A Grant to Every Citizen: Survey Evidence of the Impact of a Direct Government Payment in Israel (Working Paper 28312)
Summary
NBER Working Paper No. 28312, dated December 2020, by Naomi Feldman and Ori Heffetz, studies the one-time grants Israel disbursed to its citizens in early August 2020 during the COVID-19 pandemic, of $220 per adult and $150 per child. Using survey data collected in late August, the authors estimate that 25–45 percent had already mostly spent or planned to spend the money by year's end, and 36–52 percent mostly paid down debts. They report that about as many people mostly donated the grant or used it to help family or friends as saved it, with donations mostly originating from higher-income respondents. The paper compares its findings with past US episodes.
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
- A Grant To Every Citizen Survey Evidence Of The Impact Of A Direct Government Payment In I
- Case
- A Grant To Every Citizen Survey Evidence Of The Impact Of A Direct Government Payment In Israel
Cite this document
“A Grant to Every Citizen: Survey Evidence of the Impact of a Direct Government Payment in Israel (Working Paper 28312).”
Original at www.nber.org. Archived at Pandemic Darlings, https://pandemicdarlings.com/source-documents/a-grant-to-every-citizen-survey-evidence-of-the-impact-of-a-direct-government-payment-in-israel/
, SHA-256 bf89892284a8b0278c6859eb2829dc41cee599a78de43e1d8f4440c86efb1d88.Full text
NBER WORKING PAPER SERIES
A GRANT TO EVERY CITIZEN:
SURVEY EVIDENCE OF THE IMPACT OF A DIRECT GOVERNMENT PAYMENT IN ISRAEL
Naomi Feldman
Ori Heffetz
Working Paper 28312
http://www.nber.org/papers/w28312
NATIONAL BUREAU OF ECONOMIC RESEARCH
1050 Massachusetts Avenue
Cambridge, MA 02138
December 2020
We thank Joel Slemrod, Claudia Sahm and Jonathan Parker for generously sharing their original
materials, survey scripts, and calculations; seminar participants at the Hebrew University and
IDC Herzliya; Aharon Haver, Lev Maresca, and Doron Zamir for excellent research assistance;
and Israel’s National Economic Council and Prime Minister’s Office for funding the survey. The
views expressed here are of the authors only; they have no financial or other material interests
related to this research to disclose. 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.
© 2020 by Naomi Feldman and Ori Heffetz. 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.
A Grant to Every Citizen: Survey Evidence of the Impact of a Direct Government Payment
in Israel
Naomi Feldman and Ori Heffetz
NBER Working Paper No. 28312
December 2020
JEL No. D14,D91,E21,E62,E65,H24,H31
ABSTRACT
In early August, 2020, during the COVID-19 pandemic, Israel disbursed one-time, universal
grants to its citizens, of $220 per adult and $150 per child. Using survey data, we estimate that
25–45 percent either had already mostly spent or were planning to spend the money by year’s end
and 36–52 percent mostly paid down debts. Interestingly, about as many people reported mostly
donating the grant or using it to help family or friends as reported saving it (10–18 percent), with
donations mostly originating from higher-income respondents. This voluntary rerouting of
governmental assistance may help alleviate the trade-off between targeting and simplicity/speed
of disbursement.
Naomi Feldman
Bogen Family Department of Economics
The Hebrew University of Jerusalem
naomi.feldman@mail.huji.ac.il
Ori Heffetz
S.C. Johnson Graduate School of Management
Cornell University
324 Sage Hall
Ithaca, NY 14853
and The Hebrew University of Jerusalem
and also NBER
oh33@cornell.edu
In mid-July, 2020, the Israeli government announced one-time grants of roughly $220 per
adult and $150 per child to all eligible citizens in response to the COVID-19 crisis. Thus, a
family of four would have received about $740, nearly 18 percent of median monthly house-
hold income in 2018. By early August, the payments had been disbursed. The government
argued that the grants would provide much needed stimulus to the domestic economy and
encouraged those receiving the payments to go out and spend on domestically produced
goods and services.
