Bauer, Ruffini, Schanzenbach, The Effects of Lump-Sum Food Benefits during the COVID-19 Pandemic on Spending, Hardship, and Health (NBER w33199, 2024-11)
Summary
NBER Working Paper 33199, dated November 2024, by Lauren L. Bauer, Krista J. Ruffini and Diane Whitmore Schanzenbach. It studies P-EBT, a program that provided grocery vouchers worth approximately $300 per student during spring and summer 2020, using cross-state variation in program timing. The abstract states that families spent $18-42 per student per week in the 6 weeks after benefit receipt. It reports that household food insufficiency and children's food insecurity among low-income families declined by 27-49% in the month following receipt, and that maternal mental health improved by 0.9 standard deviation.
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
Cite this document
“Bauer, Ruffini, Schanzenbach, The Effects of Lump-Sum Food Benefits during the COVID-19 Pandemic on Spending, Hardship, and Health (NBER w33199, 2024-11).”
Original at www.nber.org. Archived at Pandemic Darlings, https://pandemicdarlings.com/source-documents/bauer-ruffini-schanzenbach-the-effects-of-lump-sum-food-benefits-durin-3aa4e1b61f/
, SHA-256 4ed457d0970e4cdee83457cbe27e0bb9fd7859946c2cf552bde3be3af5a4692c.Full text
NBER WORKING PAPER SERIES
THE EFFECTS OF LUMP-SUM FOOD BENEFITS DURING THE COVID-19 PANDEMIC ON
SPENDING, HARDSHIP, AND HEALTH
Lauren L. Bauer
Krista J. Ruffini
Diane Whitmore Schanzenbach
Working Paper 33199
http://www.nber.org/papers/w33199
NATIONAL BUREAU OF ECONOMIC RESEARCH
1050 Massachusetts Avenue
Cambridge, MA 02138
November 2024
The authors thank Jess Belledonne, Eliana Buckner, Ronnie Clevenstine, Kirsten dela Cruz, Moriah
Macklin, Emily Moss, Abigail Pitts, Elisabeth Raczek, Beatriz Rivera, Eric Schroeder, Bree Seiler,
Payal Soneja, Aaron Till, Natalie Tomeh, Jennifer Umanzor, Sarah Wheaton, and Winnie Yee for
exceptional diligence and research assistance over the course of this project. We are grateful to
Raheem Chaudhry, Sheila Fleischhacker, Hilary Hoynes, Zoë Neuberger, Dottie Rosenbaum, and
federal and state officials for sharing program implementation information with us. Lindsey
Bullinger, Kristin Butcher, Wendy Edelberg, Poonam Gupta, Emily Gutierrez, Joseph Llobera,
Michah Rothbart, Dottie Rosenbaum, Elaine Waxman, and participants at the Association for
Education Finance, Association for Public Policy and Management, Southern Economics
Association, Policy and American Society of Health Economists, NBER Summer Institute, Virtual
Economics of Poverty and Policy Seminar, NOPREN working group, and Economic Studies at
Brookings seminars provided helpful feedback on this work. Bauer and Ruffini acknowledge support
from Healthy Eating Research (Grant 111326). The views expressed in this piece are the authors’
alone. The views expressed herein are those of the authors and do not necessarily reflect the views of
the National Bureau of Economic Research.
NBER working papers are circulated for discussion and comment purposes. They have not been peer-
reviewed or been subject to the review by the NBER Board of Directors that accompanies official
NBER publications.
© 2024 by Lauren L. Bauer, Krista J. Ruffini, and Diane Whitmore Schanzenbach. All rights
reserved. Short sections of text, not to exceed two paragraphs, may be quoted without explicit
permission provided that full credit, including © notice, is given to the source.
The Effects of Lump-Sum Food Benefits during the COVID-19 Pandemic on Spending, Hardship,
and Health
Lauren L. Bauer, Krista J. Ruffini, and Diane Whitmore Schanzenbach
NBER Working Paper No. 33199
November 2024
JEL No. H51, H53, I18, I38
ABSTRACT
This paper examines how providing families with lump-sum in-kind assistance during the pandemic
affected food hardship, economic well-being, and maternal health. We study the introduction of a
new program, P-EBT, that provided grocery vouchers worth approximately $300 per student during
spring and summer 2020. Using cross-state variation in program timing, we find that families spent
$18-42 per student per week in the 6 weeks after benefit receipt. Household food insufficiency and
children’s food insecurity among low-income families declined by 27-49% in the month following
receipt, and maternal mental health improved by 0.9 standard deviation.
Lauren L. Bauer Diane Whitmore Schanzenbach
Brookings Institution Institute for Policy Research
lbauer@brookings.edu Northwestern University
2040 Sheridan Road
Krista J. Ruffini Evanston, IL 60208
McCourt School of Public Policy and NBER
Georgetown University dws@northwestern.edu
125 E St. NW
Washington, DC 20001
and NBER
kr333@georgetown.edu
1. Introduction
Economic downturns frequently lead to increased food insecurity as the number of households
being unable to afford basic necessities rises. The experience during the COVID-19 pandemic was
no different. During spring and summer 2020 when unemployment soared to unprecedented levels,
about 27% of households with children reported sometimes or often being unable to afford food
(U.S. Census Bureau Household Pulse Survey 2020). While this high level of insecurity partially
reflected the severe downturn, factors unique to the pandemic setting, such as stay-at-home orders,
school closures, and business restrictions likely contributed to this hardship. Over this period,
federal and state governments implemented numerous innovative programs in an effort to stabilize
the economy and households’ balance sheets (Ruffini and Wozniak 2021). In many cases,
however, empirical evidence on the effects of these programs remains unknown, even years later.
Yet the pandemic-era policy environment provides a valuable opportunity to examine the
effectiveness of safety net expansions, which may inform how policymakers design responses to
future economic crises.
Nutrition assistance programs are particularly important to evaluate as food hardship
presents both short- and long-term costs (Hoynes and Schanzenbach 2019). For children, food
insecurity is associated with worse health (Case, Fertig, and Paxson 2005) and poorer academic
outcomes (Jyoti, Frongillo, and Jones 2005) such as lower test scores and difficulty getting along
with others (Howard 2011). In the long-term, exposure to adverse economic shocks during
childhood negatively affects health and economic outcomes into adulthood (Hoynes and
Schanzenbach 2018). Therefore, the food insecurity patterns we observe today shape well-being
and economic performance not just now, but potentially for decades to come.
2
While many permanent programs, such as the Supplemental Nutrition Assistance Program
(SNAP, formerly food stamps) are countercyclical (Bitler and Hoynes 2016, Ganong and Liebman
2018), there are reasons to expect that the existing safety net may be inadequate in helping
households weather an economic downturn, particularly one in which schools and businesses are
closed. Households frequently exhaust monthly SNAP benefits in the first weeks after receipt,
leading to fluctuations in both consumption and nutritional intake over the month (Hastings and
Washington 2010, Wilde and Ranney 2000, Shapiro 2005, Seligman et al. 2014, Hamrick and
Andrews 2016). Increasing SNAP payments has been shown to reduce this volatility (Todd 2015,
Todd and Gregory 2018).
The existing literature investigates consumption responses to nutrition assistance examines
programs that are recurrent monthly in-kind benefits, like SNAP. More broadly, the literature on
consumption responses to safety net programs has focused on either recurrent in-kind benefits or
intermittent one-time cash transfers (such as the Earned Income Tax Credit (EITC) or stimulus
checks) and has not included measures of subjective well-being. Therefore, it is unclear how
families respond to one-time, lump-sum increases in nutrition assistance as such programs are
unprecedented.
In this paper, we provide new information on the spending responses of a one-time, in-kind
transfer, and how this influx of resources maps onto measures of reported well-being, including
food hardship and maternal mental health. We examine these questions in the context of additional
nutrition assistance during the spring and summer of 2020 when widespread school closures due
to the onset of the COVID-19 pandemic cut off access to school meals for more than 20 million
students who had been receiving free or reduced-price school meals. In response, Congress
authorized a new program, Pandemic Electronic Benefit Transfers (Pandemic EBT, or “P-EBT”),
3
to provide grocery vouchers equivalent to the dollar value of missed school meals to families that
were affected by COVID-related school closures and whose children were eligible to receive free
school meals. P-EBT was different from many existing programs in that it provided lump-sum, in-
kind benefits; in contrast, other assistance programs are recurrent cash or in-kind transfers or are
lump sum cash payments.
As a state-led program operating for the first time during a crisis, there was substantial
idiosyncratic variation across states in the implementation of P-EBT that was uncorrelated with
other state policy or political conditions. We leverage variation in the initial disbursement of
benefits from April through August 2020—a period with universal school closures—combined
with biweekly data on food hardship and household well-being, as well as weekly data on EBT
spending to provide some of the first evidence on how P-EBT affected spending, food hardship,
and household well-being. 3
We find that that P-EBT benefits are spent across many weeks. Average weekly state-level
EBT spending per student eligible for free or reduced-price meals increased by $18-42 in each of
the subsequent six weeks after P-EBT payments are first made in a state. This spending pattern is
both more modest and more sustained than the sharp, short-lasting increase in spending
immediately after households receive other forms of assistance like SNAP (Hastings and
Washington 2010; Wilde and Ranney 2000, Todd 2015, Todd and Gregory 2018, Shapiro 2005)
or the EITC (Alagangady et al. 2023). Moreover, volatility in SNAP redemptions is less
pronounced in the month after families receive P-EBT. These spending patterns suggest that
families spend one-time nutrition assistance differently than routine benefits and that such transfers
3
See Bonomo et al. (2024) for a more detailed investigation of household shopping patterns.
4
can smooth short-term spending volatility, consistent with earlier work finding greater
consumption smoothing with increased SNAP benefits (Todd and Gregory 2018).
In the month following disbursement, families also report lower levels of food hardship,
with household reports of food insufficiency falling by 30 percent and very low food security
among children falling by approximately half. We also find short-term improvements in mothers’
mental health. While the improvements in mental health are concentrated in the weeks
immediately after receipt, the reduction in reported food insufficiency persists for at least a month
after disbursement—the period for which we observe higher benefit spending. That reduced food
hardship coincides with the period over which spending increases points to the potential of
additional assistance to reduce nutritional fluctuations over the benefit month, with potential
consequences for healthcare utilization (Hamrick and Andrews 2016, Seligman 2014, Cotti,
Gordanier and Ozturk 2020). While annual cash-based assistance has been shown to improve
maternal mental health (Evans and Garthwaite 2014), our results provide novel evidence that in-
kind nutrition assistance can yield similar benefits. Both the improvement in maternal mental
health and reported reduction in food hardship dissipate more quickly than the spending response.
2. The COVID-19 pandemic and the P-EBT program
2.1 Early COVID pandemic
In spring 2020, the U.S. economy experienced the sharpest contraction on record as businesses
and schools closed in the wake of the COVID-19 pandemic. The economy hit a trough in April
2020, when the unemployment rate soared to 14.8%. Although it retrenched quickly from this
high, the unemployment rate remained elevated throughout the summer, standing at 7.8% in
September 2020 compared to 3.5% a year prior. Accordingly, household income fell due to job
5
loss and reduced working hours, particularly at the lower end of the income distribution (Shrider
et al. 2021).
Some existing safety net programs, such as SNAP and Unemployment Insurance (UI),
served as automatic stabilizers and partially replaced this lost income (Hembre et al. 2024, Bitler
et al. 2023a). In addition, these programs were modified in an attempt to stabilize the economy—
for example, UI was expanded in benefit amount and eligibility and SNAP benefits were increased
for most households and the requirement to recertify for benefits was temporarily suspended for
current participants (Hembre et al. 2024, Ruffini and Wozniak 2021, Bitler et al. 2020a, 2020b,
2020c). In addition, Congress created new programs and passed new rules. In addition to P-EBT,
more than 160 million families received stimulus checks (“Economic Impact Payments”) of up to
$1,200 per adult and $600 per child prior to our analysis period in April 2020 (U.S. Internal
Revenue Service 2024). There were also eviction moratoria in place in various forms throughout
our sample period from March 2020 through August 2021 (Keene et al. 2023). Appendix Figure
A1 shows the timing of several of the largest programs implemented or expanded during the period
using daily Treasury disbursement data (throughout this period, P-EBT is reported with SNAP).
