Sick, Hungry, and Vulnerable: Federal Stimulus and Food Security on Marginalized Populations
Summary
A journal article, "Sick, Hungry, and Vulnerable: Federal Stimulus and Food Security on Marginalized Populations During the COVID-19 Pandemic," by Molly Jacobs, Timothy R. McDade, Mateo Villamizar Chaparro and Michelle Corea, in the Journal of Racial and Ethnic Health Disparities (2023) 10:2685–2703, published online 15 November 2022. The abstract says the study uses the Household Pulse Survey and the 2017–2018 Current Population Survey Food Security Supplement to test whether the initial economic stimulus payment improved the nutritional well-being of vulnerable populations. It reports that marginalized groups faced greater food insecurity and that Blacks, Hispanics, and individuals living in rural areas became less food secure as the pandemic progressed. It adds that receipt of a stimulus payment appears to have improved conditions.
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
- Issuer
- Source documents
- Document type
- Sick Hungry And Vulnerable Federal Stimulus And Food Security On Marginalized Populations
- Case
- Sick Hungry And Vulnerable Federal Stimulus And Food Security On Marginalized Populations During The Covid 19 Pandemic
Cite this document
“Sick, Hungry, and Vulnerable: Federal Stimulus and Food Security on Marginalized Populations.”
Original at doi.org. Archived at Pandemic Darlings, https://pandemicdarlings.com/source-documents/sick-hungry-and-vulnerable-federal-stimulus-and-food-security-on-marginalized-populations-during-the-covid-19-pandemic/
, SHA-256 54d01416a83702225a86494517679cda9cc27218c9befb6539f23d138b59fe88.Full text
Journal of Racial and Ethnic Health Disparities (2023) 10:2685–2703
https://doi.org/10.1007/s40615-022-01447-8
Sick, Hungry, and Vulnerable: Federal Stimulus and Food Security
on Marginalized Populations During the COVID‑19 Pandemic
Molly Jacobs1 · Timothy R. McDade2 · Mateo Villamizar Chaparro2 · Michelle Corea3
Received: 16 August 2022 / Revised: 21 October 2022 / Accepted: 1 November 2022 / Published online: 15 November 2022
© W. Montague Cobb-NMA Health Institute 2022
Abstract
In addition to the threat of serious illness, COVID-19 and subsequent restrictions had devastating economic consequences
for many US citizens. This study examines the evolution of food security over the first months of the COVID-19 pandemic
testing whether the initial economic stimulus payment improved the nutritional well-being of vulnerable populations. We
use data from phase 1 of the Census Bureau’s Household Pulse Survey among a nationally representative sample of adults
and the 2017–2018 Current Population Survey Food Security Supplement. Using an ordered logistic regression, we assess
differences in the incidence and severity of food security across demographic, income, geographic, and employment status
cohorts and assess the effects of the first economic stimulus payment. Our results show that marginalized groups faced
greater food insecurity and had food-related outcomes worsen over time. Blacks, Hispanics, and individuals living in rural
areas became less food secure as the pandemic progressed. However, receipt of a stimulus payment appears to have improved
conditions. Rising food prices and persistent high unemployment have the potential to exacerbate food insecurity among
marginalized and at-risk groups.
Keywords COVID-19 · Stimulus · Inequality · Racial differences · Black · Rural · Food security
Introduction effects have been felt in high-income [2], middle-income,
and low-income nations [1, 3].
In their 2020 report, the Food and Agricultural Organization International organizations like the International Mone-
(FAO) estimated that nearly 8.9% of the world’s popula- tary Fund (IMF) suggested that countries use direct transfers
tion lives in hunger [1]. Moreover, the FAO also estimated to ease the economic burden of the pandemic [4]. Accord-
that the number of people affected by severe food insecu- ingly, the US government began the disbursement of eco-
rity is increasing, decreasing the likelihood of achieving the nomic stimulus checks to qualified households in April
zero-hunger sustainable development goal. The pandemic 2020. The dollar amount of this payment varied by the num-
caused by severe acute respiratory syndrome coronavirus 2 ber and status of the household members and was intended
(COVID-19) has created harsh economic conditions that left to help individuals bridge any income gap that may have
many individuals unemployed, increased food prices, and arisen due to the pandemic. This stimulus was an uncondi-
increased the probability of becoming food insecure. These tional cash transfer. An average household with two adults
and two children received $1200 per adult and an extra $500
per child. The timing of the receipt of the stimulus payment
varied depending on if the household was enrolled in direct
* Molly Jacobs
mollyjacobs@ufl.edu deposit with the Internal Revenue Service (IRS). However,
most recipients received their direct deposit on April 15th
1
Department of Health Services Research, Management and the two following weeks. For those who did not typically
and Policy, University of Florida, 3118 HPNP, 1225 Center file tax returns or opted to receive paper checks, mailing
Drive, Gainesville, FL 32603, USA
began in May and continued throughout the month.
2
Department of Political Science, Duke University, Durham, In addition to its immediate consequences, such as hunger
NC, USA
and psychological distress, food insecurity is also associ-
3
Department of Political Science, University of North ated with adverse long-term health outcomes. In particular,
Carolina, Chapel Hill, NC, USA
1 3
Vol.:(0123456789)
2686 Journal of Racial and Ethnic Health Disparities (2023) 10:2685–2703
studies over the last several decades have demonstrated that flows may be an effective tool in helping to combat food
food insecurity is associated with worse outcomes on health insecurity.
exams [5], being in poor health [6, 7], increased odds of Previous research has shown that COVID-related stimu-
being hospitalized [8], decreased nutrient intakes [9, 10], lus payments in developed democracies allowed individuals
and chronic diseases and conditions such as diabetes [11, to buy more food once received [31–35]. Thus, we expect
12], hypertension [5], obesity [13], and childhood asthma that the most vulnerable households will suffer from less
[14, 15]. Furthermore, several of these conditions impair food insecurity after the distribution of stimulus payments.
immune function and weaken the immune response, making However, this expectation might be mitigated by the pan-
it more difficult for individuals to fight infections [16–18]. demic’s differential across race and urbanization since this
This concern becomes even more critical in the wake of the has been highlighted by previous studies [22]. In fact, news
COVID-19 pandemic, as severe outcomes, including death, reports worldwide illustrate that more vulnerable groups
are already associated with these exact underlying health face higher levels of food insecurity. This paper analyzes the
conditions [19]. relationship between economic stimulus and food insecurity
We focus on the USA for a couple of reasons. Prior to in certain population subgroups.
the COVID-19 pandemic, food insecurity (measured as a This paper’s main objective is to analyze if the stimu-
lack of food consumption) among US citizens increased lus packages reduced the incidence of food insecurity in
after the 2008 financial crisis and remained relatively con- marginalized households during the COVID-19 pandemic.
stant through 2014 [20]. In 2015, food insecurity began to Given the rising food prices and the persistent high unem-
decrease, but the recovery was stalled by the onset of the ployment that accompanied the pandemic, it is plausible that
COVID pandemic in 2020. Rates of food insecurity in the the stimulus packages were unable to ameliorate the lack of
USA have historically been higher than the national aver- food consumption during the first months of the pandemic.
age for households with incomes near or below the federal We use nationally representative data from the USA and a
poverty line, households with children (particularly those series of regression analyses to assess the differences in the
headed by single adults), adults living alone, Black- and incidence and severity of food security across demographic,
Hispanic-headed households, and households in principal income, geographic, and employment status cohorts.
cities [21, 22].
As the number of COVID-19 cases in the USA rap-
idly rose in early 2020, communities already hard-hit by Materials and Methods
food insecurity became further strapped for resources, as
mandated business shutdowns, increasing unemployment, Data Sources and Variable Operationalization
and food supply chain shortages added to the challenges
in acquiring sufficient food [23–25]. Historically, job loss This paper’s empirical analysis centers on the US Census
coupled with the loss of childcare is associated with lower Bureau Household Pulse Survey (HPS): a quick deployment
food security. Low-income earners have been disproportion- cross-sectional data collection instrument designed to col-
ately hit by the COVID-19 pandemic, being more likely to lect information on a range of topics that have impacted US
lose their income and be food insecure. For example, Cen- households during the pandemic. The HPS asked household
sus Bureau surveys from the summer of 2020 show that heads about their experiences in terms of employment sta-
food insecurity among Black households with children has tus, spending patterns, food security, housing, physical and
increased from 25 to 39% since 2018, and food insecurity mental health, access to health care, and educational dis-
among Hispanic households with children has increased ruption. In weeks 1 and 2 of the survey, the Census Bureau
from 17 to 37% in the same period [26]. Moreover, school sampled nearly 1.9 million and 1.0 million housing units,
closures have removed children’s access to meals provided respectively. It added an additional 1.1 million addresses
for free or at reduced costs, which increases food pressure on each week thereafter to ensure that the survey remained rep-
larger families [27]. Furthermore, many of the communities resentative of the US population. In total, the survey sam-
already disproportionately experiencing food insecurity are pled approximately 13.8 million housing units collecting
the same ones disproportionately affected by the COVID-19 roughly 108,000 responses per week, for an approximate
disease itself [28–30]. 5.0% response rate. This study utilized Phase I of the HPS
With the wide-ranging effects of the COVID-19 pan- collected between April 23 and July 21, 2020. Twelve cross-
demic, it is increasingly critical that governments roll out sectional sample which were collected roughly bimonthly
the most effective responses to communities in need. The were assembled to create the full Phase I HPS data file.
stimulus check of April 2020, received by over 80 million HPS was designed to produce estimates at three different
US citizens [31, 32], provides a unique opportunity to study geographical levels—metropolitan statistical areas (MSAs),
the efficacy of this government response and whether cash state, and nation. Sampling for the HPS was drawn from the
13
Journal of Racial and Ethnic Health Disparities (2023) 10:2685–2703 2687
Census Bureau Master Address File (MAF) and was applied stage of the analysis to only estimating the effect of stimulus
through four adjustments within each state to account for on food insecurity conditional on receiving the stimulus.