How were these grants used? This paper reports results from a survey conducted in late
August, designed to answer this question. We view the Israeli episode as an opportunity to
study the impact of such direct governmental transfers in a non-US setting and in an eco-
nomic crisis of historic proportions. Moreover, the payments in Israel were, for all practical
purposes, universal, thus providing a case study into the economic and political implications
in a situation where everyone, needy or not, receives government handouts. Finally, by using
a survey instrument modeled after those used to study past US episodes, and by replicating
the calibration methods used in past research to estimate marginal propensities to consume
out of the grant, we can compare our findings with past findings, and discuss similarities
and differences across time, location, and context.
The COVID-19 pandemic spread quickly across the globe in the first quarter of 2020.
Many countries began shutting down their economies in order to mitigate the health implica-
tions of a little-understood contagious virus. As part of their response to the severe economic
consequences of these shutdowns, a number of countries turned to direct grant payments to
citizens. The US provided stimulus payments of $1,200 per person and $500 per child up
to a maximum of $3,400 per household. South Korea and Japan provided universal direct
payments to each of its citizens of $340-$680 and $960, respectively.1 A number of other
countries provided one-time child supplements (e.g., Germany and Canada), while Spain
1
The payments have a number of names, reflecting the different governments’ views. They are called
“stimulus payments” in the US, “emergency relief payments” in South Korea, and simply “special cash
payments” in Japan. As of the time of this writing, to our knowledge, these non-US payments have not yet
been studied in the academic literature.
1
appeared set to establish permanent basic income for its poorest residents. These direct
stimulus payments are not new for the US, where they were also disbursed most recently in
2001 as part of a tax rebate pre-payment and in 2008 during the Great Recession.
Survey data from these earlier episodes suggest that the payments provided modest stim-
ulus, with average marginal propensities to consume (MPC) estimated at around one-third
(Shapiro and Slemrod (2003), Shapiro and Slemrod (2009), Sahm, Shapiro and Slemrod
(2010)), though, consistently, around half of respondents report mostly using the payment
to pay down debt. Recently, Parker and Souleles (2019) replicate these survey findings and
show that they yield MPC estimates that are similar to those estimated from expenditure
data. They conclude that “reported spending is highly informative about revealed-preference
propensities to spend” and that their findings “lend credence to the use of reported causal
effects to measure the effect of economic policy or to estimate model parameters.” Most
recently, studies of the 2020 COVID-19 stimulus payments in the US estimate MPCs in a
similar range—25–42 percent—using a variety of data sources and methods (Baker et al.
(2020) use individual-level financial-app data, Misra, Singh and Zhang (2020) use debit-
card transaction data, and Coibion, Gorodnichenko and Weber (2020) use survey data). To
our knowledge, little is known on such payments outside of the US, with the exception of
Australia’s direct payments of 2008–2009.2
We use a survey that is generally similar to those used in past work, but is conducted
online. Akin to past findings, we find that the main reported short-term use of the money
was paying down debt (42–52 percent of respondents), followed by spending (25–31 percent)
and saving (15–18 percent). We further estimate a somewhat short-run MPC mostly within
the range of 33-50 percent, although, depending on assumptions, calibration method, and
time horizon, we also report slightly lower and higher estimates. Thus, to the extent that our
2
During the global financial crisis of 2008–2009, Australia sent targeted direct payments to its taxpayers,
focusing on lower income families with children and the elderly. Leigh (2012) estimates that about 40
percent of households mostly spent the payment—higher than in the US. He attributes the difference in part
to marketing, noting that the Australian payment was called a “bonus” while the US payment was called a
“rebate.”
2
survey findings generalize, we cautiously conclude that in per-dollar terms, the government’s
stated goal of stimulating the economy was modestly achieved.3
New to our study, we also find that around 14 percent report either donating the money or
giving it to family or friends, and around three percent report that the grant had no effect on
them. Although our online survey—which did less to discourage response options other than
spending, saving, and paying down debt—may have helped, these two findings may be unique
to the 2020 Israeli context. In response to the grant announcement, grassroots campaigns
sprung up mostly among secular, educated Israelis to collectively donate the money or even
find ways to refuse to accept it (the latter, to our knowledge, was not possible). Reflecting
a certain political zeitgeist, such campaigns were part of a public backlash against making
the grants universal and thereby giving money to those who did not need it and were not
directly financially harmed by the COVID-19 crisis. Consistently, we find that higher-income
households are significantly more likely to mostly donate their grant, and we find suggestive
evidence of inter-generational transfers in the flavor of Ricardian equivalence; we also find
that the ultra-orthodox are less likely to report having done either. Finally, we find that the
non-negligible number of respondents who refused to acknowledge that they did anything
with the grant—whom we term “protest” responders—are secular and highly educated but,
interestingly, are not of higher income.