The figure shows that of these programs, UI was the largest during our analysis period. The large
cash assistance payments through the Economic Impact Payments (EIP) and expanded Child Tax
Credit (CTC) were made outside of our sample window.
The net effect of these policy changes was a reduction in the annual poverty rate under the
Supplemental Poverty Measure (SPM), a metric that includes taxes and transfers (Fox and Burns
2021). However, when using a monthly measure of income, SPM poverty increased beginning in
April 2020 after most of the initial round of stimulus checks had been paid (Parolin et al. 2022).
6
Despite the robust policy response, many families faced economic challenges. UI payments
were delayed in many states (Ruffini and Wozniak 2021) and some families were ineligible for
additional assistance, including the lowest-income families already receiving the maximum SNAP
amount, immigrants, and those who did not qualify for UI (Bitler et al. 2020a, 2020b). For many
families, the hardship from lost income was compounded by widespread childcare and school
closures and rising food prices.
One factor that likely contributed to this hardship was the loss of school meals related to
school closures. The school meals programs (collectively the National School Lunch Program and
the School Breakfast Program) are a substantial source of nutrition assistance to school-aged
children. In 2019, approximately 20 million children – 38 percent of the population aged 5-17 –
received a free school meal on a typical school day; for comparison about one in five receive SNAP
benefits (USDA 2024, King and Giefer 2021). The school meals programs are also a sizable in-
kind transfer to low-income families with a fungible value of approximately $6.50 per student per
school day in 2023, compared to a maximum per-person SNAP benefit of about $8 per (calendar)
day for a family of three (USDA 2022). To receive free school meals, students’ families must
certify that their family income is no more than 130 percent of the federal poverty line (in 2023,
$32,318 for a single parent with two children and $39,000 for a family of four), receive SNAP or
Temporary Assistance for Needy Families (TANF), or attend a school that offers a schoolwide
free meals program. 4
4
Schoolwide free meals programs include Provisions I-III, and the Community Eligibility Provision (CEP). In 2019,
approximately 14 million students attended a CEP school (Billings and Carter 2020), most of whose families also
received SNAP.
7
To compensate families for the value of forgone school meals, Congress authorized a new
program in March 2020, P-EBT, to provide families with benefits in the amount of the value of
the free school breakfasts and lunches missed due to pandemic-related school closures. 5
2.2 Design of P-EBT
P-EBT was a new program that functioned similarly to SNAP: families that were eligible for free
and reduced-price school meals received benefits on an EBT debit card, which could then be used
to purchase food at most grocery stores. 6 P-EBT was a voluntary, federally funded, state-led
program: states and territories had to apply to USDA by submitting an implementation plan. While
every state and territory eventually submitted a plan and received approval to implement P-EBT
for the 2019-20 school year, states implemented the program on different schedules (Gupta et al.
2021). Crucial for this analysis, states varied in how much time it took to develop an
implementation plan (which was driven in large part by whether student-level data were readily
available and could be integrated with SNAP EBT systems), obtain approval from USDA, and
begin disbursing benefits (Koné Consulting, 2020).
There is relatively little variation in the per-child amount awarded through the first P-EBT
disbursement. By formula, payments per student were calculated as the daily federal
reimbursement to schools for free-price breakfasts and lunches ($5.70 in the 48 contiguous states
and D.C.) multiplied by the average number of school days missed due to pandemic-related school
closures. 7 Most states closed schools within days of March 15, 2020, and were closed for the rest
5
Because we examine a period in which school was not in session for nearly all students, we are unable to speak to
the relative efficiency of providing nutrition assistance through school-based or household programs. Bauer et al.
(2023) note that when children are in primarily eating meals at home, providing resources directly to families is
likely more efficient than when students are in congregate settings. Some evidence on this point is provided by the
low take-up rates of “grab-and-go” brownbag school meals in the early months of the pandemic (Bauer et al. 2020).
6
In particular, P-EBT could be used to purchase SNAP-eligible foods at food retailers that accept SNAP benefits.
7
Rates were $6.66 and $9.16 per day in Hawaii and Alaska, respectively.
8
of the academic year, so cross-state differences in the payment amount was determined by planned
vacation days and timing of the end of the school year. Almost all states made a single lump-sum
payment for the number of school days missed from March 2020 through the end of the school
year. 8 Based on information in state plans issued to USDA, the average payment was $311 per
child, with a difference in disbursement amounts between the 25th ($268) and 75th ($332) percentile
of states of $64 per student. This average P-EBT payment amount is nearly twice the maximum
per-person monthly SNAP benefit for a family of 3 in 2020 ($170).
Families that were participating in SNAP did not need to apply to the program or go
through other administrative burdens to obtain the benefits. For these families, P-EBT was
automatically added to their existing EBT cards. Families that were not on SNAP but were
receiving free school meals were issued new EBT cards loaded with the P-EBT benefits. In states
like California, Florida, Missouri, and Texas, non-SNAP families had to proactively apply for the
benefits through portals that states or schools developed. Other states, such as Minnesota, Nevada,
and Ohio, relied on existing administrative linkages across programs when possible. Yet other
states, such as Louisiana, developed these administrative linkages to process P-EBT.
No matter how strong the administrative linkage was, some eligible non-SNAP students
could not be identified immediately, requiring additional follow-up and retroactive payments at
later dates. This process and the need to issue new EBT cards resulted in later and more varied
disbursements for SNAP non-participants. While many states issued at least some payments to
non-SNAP households the same day as SNAP households, in other states, the first non-SNAP
households received payments 3-4 weeks later. Because administrative burdens and the timing of
receipt varied substantially across and within states for the non-SNAP population, we focus on
8
About 84% of students lived in states that reported making one payment to SNAP families during summer 2020
(U.S. Department of Education 2023).
9
families that were income-eligible for SNAP and therefore less likely to be affected by such
application processes and retroactive payments and for whom we can better identify the timing of
receipt. This population is a large share of children who eventually received P-EBT, as
approximately 60% of students who receive free or reduced-price meals participate in SNAP
(Federal Register 2016).
2.3 Timing of P-EBT disbursements
Below we estimate the impact of P-EBT payments on a series of outcomes using event-study and
difference-in-differences frameworks. We combine several sources of information to construct a
database that captures the first date that P-EBT benefits were paid to SNAP-participating families
in each state from April through August 2020. Our first source is drawn from correspondence with
state and D.C. school nutrition officials between June 2020 and August 2022. We contacted each
state at least three times and we received information on P-EBT disbursement dates from 46 states
(of the remaining five states, four reported that they had not disbursed benefits by the end of August
2020). Subsequently we issued Freedom of Information Act (FOIA) requests to all 50 states and
D.C. requesting more detailed information about the dates of disbursement, the populations
receiving each disbursement (SNAP recipient households, non-SNAP recipient households,
school-aged children, and childcare populations), the number of children or families receiving each
disbursement, and any application processes. Forty-one states and D.C. provided specific date and
population information through the FOIA process. From these sources we have reports of exact
dates of initial P-EBT payments to SNAP-participating families. We collect a third source of P-
EBT payment timing from USDA public reports that include only the month(s) (but not specific
days) in which payments were made.
10
In eleven cases, there were discrepancies between the payment dates provided through the
FOIA requests and those obtained via the earlier correspondence. We reconciled both sources with
the state-by-month disbursement information published by USDA and selected the date that
corresponded to the first month that states began issuing benefits. 9 When there was disagreement
within the month between the FOIA and initial requests, we use the date provided through the
FOIA request. We drop five states (AK, IN, UT, WI, and WY) with conflicting dates across all
three sources. Appendix Table A1 summarizes the date provided through each measure and the
states used in the event-study and difference-in-differences analyses. 10
Figure 1 illustrates timing in initial P-EBT disbursements across states over time. The first
states issued benefits in April 2020 and the last states that made payments during our analysis
period issued them in August 2020. Note that the five states that are dropped due to irreconcilable
dates are in gray.
9
We include Michigan even though the reported dates fall in different months because they are within 3 days of
each other and include Idaho as all sources report payments were issued after the end of our analysis period.
10
Appendix Table A8 replicates our analysis using the first month that P-EBT appears in the USDA data without
leveraging any information collected from state officials. For both measures of food hardship, improvements are
sharply attenuated under this approach, highlighting the importance of more precise daily payment data, particularly
when considering safety net programs where changes are immediate or short-lived (Aladangady et al. 2023).
11
Figure 1: P-EBT initial payment timing
Note: Figure shows when each state issued the first P-EBT payment. States shaded in gray are excluded from the
analysis; states in dark purple had not disbursed the first payment by mid-August 2020.
The identifying assumption for a causal interpretation in our event-study and difference-
in-differences models below is that the timing of states’ initial payment of P-EBT benefits is not
correlated with the timing of other factors affecting food hardship, spending, or family well-being.
While this assumption is not directly testable, Appendix Table A2 presents results from a
regression where the dependent variable is either the day (column 1) or HPS survey wave
(column 2) of initial P-EBT disbursement among states that made payments between April through
August 2020. States with a larger school-age population (as a share of the total population) tended
to implement P-EBT later in the period. However, no other state demographic, economic, policy
or political factor in the model predicts P-EBT timing, including measures of welfare state
generosity (minimum wage, EITC), administrative burden (proxied by broad-based categorical
eligibility for SNAP and UI processing delays), or COVID responses (mask mandate or stay-at-
home-order). Altogether, these observable characteristics account for less than 40% of the timing
of P-EBT disbursement, supporting the assumption that implementation timing is as-good-as-
12
random across states. To further ensure that our results are not capturing COVID policy responses
changing at the same time, all of our results control for whether a stay-at-home order, mask
mandate, or indoor restaurant closure was in effect, as well as the share of UI claims processed
during the survey month that were within 1 week or 2 weeks of claim filing, and the share delayed
by more than 10 weeks.
3. Data
We combine survey and administrative data with timing of P-EBT payments to measure the impact
of payments on spending and well-being. We describe each data source below.
3.1 Administrative data on state-level benefit spending and economic conditions
To determine how quickly P-EBT benefits were spent after states issued payments to families, we
obtain administrative data on benefit spending from the USDA’s Store Tracking and Redemption
System (STARS) for 2015 through 2020. Establishing such a “first-stage” is crucial, particularly
in the policy environment discussed in Section 2.1 when there were a number of policy changes
occurring in spring and summer 2020. STARS provides the dollar amount of all EBT spending
(SNAP until 2020 and P-EBT and SNAP combined from 2020) at the state-by-week level. These
aggregate data do not separate P-EBT from SNAP spending or distinguish spending by SNAP
households from those who received P-EBT but not SNAP. With these caveats in mind, the
STARS data allow us to examine how total EBT spending responds following initial P-EBT
payments. We divide the total redemptions by the number of students eligible for free- or reduced-
price meals in order to provide a measure that is comparable across states and use data covering
the same period for which we have HPS data (described below, April 23 through August 31, 2020).
13
We also use state-level administrative data on the unemployment rate and number of
workers claiming UI to test whether P-EBT timing was correlated with other state-level economic
factors. These data are provided by the federal Department of Labor and show the insured
unemployment rate (the number of workers receiving UI benefits divided by the number of jobs
covered by UI) and the UI claimant rate per FRP student (the number of UI claims divided by the
number of students eligible for free- or reduced-price meals). Again, we use data for the period
covered by the HPS data.