nonresponse and coverage of the interview persons demo- They are therefore unsuitable for analyses of the quantity of
graphic characteristics. These adjustments were as follows: interest. In order to ensure that these effects are accurately
(1) household nonresponse adjustment to account for house- estimated, we perform additional robustness tests on a data
holds that do not respond to the survey; (2) an adjustment set matched by stimulus receipt.
to control the weights to the occupied housing unit counts To measure food security, we leverage a question on
using the American Community Survey (ACS) occupied phase 1 of the HPS asking respondents how to best describe
housing unit estimates based on the 2014–2018 5-year esti- “household food consumption in the last 7 days.” Respond-
mates; (3) an adjustment to account for the number of adults ents could choose one of the following 4 statements:
within the housing unit; and (4) a two-step iterative raking
procedure which rakes the demographics of the interviewed 1. Enough of the kinds of food (I/we) wanted to eat,
persons to known educational attainment/sex/age distribu- 2. Enough, but not always the kinds of food (I/we) wanted
tions and ethnicity/race/sex/age population distributions. to eat,
Within each state, demographic groups of interviewed per- 3. Sometimes not enough to eat,
sons were assessed prior to the raking procedure to deter- 4. Often not enough to eat.
mine if collapsing was necessary. The base weights reflect
these sampling rates and adjustment procedures. We treat this variable as a categorical/count variable
The federal payment was a cash transfer to US citizens, given the different levels of food security presented for each
permanent residents, or qualifying resident aliens. The only response. For robustness in the empirical estimation, we also
qualification was that an individual must have a Social Secu- use two more questions that ask respondents whether food
rity Number and not be claimed as a dependent on another insecurity arises from lack of ability to pay for food and
taxpayer’s tax return. Individuals received $1200 per per- whether the respondent is confident in her ability to secure
son if they filed individual tax returns with annual adjusted food. Morales, Morales, and Beltrán use a similar strategy
gross income (AGI) of $75,000 or less, if they filed as head in their analysis of food insecurity during the first week of
of household with AGI less than $112,500, or if they filed HPS collection [37].
married jointly with AGI less than $150,000. There was an The HPS asked a new question beginning in week seven
income cap on the amount received: individuals with no to evaluate household use of the federal stimulus payment.
qualifying children did not receive an Economic Impact Pay- We use this question to create an indicator variable if the
ment (EIP) if their AGI was more $99,000, $136,500, or respondent received such payment. Furthermore, we include
$198,000, respectively. a range of socioeconomic and demographic variables to
Importantly, the presence of children in a household serve as controls. These include income, age, employment
affected the stimulus amount received. Recipients got up status, marital status, gender, and household size. We create
to an additional $500 per qualifying child (subject to the an indicator variable for having an educational attainment
same qualifications as above and were less than 17 years of level equal to or greater than a high school degree. We also
age when the latest tax return was filed). The total-phaseout create a variable for household size based on respondent
cutoff AGI levels increased by $10,000 for each qualifying indication of the number of individuals in their household
child [36]. under the age of 18 and dichotomous variables for respond-
To account for the relative importance of the stimulus ent racial, ethnic, and residential self-classification. Finally,
to the respondent, we include an explanatory variable for we create an indicator of “rural” status for households
household income. The transfer itself was first distributed located outside one of 15 classified MSAs.
electronically to those who had filed their taxes online in the HPS is a cross-sectional survey containing very few ques-
prior year and was then subsequently distributed via mail to tions about events that occurred in the past or the recipient’s
all remaining others. Because it applied to all US persons state at points in the past. However, the literature has shown
and was agreed upon on a short-term basis by elected offi- that families that suffered food security in the past are more
cials, the transfer can be treated exogenously. likely to suffer from it again. We take two steps to meas-
Despite these arguments for its exogeneity, it could be ure whether the household has a history of food insecurity
argued that EIP receipt was endogenously related to whether before the survey was answered. First, we consider the HPS
the respondent filed taxes online, and therefore potentially measure of respondent household food sufficiency before
to previous food insecurity levels. However, the quantity of March 13, 2020, as a baseline measure of food insecurity
interest is the effect of household-level stimulus receipt on before the stimulus was disbursed. Second, to compare
food insecurity. Methods to correct for endogenous selection food security during the COVID-19 pandemic to measures
bias, such as a Heckman style model, would limit the second of food security during previous time periods, we match
13
2688 Journal of Racial and Ethnic Health Disparities (2023) 10:2685–2703
Table 1 CPS food security and Rank CPS HPS
HPS food sufficiency categories
0 High food security Enough of the kinds of food (I/we) wanted to eat
1 Marginal food security Enough, but not always the kinds of food (I/we)
wanted to eat
2 Low food security Sometimes not enough to eat
3 Very low food security Often not enough to eat
the HPS households with entries in the Current Population matched HPS cohort to ensure that the standardized variable
Survey’s Food Security Supplement (CPS-FSS).1,2,3 We use differences were within an acceptable differential so not to
adult food security status reported in the CPS, which is the alter the covariate distribution.
closest analog to the HPS measure. Moreover, like the HPS
measure, the CPS measure consists of four response catego- Empirical Estimations
ries reflecting an ordered level of difficulty obtaining food,
making matching logically consistent. Table 1 shows cor- We use a three-part empirical approach. We first assess the
respondence between the CPS and HPS ordered categories. effect of the pandemic on food security over all 12 weeks
To create this second historical food security measure, of data. In the second and third parts of the analysis, we
we matched HPS respondents with similar individuals in divide the 12-week phase 1 survey into two 6-week panels—
the 2018 CPS using propensity score matching (PSM).4 We pre-stimulus and post-stimulus—and estimate the statistical
matched on demographic variables common to both the HPS model separately for each. This three-part approach separates
and CPS: race, ethnicity, gender, age, marital status, cur- the effect of the pandemic over time from the effects of the
rent employment status, household size, state of residence, stimulus. Block bootstrapping (as recommended by [36]) was
and household income. To produce high-quality matches but then performed to correct the standard errors and confidence
ensure that final estimates were not biased, we chose two-to- intervals generated from the matched sample regression.
one greedy matching with replacement.5 This method mini- To identify different population subgroups, we rely on
mizes the propensity score distance between the matched HPS respondent self-classification by race (White, Black,
pairs such that each HPS respondent is matched to the two Asian, any other race alone, or other combination) and eth-
closest CPS respondents even if the CPS respondent has nicity (Hispanic or non-Hispanic). Based on these responses,
been selected before. To avoid bad matches, CPS and HPS we created four racial/ethnic subgroups—White non-His-
respondents are only matched if the propensity score match panic, Black non-Hispanic, Hispanic, and other. We include
lies within ± 0.5%. We assessed match quality by compar- dichotomous indicators for Black non-Hispanic, Hispanic,
ing the characteristics of the original HPS cohort and the and other race, leaving White non-Hispanic as the reference
category. We also include dichotomous demographic control
variables for whether a respondent identifies as female, mar-
1
This is the source of the US Department of Agriculture’s annual
ried, employed, and having less than a high school education.
reports. To capture the trend in food security over time, we include
2
We acknowledge that the recall periods between both measures are a variable for the survey week and the prior level of food
different. However, since we want to control for previous states of security measured from the matched 2018 CPS to account
food insecurity, this difference does not impede the conceptual valid- for intertemporal variation. Finally, we include time-subgroup
ity of the new control.
3
interaction terms (Black, Hispanic, and rural indicators) to test
PSM was performed in SAS 9.4 (Cary, NC) using Proc Psmatch.
4
for differences during the sample among different groups. We
Since the validity of estimates depends largely on deriving a high-
quality matched sample, is it important to ensure that the matching weighted all estimates using person-level sampling weights to
technique did not bias analyses. To test the validity of results, we conform to a nationally representative sample.
supplement the original two-toone greedy matching with various Because our food security dependent variable is
greedy, nearest neighbor, replacement, and match count conditions. expressed as a, ranked, discrete numeric value, we use
We also adjust the characteristics on which respondents were matched
to include the appropriate number and type of characteristics. We ordinal logistic (OL) regression as our empirical model.6
also exclude various groups of participants using each of these
6
approaches. We form conclusions under each These estimates provide a more intuitive contextualization of the
additional match specification to ensure that results were not due to regression results. While OLS is not the best
group differences or the statistical technique. unbiased estimator for non-continuous dependent variables, its coef-
5
(0 = enough of the kinds of food (I/we) wanted to eat, 1=enough, ficient estimates are straightforward and widely
but not always the kinds of food (I/we) wanted to eat, 2=sometimes understood, but the standard errors that accompany these estimates
not enough to eat, and 3=often not enough to eat. are not considered reliable.