We cautiously draw two lessons from these findings. The first is with regards to the
perceived trade-off in governmental transfers between getting the money to where it is needed
and getting it out quickly, transparently, with minimum bureaucracy and with minimum
distortion to incentives. Our findings raise the possibility that, at least in a socially and
politically aware society, universal or near-universal transfers may end up more targeted
3
Stating the obvious, our data do not allow us to assess whether stimulating consumption created or saved
jobs, nor any possible Keynesian multiplier effects, nor the per-dollar effectiveness of the grants compared
with alternative stimulus policies. See Guerrieri et al. (2020) for analysis of these questions using a multiple-
sector-economy model with pandemic-like lockdowns. We note however that the Israeli grants arrived at a
time when much of the economy had already reopened. (Following the first lockdown, most sectors reopened
during May, with limitations on number of customers/patrons; a second lockdown would only be announced
in mid September.)
3
than they appear on paper. It may therefore, under certain social conditions, be possible for
the government to disburse a universal grant quickly; and for society, acting as a decentralized
allocation mechanism, to redistribute the grants to where they are most needed. Our second
lesson is methodological. Survey data may be limited in their ability to measure the impacts
of such grants when they encourage respondents, if only implicitly, to stick to the response
options suggested by economic theory. Such limitations may be amplified when the grants
are controversial or, more generally, when respondents are opposed to a government’s actions.
Future surveys may do well to give respondents more flexibility in responding, for example
by explicitly offering response options that go beyond spend, save, and pay down debt.
1 Background
On July 15, 2020, the Israeli government announced a plan called “A Grant to Every Citi-
zen.” It provided for a universal grant to Israeli citizen-residents, of 750 New Israeli Shekels
(NIS), or roughly $220 ($1 USD = 3.5 NIS), per adult (over the age of 18) and 500 NIS,
or roughly $150, per child, with a maximum of three children per household. The plan was
announced on prime-time TV by the Israeli Prime Minister, and became the topic of much
local publicity and controversy.4 After subsequent heated debates in the Knesset (Israeli
Parliament) and the public sphere, the grant was slightly modified: it was ultimately capped
at annual earnings of 650,000 NIS (in 2019) with no phase-out range;5 the number of eligible
children per family was uncapped with the first four children receiving 500 NIS each and
subsequent children 300 NIS each; and certain individuals who were already receiving gov-
ernment benefits due to a disability or old-age, as well as new immigrants, victims of terrorist
attacks, Holocaust survivors, and several others, were eligible to receive an augmented grant
of 1,500 NIS.
4
The government subsequently created a dedicated web page for the program, in Hebrew, Arabic, and
English: https://govextra.gov.il/mof-corona-eng/grantscitizens/.
5
In practice, less than 2 percent of citizens were ineligible due to either having income above the ceiling
or not having filed a tax return for 2018 in spite of being obligated to do so (the large majority of taxpayers
do not have to file tax returns). Thus, the grant was practically universal.
4
Eligibility was at the individual level, rather than the household level as is typical in
the US. In particular, children were universally eligible, even the children of very wealthy
households. Similarly, a low-earning spouse of a very high-earning individual who is ineligible
may be eligible for the grant, in spite of their joint household income being at the top of
the income distribution. According to the Bank of Israel, the final cost of the grants was
estimated at the time to be 6.5 billion NIS (0.46 percent of 2019 GDP).6
The time from announcement of the grant in mid July to its disbursement was short, and
by the first week of August, 2020, nearly all eligible citizens had received the direct deposit.7
The grant was first disbursed to children (typically deposited into the parents’ bank account);
it was next disbursed to the elderly and those receiving the augmented grant; and finally to
the remaining eligible adults. Thus, a married household with children potentially saw up to
three deposits in a joint bank account—one for the children, and one for each adult, within
the span of a week.