3.2 Household Pulse Survey data
Data on food hardship and reported household well-being come from the Census Bureau’s
Household Pulse Survey (HPS). We use the first 13 waves of HPS data collection spanning
April 23 through August 31, 2020. During this period, HPS was collected approximately every
two weeks. HPS data include standard demographic information, the respondent’s state of
residence, and questions about households’ economic and health status. The nature of these
questions, combined with the high frequency of the data, provides a unique opportunity to identify
the short-term effects of policy changes across the country.
We examine the effects of P-EBT on two measures of food hardship. First, “food
insufficiency” is defined as whether the respondent reports that their household sometimes or often
did not have enough to eat over the previous seven days. Second, very-low food security among
children is whether the respondent reports that in the last seven days the children in the household
sometimes or often did not eat enough because the household could not afford food.
Respondents are also asked to report the frequency over the past week that they experienced
a range of poor mental health symptoms, with options ranging from not at all, to several days, to
14
nearly every day. Outcomes include the four following domains: being nervous, anxious, or on
edge; being unable to stop worrying; having little interest or pleasure in doing things; and being
down, depressed or hopeless. We combine responses to these four questions, standardizing them
into a z-score to measure mothers’ current mental health so that higher values indicate worse
mental health. We also explore a “stock” version of health measures as the response to where one’s
health “in general” ranges from excellent to poor. As a measure of general household financial
health, we use the response to a question on the respondent’s confidence their household will be
able to afford the kinds of food they need over the next four weeks.
Respondents are also asked to report their total spending on food to prepare and eat at home
over the past week, including purchases made with SNAP. 11 We scale this spending by the number
of children in the household, and take its log. This measure of spending is complementary to the
STARS data; while it has the advantage of being available for our target population (SNAP-
eligible households), it is prone to concerns of measurement error: Appendix Figure A2 shows the
distribution of reported spending prior to receiving P-EBT. During this period, nearly 15% of
households in our sample reported spending exactly $200 and 70% report food spending in
multiples of $50. 12 Given the pronounced heaping, this measure of spending, while having the
advantage of focusing on our low-income population with children, is prone to concerns of
attenuation bias.
SNAP benefits typically do not fully cover households’ food budgets and represent less
than 67% of grocery spending for 26% of SNAP recipients (Hoynes, McGranahan, and
Schanzenbach 2015). Although P-EBT can only be used for grocery purchases, for some families
Respondents were instructed to exclude any spending on alcoholic beverages.
11
In contrast, only 4% of households overall (3.6% of households reporting SNAP receipt) in the 2019 December
12
Current Population Survey report spending exactly $200 on food in the prior week, and 35% (overall and among
SNAP recipients) report spending in multiples of $50.
15
this assistance is inframarginal in the sense that they would normally spend more than their SNAP
benefit plus P-EBT within several months. 13 As a “near-cash” benefit, P-EBT could have
alleviated other financial pressures and allowed families to increase spending on housing, utility
bills, or other expenses. However, other measures of consumption hardship were not routinely
asked over our sample period, and questions related to making housing payments are confounded
by national and state eviction moratoria over our sample period. Therefore, with the available data,
we are unable to directly speak to more general spending patterns.
4. Empirical strategy
When investigating outcomes from the HPS, we leverage cross-state variation in the timing of the
first P-EBT disbursement in a difference-in-differences or event-study framework. Starting with
the simpler difference-in-differences approach, for each outcome y for family i living in state s at
time t, we estimate:
′
𝑦𝑦𝑖𝑖𝑖𝑖𝑖𝑖 = 𝛽𝛽𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑠𝑠𝑠𝑠 + 𝑋𝑋𝑖𝑖𝑖𝑖𝑖𝑖 𝜃𝜃 + 𝛿𝛿𝑠𝑠 + 𝛾𝛾𝑡𝑡 + 𝜑𝜑𝑠𝑠′ 𝑡𝑡 + 𝜀𝜀𝑖𝑖𝑖𝑖𝑖𝑖 (1)
where the coefficient of interest, 𝛽𝛽, measures whether P-EBT was first disbursed in the 2 weeks
prior to the start of the survey wave. During the period studied, the HPS data only include
information on the total number of children and whether there are any school-age children (but not
the number of school-age children). Since we cannot determine the actual amount families
received (number of school-aged children times per student amount), our approach uses a binary
treatment measure equal to one if P-EBT was disbursed within 2 weeks before the start of the HPS
wave. Since the HPS is conducted on an approximately bi-weekly basis over our analysis period,
13
USDA did not set a standard expungement date for P-EBT but advised states to cancel any benefits that had not
been spent after 365 days (USDA 2020).
16
our measure corresponds to receiving P-EBT in the previous 0-4 weeks. 14 To avoid comparisons
between newly treated and previously treated states (Goodman-Bacon 2021, Roth et al. 2023), we
exclude states that made a disbursement more than 2 weeks before the start of the survey wave.
Accordingly, each state is “treated” for up to two HPS waves and the “control” observations are
states that had not yet made P-EBT payments by the start of the survey wave. We further explore
dynamic responses in an event-study framework and using shorter (1 week prior to the start of the
survey wave) and longer (3 week) treatment windows (Appendix Table A4).
′
𝑋𝑋𝑖𝑖𝑖𝑖𝑖𝑖 is a vector of standard control variables, including respondent age, race/ethnicity,
educational attainment, marital status, the number of children and adults in the household, and the
state unemployment rate during the reference period. In addition, we control for other state-specific
policy responses to the pandemic that may have affected food hardship, specifically whether the
state of residence made SNAP Emergency Allotment payments in the month of observation, the
monthly share of usual SNAP disbursements paid during in the 2 weeks leading up to the survey
wave, whether a state had a mask mandate or stay-at-home-order at the start of the survey wave,
and the share of UI claims processed within 1, 2, or more than 10 weeks of the claiming date. 15 𝛿𝛿𝑠𝑠 ,
𝛾𝛾𝑡𝑡 , and 𝜑𝜑𝑠𝑠′ 𝑡𝑡 are state fixed effects, survey-week fixed effects, and state-specific linear time trends,
14
At one extreme, if P-EBT was issued the day before the survey period began, households surveyed the first day
would have had P-EBT for 1 day when they answered the survey. At the other extreme, if P-EBT was disbursed 13
days before the start of the survey period, households that were interviewed towards the end of the period
(approximately 1-2 weeks after the survey period began) would have received P-EBT approximately 3-4 weeks
prior.
15
Emergency Allotment payments increased monthly SNAP benefits by an average of approximately 40% in
families with children (Bitler et al. 2023a).
17
respectively. 16 All analyses use person weights for the respondent, and standard errors are
clustered at the state level. 17
We extend the difference-in-difference set-up to an event-study framework in order to track
the evolution of hardship over time and to assess the plausibility of the parallel trends assumption
(i.e., that states were not on different trajectories in the weeks leading up to payments). The event-
study framework takes the same structure as the difference-in-difference approach, but replaces 𝛽𝛽
with a vector of event-time indicators for the j weeks leading up to and following P-EBT payments:
′
𝑦𝑦𝑖𝑖𝑖𝑖𝑖𝑖 = ∑𝑗𝑗≠−1 𝛽𝛽𝑗𝑗 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑠𝑠,𝑡𝑡=𝑗𝑗 + 𝑋𝑋𝑖𝑖𝑖𝑖𝑖𝑖 𝜃𝜃 + 𝛿𝛿𝑠𝑠 + 𝛾𝛾𝑡𝑡 + 𝜑𝜑𝑠𝑠′ 𝑡𝑡 + 𝜀𝜀𝑖𝑖𝑖𝑖𝑖𝑖 (2)
If the effect of P-EBT varies across states that issued payments at different times, the
approach in Equation (2) will yield a biased estimate of 𝛽𝛽𝑗𝑗 (Sun and Abraham 2021). To address
this concern, we additionally follow the recommendation in Roth et al. (2023) and report results
from a stacked event-study design in Appendix Figure A4, as in Cengiz et al. (2019). 18 Results
from this approach are very similar to the standard TWFE event study, suggesting that
heterogeneous effects over time are not an acute concern in this setting.
States are only included in the event study approach if there is a balanced panel—that is,
the state is observed for each of the j weeks before and after payments were made. Given the
relatively small number of time periods (13 HPS waves), the varying dates over which states first
issued payments, and differences in the length of the HPS waves, the states included in a balanced-
sample event-study approach vary by the window included in the analysis. To show sensitivity to
16
Appendix Table A7 provides results without state trends, which are somewhat attenuated relative to the main
findings.
17
Results are substantially unchanged if instead we weight by quasi-household weights, defined as the person
weight divided by the number of adults in the household. HPS began including household weights after our analysis
period.
′
18
Specifically, we estimate 𝑦𝑦𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 = ∑𝑗𝑗≠−1 𝛽𝛽𝑗𝑗 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑠𝑠𝑠𝑠,𝑡𝑡=𝑗𝑗 + 𝑋𝑋𝑖𝑖𝑖𝑖𝑖𝑖 𝜃𝜃 + 𝛿𝛿𝑠𝑠𝑠𝑠 + 𝛾𝛾𝑡𝑡𝑡𝑡 + 𝜀𝜀𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 where each cohort of treated
states c is matched to states that implemented P-EBT in ~c > c (e.g.: a later period) and all fixed effects are interacted
with the treatment cohort c. The stacked event study is conducted using corrected sample weights as in Wing et al.
(2024).
18
the analysis window, we provide three sets of event-study results: 1) survey waves starting 2 weeks
before a state made P-EBT payments through 4 weeks after (Figure 3); 2) survey waves starting 4
weeks prior to P-EBT payments through a full week after (Appendix Figure 2); 3) survey waves
starting 4 weeks prior to P-EBT through 4 weeks after in a stacked event study design following
Cengiz et al. (2019) and Roth et al. (2023). The states used in each sample are reported in Appendix
Table A1 and the DD results for each subsample, as well as the stacked DD for the stacked event
study, are provided in Appendix Table A3.
When we use the state-by-week STARS data to measure the duration of the spending
response or the UI data to examine other state-level economic conditions, we estimate similar
event-study frameworks that exclude household demographic characteristics which are not
available in those data.
5. Results
5.1 Benefit spending
Figure 2 shows results of an event-study analysis of the impact of initial P-EBT disbursement on
total state-level EBT spending divided by the number of public-school students eligible for free-
or reduced-price meals in the state, using state-week data from USDA’s administrative STARS
data. We limit the analysis to the subset of states for which we have a balanced panel (i.e., states
that initially disbursed P-EBT at least 6 weeks before September 1, 2020) and normalize EBT
spending to 0 in the week prior to P-EBT disbursement.
19
Figure 2: Timing of State-Level EBT spending
Notes: Figure shows event studies on state-week EBT spending per-FRP-eligible public-school student relative to the
week prior to disbursement using the approach in Equation 2. Sample includes a balanced panel of 16 states that
disbursed P-EBT between May 30 and July 15, 2020. Vertical bars denote 95% confidence intervals of robust standard
errors clustered by state.
The small, insignificant coefficients in weeks -4 through -1 show that EBT spending (reflecting
SNAP EBT spending) was not trending in the weeks leading up to P-EBT treatment—supporting
the plausibility of the parallel trends assumption that will be explored more below. 19 After states
first made P-EBT payments, average state-level EBT spending per student eligible for free or
reduced-price (FRP) meals promptly and significantly increased by $18 to $42 over each of the
subsequent six weeks. As the average P-EBT payment was $311 per student, these patterns
indicate that approximately 63% of benefits were spent within the six weeks after states made
payments and suggest that families may have saved some funds to draw upon over the following
19
Formal tests that the pre-period trend is equal to zero are frequently underpowered (Roth 2022).
20
weeks. This spending response is on the upper bound of the 37-66% total spending response among
low-income and low-asset households after the first Economic Impact Payment, which provided
unrestricted cash and was disbursed at the beginning of the P-EBT rollout period in mid-April
2020 (Cox et al. 2020, Meyer and Zhou 2020, Misra et al. 2021, Chetty et al. forthcoming, Cooper
and Olivei 2021, Baker et al. 2023). The spending responses are much larger than earlier stimulus
payments made during the 2001 and 2008 recessions for the full population (Johnson et al. 2006,
Parker et al. 2013), suggesting that P-EBT helped families meet substantial unmet need. More
generally, the significant increase in EBT spending after P-EBT was paid out supports a “first-
stage” of the program—that is, families spent the money after receipt.