13
Journal of Racial and Ethnic Health Disparities (2023) 10:2685–2703 2689
Table 2 Comparison of Unmatched HPS sample Matched HPS sample
matched and unmatched sample
characteristics N Mean Std dev Min Max N Mean Std dev Min Max
Income 1,066,593 4.54 2.09 1 8 941,575 4.54 2.09 1 8
Black 1,066,593 0.08 0.27 0 1 941,575 0.08 0.26 0 1
Hispanic 1,066,593 0.09 0.28 0 1 941,575 0.08 0.28 0 1
Age 1,066,593 51.57 15.72 18 88 941,575 51.65 15.49 18 88
Employed 1,066,593 0.57 0.49 0 1 941,575 0.58 0.49 0 1
Female 1,066,593 0.6 0.49 0 1 941,575 0.59 0.49 0 1
Education 1,066,593 0.02 0.14 0 1 941,575 0.02 0.13 0 1
Married 1,066,593 0.58 0.49 0 1 941,575 0.58 0.49 0 1
State FIPS code 1,066,593 27.75 16.2 1 56 941,575 27.78 16.21 1 56
Due to the ordering of dependent variable levels, the OL FS2020 =𝛽0 + 𝛽1 Age + 𝛽2 Female + 𝛽3 Married + 𝛽4 LessHSEduc+
regression accounts for discrete and ordered nature and 𝛽5 NumChildren + 𝛽6 Week + 𝛽7 FS2019 + 𝛽8 FS2018 + 𝛽9 Black+
the relationship between discrete and continuous regres-
𝛽10 Hispanic + 𝛽11 OtherRace + 𝛽12 Rural + 𝛽13 Black ∗ Week+
sors. One of the assumptions underlying OL regression is
that the relationship between each pair of outcome groups 𝛽14 Hispanic ∗ Week + 𝛽15 OtherRace ∗ Week + 𝛽16 Rural ∗ Week
is the same. In other words, OL assumes that the coef- + 𝛽17 EIP + 𝛽18 Black ∗ EIP + 𝛽19 Hispanic ∗ EIP+
ficients that describe the relationship between, say, the 𝛽20 OtherRace ∗ EIP + 𝛽21 Rural ∗ EIP + 𝜀i
lowest versus all higher categories of the response vari- (2)
able are the same as those that describe the relationship For interpretability reasons, we also present OLS results
between the next lowest category and all higher catego- as a test for robustness7.
ries, etc. The proportional odds assumption allows for the
estimation of only one set of coefficients. Application of
the Brant test indicates whether this assumption is met Results
or violated.
Given that food security also represents categories Quality of Matches
of counts of a food-related condition, a negative bino-
mial (NB) and ordinary least squares (OLS) regression A total of 1,066,593 respondents from HPS reported their level
were also estimated as a robustness tests. The NB model of food security during the first 3 months of the pandemic.
accounts for overdispersion, the presence of dispropor- Most of these respondents were female (60.0, SD = 0.49), most
tionately more zeros than any other numerical category, were married (58.1, SD = 0.49), and the average age was 51.6
which is particularly important because most respond- (SD = 12.72). Only about 2% had not graduated from high
ents during the early weeks of the pandemic did not school (1.9, SD = 0.14). Hispanics (9.0, SD = 0.28) and Blacks
experience food insecurity. The OLS regression provides (8.0, SD = 0.27) comprise less than 20% of respondents, and
highly logical and interpretable results despite unreli- 57% of respondents (SD = 0.49) were employed at the time
able standard errors. For the 12-week period, our basic of the survey. The average income level is 4.54 (SD = 2.09),
empirical model is: which corresponds to an annual income between $35,000 and
FS2020 =𝛽0 + 𝛽1 Age + 𝛽2 Female + 𝛽3 Married + 𝛽4 LessHSEduc+ $45,000. Table 2 shows the characteristics of the matched and
𝛽5 NumChildren + 𝛽6 Week + 𝛽7 FS2019 + 𝛽8 FS2018 + 𝛽9 Black+
unmatched samples, which are nearly identical.8
𝛽10 Hispanic + 𝛽11 OtherRace + 𝛽12 Rural + 𝛽13 Black ∗ Week+
𝛽14 Hispanic ∗ Week + 𝛽15 OtherRace ∗ Week + 𝛽16 Rural ∗ Week + 𝜀i
7
(1) These estimates provide a more intuitive contextualization of the
regression results. While OLS is not the best unbiased estimator for
The empirical model for the second and third parts of the non-continuous dependent variables, its coefficient estimates are
analysis is slightly different. It includes a dummy variable straightforward and widely understood, but the standard errors that
for receipt of the stimulus payment (EIP) as well as interac- accompany these estimates are not considered reliable.
8
tions between each subgroup and receipt of the payment: The average state code is 27.75 (SD = 16.20) in both cohorts, indi-
cating balanced state representation.
13
2690
13
Table 3 Descriptive statistics by racial/ethnic cohort
N Min Max Std dev Mean N Mean Std dev N Mean Std dev N Mean Std dev N Mean Std dev
Full sample Black White Hispanic Other race
Average food security levels
Food security2018 941,575 0 3 0.92 0.47 66,836 0.48 0.92 722,730 0.47 0.92 78,408 0.46 0.91 73,601 0.46 0.91
Food security2019 940,423 0 3 0.58 0.28 66,722 0.55 0.8 721,952 0.23 0.52 78,232 0.47 0.73 73,517 0.33 0.63
Food security2020 941,575 0 3 0.64 0.4 66,836 0.68 0.81 722,730 0.35 0.59 78,408 0.62 0.75 73,601 0.48 0.68
Sample statistics for individual characteristics
Age 941,575 18 88 15.49 51.65 66,836 48.03 14.02 722,730 53.09 15.5 78,408 45.71 14.55 73,601 47.22 14.65
Female 941,575 0 1 0.49 0.59 66,836 0.7 0.46 722,730 0.58 0.49 78,408 0.61 0.49 73,601 0.55 0.5
Married 941,575 0 1 0.49 0.58 66,836 0.38 0.48 722,730 0.61 0.49 78,408 0.52 0.5 73,601 0.57 0.49
HS education 941,575 0 1 0.13 0.02 66,836 0.03 0.17 722,730 0.01 0.11 78,408 0.07 0.26 73,601 0.02 0.14
# HH children 941,575 0 5 1.06 0.66 66,836 0.86 1.17 722,730 0.59 1.02 78,408 0.94 1.21 73,601 0.77 1.1
Rural 941,575 0 1 0.46 0.7 66,836 0.55 0.5 722,730 0.74 0.44 78,408 0.54 0.5 73,601 0.59 0.49
Week 941,575 1 12 3.34 6.63 66,836 6.65 3.35 722,730 6.61 3.34 78,408 6.78 3.33 73,601 6.62 3.34
Stimulus 476,047 0 1 0.5 0.56 34,393 0.69 0.46 362,986 0.53 0.5 41,380 0.64 0.48 37,288 0.58 0.49
Hispanic 941,575 0 1 0.28 0.08
White 941,575 0 1 0.42 0.77
Black 941,575 0 1 0.26 0.07
Other race 941,575 0 1 0.27 0.08
N represents total number of observations
Min represents the minimum vale
Max represents the maximum value
Std dev represents the sample standard deviation
Mean represents the sample mean
Journal of Racial and Ethnic Health Disparities (2023) 10:2685–2703
Journal of Racial and Ethnic Health Disparities (2023) 10:2685–2703 2691
Fig. 1 Standardized variable
differences
Table 4 Food security Food security categorical distribution by subgroup in 2018, 2019, and 2020
categorical representation
Total Hispanic Black White Other race
Food security2020 High 66.9 52.41 51.23 70.52 60.97
Marginal 26.88 35.4 32.81 24.96 31.24
Low 5.06 10 13 3.67 6.26
Very low 1.16 2.2 2.96 0.84 1.53
Food security2019 High 77.98 64.89 62.3 81.26 73.91
Marginal 17.11 24.76 23.3 15.42 19.97
Low 3.96 8.35 11.61 2.69 4.86
Very low 0.95 2 2.78 0.64 1.26
Food security2018 High 75.37 75.58 74.89 75.36 75.75
Marginal 9.54 9.66 9.56 9.55 9.36
Low 7.89 7.72 8.21 7.87 7.97
Very low 7.19 7.05 7.35 7.22 6.92
As expected, not all HPS respondents had a suitable residence. Figure 1 shows the difference between the
CPS match. To ensure the matched cohort provided an treatment sample from HPS and the matched sample from
accurate presentation of the original population, Table 3 CPS. The sample proportions differ by less than ± 0.2
summarizes the standardized baseline characteristics of suggesting high-quality matches and equivalent pairing
the HPS unmatched and matched samples. Demographic of sample characteristics.
proportions between the two samples were similar with
slight differences in characteristic means and variances. Food Security in the Data
We matched the samples using propensity scores cal-
culated from race, ethnicity, gender, age, marital status, For any of our food security measures, a value of zero indicates
income, employment status, and state of geographical that respondents have high food security, with enough of the
13
2692
Journal of Racial and Ethnic Health Disparities (2023) 10:2685-2703
Fig. 2. Proportion of each
subgroup reporting high food
security and very low food
security
62
57
52
47
Proportion of Subgroup Population
=== Other Race
Proportion Reporting High Food Security
° _
67
4 5 6 7 8 9 10 11 12
Week
==——_ White Black === Hispanic
Proportion Reporting Very Low Food Security
3.5
1 ee eNO
0.5
Proportion of Subgroup Population
e==== Other Race
type of foods they want to eat. A value of three indicates that
respondents have very low food security, with often not enough
to eat. Table 4 shows the distribution of all three food security
metrics for 2018, 2019, and 2020 across all four racial/ethnic
subgroups. All four subgroups have become less food secure
over time. Although the proportion of respondents at the low-
est level of food security has fallen between 2018 and 2020, the
largest change is among the second tier — those who have suf-
ficient food, but not always the types that they prefer — which
increased from 9.54 to 26.88% of the full sample.
Figure 2 shows the portion of each subgroup reporting
very low food security in each of the 12 weeks of HPS Phase
I. Food security is stable over time among Whites but varies
more among Blacks and Hispanics. These fluctuations could
a shift in the distribution of FS categories within subgroups.
g) Springer
4 5 6 7 8 9 10 11 12
Black e=== Hispanic
Chi-square tests indicate that the distributions of
food security are indeed statistically significantly dif-
ferent between Whites and Blacks Ca = 16,573.74,
p<0.001), Hispanics Cas = 9899.25, p<0.001), and
other races (y?= 1306.3521, p< 0.0001). Rural resi-
dents (y’ = 169.61, p < 0.001) and women (7? = 3238.93,
p<0.001) also differ significantly compared to their ref-
erence category (urban residents and men, respectively).