When the grant was announced, Israel—like other countries—was facing record numbers
of unemployed and furloughed workers and distressed businesses. In addition, Israel’s polit-
ical situation was unprecedentedly fragile, with a newly formed government that many were
frustrated with and predicted would not last long.8 The government was thus facing a mix of
political and economic discontent, and certain segments of the population held it responsible
for a perceived lack of coherent plan for dealing with the COVID-19 crisis, among other
things.
6
“Israel Central Bank Chief Notes Progress on Cash Grant Plans.” July 26, 2020. https://www.bloomb
ergquint.com/global-economics/israel-cabinet-approves-grant-for-every-citizen-program.
7
A very small number of eligible citizens without an established bank account experienced delays in
receiving their grants. Instructions were provided by the National Insurance Institute as to how to obtain
the money in cash at a post-office if obtaining a bank account was not feasible. However, after August 23, the
remaining option is to provide a bank account number where the money will be deposited by mid-October,
2020. See https://www.btl.gov.il/English%20Homepage/Pages/each-citizen.aspx for further details.
8
Israel held three divisive rounds of general elections—typically held every four years—within a single
year, failing to resolve a political deadlock. An “emergency” national-unity government was eventually
reluctantly sworn in only two months prior to the grant announcement. It brought in political opponents
with little apparent mutual trust or willingness to work together. The new government could not seem to
agree on a budget—the timely passing of which is a condition for avoiding a fourth round of elections. General
frustration was visibly growing; large street protests, for example, were becoming increasingly common.
5
In addition to the grant, a number of other policies were announced in the preceding
days—including extending unemployment benefits until June 2021, various safety net mea-
sures for the unemployed and a number of policies aimed at helping businesses, such as
property tax abatement, loans, extending business licenses, and other stipends.
2 Data
We use data collected by an independent, third-party survey company (https://www.pane
l4all.co.il/) that administered the anonymous survey from August 26 to August 30, 2020.
The respondent pool available to the survey company consists of active participants in its
internet panel—individuals who voluntarily sign up to “make an impact and make money,”
according to the firm’s website. While we have no expectations that panel participants are
nationally representative of the Israeli population, especially with regard to unobservable
characteristics, the company’s large database of potential respondents can be sampled to
match the Israeli 18+ population on five observable variables: sex, age, geographic region,
level of religiosity, and income.
The survey was conducted in Hebrew, limiting responses by Israel’s Arab minority and
large immigrant population to those sufficiently proficient in the language. The company
sent out 8,867 initial invitations and 1,128 reminders to its panel participants, of whom
1,182 began the survey, and 1,002 completed it, at which point the survey was closed.9 This
partial coverage of the Israeli population and non-response among panel participants should
be kept in mind when generalizing from our survey findings.
The survey itself was based on the surveys used in Shapiro and Slemrod (2003), Shapiro
and Slemrod (2009) and Sahm, Shapiro and Slemrod (2012), translated and adapted to
9
The survey was promptly closed after four live days, upon reaching the requested target of 1,000 obser-
vations relatively quickly. While we speculate that response rate would have been somewhat higher had the
survey remained open longer, unit nonresponse would likely remain quite high. Past research shows that
easier-to-reach respondents may differ systematically from harder-to-reach ones, even after controlling for
sample composition on observables, suggesting non-random nonresponse (Heffetz and Reeves, 2019).
6
better fit the nature of the Israeli grant, society and Hebrew language.10
Table 1 presents demographic summary statistics for our respondents, and compares them
to national statistics. In spite of being relatively small, our sample is remarkably close to
national population statistics collected by the Central Bureau of Statistics (CBS) in terms of
age, religion, religiosity, and gender. While the income distribution of survey respondents is
difficult to directly compare to the national distribution due to different bracketing, roughly
63 percent are below the national median income and roughly 88 percent of respondents are
around or below the national 80th percentile, suggesting some undercoverage at the high
end, though arguably very reasonable for an internet survey. In terms of education, our
survey sample is significantly more educated than the Israeli population, possibly in part
due to the survey being offered only in Hebrew and only online.11
Average household size in our survey is close to the national average; however, conditional
upon having at least one child, our sample has substantially fewer children per household
compared to the national distribution.