5.2 Assessing Pre-Trends from Administrative Unemployment Data
Figure 2 provides some visual evidence in support of the parallel trends assumption by illustrating
that EBT redemptions were not differently trending prior to when families received P-EBT
payments. We further explore the parallel trends assumption by assessing whether food hardship
or related economic factors were trending prior to P-EBT payments in a manner that may confound
our ability to separately identify the effect of P-EBT. We start by showing event-study results from
other large administrative datasets that provide a balanced panel of state-week observations. As
with the STARS data, since the unemployment administrative data are reported each week, the
event studies with these data include a longer timeframe and more states than event study analyses
with the HPS. In the next section we also explore pre-trends in the more limited HPS sample that
includes our food hardship and health outcomes.
Figure 3 examines pre-trends in the unemployment rate and UI claims in the weeks prior
to P-EBT payments. If these measures of unemployment declined prior to (or coincident with)
21
P-EBT payments, it would signal an improving state economy that may itself contribute to declines
in food hardship. Instead, we do not observe any significant changes in unemployment or UI claims
in the weeks leading up to or following P-EBT receipt. Therefore, these patterns bolster confidence
in the parallel trends identifying assumption by suggesting that the timing of P-EBT was unrelated
to other concurrent changes in economic conditions .
Figure 3: Event study of administrative unemployment outcomes
a. Insured unemployment rate b. Continuing claims per student
Notes: Figure shows event studies on state-week unemployment data using the approach in Equation 2. Sample
includes a balanced panel of 16 states that disbursed P-EBT payments between May 30 and July 15, 2020. Vertical
blue (gray) bars denote 90 (95) percent confidence intervals with robust standard errors clustered by state.
5.3 Dynamics of food hardship and household well-being
Figure 4 presents event-study analyses of the HPS data, examining dynamics in food
hardship and other measures of household well-being in the 18 states that made payments between
May 15 and June 16 (for which we have a balanced panel of HPS observations two weeks prior to
P-EBT payments through four weeks after). 20 We restrict the analysis to respondents with children
who were enrolled in school during the 2019-20 school year and that had household incomes
20
Since the HPS waves are not exactly weekly waves and there are long gaps between the first and second and 12th
and 13th waves, the length of the event study period and the number of states used in these analyses is more limited
than the event studies with either the STARS or UI data.
22
approximately less than 130 percent of the poverty line in 2019 to proxy eligibility for SNAP
(sensitivity to this definition is tested in the appendix). 21
Figure 4: Household well-being event study
a. Food insufficiency in HH
b. Poor mental health index (mother) c. Poor or fair physical health (mother)
d. Very confident can afford food in next 4 e. Log(grocery spend)/child
weeks
Notes: Figure shows event studies on a balanced panel of 18 states that made P-EBT payments between May 15 and
June 16, 2020. Event time indicates the number of weeks between the P-EBT payment and the end of the Pulse
survey wave. Specifications include controls for respondent race/ethnicity, gender, educational attainment, and age,
21
Household income in the Pulse survey is reported in $10,000 to $50,000 increments, ranging from less than
$25,000 to $200,000. We measure the income-to-poverty ratio by taking the midpoint income in a respondent’s
reported income category divided by the poverty threshold for its household size and exclude households with a
ratio that is greater than 130 percent (similar to Han et al., 2020). Appendix Table A6 shows results using alternative
income thresholds.
23
as well as controls for household size, the number of children in the household, state and time-period fixed effects,
and state trends. Sample restricted to households with children and with income below 130 percent of the federal
poverty line. Regressions weighted by respondent sample weights. Vertical blue (gray) bars denote 90 (95) percent
confidence intervals with robust standard errors clustered by state.
As shown in panel a, in the three weeks after P-EBT payments were made, the share of
adults reporting food insufficiency in their household over the prior week dropped by 4.8 to
7.6 percentage points. Consistent with the parallel trends assumption, there were no differences in
food insufficiency prior to P-EBT payments.
Because of sample limitations, our balanced panel event study that includes data 4 weeks
after payment can only include 2 weeks before payment (see Appendix Table 1), limiting the
ability to investigate pre-trends. We can observe a longer, 4-week pre-period for the 16 states that
implemented P-EBT between May 27 and July 7 and thus have a balanced panel of observations
for 4 weeks prior through a full week after P-EBT disbursement. These results, shown in Appendix
Figure A3, show no substantial trends in family well-being in the weeks leading up to when
families received payments.
Panels b and c examine impacts on health outcomes. Although P-EBT is targeted to
children, for many reasons including the fungibility of money and the fact that food is generally
shared within the household, it is reasonable to investigate impacts on other members of the
household. Indeed, prior work documents within-family spillovers of child-specific nutrition
assistance (Bitler et al. 2023b; Bhattacharaya, Currie, and Haider 2006). Panel b shows that the
index of poor mental health among mothers declines after initial P-EBT payments. There is no
significant change in the share of mothers reporting poor or fair physical health (panel c)—an
outcome unlikely to be malleable in the short run, and likely best interpreted as a placebo test.
Although P-EBT provided families with a meaningfully large transfer (approximately $300
per student), we do not observe broader improvements in their perceived financial security as
24
measured by the respondent’s high confidence in their ability to afford the food they need over
next four weeks (panel d). Note that only 13% of the sample reports high confidence prior to
P-EBT payment. One potential explanation for the lack of future confidence, despite reported
reductions in food hardship, is that P-EBT was perceived as a temporary program during summer
2020 (it was not re-authorized for the 2020-21 school year until October 2020, and implementation
guidance for further payments was not finalized until January 2021) and most states issued a single
payment. Therefore, even though the amounts they had already received reduced families’ food
hardship, since they did not anticipate another payment, they were not optimistic about future
affordability.
As shown in panel e, impacts of P-EBT payments on log grocery spending per child are
imprecisely measured. On average, the pre-treatment spending mean was $103 per child, but as
described in the data section above there is substantial measurement error in this variable
(Appendix Figure A2). Using a similar research design with the Nielsen Homescan Data, Bonono
et al. (2024) find that log grocery spending per child in SNAP-eligible households with children
increases by 0.11 after P-EBT payments are made, and the increase persists over several weeks.
With the limitations of the spending variable in the HPS and larger estimates from better data, we
are reluctant to draw strong conclusions on grocery spending from the HPS.
We also implement a stacked event-study approach that is robust to settings where not all
states receive treatment at the same time (Roth et al. 2023), following Cengiz et al. (2019) and
Wing et al. (2024) with a balanced panel of 9 “treated” states that implemented P-EBT between
25
May 27 and June 16 (a shorter timeframe to also allow 4-weeks of post-event time). Again, we do
not find any notable trends in any measure of family well-being in the month prior to receipt. 22
5.4 Food hardship and household well-being: Difference-in-differences results
We leverage more of the cross-state variation in P-EBT timing using a difference-in-differences
approach, described in equation (1). The difference-in-differences results are more than a
parametric summary of the event-study results because, by relaxing the need to create a balanced
panel, we can include all 46 states that provided information on P-EBT payments, since all of these
states made payments after April 9, 2020 (i.e. within 2 weeks prior to the start of the first HPS
survey wave or later). Difference-in-differences results for the smaller event-study samples are
broadly similar and are presented in Appendix Table A3.
In the month after P-EBT payments were made, the share of adults reporting food
insufficiency in their household over the prior week declined by 8.3 percentage points, or 27% of
the pre-disbursement mean (column 1). HPS began collecting information on very low food
security among children in the week of June 4, 2020 (wave 7). Results for this outcome are
necessarily restricted to waves after this point; 27 states—including nearly all of the states included
in the event-study analyses above (Figure 4)—had already issued benefits by then. We find a larger
reduction in very low food security among children, which fell by 17.7 percentage points, or
approximately 49% (column 2).
22
Appendix Table A3 shows that food insufficiency and maternal mental health is qualitatively similar with the
smaller subsample of states used in each event study sample under the approach in Equation 1, albeit less precisely
estimated for the sample in Appendix Figure A3. The confidence intervals for physical health, food confidence, and
grocery spending also substantially overlap, although results are not significant in any of the samples.
26
Table 1. P-EBT impacts on household food hardship, well-being, and spending
(1) (2) (3) (4) (5) (6)
Very
Children Poor Poor or confident
Sometimes/often have mental fair can Log
not having very low health physical afford grocery
enough to eat in food index health food next spending
HH security (mothers) (mothers) 4 weeks per child
P-EBT disbursed within 2 weeks -0.0831** -0.1768* -0.0927* 0.0006 0.0028 0.0147
of start of survey wave (0.0399) (0.0895) (0.0533) (0.0414) (0.0278) (0.0426)
N 20306 7595 14510 14572 20320 19263
DV mean prior to disbursement 0.2749 0.3625 0.2602 0.2906 0.1326 4.6305
State linear time trends X X X X X X
Notes: Table shows effect of being in a state that disbursed the first P-EBT payment within 2 weeks of the beginning
of the survey wave. States are dropped from the analyses after treatment so that the control group in each period is
states that have not yet disbursed payments. All specifications include controls for respondent race/ethnicity, gender,
educational attainment, and age, as well as controls for household size; the number of children in the household, the
state unemployment claimant rate; the share of SNAP Emergency Allotment payments and SNAP benefits disbursed
in 2 weeks before the survey period; whether there is an active stay-at-home order, mask mandate, or indoor restaurant
closure at the start of the survey wave; share of UI claims processed within 1, 2, and more than 10 weeks of claiming
date; state and time-period fixed effects; and state linear time trends. Sample restricted to households with children
who attended school and with income below 130 percent of the federal poverty line. Regressions weighted by
respondent sample weights. Robust standard errors clustered by state. Data from Household Pulse Survey waves 1-13
(columns 1, 3-6) or waves 7-13 (column 2). * p < 0.1, ** p < 0.05, *** p < 0.01.
Columns (3) and (4) examine impacts on mothers’ health. The index of poor mental health among
mothers (column 3) declines by 0.09 standard deviation after initial P-EBT payments. Note that
this effect fades and is no longer statistically significant 3 weeks after payment, as shown in
Appendix Table A4. There is no change in the share of mothers reporting poor or fair physical
health (column 4). Consistent with the event-study results presented previously, we do not find
broader improvements in the respondent’s confidence in their ability to afford the food they need
over next four weeks (column 5). Recalling the substantial heaping in the self-reported grocery
spending, log grocery spending per child increases by a statistically insignificant 0.015 after initial
P-EBT payments are made (column 6).
27
Similar dynamics are shown in Appendix Table A4, which applies alternative reference
periods to the difference-in-differences approach for the entire sample. Here, we find that the
change in food hardship is most pronounced in the 0-4 weeks after disbursement (within 2 weeks
of the start of the survey wave) and is slightly attenuated when looking at either narrower or
broader windows. In contrast, the improvements in both physical and mental health are front-
loaded so that the effect is larger in the 0-3 weeks following payment (within 1 week of the start
of the survey wave).
Figures 2 and 3 provided visual evidence that P-EBT payments did not coincide with other
substantial changes in economic conditions or SNAP benefit redemptions and Appendix Table A2
showed that P-EBT timing was not fully explained by other state-level political, economic, or
demographic factors. To further examine whether the results in Table 1 may be reflecting broader
changes in food hardship over this period, we estimate the difference-in-difference approach on
two “placebo” samples that were ineligible to receive P-EBT. Table 2 reports results for families
without children (panel a) and families with income above 200% of the federal poverty guideline
(panel b). 23 For both groups, there is no economically or statistically significant improvement in
food hardship or health outcomes. These patterns bolster confidence that our results are capturing
the effect of P-EBT, rather than unobserved factors affecting all families at the same time as
disbursement.