Regression Analysis: Base Model Weeks 1-12
Table 5 shows results for the regression without the
stimulus indicator or interaction terms included. Posi-
tive coefficients correspond to lower food security. Esti-
mates indicate significantly lower food security among
Journal of Racial and Ethnic Health Disparities (2023) 10:2685–2703 2693
Table 5 Food security and race: base model ordered logit estimation
Weeks 1–12 Weeks 1–6 Weeks 7–12
Log likelihood 1,093,796 519,950 572,213
AIC 1,093,840 519,994 572,257
Estimate Std err t-value Estimate Std err t-value Estimate Std err t-value
Intercept(1) − 3.05145 0.0176517 − 172.86996 − 3.53995437 0.0269228 − 131.48536 − 2.60585 0.0456886 − 57.034893
Intercept(2) 0.414236 0.0172961 23.949695 0.092008412 0.026264 3.5032141 0.726198 0.0455041 15.958952
Intercept(3) 3.207183 0.0192434 166.66445 2.965413331 0.0290619 102.03798 3.453968 0.0469393 73.583642
Food security2019 0.595203 0.0191383 311.00139 0.629322706 0.0275149 228.7203 0.564339 0.0268669 210.05015
Food security2018 0.07266 0.0059203 12.272942 0.063046535 0.0085966 7.3339163 0.081202 0.0081768 9.9307579
Age − 0.01151 0.0001793 − 64.170977 − 0.0121812 0.0002616 − 46.560643 − 0.01079 0.0002467 − 43.745577
Female 0.145847 0.0052526 27.766831 0.14308781 0.0076109 18.800482 0.147334 0.0072697 20.266904
HS education 0.261232 0.017151 15.231269 0.24272427 0.025332 9.5817383 0.28256 0.0233144 12.119556
Rural 0.03378 0.0124446 2.7144665 0.075214095 0.0199186 − 3.7760677 0.066658 0.0457077 2.4583483
Week − 0.0083 0.0015305 − 5.4236384 − 0.08636938 0.0046301 − 18.65389 0.019062 0.0043139 4.4188025
# HH children 0.038388 0.002457 15.624308 0.02983655 0.0035608 8.379062 0.044761 0.0034009 13.161492
Black 0.18641 0.0202282 9.215326 0.112232379 0.0324009 3.4638616 0.414683 0.0724313 5.7251911
Hispanic 0.154525 0.0196345 7.8700911 0.179951532 0.0318909 5.642717 0.1370978 0.0688497 1.9912622
Other race 0.147121 0.0188823 7.7914337 0.13188391 0.0304871 4.3258927 0.286427 0.0679534 4.2150591
Black * Week 0.015316 0.0026941 5.6850541 0.031192144 0.0081907 3.8082489 − 0.0075672 0.0075079 − 1.007903
Hispanic * Week 0.012541 0.0025882 4.8454944 0.005669595 0.0080309 0.7059769 0.014389 0.0071037 2.0256184
Other race * Week 0.008432 0.0025259 3.3380544 0.010647632 0.0077424 1.3752317 − 0.0054907 0.007035 − 0.7804853
Rural * Week 0.012689 0.0016826 7.5414042 0.024602557 0.0050978 4.8261288 0.0019713 0.0047332 0.4164862
Odds ratios
Estimate 95% CI Estimate 95% CI Estimate 95% CI
Food security2019 384.53221 370.37545 399.23007 540.8960191 512.499 570.86649 282.41784 267.93103 297.68794
Food security2018 1.0753647 1.0629587 1.0879156 1.065076402 1.0472813 1.0831739 1.0845896 1.0673463 1.1021114
Age 0.9885579 0.9882105 0.9889054 0.987892688 0.9873863 0.9883994 0.9892667 0.9887886 0.9897452
Female (1 vs 0) 1.1570191 1.1451689 1.1689919 1.153831115 1.1367471 1.1711718 1.1587407 1.1423477 1.175369
HS education (1 1.2985286 1.2556036 1.3429211 1.274717098 1.2129732 1.339604 1.3265209 1.2672692 1.3885429
vs 0)
Rural (1 vs 0) 1.9667837 1.9434882 1.9906545 1.927544882 1.8920314 1.9644722 1.0689295 0.9773332 1.1691104
Week 0.9917336 0.9887632 0.994713 0.917255353 0.9089691 0.9256172 1.0192452 1.0106637 1.0278995
# HH children 1.0391345 1.0341425 1.0441506 1.03028612 1.0231207 1.0375018 1.0457775 1.0388299 1.0527715
Black (1 vs 0) 1.2049156 1.1580795 1.253646 1.11877281 1.0499344 1.1921246 1.5138912 1.313533 1.7448108
Hispanic (1 vs 0) 1.167104 1.1230436 1.2128929 1.197159338 1.1246214 1.274376 1.1469404 1.0021572 1.3126405
Other race (1 vs 0) 1.1584936 1.1164029 1.2021712 1.140975856 1.0747954 1.2112314 1.3316615 1.1656062 1.5213735
Black * Week 1.015434 1.0100863 1.02081 1.031683716 1.0152539 1.0483794 0.9924614 0.977964 1.0071736
Hispanic * Week 1.0126199 1.0074962 1.0177697 1.005685697 0.98998 1.0216406 1.0144935 1.0004665 1.0287172
Other race * Week 1.0084672 1.003487 1.0134722 1.01070452 0.995483 1.0261588 0.9945243 0.9809055 1.0083322
Rural * Week 1.0127701 1.0094356 1.0161156 1.024907697 1.0147183 1.0351994 1.0019733 0.992721 1.0113118
Dependent variable: food security (0 = high security, 1 = marginal food security, 2 = low food security, 3 = very low food security)
Indicates significant at the 95% confidence level
Estimates weighted using survey sampling weights
females (0.1458, se = 0.0053), Hispanics (0.1545, se = 0.0124), and those who previously experienced food
se = 0.0196), Blacks (0.1864, se = 0.0202), other racial insecurity in 2019 (0.5952, se = 0.0191) or 2018 (0.0727,
groups (0.1471, se = 0.0189), household with children se = 0.0059) compared to the reference group. Conversely,
(0.0384, se = 0.0025), those without a high school diploma older respondents (− 0.0115, se = 0.0002) had higher
(0.2612, se = 0.0172), those living in rural areas (0.0338, food security than younger respondents. Interaction terms
13
2694 Journal of Racial and Ethnic Health Disparities (2023) 10:2685–2703
for Black (0.0153, se = 0.00230), Hispanic (0.01254, exactly when payments were received, or the precise amount
se = 0.0036), other racial groups (0.0084, se = 0.0025), and received. However, we accounted for individuals who
rural residents (0.0127, se = 0.0017) with the weekly indi- received a stimulus payment at some point during weeks
cator are positive and statistically significant suggesting seven through 12. While the average food security level did
that these groups had comparatively lower food security not differ significantly between those that did and did not
than the reference group and the level of disparity changed receive a stimulus (0.0030, se = 0.0048), receipt of a stimu-
over time. lus by identified subgroups did show a comparative differ-
ence. Blacks (− 0.2933, se = 0.0415), Hispanics (− 0.2917,
Regression Analysis: Base Model Weeks 1–6 se = 0.0344), other racial groups (− 0.1060, se = 0.0334),
and 7–12 and rural residents (− 0.1595, se = 0.0230) who received a
stimulus payment had statistically different food security
Regression results from the first 6 weeks of the sample levels than individuals in their cohort who did not receive a
mirror the full sample results. The week during which the payment. In other words, receipt of a stimulus payment was
individual was surveyed is negative (− 0.0863, se = 0.0046) associated with lower levels of food insecurity among all
suggesting that within this time frame, individuals experi- three-minority racial and residential groups.
enced relatively lower levels of food security in later weeks Figure 3 shows the predicted food security for each ethnic
compared to earlier weeks. Regression results from the sec- group in rural and non-rural households. Rural households
ond half of the sample are also like those presented for the experienced higher predicted levels of food insecurity than
full 12-week duration. However, neither indicators for rural their non-rural counterparts, suggesting that the existing dis-
residence nor Hispanic ethnicity appear to be statistically parity is increasing over time.
significant from the reference group. Interaction effects vary Figure 4 shows predicted food insecurity for all racial
slightly from those presented above. In the first half of the groups in rural and non-rural areas for weeks one through
sample, only interactions with Black (0.0312, se = 0.0082) six. Between-group racial and urban/rural disparities in food
and rural (0.0246, se = 0.0051) remain statistically signifi- insecurity are present even early in the pandemic suggesting
cant. In the second half of the sample, only the interaction that they likely existed prior to the observation period.
with Hispanic (0.0144, se = 0.0071) is significantly different Figure 5 shows predicted food insecurity level for each
from the reference group. The significant interaction terms subgroup in weeks 7–12 split by receipt of the EIP.
suggest that these groups experience even greater differences Overall, recipients who identify as members of marginal-
in food security compared to the reference group in the later ized groups are more food insecure than members of non-
weeks of observation than they did in the earlier weeks. marginalized groups. The positive coefficient for Stimulus
indicates that the baseline effect is that EIP recipients are
Regression Analysis: Full Model Weeks 1–6 and 7–12 more food insecure, which makes sense because program
(OL) eligibility was restricted by income. As a result of receiv-
ing the stimulus, members of marginalized groups experi-
Table 6 shows the OL regression results including all enced an overall reduction in this food insecurity, shown in
interaction terms. Estimates for demographic and contex- each subgroup’s negative interaction coefficient. However,
tual characteristics are similar to those presented above. In because the magnitudes of the interaction coefficients are
the first 6 weeks, comparatively lower food security was less than the magnitude of the positive baseline Stimulus
observed among females (0.1431, se = 0.0.0076), Hispan- coefficient, this reduction did not entirely alleviate the food
ics (0.1800, se = 0.0319), Blacks (0.1122, se = 0.0324), insecurity that marginalized groups generally experience.