Table 1’s bottom panel reports two variables derived from questions in our survey. Finan-
cially damaged: a Yes/No question asks “Because of the Coronavirus crisis, has your or your
household’s income declined?” Sixty-one percent responded affirmatively. Next, we define
Liquidity constrained: a Yes/No question asks “Did you and/or your household fall behind
on mortgage, rent or other payments due to the Coronavirus crisis?” Twenty-two percent
responded affirmatively. Finally, as a check for the salience of the grant, we also asked house-
holds to self-report the amount that they received. Respondents report 1,774 NIS (roughly
$520) on average. We use the government’s formula and respondents’ reported household
composition to predict how much they were supposed to get; however, our predicted amount
misses eligibility for the augmented grant because we do not have the required information
on the long list of criteria determining eligibility. This would lead us to under-predict the
10
Appendix A contains the English translation of the survey. The original Hebrew version is available
from the authors upon request.
11
We note that education was not one of the categories on which the survey company seeks to balance its
sample.
7
grant amount for some households. In addition, there could be some confusion in family sit-
uations of divorced or separated parents. Nonetheless, sixty percent of respondents provide
a figure that exactly matches our predicted amount. The discrepancies between what our
implementation of the formula predicts and what survey respondents claim increase with the
number of children in the household.
3 Results
Of the 1,002 respondents who completed the survey, four reported they were not eligible
for the grant; our analysis is based on the remaining 998 respondents. In our sample, 975
respondents reported having already confirmed receipt of the grant, 15 believed that they
had already received it but did not yet check their bank accounts, and eight had either not
received it yet or had not checked their bank accounts but believed they would receive it
soon.
Asked “How soon, in your estimation, will you / your household use or did use most (over
half) of the grant money?,” 30 percent responded “Even before actually receiving the grant,”
another 37 percent responded “Within two weeks of receiving the grant,” and another 24
percent indicated between two weeks and three months. Put differently, less than 9 percent
indicated that it will take them more than three months to use most of the grant.
3.1 How Israelis used the grant
Table 2 summarizes the survey results. We start with panel A, which reports response
frequencies and percentages. Columns 1–6 are based solely on our main survey question:
Compare the current situation in which you / your household received or will
receive the grant to a situation in which no grant existed.
What do you think is the grant’s primary effect on you and your household?
Respondents could choose one of seven responses, presented in this order: “Spend more
8
money,” “Increase saving,” “Reduce debt,” “Donate,” “Help a friend or family member,”
“Don’t know,” and “Other. Specify: .” All but two of the respondents who chose the
“Other” option provided an explanation.12
Columns 1–3 report the raw data: column 1 shows the by-response frequencies of the
998 responses; column 2 reports the same responses as percentages of the 804 responses
excluding “Don’t know” and “Other”; and column 3 reports them as percentages of the 685
responses that correspond to the first three options, in order to facilitate comparisons with
previous studies. Columns 2 and 3 thus implicitly redistribute the responses in the excluded
categories proportionally among the first six and three categories, respectively.
Columns 4–6 replicate columns 1–3 after we recoded 68 of the 76 “Other” responses, based
on the detail provided by the respondents (in an open textbox). Of course, such recoding
involves judgement. That said, we only recoded when we felt that the provided text was
sufficiently unambiguous. This recoding also resulted in a new category, “No/negligible
effect,” which did not originally exist as a response option but which 27 of the 76 “Other”
respondents wrote in (e.g., “no effect,” “did not have any effect”). As we will discuss in
Section 4, one may interpret these “No/negligible effect” responses as protests against the
government’s universal grant. Another 41 responses were recoded into existing categories
(e.g., “Mostly spend”). The remaining eight were empty or sufficiently ambiguous.13
12
The original Michigan Surveys of Consumers’ telephone question in Shapiro and Slemrod (2003), Shapiro
and Slemrod (2009) and Sahm, Shapiro and Slemrod (2010) read:
Thinking about your (family’s) financial situation this year, will the rebate lead you mostly to
increase spending, mostly to increase saving, or mostly to pay off debt?