23
Higher-income households with children were eligible to receive P-EBT if they attended a school offering
schoolwide free meals; however, due to application processes and the need to identify these students, higher-income
students typically received P-EBT 4-6 weeks after SNAP families.
28
Table 2. Placebo Test: P-EBT impacts on household food hardship, well-being, and
spending among ineligible populations
(1) (2) (3) (4) (5) (6)
Poor Poor or
mental fair
health physical
Sometimes/ Children index health Very
often not have (mothers/ (mothers/ confident Log
having very low women women can afford grocery
enough to food ages ages food next spending
eat in HH security 18-54) 18-54) 4 weeks per child
Panel a: Families without children
P-EBT disbursed within 2 weeks -0.0102 -0.0209 -0.0242 0.0007 -0.0378**
of start of survey wave (0.0069) (0.0459) (0.0231) (0.0099) (0.0175)
N 315724 65804 66055 296155 279159
DV mean prior to disbursement 0.0844 0.2970 0.1813 0.5095 4.0731
Panel b: Families with children, income > 200% poverty line
P-EBT disbursed within 2 weeks 0.0100 0.0203 0.0007 -0.0012 0.0123 0.0030
of start of survey wave (0.0104) (0.0227) (0.0506) (0.0188) (0.0218) (0.0314)
N 89761 33673 48610 48791 89771 85959
DV mean prior to disbursement 0.0518 0.0753 -0.0407 0.0862 0.5302 4.8617
State linear time trends X X X X X X
Notes: Table shows effect of being in a state that disbursed the first P-EBT payment within 2 weeks of the beginning
of the survey wave. States are dropped from the analyses after treatment so that the control group in each period is
states that have not yet disbursed payments. All specifications include controls for respondent race/ethnicity, gender,
educational attainment, and age, as well as controls for household size; the number of children in the household, the
state unemployment claimant rate; the share of SNAP Emergency Allotment payments and SNAP benefits disbursed
in 2 weeks before the survey period; whether there is an active stay-at-home order, mask mandate, or indoor restaurant
closure at the start of the survey wave; share of UI claims processed within 1, 2, and more than 10 weeks of claiming
date; state and time-period fixed effects; and state linear time trends. Sample restricted to households without children
(panel a) and households with children who attended school and with income above 200 percent of the federal poverty
line (panel b). Regressions weighted by respondent sample weights. Robust standard errors clustered by state. Data
from Household Pulse Survey waves 1-13 (columns 1, 3-6) or waves 7-13 (column 2). * p < 0.1, ** p < 0.05, *** p <
0.01.
5.5 Robustness
Our results have focused on families likely to be eligible for SNAP based on their reported income
bin relative to the poverty threshold because we can better identify P-EBT payment timing for this
29
sample. In Appendix Table A6 columns 1-4 we probe the robustness of our food insufficiency
results across alternative “low-income household” definitions: requiring the maximum (instead of
median) income in the bins to be below 130% of the poverty threshold (column 1), expanding the
sample to include households that are income-eligible for reduced-price meals (185% of the
poverty threshold, column 2), using a fixed income threshold (column 3), or a proxy (educational
attainment) for low-income (column 4). Results are similar, albeit less precise, when using the
smaller sample obtained from implementing a more restrictive income threshold (column 1). For
other low-income samples (columns 2-4), the estimated reduction in food insufficiency is
somewhat smaller, but the reductions in very low food security among children and improvements
in maternal mental health are generally similar.
Appendix Table A7 shows results excluding state trends, under which we observe similar
reductions food insufficiency in the household (7.1 percentage points, 26% of the pre-payment
mean) and slightly attenuated, but more precise, reductions in children having very low food
security of 29% (10.4 percentage points).
6. Discussion
An unusual feature of P-EBT is that it was provided as a sizeable, one-time, lump-sum
payment, in contrast to other forms of nutrition assistance (SNAP, WIC, summer benefits) that are
provided as recurrent, monthly benefits. Although the generalizability of our findings may be
limited due to their setting within the COVID-19 pandemic, this novel policy experiment can give
us insights into the dynamics of food hardship as well as how spending responds to resource
shocks.
30
We find that P-EBT benefits, which averaged $311 per child, reduce food insufficiency
among households eligible for SNAP by 27 percent. This magnitude is broadly consistent with
other findings for benefits paid to low-income families with children. For example, experimental
evidence from Summer-EBT pilot programs conducted in 2011-13 found that $230 per child in
grocery vouchers (in 2020 dollars) reduced very low food security—the concept closest to food
insufficiency as measured in the HPS—by 31 percent and food insecurity by 16 percent (Klerman
et al. 2017). Further, towards the end of the COVID-19 pandemic in July-December 2021, the fully
refundable Child Tax Credit (CTC) provided $250 per month to families for each child age 17 or
younger ($300 for each child younger than 6), a similar amount to the grocery vouchers provided
through P-EBT. Parolin et al. (2021) find that these CTC payments reduced food insufficiency
among low-income families with children by 27 percent relative to those without children. 24
Another consideration is how P-EBT compares with SNAP spending and how a new
program interacted with existing nutrition assistance programs. There is a strong pattern in SNAP
spending over the “benefit month” in which spending peaks shortly after benefits are issued each
month (Franckle et al. 2019, Hastings and Washington 2010, Goldin et al. 2022, Kuhn 2021, Todd
2015, Todd and Gregory 2018). In contrast, the spending response to P-EBT was both more modest
and sustained, as shown in Figure 2. We directly compare benefit redemption patterns across these
two programs in Appendix Table A5. While SNAP recipients receive payments once per month,
most states stagger benefit payment dates across their caseloads over the month making it
impossible for us to estimate general SNAP spending patterns in the state-week STARS data.
However, a few low-population states (New Hampshire, North Dakota, Rhode Island, South
24
Using a non-experimental approach, Gundersen, Kreider and Pepper (2018) estimate that an additional recurring
$183 per month for SNAP families with children would reduce food insecurity by 57 percent.
31
Dakota, and Vermont) issue all SNAP benefits within a single-week period, allowing us to identify
the timing of SNAP payments.
As shown in column (1), relative to the week when benefits are issued, SNAP spending in
the 2015-19 period sharply increases in the week following disbursement, followed by a rapid
decrease in subsequent weeks. Columns 2-3 show the equivalent “cycle” of EBT spending in the
month after P-EBT receipt, indicating a more muted pattern. Although these estimates are
imprecisely estimated due to small sample sizes, point estimates echo the pattern in Figure 2 and
do not show a decrease for at least 6 weeks. This comparison shows that lump-sum P-EBT
payments are spent differently than monthly recurrent SNAP payments. Finally, to look at
interactions between the two programs, column 4 replicates column 1 for the SNAP month in the
first SNAP disbursement cycle following P-EBT payments. In this month, families spent about
$14 more per child in the week they received SNAP benefits, spending remained elevated over the
following three weeks, and the drop-off in spending was less pronounced than prior to receiving
P-EBT. Altogether, these findings confirm prior work that additional assistance leads to both
increased spending and smoother spending patterns over the benefit month (Todd and Gregory
2018).
Although we find reductions improvements in food hardship and increases in benefit
redemption, households do not report spending significantly more in the HPS data. With big
caveats in mind, taking the imprecisely estimated coefficient in Table 1 at face value implies a
$1.50 per child per week increase in grocery spending, substantially smaller than the average $29
per student increase in EBT spending in the first 4 weeks after disbursement (Figure 2). There are
several potential explanations for the differences we find for P-EBT benefit redemptions and
grocery spending. First, as discussed in Section 3.2, we view the HPS grocery spending results
32
with substantial skepticism due to measurement error. For example, related work that uses more
accurate retail scanner data finds log grocery spending increases by more than 0.10 (Bonomo et
al. 2024).
Another hypothesis for the divergence between benefit redemptions and reported grocery
spending is that the spending response was much larger among higher-income households that
received P-EBT because their children attended a school that offered schoolwide free meals (who
are excluded from this analysis). However, Bonomo et al. (2024) find that higher-income families
in areas with schoolwide free meals increased their log grocery spending per student by 0.05, a
coefficient smaller than but not statistically significantly different from the spending impact among
SNAP-eligible households. Since most students receiving a free meal are income-eligible for
SNAP, these results suggest that responses among higher-income households cannot reconcile the
differences between the redemption and self-reported spending data.
Finally, another potential explanation is that most P-EBT spending could have gone to
finance families’ existing grocery budgets so of the $29 redemptions, only $1.50 financed
additional purchases and $27.50 went to usual spending. This implied marginal propensity to
consume (0.04) is smaller than estimates from Hastings and Shapiro (2018) who examine changes
in spending at a single grocery chain after the onset or termination of SNAP receipt (0.50) but is
within the range of the (imprecise) estimates from Hoynes and Schanzenbach (2009) who examine
consumption responses to the introduction of SNAP.
7. Conclusion
We find that lump-sum, in-kind assistance through grocery vouchers can protect families against
food hardship and improve maternal mental health. Our analysis leverages weekly data on
33
households along with the exact date of the introduction of the P-EBT program during the first
summer of the pandemic and shows that the program reduced food hardship among low-income
families across the month after they received benefits. We find that food insufficiency declines by
approximately 30 percent after benefit receipt—a magnitude that is consistent with results from
studies of other benefit programs aimed at low-income families with children including the
expanded CTC payments made during COVID-19 and experimental evidence from Summer EBT
demonstration projects. We further find that families appear to have smoothed their consumption
by spending down their P-EBT by $18-42 each week for at least six weeks after disbursement. The
reductions in food hardship that we observe are especially pronounced given the large increase in
economic vulnerability that many low-income families experienced during the pandemic (Bitler
et al. 2020a). These findings indicate that providing additional nutrition assistance during times of
crisis can alleviate hardship.
34
References
Aladangady, Aditya, Shifrah Aron-Dine, David Cashin, Wendy Dunn, Laura Feiveson, Paul
Lengermann, Katherine Richard, and Claudia Sahm. "Spending Responses to High-
frequency Shifts in Payment Timing: Evidence from the Earned Income Tax
Credit." American Economic Journal: Economic Policy 15, no. 3 (2023): 89-114.
Baker, Scott R., Robert A. Farrokhnia, Steffen Meyer, Michaela Pagel, and Constantine
Yannelis. "Income, Liquidity, and the Consumption Response to the 2020 Economic
Stimulus Payments." Review of Finance 27, no. 6 (2023): 2271-2304.
Bauer, Lauren, Abigail Pitts, Krista Ruffini, and Diane Whitmore Schanzenbach. "The Effect of
Pandemic EBT on Measures of Food Hardship." The Hamilton Project, Brookings
Institution, Washington, DC (2020).
Bauer, Lauren, Krista Ruffini, Diane Whitmore Schanzenbach. “The Case for and Challenges of
Delivering In-Kind Nutrition Assistance to Children.” The Hamilton Project, Brookings
Institution, Washington, DC (2023).
Bhattacharya, Jayanta, Janet Currie, and Steven J. Haider. "Breakfast of Champions?: The
School Breakfast Program and the Nutrition of Children and Families." Journal of Human
Resources 41, no. 3 (2006): 445-466.
Billings, Kara, and Jameson Carter. "Serving Free School Meals through the Community
Eligibility Provision (CEP): Background and Participation." Congressional Research
Service 20 (2020): 529-550.
Bitler, Marianne P., and Hilary W. Hoynes. “The More Things Change, the More They Stay the
Same? The Safety Net and Poverty in the Great Recession.” Journal of Labor Economics
34, no. S1 (2016): S403–S444.
Bitler, Marianne P., Hilary W. Hoynes, and Diane Whitmore Schanzenbach. "Suffering, the
Safety Net, and Disparities During COVID-19." RSF: The Russell Sage Foundation Journal
of the Social Sciences 9, no. 3 (2023a): 32-59.