other racial groups (0.1319, se = 0.0305), household with
children (0.0298, se = 0.0036), those without a high school Robustness Tests
diploma (0.2427, se = 0.0253), and those who previously
experienced food insecurity in 2019 (0.6393, se = 0.0291) or Given the use of an ordered dependent variable, the Brant
2018 (0.0631, se = 0.0096) compared to the reference group. test was run for all models to ensure that the parallel regres-
Older respondents (− 0.0128, se = 0.0002) had lower food sion assumption was in fact valid. Results indicated that
security than younger respondents. Interaction terms show all models were valid under this specification. While the
that rural residents (0.0246, se = 0051) and Blacks (0.0057, authors believe that the above model meets all assumptions
se = 0.0080) experience comparatively different food secu- of the OL model, OLS regression and NB regression are
rity changes compared to the reference groups during the specified for Eqs. 1 and 2. Results are listed in Tables 7
first 6 weeks. and 8 respectively validate the results presented above and
Following week six of the HPS, the $1200 federal ensure that the model selection did not greatly influence the
stimulus payment was announced. Data is not available on results. While standard errors are unreliable when OLS is
13
Journal of Racial and Ethnic Health Disparities (2023) 10:2685–2703 2695
Table 6 Food security, stimulus payments, and race: full model ordered logit estimation
Weeks 1–6 Weeks 7–12
Log likelihood 519,950 568,225
AIC 519,994 568,279
Estimate Std err t-value Estimate Std err t-value
Intercept(1) − 3.53995437 0.0269228 − 131.48536 − 1.914339337 0.049070567 − 39.01196711
Intercept(2) 0.092008412 0.026264 3.5032141 1.430223734 0.048975534 29.20282059
Intercept(3) 2.965413331 0.0290619 102.03798 4.159100309 0.050321802 82.65006681
Food security2019 0.629322706 0.0275149 228.7203 0.565832251 0.02699385 209.6152471
Food security2018 0.063046535 0.0085966 7.3339163 0.05651632 0.008192328 6.898688476
Age − 0.012181203 0.0002616 − 46.560643 − 0.010335994 0.00024699 − 41.84784549
Female 0.14308781 0.0076109 18.800482 0.12716756 0.007299693 17.42094615
HS education 0.24272427 0.025332 9.5817383 0.293914526 0.023318254 12.60448277
Rural 0.075214095 0.0199186 3.7760677 0.143287015 0.050113895 2.859227252
Week − 0.08636938 0.0046301 − 18.65389 0.018676384 0.004349617 4.29379921
# HH children 0.02983655 0.0035608 8.379062 0.045125615 0.003408669 13.23848576
Black 0.112232379 0.0324009 3.4638616 0.64957343 0.081761781 7.944707457
Hispanic 0.179951532 0.0318909 5.642717 0.384415099 0.075236903 5.109395585
Other race 0.13188391 0.0304871 4.3258927 0.391610479 0.074047594 5.288632028
Black * Week 0.031192144 0.0081907 3.8082489 0.824570014 0.019977396 41.27514915
Hispanic * Week 0.005669595 0.0080309 0.7059769 − 0.007093174 0.007514718 − 0.943904137
Other race * Week 0.010647632 0.0077424 1.3752317 0.014652384 0.00710797 2.061402174
Rural * Week 0.024602557 0.0050978 4.8261288 − 0.005419017 0.007060354 − 0.767527672
Stimulus 0.003042242 0.004762261 0.638823106
Black * Stimulus − 0.293257943 0.041546945 − 7.05847186
Hispanic * Stimulus − 0.291669634 0.034460361 − 8.463916977
Other race * Stimulus − 0.10599822 0.033446012 − 3.169233476
Rural * Stimulus − 0.159486795 0.022988053 − 6.937812207
Odds ratios
Estimate 95% CI Estimate 95% CI
Food security2019 540.8960191 512.499 570.86649 286.6673562 271.8948831 302.2424408
Food security2018 1.065076402 1.0472813 1.0831739 1.058143883 1.041289295 1.075271285
Age 0.987892688 0.9873863 0.9883994 0.989717239 0.989238241 0.990196468
Female (1 vs 0) 1.153831115 1.1367471 1.1711718 1.135607285 1.119475671 1.151971354
HS education (1 vs 0) 1.274717098 1.2129732 1.339604 1.341669221 1.281731104 1.40441025
Rural (1 vs 0) 1.927544882 1.8920314 1.9644722 1.154060986 1.046096466 1.273168206
Week 0.917255353 0.9089691 0.9256172 1.018851878 1.01020299 1.027574815
# HH children 1.03028612 1.0231207 1.0375018 1.046159265 1.039193308 1.053171917
Black (1 vs 0) 1.11877281 1.0499344 1.1921246 1.91472389 1.631211979 2.247511435
Hispanic (1 vs 0) 1.197159338 1.1246214 1.274376 1.468754995 1.267382157 1.702123722
Other race (1 vs 0) 1.140975856 1.0747954 1.2112314 1.479361357 1.279513415 1.710423665
Black * Week 1.031683716 1.0152539 1.0483794 2.2808998 2.19331706 2.371979863
Hispanic * Week 1.005685697 0.98998 1.0216406 0.992931923 0.978414621 1.007664628
Other race * Week 1.01070452 0.995483 1.0261588 1.014760257 1.00072128 1.028996184
Rural * Week 1.024907697 1.0147183 1.0351994 0.994595639 0.980927176 1.008454562
Stimulus 1.003046874 0.993728134 1.012453002
Black * Stimulus 0.745829735 0.687503431 0.809104315
Hispanic * Stimulus 0.747015283 0.698227219 0.799212375
Other race * Stimulus 0.899426249 0.84235714 0.960361751
Rural * Stimulus 0.852581226 0.815020061 0.891873441
Dependent variable: food security (0 = high security, 1 = marginal food security, 2 = low food security, 3 = very low food security)
Indicates significant at the 95% confidence level
Estimates weighted using survey sampling weights
13
2696 Journal of Racial and Ethnic Health Disparities (2023) 10:2685–2703
Fig. 3 Weeks 1–12 food secu-
rity in rural and non-rural areas
Fig. 4 Weeks 1–6 food security
in rural and non-rural areas
used to estimate a discrete dependent variable model, coef- findings. These results remained largely consistent when the
ficient estimates approximate the linear association between number of matches, matching characteristics, and matching
food security and the respective regressors. The OLS coef- method were changed. Additionally, to ensure that struc-
ficient estimates reinforce the findings from the OL model tural differences between age and gender cohorts did not
and similar levels of statistical significance support earlier bias these findings, the sample was divided by sex and age
13
Journal of Racial and Ethnic Health Disparities (2023) 10:2685–2703 2697
Fig. 5 Weeks 7–12 food secu-
rity with the stimulus
to test within the sample heterogeneity. Both genders and all Our findings complement the studies focusing on how
age cohorts showed similar results. the COVID-19 pandemic has brought about all kinds of
Finally, individuals who received a stimulus were income shocks that affect food security by specifically aim-
matched to those who did not receive a stimulus based on ing at the differential impact across race and ethnic groups
age, race, ethnicity, gender, marital status, state of residence, [40, 41]. More specifically, the findings of this study show
week of survey response, and previous food security. The that young female non-educated individuals living in rural
matched sample thus included individuals who were similar areas are more food insecure. Another consistent pattern is
except for their receipt/non-receipt of a stimulus which was that households with more children usually are more food
not included in the HPS until week 7. Table 9 shows OL insecure than households with fewer children. This result
model estimation results using only weeks 7 through 12. has been found in both high-income [42, 43] and develop-
Specifications both with and without the stimulus interaction ing contexts [20, 21, 26]. We also find that Black and His-
terms using allow for the comparison of the food security panic individuals tend to have lower levels of food security
levels of these matched individuals. Similar coefficients in when compared to their White counterparts and that they do
both specifications ensure consistency and model sensitivity. worse as weeks pass. When looking at the descriptive data,
we also find that all ethnic subgroups become less food
secure over time. Other work in the USA has highlighted
Discussion that there are differential rates of food security across races
where Blacks and Hispanic households tend to be more food
While food security in the USA has been slowly increas- insecure than their White counterparts [20, 21, 26, 37, 44].
ing since 2008, the ongoing COVID-19 pandemic cri- We also find that the largest change was from having suffi-
sis has made it more difficult for households to obtain cient food to having sufficient food but not always the types
enough nutritious food as illustrated in our analysis and that respondents prefer. This indicates that the pandemic
some others [20]. The pandemic and social distancing decreased the food options available for consumption prob-
efforts implemented to slow its spread have disrupted ably through lack of purchasing power [39] which aligns to
global and local food supply chains [38]. Due to reces- the idea that poorer households tend to be the most prone
sionary pressure, an unprecedented rise in unemploy- to food insecurity and have less nutritious options available
ment, and loss of income exacerbated nutritional to them [20, 21, 45].