See Shapiro and Slemrod (2003), section “4.3. Converting Survey Responses into an Aggregate MPC ” for the
considerations behind the design of this question (as opposed to asking, e.g., “What fraction of the rebate
did you spend?”) for the purpose of MPC estimation. We modified the original question slightly to better
fit the Israeli context and Hebrew language, and we added the response option “Other. Specifiy:,” which did
not exist in the original script, due to our survey mode. (A telephone interviewer can ask for clarification,
and classify in real time would-be “Other” responses into one of the other six response options.) For a similar
reason, while the original telephone question explicitly mentions only the first three options—spend, save,
and pay off debt—and allows for the other three options—donate, help, and don’t know—only if volunteered
by the respondent, in our online survey all response options are explicitly listed.
13
We follow previous work in recoding “pay bills” as paying off debt. However, as noted e.g. by Sahm,
Shapiro and Slemrod (2010) and Leigh (2012), the issue is not clear cut. On the one hand, the bills themselves
may have increased due to increased consumption as a result of the grant, and/or paying them with the
9
Our findings in columns 1–6 show both similarities with, and differences from, findings
from telephone interviews in the U.S. When looking at the only three response options
explicitly mentioned by the interviewer in the US surveys—spend, save, and pay off debt—
the percentages in columns 3 and 6 closely replicate US percentages. To see this, Table 3
compares the shares from column 6 in Table 2 with findings from all other past studies we
are aware of. As the columns “Mostly spend,” “Mostly save,” and “Mostly pay off debt”
show, across studies and US episodes, the three shares are almost always in the respective
ranges 15–32, 17–37, and 46–55 percent.14 The shares we find in Israel are within these
ranges, with Mostly spend at the top of the range and Mostly save at its bottom.
Interestingly, our findings in Israel are almost identical to the CEX findings in Parker
and Souleles (2019), whose calibration methodology we use below for estimating marginal
propensities to consume. Studying the 2008 episode in the US, they find that 32, 17, and 51
percent of their respondents report mostly spending, saving, and paying off debt, respectively,
compared with our 31, 18, and 51 percent (or, in our Table 2 column 3, 29, 18, and 52
percent).
At the same time, when looking at the full distribution of responses in Table 2, we find
in columns 1 and 4 more donate, help, don’t know, and other than in the US surveys. While
likely at least partly a result of the differences in survey mode (see footnote 12), the high
shares of donate and help friends or family (7 percent each in columns 2 and 5) and of
the new category no/negligible effect (3 percent in column 5) may be unique to the Israeli
context. We return to this point below.
Finally, columns 7–9 provide estimates that also include delayed spending from the grant.
They are based on columns 4–6, but additionally recode some of the “Mostly save” and
“Mostly pay off debt” responses as “Mostly spend” responses, based on two follow-up ques-
grant may have released other funds, increasing spending. On the other hand, if the bills had already been
outstanding, then paying them is paying down debt. We also recode “routine payments” as paying off debt,
but we recode “routine expenditures” as increased spending. Among the eight responses that we did not
recode are “help in lieu of income,” “I did not understand the question,” and “peace of mind.”
14
Leigh’s 2012 study of the Australian episode stands out with more Mostly spend and less Mostly pay
off debt).
10
tions. Specifically, those who responded “Increasing saving” were asked:
Do you / your household plan to use the money saved, including the grant money,
by the end of the year?
Similarly, those reporting “Reducing debt” were asked:
As a result of reducing your debt with the assistance of the grant money – do
you think that you will spend more money later this year compared to what you
would have spent without the grant?15
Those who responded positively to (one of) these questions were moved in column (7) to
“Mostly spend” relative to column (4). This recoding only affected the top three rows of the
table. The resulting estimates of delayed behavior are thus more reliant on reported planned
future behavior—albeit near future, as the end of the year was only four months away.
These estimates also rely on the assumption that respondents interpreted the two follow-up
questions as again asking them—like the main survey question—about the primary effect,
although these follow-up questions did not explicitly repeat that language.16
3.2 Estimating marginal propensities to spend
Panel B of Table 2 translates the estimated percentages of mostly spenders from panel A
into estimates of a marginal propensity to consume (MPC) out of the grant. Its first row
15
The original Michigan Surveys of Consumers questions read:
Do you plan to keep your savings higher until the end of this year? Or do you plan to spend
the rebate later this year?