Bitler, Marianne, Hilary Hoynes, and Diane Whitmore Schanzenbach. “Why the safety net might
not respond as effectively to COVID-19 as it should.” Milbank Quarterly Opinion 30
(2020b).
Bitler, Marianne P., Hilary W. Hoynes, and John Iselin. “Cyclicality of the US safety net:
Evidence from the 2000s and implications for the COVID-19 crisis.” National Tax
Journal 73, no. 3 (2020c): 759-780.
Bitler, Marianne, Janet Currie, Hilary Hoynes, Krista Ruffini, Lisa Schulkind, and Barton
Willage. "Mothers as insurance: Family spillovers in WIC." Journal of Health
Economics 91 (2023b): 102784.
35
Bonomo, Therese, Krista Ruffini, and Diane Whitmore Schanzenbach. “Structuring
Consumption Subsidies: The Impact of Recurrent and One-shot Food Vouchers and Cash
Transfers on Grocery Purchases.” Working paper, (2024).
Case, Anne, Angela Fertig, and Christina Paxson. "The Lasting Impact of Childhood Health and
Circumstance." Journal of Health Economics 24, no. 2 (2005): 365-389.
Cengiz, Doruk, Arindrajit Dube, Attila Lindner, and Ben Zipperer. “The Effect of Minimum
Wages on Low-wage Jobs.” Quarterly Journal of Economics 134, no. 3 (2019): 1405–54.
Chetty, Raj, John N. Friedman, and Michael Stepner. “The Economic Impacts of COVID-19:
Evidence from a New Public Database Built Using Private Sector Data.” The Quarterly
Journal of Economics, forthcoming.
Cooper, Daniel, and Giovanni Olivei. "High-frequency Spending Responses to Government
Transfer Payments." Federal Reserve Bank of Boston, Working Papers 21-10 (2021).
Cotti, Chad D., John M. Gordanier, and Orgul D. Ozturk. "Hunger pains? SNAP Timing and
Emergency Room Visits." Journal of health economics 71 (2020): 102313.
Cox, Natalie, Peter Ganong, Pascal Noel, Joseph Vavra, Arlene Wong, Diana Farrell, Fiona
Greig, and Erica Deadman. "Initial Impacts of the Pandemic on Consumer Behavior:
Evidence from Linked Income, Spending, and Savings Data." Brookings Papers on
Economic Activity 2020, no. 2 (2020): 35-82.
Evans, William N., and Craig L. Garthwaite. "Giving mom a break: The impact of higher EITC
payments on maternal health." American Economic Journal: Economic Policy 6, no. 2
(2014): 258-290.
Federal Register. “National School Lunch Program and School Breakfast Program: Eliminating
Applications Through Community Eligibility as Required by the Healthy, Hunger-Free Kids
Act of 2010.” 2016. Available at
https://www.federalregister.gov/documents/2016/07/29/2016-17232/national-school-lunch-
program-and-school-breakfast-program-eliminating-applications-through.
Fox, Liana, and Kalee Burns. “The Supplemental Poverty Measure: 2020.” U.S. Census Bureau
(2021). Available at https://www.census.gov/library/publications/2021/demo/p60-275.html.
Franckle, Rebecca L., Michele Polacsek, Sara N. Bleich, Anne N. Thorndike, Mary TG Findling,
Alyssa J. Moran, and Eric B. Rimm. "Support for supplemental nutrition assistance program
(SNAP) policy alternatives among US adults, 2018." American Journal of Public
Health 109, no. 7 (2019): 993-995.
36
Ganong, Peter, and Jeffrey B. Liebman. “The Decline, Rebound, and Further Rise in SNAP
Enrollment: Disentangling Business Cycle Fluctuations and Policy Changes.” American
Economic Journal: Economic Policy 10, no. 4 (2018): 153-176.
Goldin, Jacob, Tatiana Homonoff, and Katherine Meckel. "Issuance and incidence: SNAP
benefit cycles and grocery prices." American Economic Journal: Economic Policy 14, no. 1
(2022): 152-178.
Goodman-Bacon, Andrew. "Difference-in-differences with variation in treatment
timing." Journal of Econometrics 225, no. 2 (2021): 254-277.
Gundersen, Craig, Brent Kreider, and John V. Pepper. “Reconstructing the Supplemental
Nutrition Assistance Program to More Effectively Alleviate Food Insecurity in the United
States.” RSF: The Russell Sage Foundation Journal of the Social Sciences 4, no. 2 (2018):
113-130.
Gupta, Poonam, Elaine Waxman, Eleanor Pratt, Matt Lyons, and Chloe Green. "Documenting
Pandemic EBT for the 2020-21 School Year: State Perspectives on Implementation
Challenges and Lessons for the Future. Research Report." Urban Institute (2021).
Hamrick, Karen S., and Margaret Andrews. "SNAP participants’ eating patterns over the benefit
month: a time use perspective." PloS one 11, no. 7 (2016): e0158422.
Han, Jeehoon, Bruce D. Meyer, and James X. Sullivan. “Income and Poverty in the COVID-19
Pandemic.” NBER Working Paper, no. 27729 (2020).
Hastings, Justine, and Jesse M. Shapiro. "How are SNAP benefits spent? Evidence from a retail
panel." American Economic Review 108, no. 12 (2018): 3493-3540.
Hastings, Justine, and Ebonya Washington. "The first of the month effect: consumer behavior
and store responses." American Economic Journal: Economic Policy 2, no. 2 (2010): 142-
162.
Hembre, Erik, Katherine McElroy, and Shogher Ohannessian. “Supplemental Nutrition
Assistance Program and Food Expenditures: Evaluating California's Cash‐Out
Policy.” Contemporary Economic Policy 42, no. 3 (2024): 544-573.
Howard, Larry L. "Does food insecurity at home affect non-cognitive performance at school? A
longitudinal analysis of elementary student classroom behavior." Economics of Education
Review 30, no. 1 (2011): 157-176.
Hoynes, Hilary W., and Diane Whitmore Schanzenbach. “Consumption Responses to In-Kind
Transfers: Evidence from the Introduction of the Food Stamp Program.” American
Economic Journal: Applied Economics 1, no. 4 (2009): 109-139.
37
Hoynes, Hilary W., and Diane Whitmore Schanzenbach. "Safety Net Investments in Children."
Brookings Papers on Economic Activity 2018, no. 1 (2018): 89-150.
Hoynes, Hilary W., and Diane Whitmore Schanzenbach. "Strengthening SNAP as an automatic
stabilizer." The Hamilton Project (2019).
Hoynes, Hilary W., Leslie McGranahan, and Diane W. Schanzenbach. “SNAP and Food
Consumption.” In SNAP Matters: How Food Stamps Affect Health and Well-Being, edited
by Judith Bartfeld, Craig Gundersen, Timothy Smeeding, and James P. Ziliak, Redwood
City: Stanford University Press, 2015.
Johnson, David S., Jonathan A. Parker, and Nicholas S. Souleles. "Household expenditure and
the income tax rebates of 2001." American Economic Review 96, no. 5 (2006): 1589-1610.
Jyoti, Diana F., Edward A. Frongillo, and Sonya J. Jones. "Food Insecurity Affects School
Children’s Academic Performance, Weight Gain, and Social Skills–." The Journal of
Nutrition 135, no. 12 (2005): 2831-2839.
Keene, Danya E., Whitney Denary, Annie Harper, Anna Kapolka, Emily A. Benfer, and Peter
Hepburn. “A Little Bit of a Security Blanket”: Renter Experiences with COVID-19–Era
Eviction Moratoriums.” Social Service Review 97, no. 3 (2023): 423-455.
King, Michael D. and Katherine G. Giefer. “Nearly a Third of Children Who Receive SNAP
Participate in Two or More Additional Programs.” US Census Bureau: America Counts:
Stories (2021). Available at https://www.census.gov/library/stories/2021/06/most-children-
receiving-snap-get-at-least-one-other-social-safety-net-benefit.html.
Klerman, Jacob A., Anne Wolf, Ann Collins, Stephen Bell, and Ronette Briefel. “The Effects the
Summer Electronic Benefits Transfer for Children Demonstration has on Children’s Food
Security.” Applied Economic Perspectives and Policy, 39 (2017): 516-532.
Koné Consulting, “Pandemic EBT Implementation Documentation Project,” October 7, 2020.
Available at https://www.cbpp.org/sites/default/files/atoms/files/10-7-20fa-kone.pdf.
Kuhn, Michael A. "Electronic benefit transfer and food expenditure cycles." Journal of Policy
Analysis and Management 40, no. 3 (2021): 744-773.
Meyer, Michelle, and Anna Zhou. "COVID-19 and the consumer: data through April 16." Bank
of America Global Research Data Analytics 22 (2020).
Misra, Kanishka, Vishal Singh, and Qianyun Poppy Zhang. "Impact of Stay-at-home-orders and
Cost-of-living on Stimulus Response: Evidence from the CARES Act." NYU Stern School of
Business Working Paper. (2021).
38
Parker, Jonathan A., Nicholas S. Souleles, David S. Johnson, and Robert McClelland.
"Consumer spending and the economic stimulus payments of 2008." American Economic
Review 103, no. 6 (2013): 2530-2553.
Parolin, Zachary, Megan Curran, Jordan Matsudaira, Jane Waldfogel, and Christopher Wimer.
"Estimating monthly poverty rates in the United States." Journal of Policy Analysis and
Management 41, no. 4 (2022): 1177-1203.
Parolin, Zachary, Elizabeth Ananat, S. M. Collyer, Megan Curran, and Christopher Wimer. “The
Initial Effects of the Expanded Child Tax Credit on Material Hardship.” NBER Working
Paper, no. 29285 (2021).
Roth, Jonathan. “Pretest with caution: Event-study estimates after testing for parallel
trends” American Economic Review: Insights 4, no. 3 (2022): 305-322.
Roth, Jonathan, Pedro HC Sant’Anna, Alyssa Bilinski, and John Poe. “What’s trending in
difference-in-differences? A synthesis of the recent econometrics literature.” Journal of
Econometrics 235, no. 2 (2023): 2218-2244.
Ruffini, Krista, and Abigail Wozniak. “Supporting Workers and Families in the Pandemic
Recession.” Brookings Papers on Economic Activity (2021): 111-139.
Seligman, Hilary K., Ann F. Bolger, David Guzman, Andrea López, and Kirsten Bibbins-
Domingo. "Exhaustion of food budgets at month’s end and hospital admissions for
hypoglycemia." Health Affairs 33, no. 1 (2014): 116-123.
Shapiro, Jesse M. "Is there a daily discount rate? Evidence from the food stamp nutrition
cycle." Journal of Public Economics 89, no. 2-3 (2005): 303-325.
Shrider, Emily A., Melissa Kollar, Frances Chen, and Jessica Semega. “Income and Poverty in
the United States: 2020.” U.S. Census Bureau (2021). Available at
https://www.census.gov/library/publications/2021/demo/p60-273.html.
Sun, Liyang, and Sarah Abraham. "Estimating dynamic treatment effects in event studies with
heterogeneous treatment effects." Journal of Econometrics 225, no. 2 (2021): 175-199.
U.S. Census Bureau. “Household Pulse Survey.” 2024. Available at
https://www.census.gov/programs-surveys/household-pulse-survey/data.html.
U.S. Department of Agriculture. (USDA) “Child Nutrition Tables.” (2024). Available at
https://www.fns.usda.gov/pd/child-nutrition-tables.
—. “Pandemic EBT (P-EBT) Questions and Answers.” (2020). Available at https://fns-
prod.azureedge.us/sites/default/files/resource-files/SNAP-COVID-PEBTQA.pdf.
39
—. “SNAP FY2023 Cost-of-Living Adjustments.” 2022. Available at
https://www.fns.usda.gov/snap/fy-2023-cola.
U.S. Department of Education. “Table 203.20: Enrollment in Public Elementary and Secondary
Schools, by Region, State, and Jurisdiction: Selected Years, Fall 1990 through Fall 2029.”