hardships for millions of households [39] and made vul- Our pre-stimulus results mirror those from the 12-week
nerable groups more vulnerable as our results show. analysis except for a lack of statistical significance for the
13
2698 Journal of Racial and Ethnic Health Disparities (2023) 10:2685–2703
Table 7 OLS robustness
Food security, stimulus payments, and race: OLS estimation
Weeks 1–12 Weeks 1–6 Weeks 7–12
R-square 0.5095 0.535 0.4988
Adj R-sq 0.5095 0.535 0.4988
F value 65,117.4 35,533.3 23,660.6
Parameter estimates
Estimate Std err t-value Estimate Std err t-value Estimate Std err t-value
Intercept 0.35587 0.00296 120.37 0.38423 0.0043 89.3 0.2434 0.00912 26.69
Food security2019 0.72163 0.000794 908.55 0.73553 0.00109 673.16 0.68684 0.00115 594.76
Food security2018 0.0049 0.000576 8.51 0.00508 0.000793 6.41 0.00305 0.000825 3.7
Age − 0.00306 3.33E-05 − 92.02 − 0.00302 4.68E-05 − 64.62 − 0.00315 4.68E-05 − 67.38
Female 0.02277 0.00106 21.55 0.01936 0.00147 13.21 0.01991 0.0015 13.24
HS education 0.05134 0.00206 24.89 0.04637 0.00294 15.76 0.05147 0.00286 17.97
Rural − 0.00298 0.00243 − 1.23 − 0.00599 0.0036 − 1.66 − 0.0002 0.00938 − 0.02
Week 0.000873 0.00031 0.28 − 0.00986 0.000873 − 11.3 0.00174 0.000894 1.95
# HH children 0.00856 0.000501 17.06 0.00736 0.000696 10.58 0.00587 0.000714 8.22
Black 0.05013 0.0037 13.53 0.05593 0.00549 10.19 0.07104 0.01444 4.92
Hispanic 0.0433 0.00321 13.49 0.03575 0.00473 7.55 0.01798 0.01238 1.45
Other race 0.01093 0.00407 2.69 0.01086 0.00597 1.82 − 0.02584 0.01574 − 1.64
Black * Week 0.000893 0.000499 1.79 − 0.00271 0.00142 − 1.91 0.00197 0.00144 1.37
Hispanic * Week 0.0014 0.000433 3.24 0.00376 0.00122 3.09 0.00791 0.00125 6.33
Other race * Week 0.00255 0.000551 4.62 0.00246 0.00155 1.59 0.00642 0.00159 4.03
Rural * Week 0.00123 0.00033 3.72 0.00196 0.00093 2.1 0.00134 0.000953 1.41
Stimulus 0.20545 0.00312 65.81
Black * Stimulus − 0.06254 0.0055 − 11.36
Hispanic * Stimulus − 0.06787 0.00454 − 14.96
Other race * Stimulus − 0.00609 0.00561 − 1.08
Rural * Stimulus − 0.01339 0.00335 − 3.99
Dependent variable: food security (0 = high security, 1 = marginal food security, 2 = low food security, 3 = very low food security)
Indicates significant at the 95% confidence level
Estimates weighted using survey sampling weights
Black racial indicator variable and the interaction between looking at the coefficients for the race indicator variables,
time and being Hispanic. These results suggest two things. we find that Hispanics and Blacks have higher levels of
First, that before the stimulus, Whites and Blacks had similar food insecurity when compared to their White counterparts.
levels of food security at baseline controlling for a diverse Despite this, we do find that Black individuals are becoming
array of factors. With the pass of time, the interaction between more food insecure with time. The fact that we do not find
survey week and Blacks shows that Black respondents are statistically significant results for the interaction between
more food insecure than Whites. Similarly, rural households the time and race indicator variables for Hispanics and other
get more food insecure with time. Second, our results show races seems to suggest that food insecurity is not getting
that food security for Hispanics remains high and relatively worse with time for people in these groups. Similarly, that
constant across time. Again, this relationship can be traced to rural households are more food insecure than urban ones
the lack of purchasing power [39], type of economic activity but they are not getting worse with time once the stimulus
[46], and employment status [26, 47], phenomena that have is taken into account.
been exacerbated during the pandemic [48, 49]. In terms of the stimulus, our results suggest that there is
Post-stimulus, our results indicate higher levels of food no statistically significant difference in the baseline levels
insecurity among young female non-educated individu- of those individuals who got the stimulus and those who
als living in rural areas and with multiple children. When did. However, the interactions between our race and rural
13
Journal of Racial and Ethnic Health Disparities (2023) 10:2685–2703 2699
Table 8 Negative binomial Food security, stimulus payments, and race: negative binomial model estimation
robustness
Weeks 1–12 Weeks 1–6 Weeks 7–12
Log likelihood − 584,109 − 281,512 − 293,684
AIC 1,273,204 613,417.2 641,812.9
Maximum likelihood estimates
Estimate Std err Estimate Std err Estimate Std err
Intercept − 1.1768 0.0101 − 1.0386 0.0154 − 1.73 0.0306
Food security2019 0.9292 0.0017 0.9455 0.0024 0.8522 0.0025
Food security2018 0.0138 0.0017 0.0105 0.0025 0.0116 0.0024
Age − 0.007 0.0001 − 0.0072 0.0002 − 0.0067 0.0002
Female 0.0797 0.0034 0.0807 0.0049 0.0551 0.0047
HS education 0.0334 0.0083 0.0262 0.012 0.0501 0.0114
Rural − 0.0025 0.008 − 0.0302 0.0127 0.0931 0.0309
Week − 0.002 0.001 − 0.0419 0.0031 0.0059 0.0029
# HH children 0.0198 0.0015 0.0154 0.0021 0.0151 0.002
Black 0.0638 0.0116 0.0159 0.0181 0.4692 0.0447
Hispanic 0.1225 0.0115 0.0995 0.0184 0.4102 0.0423
Other race 0.0927 0.0127 0.0703 0.0202 0.2978 0.0485
Black * Week 0.0091 0.0015 0.0178 0.0046 0.0002 0.0044
Hispanic * Week 0.0048 0.0015 0.0095 0.0046 0.0064 0.0042
Other race * Week 0.0041 0.0017 0.0081 0.0051 0.0005 0.0048
Rural * Week 0.0065 0.0011 0.0143 0.0033 0.0019 0.0031
Stimulus 0.8069 0.0111
Black * Stimulus − 0.4313 0.0182
Hispanic * Stimulus − 0.399 0.0158
Other race * Stimulus − 0.2351 0.0186
Rural * Stimulus − 0.0959 0.0118
Dependent variable: food security (0 = high security, 1 = marginal food security, 2 = low food security,
3 = very low food security)
Indicates significant at the 95% confidence level
Estimates weighted using survey sampling weights
indicator variables with the stimulus imply that Black, et al. [52] show that increasing income to low-income house-
Hispanic, and rural individuals are more food secure after holds can increase the purchase of healthier food. However,
receiving this money. These results indicate that households in the USA, Hut and Oster [53] present some evidence that
were using their stimulus to tackle food insecurity and that dietary changes are hard and often do not change in the
the stimulus is helping more vulnerable groups to insulate face of household shocks like job loss and childbirth. In
from the effects of the COVID-19 pandemic. However, when our categorical analysis, the fact that there is an increase in
comparing the magnitude of our coefficients, we find that respondents saying they have enough food but not the type
the stimulus is not able to completely counteract the effects they want agrees with these studies.
of the pandemic on food security. Hence, the US federal Despite the robustness of our results, this study faces
stimulus response to the COVID-19 pandemic was effective several limitations. First, the HPS is a temporary, short-
at alleviating but not eradicating food insecurity. term data collection instrument that collects data at a single
Our findings also align with previous work suggesting point in time. Therefore, no historical data exists, and indi-
that unconditional cash transfers (UCT) improve food secu- viduals were not repeatedly surveyed in successive weeks.
rity more than conditional cash transfers (CCT) because they Because the HPS questions change every 3 months, survey
do not constrain consumer choice [50, 51]. Although the items concerning food security were removed after Phase
literature, including this study, assesses the effect of stimu- 1. Second, the cross-sectional nature of the data makes
lus on food security, the effects on the type of food being it impossible to determine causality. Each week of the
eaten are still unclear. Both Allcott et al. [45] and Bartlett HPS included a different group of respondents making
13
2700 Journal of Racial and Ethnic Health Disparities (2023) 10:2685–2703
Table 9 Stimulus matched Food security, stimulus payments, and race: ordered logit model estimation with stimulus matched sample
sample
Weeks 7–12 Weeks 7–12
Log likelihood − 919,487 − 905,224
AIC 1,838,974 1,810,449
Maximum likelihood estimates
Estimate Std err Estimate Std err
Intercept(1) − 5.1549 0.001 − 5.8433 0.0007
Intercept(2) − 2.8256 0.0009 − 3.5335 0.0006
Intercept(3) − 0.2178 0.0009 − 0.8979 0.0006
Food security2019 1.548 0.0001 1.5027 0.0001
Food security2018 0.0277 0.0001 0.0276 0.0001
Age − 0.0122 0.005 − 0.0135 0.0112
Female − 0.0511 0.0001 0.0978 0.0001
HS education 0.3497 0.0002 0.3438 0.0002
Rural − 0.0444 0.0004 0.0602 0.0004
Week − 0.0251 0.0001 0.0221 0.0001
# HH children 0.0636 0.0001 0.0489 0.0001
Black 0.2031 0.0005 0.3456 0.0006
Hispanic 0.0678 0.0005 0.5511 0.001
Other race 0.0441 0.0006 0.1382 0.0014
Black * Week 0.0116 0.0001 − 0.0065 0.0001
Hispanic * Week 0.0145 0.0001 0.0188 0.0001
Other race * Week 0.0073 0.0001 0.0142 0.0001
Rural * Week − 0.0037 0.0001 − 0.0041 0.0001
Stimulus 1.0212 0.0003
Black * Stimulus − 0.5524 0.0004
Hispanic * Stimulus − 0.4129 0.0005
Other race * Stimulus − 0.1583 0.0006
Rural * Stimulus − 0.043 0.0004
Odds ratios
Estimate 95% CI Estimate 95% CI
Food security2019 4.701 4.7 4.702 4.493 4.493 4.494
Food security2018 1.028 1.028 1.028 1.028 1.028 1.028
Age 0.988 0.988 0.988 0.987 0.987 0.987
Female (1 vs 0) 0.108 0.107 0.108 1.103 1.102 1.103
HS education (1 vs 0) 1.419 1.418 1.419 1.41 1.41 1.411
Rural (1 vs 0) 0.093 0.091 0.095 1.128 1.126 1.13
Week 0.025 0.025 0.026 1.022 1.022 1.022
# HH children 1.066 1.066 1.066 1.05 1.05 1.05
Black (1 vs 0) 1.501 1.498 1.504 1.996 1.991 2.001
Hispanic (1 vs 0) 1.145 1.143 1.147 1.735 1.732 1.739
Other race (1 vs 0) 1.092 1.089 1.095 1.148 1.145 1.151
Black * Week 1.989 1.988 1.989 0.994 0.993 0.994
Hispanic * Week 1.015 1.014 1.015 1.019 1.019 1.019
Other race * Week 1.007 1.007 1.008 1.014 1.014 1.015
Rural * Week 0.996 0.996 0.996 0.996 0.996 0.996
Stimulus 2.777 2.775 2.778
Black * Stimulus 0.576 0.575 0.576
Hispanic * Stimulus 0.662 0.661 0.662
Other race * Stimulus 0.854 0.853 0.855
Rural * Stimulus 0.958 0.957 0.959
Dependent variable: food security (0 = high security, 1 = marginal food security, 2 = low food security,
3 = very low food security)
Indicates significant at the 95% confidence level
Estimates weighted using survey sampling weights
13
Journal of Racial and Ethnic Health Disparities (2023) 10:2685–2703 2701
the evolution of individual food security and responses This study contributes to an important adjudication of
to external stimuli impossible to track. While the average policy. However, its contribution goes further: the economic
number of respondents who reported lower food security stimulus payments were a test of a different kind of redistri-
increased, this study cannot definitively attribute these bution from one currently practiced in the USA. The USA
increases to the COVID-19 pandemic or the economic currently redistributes resources to individuals based on a
manifestations thereof. Third, propensity score match- patchwork quilt of means-tested programs targeted at spe-
ing cannot account for all sources of potential cofounders cific services like housing choice vouchers and food stamps.