Do you plan to keep your debt paid off until the end of this year? Or do you plan to spend the
rebate later this year?
We again slightly modified the original text to better fit into the Israeli context, language and interview
mode.
16
We note that the 11 respondents who in the main survey question responded “Other” (column 1) and
were recoded by us into the saving/debt categories (column 3) were not asked one of the two follow-up
questions. We also note that some respondents may have interpreted the follow-up questions as referring
to the end of the Jewish year, which was on September 19, i.e., only three weeks into the future. However,
other respondents may have interpreted “later this year” more generally as “later on.”
11
(“Input from A: Share mostly spend”) reproduces these percentages from panel A, for ease
of reference. These shares are the only input into the next two rows, which turn them into
MPC estimates using two different methods, both taken from Parker and Souleles (2019).17
The first method follows the literature on reported spending, and was originally used by
Shapiro and Slemrod (2003) to estimate the MPC out of the US tax rebate of 2001. While
relying on arguably arbitrary—yet not unreasonable—assumptions, this method has become
somewhat of a standard: it was again used by Shapiro and Slemrod (2009), Sahm, Shapiro
and Slemrod (2010) to estimate the MPC out of the 2008 stimulus package, and by Leigh
(2012) for the Australian 2009 rebate. The method is explained in detail in Shapiro and
Slemrod (2003, Appendix), and again in Parker and Souleles (2019, online Appendix A.7)—
whose version we implement. Briefly, it relies on two assumptions: (i) that those reporting
mostly spending spend at least one-half of the transfer amount; and (ii) that the density
function of propensities is piecewise linear, starting at a height of A at zero propensity,
increasing to a maximum height of B at the share of respondents reporting mostly spending,
and decreasing to zero at propensity one. Given (i) and that the density must integrate to
one, one can solve for A and B as a function of the share who mostly spend, thus uniquely
determining the density and, with it, average propensity in the population.
The second method used by Parker and Souleles (2019) starts with the first method but
modifies its assumptions in a way that results in a simple calculation, equivalent to assuming
that the average spending propensity of households that report mostly spending is 80 percent,
and that of all other households is 25 percent. Parker and Souleles get the first assumption
from a simple calibration they make (see their online Appendix A.7); alternatively, one
trivially gets it by assuming that propensity among mostly spenders is uniformly distributed
between 50 and 110 percent.18 They get the second assumption by explicitly assuming that
17
Parker and Souleles (2019) use three MPC calibration methods. We use the first two; the third is based
on dollar amounts that people report spending in one of their surveys, and that we did not elicit.
18
Parker and Souleles (2019) allow for the the spending propensity of some mostly spenders to be higher
than 100 percent “to account for the possibility that some households purchase durable goods and so spend
more than the payment amount.” Indeed, their estimates using the CEX expenditure data suggest that
mostly spenders spent on average 122 percent of the payment during the three-month period following
12
propensity among mostly non-spenders is uniformly distributed between 0 and 50 percent.
Panel B shows that these two methods yield MPCs that are broadly in line with each
other, though the Shapiro and Slemrod (2003) estimates are lower in all columns. The
estimates range from 28 to 42 percent in columns 1–6, and from 40 to 50 percent in columns
7–9.
Panel B’s bottom line uses a third MPC-calibration method for columns 3, 6, and 9.
Recall that the above two methods take a single input from panel A—the share of mostly
spenders—and plug it into assumed distribution functional forms from the literature to
calculate population-average MPC. In contrast, this third method takes also the shares of
mostly debt payers and mostly savers as additional inputs, and combines them with average
MPC estimates for these three groups—84, 42, and 21 percent, respectively—from Coibion
et al.ś (2020, Table 3) study of the US 2020 stimulus grants, to calculate population-average
MPCs. This method is thus based on estimates that we do not have for Israel; it results
in MPCs that are 8–13 percentage points higher than the Shapiro-Slemrod and the Parker-
Souleles methods. We report them mostly to facilitate comparing our findings with those of
Coibion, Gorodnichenko and Weber (2020) in Table 3. As far as we know, that study is the
only one estimating MPCs during COVID-19 using comparable survey methodology.19
3.3 How spenders spent and debtors paid off debt
The top panel of Figure 1 breaks down how “Mostly spend” respondents and those who
report eventually spending (column 7 of Table 2) report spending or intending to spend the
money. Respondents were shown a list of spending categories and asked to mark all relevant
ones; thus, the figure sums to more than 100 percent. Food (and nonalcoholic beverages)
payment receipt; see their Table 3.