2023. Available at https://nces.ed.gov/programs/digest/d20/tables/dt20_203.20.asp.
U.S. Internal Revenue Service. “SOI Tax Stats - Coronavirus Aid, Relief and Economic Security
Act (CARES Act) Statistics.” 2024. Available at https://www.irs.gov/statistics/soi-tax-stats-
coronavirus-aid-relief-and-economic-security-act-cares-act-statistics.
Todd, Jessica E. "Revisiting the Supplemental Nutrition Assistance Program cycle of food
intake: Investigating heterogeneity, diet quality, and a large boost in benefit
amounts." Applied Economic Perspectives and Policy 37, no. 3 (2015): 437-458.
Todd, Jessica E., and Christian Gregory. "Changes in Supplemental Nutrition Assistance
Program real benefits and daily caloric intake among adults." Food Policy 79 (2018): 111-
120.
Wilde, Parke E., and Christine K. Ranney. "The monthly food stamp cycle: shopping frequency
and food intake decisions in an endogenous switching regression framework." American
Journal of Agricultural Economics 82, no. 1 (2000): 200-213.
Wing, Coady, Seth M. Freedman, and Alex Hollingsworth. “Stacked Difference-in-Differences.”
NBER Working Paper, no. 32054 (2024).
40
The Effects of Lump-Sum Food Benefits during the COVID-19 Pandemic
on Spending, Hardship, and Health
Appendix
Lauren Bauer
Krista Ruffini
Diane Whitmore Schanzenbach
41
Appendix Figures
Appendix Figure A1: Analysis period and federal safety net spending, 2020-2021
Notes: Figure shows the daily outlays from the Treasury Department on CTC, SNAP and P-EBT (combined), UI,
and EIP programs. Shaded gray area denotes analysis period.
42
Appendix Figure A2: Distribution of reported weekly grocery spending
Notes: Figure displays the reported dollar amount spent on food in the past month among the analysis sample
(households with children who attended school and with income below 130 percent of the federal poverty line) in
HPS waves 1-13.
43
Appendix Figure A3: Household well-being event study, extended pre-period
a. Food insufficiency in HH
b. Poor mental health index (mother) c. Poor or fair physical health (mother)
d. Very confident can afford food in next 4 e. Log(grocery spend)/child
weeks
Notes: Figure shows event studies on a balanced panel of 16 states that made P-EBT payments between May 27 and
July 7, 2020. Event time indicates the number of weeks between the P-EBT payment and the end of the Pulse survey
wave. Specifications include controls for respondent race/ethnicity, gender, educational attainment, and age, as well
as controls for household size, the number of children in the household, state and time-period fixed effects, and state
trends. Sample restricted to households with children and with income below 130 percent of the federal poverty line.
Regressions weighted by respondent sample weights. Vertical blue (gray) bars denote 90 (95) percent confidence
intervals with robust standard errors clustered by state.
44
Appendix Figure A4: Household well-being event study, stacked design
a. Food insufficiency in HH
b. Poor mental health index (mother) c. Poor or fair physical health (mother)
d. Very confident can afford food in next 4 e. Log(grocery spend)/child
weeks
Notes: Figure shows event studies on a balanced panel of 9 states that made P-EBT payments between May 27,
2020 and June 16, 2020, following the stacked event study approach in Cengiz et al. (2019) and Wing et al. (2024).
Event time indicates the number of weeks between the P-EBT payment and the end of the Pulse survey wave.
Specifications include controls for respondent race/ethnicity, gender, educational attainment, and age, as well as
controls for household size, the number of children in the household, state and time-period fixed effects, and state
trends. Sample restricted to households with children and with income below 130 percent of the federal poverty line.
Control group consists of states that will make payments at least 5 weeks after each “cohort.” Regressions weighted
by respondent sample weights, adjusted for cohort size as in Wing et al. (2024). Shaded area denotes 95 percent
confidence intervals with robust standard errors clustered by stateXstack.
45
Appendix Tables
Appendix Table A1: Reported First Pandemic EBT Disbursement, by Source
Used in event study analyses
Original FOIA Balanced Balanced Stacked
correspondence request -4 to +1 -2 to + 4 -4 to + 4
(Bauer et al. (Bauer et Reported In DD (5/27/2020- (5/15/2020- (5/27/2020-
State 2020) al. 2024) to USDA analysis 7/7/2020) 6/16/2020) 6/16/2020)
Alaska 7/7/20 10/20/21 8/20
Alabama 5/3/20 5/15/20 5/20 X X
Arkansas 6/30/20 n/a 6/20 X X
Arizona 4/23/20 4/23/20 4/20 X
California 5/11/20 n/a 5/20 X
Colorado 7/15/20 5/25/21 7/20 X
Connecticut 5/24/20 5/24/20 5/20 X X
District of Columbia 5/22/20 5/1/20 5/20 X
Delaware 5/11/20 n/a 5/20 X
Florida 6/30/20 6/25/20 6/20 X X
Georgia N/A 7/18/20 7/20 X
Hawaii 6/24/20 6/29/20 6/20 X X
Iowa 7/15/20 5/11/21 7/20 X
Idaho N/A 6/22/21 8/20 X[1]
Illinois 4/20/20 4/19/20 4/20 X
Indiana 5/20/20 n/a 3/20
Kansas 5/21/20 n/a 5/20 X X
Kentucky 5/22/20 5/20/20 5/20 X X
Louisiana 5/15/20 5/27/20 5/20 X X X X
Massachusetts 4/30/20 4/30/20 4/20 X
Maryland 6/4/20 n/a 6/20 X X X X
Maine 5/5/20 5/8/20 5/20 X
Michigan 5/1/20 4/28/20 3/20 X
Minnesota 5/30/20 5/30/20 5/20 X X X X
Missouri 5/19/20 5/20/20 5/20 X X
Mississippi 6/16/20 10/2/21 6/20 X X X X
Montana N/A 7/15/20 7/20 X
North Carolina 5/12/20 5/11/20 5/20 X
North Dakota 5/9/20 5/8/20 5/20 X
Nebraska 7/28/20 4/20/20 7/20 X
New Hampshire 6/1/20 9/1/20 6/20 X X X X
New Jersey 6/16/20 7/7/20 7/20 X X
New Mexico 5/18/20 n/a 5/20 X X
Nevada N/A 8/29/20 8/20 X
46
New York 5/1/20 n/a 5/20 X
Ohio 5/28/20 5/28/20 5/20 X X X X
Oklahoma 7/9/20 8/21/20 8/20 X
Oregon 6/1/20 n/a 6/20 X X X X
Pennsylvania 5/26/20 5/28/20 5/20 X X X X
Rhode Island 4/20/20 4/19/20 3/20 X
South Carolina 7/7/20 6/7/20 7/20 X X
South Dakota 7/1/20 6/28/20 6/20 X X
Tennessee 6/12/20 n/a 6/20 X X X X
Texas 5/22/20 6/3/21 5/20 X X
Utah N/A 3/1/21 8/20
Virginia 5/13/20 5/15/20 5/20 X X
Vermont 5/27/20 5/2/20 5/20 X
Washington 6/28/20 6/27/20 6/20 X X
Wisconsin 4/27/20 5/3/22 3/20
West Virginia 5/20/20 n/a 5/20 X X
Wyoming 6/8/20 5/10/21 3/20
Notes: Table shows the date that each state first issued Pandemic EBT payments, provided from correspondence with
state officials used in Bauer et al. 2020 (column 2); FOIA requests conducted September 2022 through 2024 (column
3); as reported to USDA (column 4). Michigan is included in the analysis as the reported dates are 3 days apart. Idaho
is included in the analysis despite the disagreement in disbursement date because all reported disbursement dates occur
after summer 2020. Columns 6-8 report the subset of states used in each event study analysis.
47
Appendix Table A2: Timing of First Pandemic EBT Disbursement and State
Characteristics
(1) (2)
Payment date HPS wave of payment
Log(population) -3.377 -0.382
(5.5850) (0.7306)
% age 5-17 1005.177** 131.549**
(473.751) (59.162)
% age 65+ 412.723 46.629
(593.256) (81.843)
Minimum wage -1.806 -0.107
(3.620) (0.471)
State EITC rate -43.174 -5.667
(36.977) (4.751)
Legislature Republican -18.198 -1.504
controlled (20.878) (2.601)
Governor Republican -1.810 -0.248
(9.804) (1.328)
Implemented Medicaid -0.896 -0.027
expansion (14.403) (1.857)
% students FRP 2.411 0.230
(2.006) (0.247)
Median income 0.001 0.000
(0.001) (0.000)
SNAP BBCE -6.577 -0.534
(14.494) (1.580)
Mask mandate -4.068 -0.183
Apr-Sep 2020 (13.902) (1.702)
Ever stay-at-home order -16.674 -2.274
(14.158) (1.866)
April 2020 % UI claims -59.256 -8.112
processed w/in 14 days (49.053) (6.172)
N 45 45
R^2 0.395 0.389
Notes: Table shows results from a regression where the dependent variable is the calendar date (column 1) or
Household Pulse Survey wave (column 2) a state made the first Pandemic EBT disbursement, as reported in Appendix
Table A1. Sample includes all states included in the analysis, with the exception of ID since ID’s disbursement date
is not consistently reported (see Appendix Table A1).
48
Appendix Table A3: P-EBT impacts on household food hardship, well-being, and spending,
event study samples
(1) (2) (3) (4) (5)
Very
Sometimes/often Poor or fair confident
not having Poor mental physical can afford Log grocery
enough to eat in health index health food next 4 spending
HH (mothers) (mothers) weeks per child
Panel a: Event study sample, balanced [-2,4]
P-EBT disbursed within 2 weeks -0.1481** -0.2256* -0.0499 -0.0572 0.1186
of start of survey wave (0.0681) (0.1081) (0.1045) (0.0506) (0.1043)
N 7416 5520 5543 7422 7006
DV mean prior to disbursement 0.2731 0.2704 0.2717 0.1409 4.5735
Panel b: Event study sample, balanced [-4,1]
P-EBT disbursed within 2 weeks -0.0674 -0.0985 -0.0984 0.0150 -0.0368
of start of survey wave (0.0645) (0.1249) (0.0911) (0.0457) (0.0915)
N 8978 6522 6555 8992 8510
DV mean prior to disbursement 0.2880 0.2712 0.2826 0.1371 4.6232
Panel c: Stacked DD sample, balanced [-4,4], Equation 1
P-EBT disbursed within 2 weeks -0.2336*** -0.0830 0.0780 0.0744 0.0988
of start of survey wave (0.0415) (0.1011) (0.0871) (0.0573) (0.0557)
N 5176 3813 3832 5181 4906
DV mean prior to disbursement 0.2729 0.1993 0.2405 0.2445 4.4313
Panel d: Stacked DD sample, balanced [-4,4], Cengiz et al. (2019) DD
P-EBT disbursed within 2 weeks -0.1120*** 0.0333 -0.0640 0.0593* -0.1893***
of start of survey wave (0.0328) (0.0858) (0.0475) (0.0323) (0.0418)
N 6339 4614 4635 6342 6024
DV mean prior to disbursement 0.2665 0.2007 0.2177 0.2528 4.4036
Notes: Table shows effect of being in a state that disbursed the first P-EBT payment within 2 weeks of the beginning
of the survey wave for the subsample of states used in each event study analysis: Figure 2 (panel a), Appendix Figure
A3 (panel b), and Appendix Figure A4 (panels c-d). Panels a-c exclude states after the treatment period so that the
control group in each period is states that have not yet disbursed payments. Panel d adopts the stacked DD approach
as in Cengiz et al. (2019) where control states are those that made payments five or more weeks after each treatment
“cohort”. All specifications include controls for respondent race/ethnicity, gender, educational attainment, and age, as
well as controls for household size; the number of children in the household, the state unemployment claimant rate;
the share of SNAP Emergency Allotment payments and SNAP benefits disbursed in 2 weeks before the survey period;
whether there is an active stay-at-home order, mask mandate, or indoor restaurant closure at the start of the survey
wave; share of UI claims processed within 1, 2, and more than 10 weeks of claiming date; state and time-period fixed
effects; and state linear time trends. Sample restricted to households with children who attended school and with
income below 130 percent of the federal poverty line. Regressions weighted by respondent sample weights (panels a-
c), adjusted for cohort size as in Wing et al. (2024). Robust standard errors clustered by state. Data from Household
Pulse Survey waves 1-13. * p < 0.1, ** p < 0.05, *** p < 0.01.