available for analysis. While we used a careful set of char- Alternatively, under a Basic Income Guarantee (BIG)
acteristics in the matching algorithm and carefully evalu- scheme, the government would transfer a small amount of
ated the quality of the matches, it is only possible to match cash to individuals unconditionally and obligation-free.
on observed characteristics. Therefore, unobserved differ- Scholars argue that a BIG could be an ethically sound, his-
ences between matched pairs could have biased results. torically congruent, practically feasible, and economically
Finally, the data used in this paper is unable to identify if efficient means of providing a basic floor of quality of life
our results are an artifact of demand or supply factors. For and alleviating inequality [58]. Under a BIG scheme, redis-
instance, people could be more food insecure because the tribution from the government would instead find its way
prices of food increases, people cannot physically access to individuals via lump-sum payments that the individuals
food, or because the food available for them is less nutri- could use to purchase whatever goods they find necessary.
tious [46, 54]. Such a limitation leaves room for additional Hence, this study contributes important, albeit specific, evi-
work on the identification of the drivers in food insecurity. dence for the effectiveness of a lump-sum non-conditional
redistribution policy in a high-income country setting.
Conclusion
Author Contribution MJ performed the data analysis and contributed
to the writing and editing of the paper. TMcD and MVC contributed
This study illustrates the disparate monthly food security overall idea guidance, wrote content, and edited the paper. MC con-
across subgroups of society. While findings indicating that tributed the overview of relevant literature.
marginalized groups experience worse food security out-
Data Availability All data is publicly accessible.
comes are not novel, they also suggest that receipt of federal
stimulus payments was associated with improved food secu- Code Availability Upon request.
rity outcomes for these groups. Additional work is needed
to determine if improvements are directly related to EIP Declarations
receipt. While this work is limited to the first 3 months of
Ethics Approval N/A, no human subjects.
the COVID-19 pandemic, trends observed here likely con-
tinued and/or worsened for many disadvantaged Americans. Consent to Participate N/A, no human subjects.
The second wave of the HPS includes data on the SNAP
program and could form a basis for analysis of its effect Consent for Publication All authors.
during the pandemic.
Clear disparities in food security across races, ethnicities, Conflict of Interest The authors declare no competing interests.
and geographic areas are likely connected to structural fac-
tors. Therefore, sustained policy reform is needed to ensure
that households of all racial, ethnic, and socioeconomic sta- References
tus have access to sufficient nutrition. Our results suggest
1. UNICEF. The state of food security and nutrition in the world
that BIG schemes could help to improve vulnerable groups 2020. 2020. https://policycommons.net/artifacts/1421967/the-
food security in high-income countries as they have done in state-of-food-security-and-nutrition-in-the-world-2020/2036027/
lower-income settings [55]. 2. Placzek O. Socio-economic and demographic aspects of food
Future work should emphasize the political effects of security and nutrition. OECD Food, Agriculture and Fisheries
Papers, No. 150, OECD Publishing, Paris, 2021. https://doi.org/
the stimuli roll out. Previous scholarship has shown that 10.1787/49d7059f-en
politicians tend to be more responsive in places where 3. A. Amankwah, S. Gourlay, A. Zezza, Food security in the face
people are better informed and more politically active of COVID-19: evidence from Africa, World Bank. (n.d.). https://
[56]. Hence, the fact that rural individuals in our study blogs.worldbank.org/opendata/food-security-face-covid-19-evide
nce-africa (accessed October 21, 2022).
tend to be more food insecure could potentially respond 4. IMF, Managing the impact on households: assessing universal
to this phenomenon. Furthermore, the electoral effects of transfers (UT), special series on fiscal policies to respond to
the stimuli might be of interest, especially in places with COVID-19. (2020). https://www.imf.org/-/media/Files/Publicatio
sustained food insecurity [57]. ns/covid19-special-notes/en-special-series-on-covid-19-manag
13
2702 Journal of Racial and Ethnic Health Disparities (2023) 10:2685–2703
ing-the-impact-on-households-assessing-universal-transfers.ashx 2011–2017: disparities by age, sex, race/ethnicity, and income.
(accessed October 21, 2022). Popul Health Manag. 2021;24:496–501. https://doi.org/10.1089/
5. Stuff JE, Casey PH, Szeto KL, Gossett JM, Robbins JM, Simpson pop.2020.0123.
PM, Connell C, Bogle ML. Household food insecurity is associ- 23. Walmsley TL, Rose A, Wei D. Impacts on the US macroecon-
ated with adult health status. J Nutr. 2004;134:2330–5. https://doi. omy of mandatory business closures in response to the COVID-
org/10.1093/jn/134.9.2330. 19 pandemic. Appl Econ Lett. 2021;28:1293–300.
6. Vozoris NT, Tarasuk VS. Household food insufficiency is associ- 24. Elgin C, Basbug G, Yalman A. Economic policy responses to
ated with poorer health. J Nutr. 2003;133:120–6. https://doi.org/ a pandemic: Developing the COVID-19 economic stimulus
10.1093/jn/133.1.120. index. Covid Economics. 2020;1(3):40–53.
7. Weinreb L, Wehler C, Perloff J, Scott R, Hosmer D, Sagor L, Gun- 25. S. Chakraborty, M.C. Biswas, Impact of COVID-19 on the
dersen C. Hunger: its impact on children’s health and mental health. textile, apparel and fashion manufacturing industry supply
Pediatrics. 2002;110: e41. https://doi.org/10.1542/peds.110.4.e41. chain : case study on a ready-made garment manufacturing
8. Kushel MB, Gupta R, Gee L, Haas JS. Housing instability and industry, Journal of Supply Chain Management, Logistics and
food insecurity as barriers to health care among low-income Procurement. (2020). https://hstalks.com/article/6059/impact-
Americans. J Gen Intern Med. 2006;21:71–7. https://doi.org/10. of-covid-19-on-the-textile-apparel-and-fash/ (accessed October
1111/j.1525-1497.2005.00278.x. 21, 2022).
9. Dixon LB, Winkleby MA, Radimer KL. Dietary intakes and serum 26. Schanzenbach D, Pitts A. Food insecurity in the census household
nutrients differ between adults from food-insufficient and food- pulse survey data tables. Institute for Policy Research. 2020;1–15.
sufficient families: Third National Health and Nutrition Examina- 27. Alvi M, Gupta M. Learning in times of lockdown: how COVID-
tion Survey, 1988–1994. J Nutr. 2001;131:1232–46. https://doi. 19 is affecting education and food security in India. Food Sec.
org/10.1093/jn/131.4.1232. 2020;12:793–6. https://doi.org/10.1007/s12571-020-01065-4.
10. Kirkpatrick SI, Tarasuk V. Food insecurity is associated with 28. Wolfson JA, Leung CW. Food insecurity during COVID-19:
nutrient inadequacies among Canadian adults and adolescents. J an acute crisis with long-term health implications. Am J Public
Nutr. 2008;138:604–12. https://doi.org/10.1093/jn/138.3.604. Health. 2020;110:1763–5. https://doi.org/10.2105/AJPH.2020.
11. Seligman HK, Bindman AB, Vittinghoff E, Kanaya AM, Kushel 305953.
MB. Food insecurity is associated with diabetes mellitus: results 29. Pryor S, Dietz W. The COVID-19, obesity, and food insecurity
from the National Health Examination and Nutrition Examination syndemic. Curr Obes Rep. 2022;11:70–9. https://doi.org/10.1007/
Survey (NHANES) 1999–2002. J Gen Intern Med. 2007;22:1018– s13679-021-00462-w.
23. https://doi.org/10.1007/s11606-007-0192-6. 30. Cunningham GB, Wigfall LT. Race, explicit racial attitudes,
12. Seligman HK, Laraia BA, Kushel MB. Food insecurity is associated implicit racial attitudes, and COVID-19 cases and deaths: an
with chronic disease among low-income NHANES participants. J analysis of counties in the United States. PLoS ONE. 2020;15:
Nutr. 2010;140:304–10. https://doi.org/10.3945/jn.109.112573. e0242044. https://doi.org/10.1371/journal.pone.0242044.
