of Commerce. Accessed January 3, November 30. Accessed January 3, 2023.
2023. https://www.census.gov/newsroom/press https://fiscaldata.treasury.gov/datasets/monthly
-kits/2020/acs-5-year.html. -treasury-statement/summary-of-receipts
———. 2020b. “Current Population Survey, 2020 An- -outlays-and-the-deficit-surplus-of-the-u-s
nual Social and Economic (ASEC).” Supplement -government.
conducted by the Bureau of the Census for the Whittaker, Julia M., and Katelin P. Isaacs. 2021. “Un-
Bureau of Labor Statistics. Washington: U.S. De- employment Insurance: Legislative Issues in the
partment of Commerce. Accessed January 3, 116th Congress.” CRS Report no. R45478. Wash-
2023. https://www.census.gov/data/datasets ington, D.C.: Congressional Research Service.
/time-series/demo/cps/cps-asec.2020.html#list Zalla, Lauren, Grace Mulholland, Lindsey Filatreau,
-tab-TLBI8972WSCHODVSVO. and Jessie Edwards. 2022. “Racial/Ethnic and
———. 2020c. “Source of the Data and Accuracy of Age Differences in the Direct and Indirect Effects
the Estimates for the 2020 Household Pulse Sur- of the COVID-19 Pandemic on US Mortality.”
vey.” Washington: U.S. Department of Com- American Journal of Public Health 112(1): 154–64.
merce. Accessed November 2, 2022. https:// Ziliak, James, 2015. “Why Are So Many Americans
www2.census.gov/programs-surveys/demo on Food Stamps? The Role of the Economy, Pol-
/technical-documentation/hhp/Source-and icy, and Demographics.” In SNAP Matters: How
-Accuracy-Statement-May-14-May-19.pdf. Food Stamps Affect Health and Well-Being, ed-
———. 2021a. “Current Population Survey, 2021 An- ited by Judith Bartfeld, Craig Gundersen, Timo-
nual Social and Economic (ASEC).” Supplement thy Smeeding, and James Ziliak. Palo Alto, Calif.:
conducted by the Bureau of the Census for the Stanford University Press.
r sf: t he russell sage f ou n dat ion jou r na l of t he so ci a l sciences
File and source
- File
- suffering-the-safety-net-and-disparities-during-covid-19.pdf
- Size
- 4,581,751 bytes
- SHA-256
- 9fcac319b5444a84428dacb55e636631dc155b9c95d818a4c57ad9bd58f9dd70
- Original
- doi.org