49
Appendix Table A4: Robustness to Alternative Timing Windows
(1) (2) (3) (4) (5) (6)
Sometimes/ Children Poor Poor or Very
often not have mental fair confident Log
having very low health physical can afford grocery
enough to food index health food next spending
eat in HH security (mothers) (mothers) 4 weeks per child
Panel a: P-EBT disbursement week before survey wave
P-EBT disbursed within 1 week -0.0694 0.0013 -0.1740*** -0.0453 -0.0285 0.0290
of start of survey wave (0.0434) (0.1083) (0.0521) (0.0439) (0.0226) (0.0605)
N 17688 6197 13779 12728 17702 16785
DV mean prior to disbursement 0.2749 0.3625 0.2647 0.2906 0.1326 4.6305
Panel b: P-EBT disbursement within 3 weeks before survey wave
P-EBT disbursed within 3 weeks -0.0838** -0.0669 -0.0270 0.0218 -0.0009 0.0535
of start of survey wave (0.0352) (0.0897) (0.0494) (0.0397) (0.0199) (0.0456)
N 12667 9316 17347 16463 22991 16394
DV mean prior to disbursement 0.2602 0.3625 0.2647 0.2906 0.1326 0.2602
Notes: Table shows effect of being in a state that disbursed the first EBT payment within 1 week (panel a) or 3 weeks
(panel b) of the beginning of the survey wave. States are dropped from the analyses after treatment so that the control
group in each period is states that have not yet disbursed payments. All specifications include controls for respondent
race/ethnicity, gender, educational attainment, and age, as well as controls for household size; the number of children
in the household, the state unemployment claimant rate; the share of SNAP Emergency Allotment payments and
SNAP benefits disbursed in 2 weeks before the survey period; whether there is an active stay-at-home order, mask
mandate, or indoor restaurant closure at the start of the survey wave; share of UI claims processed within 1, 2, and
more than 10 weeks of claiming date; state and time-period fixed effects; and state linear time trends. Sample restricted
to households with children and with income below 130 percent of the federal poverty line. Regressions weighted by
respondent sample weights. Robust standard errors clustered by state. Data from Household Pulse Survey waves 1-13
(columns 1, 3-6) or waves 7-13 (column 2). * p < 0.1, ** p < 0.05, *** p < 0.01.
50
Appendix Table A5: Timing of SNAP and Pandemic EBT Spending, Relative to Week of
Disbursement
(1) (2) (3) (4)
SNAP timing
month after
P-EBT
SNAP timing P-EBT timing P-EBT timing disbursement
Disbursed last week 11.5284** 4.7463 4.7463 13.8857**
(2.9291) (2.8384) (2.8702) (3.8538)
Disbursed 2 weeks -3.4671** 8.2964 8.2964 0.6945
ago (1.1472) (9.0963) (9.1983) (1.2234)
Disbursed 3 weeks -10.9085*** 1.5821 1.5821 -4.4551**
(1.5724) (4.2801) (4.3281) (1.4195)
Disbursed 4 weeks
ago 2.1917
(4.8273)
Disbursed 5 weeks
ago 1.0555
(6.5631)
Disbursed 6 weeks
ago 8.7669
(7.6631)
Constant 21.3575*** 31.1910*** 31.1910*** 34.0805***
(0.7697) (3.7684) (3.4528) (1.3309)
N 944 20 35 20
Jan 2015- April- April- April-
Years December 2019 August 2020 August 2020 August 2021
Notes: Table shows regression where the dependent variable is the amount of SNAP (column 1) or Pandemic EBT
(columns 2-3) spending per student at the state-week level relative to the week of disbursement. Robust standard
errors clustered by state. Data from USDA STARS. States included are ND, NH, RI, SD, and VT. * p < 0.1, ** p <
0.05, *** p < 0.01.
51
Appendix Table A6: Pandemic EBT impacts on household food hardship, well-being, and
spending, alternative income groups
(1) (2) (3) (4)
< 130% poverty
line < 185% poverty line Income < No more than
(conservative) (median) $50,000 HS education
Panel a: Food insufficiency in HH
P-EBT disbursed within 2 weeks -0.0769* -0.0532* -0.0527* -0.0255
of start of survey wave (0.0420) (0.0269) (0.0301) (0.0288)
N 14709 32485 33270 18594
DV mean prior to disbursement 0.2950 0.2441 0.2472 0.1977
Panel b: Children have very low food security
P-EBT disbursed within 2 weeks -0.1391 -0.1705** -0.1896** -0.1848***
of start of survey wave (0.0967) (0.0682) (0.0710) (0.0648)
N 5428 12317 12572 7017
DV mean prior to disbursement 0.3841 0.3071 0.3166 0.2461
Panel c: Poor mental health index (mothers/women ages 18-54)
P-EBT disbursed within 2 weeks -0.0163 -0.1089** -0.0702 -0.0753
of start of survey wave (0.0746) (0.0515) (0.0542) (0.0524)
N 10615 22320 23171 10976
DV mean prior to disbursement 0.2699 0.2276 0.2316 0.1752
Panel d: Poor or fair physical health (mothers/women ages 18-54)
P-EBT disbursed within 2 weeks 0.0097 0.0136 0.0119 0.0160
of start of survey wave (0.0468) (0.0356) (0.0352) (0.0310)
N 10659 22410 23266 11035
DV mean prior to disbursement 0.3010 0.2568 0.2638 0.2482
Panel e: Very confident can afford food next 4 weeks
P-EBT disbursed within 2 weeks 0.0152 -0.0125 0.0115 -0.0026
of start of survey wave (0.0236) (0.0265) (0.0265) (0.0250)
N 14719 32515 33300 18626
DV mean prior to disbursement 0.1238 0.1530 0.1451 0.2334
Panel e: Log grocery spending per child (per person, HH without children)
P-EBT disbursed within 2 weeks 0.0165 0.0051 -0.0211 -0.0153
of start of survey wave (0.0545) (0.0379) (0.0409) (0.0503)
N 14010 30950 31573 17499
DV mean prior to disbursement 4.6174 4.6277 4.6734 4.7452
Notes: Table shows effect of being in a state that disbursed the first EBT payment within 2 weeks of the beginning of
the survey wave. States are dropped from the analyses after treatment so that the control group in each period is states
that have not yet disbursed payments. All specifications include controls for respondent race/ethnicity, gender,
educational attainment, and age, as well as controls for household size; the number of children in the household, the
state unemployment claimant rate; the share of SNAP Emergency Allotment payments and SNAP benefits disbursed
in 2 weeks before the survey period; whether there is an active stay-at-home order, mask mandate, or indoor restaurant
closure at the start of the survey wave; share of UI claims processed within 1, 2, and more than 10 weeks of claiming
date; state and time-period fixed effects; and state linear time trends. Sample restricted to households with children
and with income in an income bin fully below 130% of the poverty line (column 1), an income bin with a midpoint
below 185% of the poverty line (column 2), below $50,000 (column 3), or respondent having no more than a high
school education (column 4). Regressions weighted by respondent sample weights. Robust standard errors clustered
by state. Data from Household Pulse Survey waves 1-13 (panels a, c-e) or waves 7-13 (panel b). * p < 0.1, ** p <
0.05, *** p < 0.01.
52
Appendix Table A7: Pandemic EBT impacts on household food hardship, well-being, and
spending, no state trends
(1) (2) (3) (4) (5) (6)
Very
confident
Poor Poor or can
Sometimes/often Children mental fair afford Log
not having have very health physical food grocery
enough to eat in low food index health next 4 spending
HH security (mothers) (mothers) weeks per child
P-EBT disbursed within 2 weeks -0.0711*** -0.1041** -0.0157 -0.0029 0.0206 0.0287
of start of survey wave (0.0177) (0.0443) (0.0308) (0.0309) (0.0171) (0.0390)
N 20306 7595 14510 14572 20320 19263
DV mean prior to disbursement 0.2749 0.3625 0.2602 0.2906 0.1326 4.6305
Notes: Table shows effect of being in a state that disbursed the first EBT payment within 2 weeks of the beginning of
the survey wave. States are dropped from the analyses after treatment so that the control group in each period is states
that have not yet disbursed payments. All specifications include controls for respondent race/ethnicity, gender,
educational attainment, and age, as well as controls for household size; the number of children in the household, the
state unemployment claimant rate; the share of SNAP Emergency Allotment payments and SNAP benefits disbursed
in 2 weeks before the survey period; whether there is an active stay-at-home order, mask mandate, or indoor restaurant
closure at the start of the survey wave; share of UI claims processed within 1, 2, and more than 10 weeks of claiming
date; and state and time-period fixed effects. Sample restricted to households with children and with income below
130 percent of the federal poverty line. Regressions weighted by respondent sample weights. Robust standard errors
clustered by state. Data from Household Pulse Survey waves 1-13 (columns 1, 3-6) or waves 7-13 (column 2). * p <
0.1, ** p < 0.05, *** p < 0.01.
53
Appendix Table A8: Pandemic EBT impacts on household food hardship, well-being, and
spending, monthly disbursement dates reported to USDA
(1) (2) (3) (4) (5) (6)
Poor Poor or Very
Children mental fair confident Log
Food have very health physical can afford grocery
insufficiency low food index health food next 4 spending
in HH security (mothers) (mothers) weeks per child
Panel a: P-EBT disbursed current month, as reported to USDA
P-EBT current month -0.1370 0.0096 0.1051 -0.0880* 0.0681** -0.0036
(0.1011) (0.0457) (0.0655) (0.0440) (0.0268) (0.0527)
N 7514 22099 15835 15902 22112 20978
DV mean 0.3531 0.2701 0.2386 0.2838 0.1299 4.6226
Panel b: P-EBT disbursed current or last month, as reported to USDA
P-EBT current or last month -0.0717 -0.0151 0.0962*** -0.0244 0.0095 -0.0171
(0.0731) (0.0245) (0.0344) (0.0311) (0.0171) (0.0392)
N 16072 32151 22881 22976 32182 30564
DV mean 0.3531 0.2701 0.2386 0.2838 0.1299 4.6226
Notes: Table shows effect of being in a state that disbursed the first EBT payment in the current month (panel a) or
current or next month (panel b) of the beginning of the survey week, according to monthly disbursement information
reported to USDA. States are dropped from the analyses after treatment so that the control group in each period is
states that have not yet disbursed payments. All specifications include controls for respondent race/ethnicity, gender,
educational attainment, and age, as well as controls for household size; the number of children in the household, the
state unemployment claimant rate; whether there is an active stay-at-home order, mask mandate, or indoor restaurant
closure at the start of the survey wave; share of UI claims processed within 1, 2, and more than 10 weeks of claiming
date; state and time-period fixed effects; and state linear time trends. Sample restricted to households with children
and with income below 130 percent of the federal poverty line. Regressions weighted by respondent sample weights.
Robust standard errors clustered by state. Data from Household Pulse Survey waves 1-13 (columns 1, 3-6) or waves
7-13 (column 2). * p < 0.1, ** p < 0.05, *** p < 0.01.
54
File and source
- File
- nber-w33199.pdf
- Size
- 1,267,286 bytes
- SHA-256
- 4ed457d0970e4cdee83457cbe27e0bb9fd7859946c2cf552bde3be3af5a4692c
- Our copy
- nber-w33199.pdf
- Original
- www.nber.org