13. Pan L, Sherry B, Njai R, Blanck HM. Food insecurity is associ- 31. Baker SR, Farrokhnia RA, Meyer S, Pagel M, Yannelis C. Income,
ated with obesity among US adults in 12 states. J Acad Nutr Diet. liquidity, and the consumption response to the 2020 economic
2012;112:1403–9. https://doi.org/10.1016/j.jand.2012.06.011. stimulus payments (No. w27097). National Bureau of Economic
14. Kirkpatrick SI, McIntyre L, Potestio ML. Child hunger and long- Research. https://doi.org/10.3386/w27097
term adverse consequences for health. Arch Pediatr Adolesc Med. 32. Coibion O, Gorodnichenko Y, Weber M. How did U.S. con-
2010;164:754–62. https://doi.org/10.1001/archpediatrics.2010.117. sumers use their stimulus payments? NBER Working. Paper,
15. Mangini LD, Hayward MD, Zhu Y, Dong Y, Forman MR. Timing (27693). (2020). https://doi.org/10.3386/w27693.
of household food insecurity exposures and asthma in a cohort of 33. Bounie D, Camara Y, Galbraith JW. Consumers’ mobility, expend-
US school-aged children. BMJ Open. 2019;8: e021683. https:// iture and online-offline substitution response to COVID-19: evi-
doi.org/10.1136/bmjopen-2018-021683. dence from French transaction data. SSRN Electron J. (2022).
16. Berbudi A, Rahmadika N, Tjahjadi AI, Ruslami R. Type 2 dia- https://doi.org/10.2139/ssrn.3588373
betes and its impact on the immune system. Curr Diabetes Rev. 34. Chang H-H, Meyerhoefer CD. COVID-19 and the demand for
2020;16:442–9. https://doi.org/10.2174/15733998156661910240 online food shopping services: empirical evidence from Taiwan.
85838. Am J Agr Econ. 2021;103:448–65. https://doi.org/10.1111/ajae.
17. Milner JJ, Beck MA. The impact of obesity on the immune 12170.
response to infection. Proc Nutr Soc. 2012;71:298–306. https:// 35. Chetty R, Friedman JN, Hendren N, Stepner M. The economic
doi.org/10.1017/S0029665112000158. impacts of COVID-19: Evidence from a new public database
18 Reider CA, Chung R-Y, Devarshi PP, Grant RW, Hazels-Mit- built using private sector data (No. w27431). National Bureau of
messer S. Inadequacy of immune health nutrients: intakes in economic research. (2020). https://doi.org/10.3386/w27431
US adults, the 2005–2016 NHANES. Nutrients. 2020;12:1735. 36. Internal Revenue Service, Calculating the economic impact pay-
https://doi.org/10.3390/nu12061735. ment, (2020). https://www.irs.gov/newsroom/calculating-the-
19. CDC COVID-19 Response Team, Bialek S, Boundy E, Bowen economic-impact-payment (accessed October 21, 2022).
V, Chow N, Cohn A, Dowling N, Ellington S, Gierke R, Hall A, 37. Morales DX, Morales SA, Beltran TF. Racial/ethnic disparities
MacNeil J, Patel P, Peacock G, Pilishvili T, Razzaghi H, Reed N, in household food insecurity during the COVID-19 pandemic: a
Ritchey M, Sauber-Schatz E. Severe outcomes among patients nationally representative study. J Racial Ethn Health Disparities.
with coronavirus disease 2019 (COVID-19) — United States, 2021;8:1300–14. https://doi.org/10.1007/s40615-020-00892-7.
February 12–March 16, 2020. MMWR Morb Mortal Wkly Rep. 38. Aday S, Aday MS. Impact of COVID-19 on the food supply chain.
2020;69(12):343–6. https://doi.org/10.15585/mmwr.mm6912e2. Food Qual Saf. 2020;4:167–80. https://doi.org/10.1093/fqsafe/
20. Coleman-Jensen A. Household Food Security in the United States fyaa024.
in. 2019;2018:47. 39. Béné C. Resilience of local food systems and links to food security
21. Coleman-Jensen A. Household Food Security in the United — a review of some important concepts in the context of COVID-
States in. 2020;2019:47. 19 and other shocks. Food Sec. 2020;12:805–22. https://doi.org/
22. Walker RJ, Garacci E, Dawson AZ, Williams JS, Ozieh M, 10.1007/s12571-020-01076-1.
Egede LE. Trends in food insecurity in the United States from
13
Journal of Racial and Ethnic Health Disparities (2023) 10:2685–2703 2703
40. T. Hanspal, A. Weber, J. Wohlfart, Income and wealth shocks Stud. 2015;51:660–73. https://doi.org/10.1080/00220388.2014.
and expectations during the COVID-19 pandemic, CESifo Work- 997219.
ing Paper, 2020. https://www.econstor.eu/handle/10419/216640 51. Kim MJ, Lee S. Can stimulus checks boost an economy under
(accessed October 21, 2022). COVID-19? Evidence from South Korea. Int Econ J. 2021;35:1–
41. P. Ozili, T. Arun, Spillover of COVID-19: impact on the global 12. https://doi.org/10.1080/10168737.2020.1864435.
economy, (2020). https:// m pra. u b. u ni- m uenc h en. d e/ 9 9850/ 52. S. Bartlett, J. Klerman, L. Olsho, C. Logan, M. Blocklin, M.
(accessed October 21, 2022). Beauregard, A. Enver, P. Wilde, C. Owens, Healthy incentives
42. Amare M, Abay KA, Tiberti L, Chamberlin J. COVID-19 and pilot final evaluation report | Food and Nutrition Service, USDA.
food security: panel data evidence from Nigeria. Food Policy. (2014). https://www.fns.usda.gov/snap/hip/final-evaluation-report
2021;101: 102099. https:// d oi. o rg/ 1 0. 1 016/j. foodp o l. 2 021. (accessed October 21, 2022).
102099. 53. Hut S, Oster E. Changes in household diet: determinants and pre-
43. Bauer L, Pitts A, Ruffini K, Schanzenbach DW. The effect of pan- dictability. J Public Econ. 2022;208: 104620. https://doi.org/10.
demic EBT on measures of food hardship. Washington, DC.: The 1016/j.jpubeco.2022.104620.
Hamilton Project, Brookings Institution; 2020. 54. Devereux S, Béné C, Hoddinott J. Conceptualising COVID-19’s
44. Lauren BN, Silver ER, Faye AS, Rogers AM, Woo-Baidal JA, impacts on household food security. Food Sec. 2020;12:769–72.
Ozanne EM, Hur C. Predictors of households at risk for food https://doi.org/10.1007/s12571-020-01085-0.
insecurity in the United States during the COVID-19 pandemic. 55. Haushofer J, Shapiro J. The short-term impact of unconditional
Public Health Nutr. 2021;24:3929–36. https://doi.org/10.1017/ cash transfers to the poor: experimental evidence from Kenya*. Q
S1368980021000355. J Econ. 2016;131:1973–2042. https://doi.org/10.1093/qje/qjw025.
45. Allcott H, Diamond R, Dubé J-P, Handbury J, Rahkovsky I, Sch- 56. Besley T, Burgess R. The political economy of government
nell M. Food deserts and the causes of nutritional inequality*. Q responsiveness: theory and evidence from India*. Q J Econ.
J Econ. 2019;134:1793–844. https://doi.org/10.1093/qje/qjz015. 2002;117:1415–51. https://doi.org/10.1162/003355302320935061.
46. Kansiime MK, Tambo JA, Mugambi I, Bundi M, Kara A, Owuor 57. Paarlberg R. Food politics: what everyone needs to know®.
C. COVID-19 implications on household income and food secu- Oxford: Oxford University Press; 2013.
rity in Kenya and Uganda: findings from a rapid assessment. 58. Van Parijs P, Vanderborght Y. 7. Politically Achievable? Civil
World Dev. 2021;137: 105199. https://doi.org/10.1016/j.world Society, Parties, and the Back Door. In: Basic Income: A Radi-
dev.2020.105199. cal Proposal for a Free Society and a Sane Economy. Cam-
47 Smith MD, Rabbitt MP, Coleman-Jensen A. Who are the world’s bridge: Harvard University Press; 2017. pp. 170–215. https://
food insecure? New evidence from the Food and Agriculture doi.org/10.4159/9780674978072-008
Organization’s Food Insecurity Experience Scale. World Devel-
opment. 2017;93:402–12. Publisher's Note Springer Nature remains neutral with regard to
48. Hobbs JE. Food supply chains during the COVID-19 pandemic. jurisdictional claims in published maps and institutional affiliations.
Canadian Journal of Agricultural Economics/Revue Canadienne
d’agroeconomie. 2020;68:171–6. https://doi.org/10.1111/cjag.12237. Springer Nature or its licensor (e.g. a society or other partner) holds
49. Valaskova K, Durana P, Adamko P. Changes in consumers’ pur- exclusive rights to this article under a publishing agreement with the
chase patterns as a consequence of the COVID-19 pandemic. author(s) or other rightsholder(s); author self-archiving of the accepted
Mathematics. 2021;9:1788. https://doi.org/10.3390/math9151788. manuscript version of this article is solely governed by the terms of
50. Gangopadhyay S, Lensink R, Yadav B. Cash or in-kind transfers? such publishing agreement and applicable law.
Evidence from a randomised controlled trial in Delhi, India. J Dev
13
File and source
- File
- sick-hungry-and-vulnerable-federal-stimulus-and-food-security-on-marginalized-populations-during-the-covid-19-pandemic.pdf
- Size
- 1,284,179 bytes
- SHA-256
- 54d01416a83702225a86494517679cda9cc27218c9befb6539f23d138b59fe88
- Original
- doi.org