19
Confidence intervals around our MPC estimates are straightforward to calculate. Treating the “Mostly
spend” percentages in each column as estimated means from binomial distributions (with their column’s
respective N as number of trials), their estimated standard errors range from 1.3 percent (column 1) to 1.8
percent (column 9). Similarly, the standard errors around the Parker and Souleles (2019) MPC estimates, for
example, which are each a linear function of its column’s “Mostly spend” percent, have an almost identical
range.
13
was marked by 55 percent of respondents. By the time the grants had been distributed to
households, much of the economy had already reopened, so this high share of spending on
food is not due to supermarkets having been the only stores open.
The bottom panel breaks down which debts were paid by “Mostly pay down debt” re-
spondents. Again, because respondents were shown a list of categories and asked to choose
all that apply, the sum of the categories is greater than 100 percent. The main use of the
money was for paying down bank-overdraft debt.20 Since the government transferred the
grants directly into people’s checking accounts, the transfers automatically paid down such
debt for anyone who had a negative balance (without taking any further action).
Finally, we note that “Mostly spend” and “Mostly pay off debt” respondents reported
using most (over half) of the grant faster than all other respondents: 90 and 95 percent,
respectively, indicated use within a month of receiving the grant.
3.4 Heterogeneity in grant use
Figure 2 presents further breakdowns of our baseline findings from column 4 of Table 2. Its
top-left panel presents grant use above and below median income. “Mostly spend” rates are
nearly identical, at 23 percent (rounded)—a recurring result that we return to in the next
subsection. The main difference between households above and below the median is that the
fraction who pay down debt is substantially lower among the former: 25 versus 44 percent.
In contrast, above-median households are more likely to save (17 versus 11 percent) and
twice as likely to donate or give away the grant to family or friends (18 versus 9 percent).
Finally, these same households are also more than twice as likely to respond that the grant
had no/negligible effect (4 versus 2 percent).
The top-right panel breaks down grant use by age. Middle-aged respondents (36–47
20
Traditional credit cards do not exist in Israel whereby people can hold credit card debt. Credits cards
in Israel are essentially debit cards that are paid off every month. As a result, most bank accounts come
with the ability to have a negative balance and interest is paid on this balance. In a 2018 survey by Israel’s
Central Bureau of Statistics (2019), 42 percent of households who owned a bank account reported having a
negative balance (overdraft) for at least one of the previous twelve months, and roughly half of those for at
least ten months.(Central Bureau of Statistics, 2018)
14
years old) have the largest share of “Mostly spend,” while those aged 60 or above have the
smallest share (29 versus 17 percent). In addition, and consistent with a theory of Ricardian
equivalence, 28 percent of the 60+ age group engage in inter-generational transfers, either
by donating their grant or by giving it to family or friends. This is roughly three to four
times higher compared with other age categories. Young respondents (18–28) are the most
likely to respond “Don’t know/Other” (19 percent).
The bottom-left panel breaks down grant use by an indicator of financial damage (see
Table 1). Households who report no financial damage due to the pandemic are, like those of
higher income, much less likely to pay off debt (26 versus 45 percent) and more than twice
as likely to donate and help family or friends (19 versus 8 percent), suggesting a rerouting of
the grant from those who need it less to those who may need it more. Somewhat deviating
from the income figures, no-financial-damage respondents are a little less likely to “Mostly
spend” (20 versus 25 percent).
Finally, the bottom-right panel breaks down grant use by those who are liquidity con-
strained and those who are not. Again, the biggest difference between these two groups is
in the fraction that choose to mostly pay down debt (54 versus 32 percent). Non-liquidity-
constrained households are also much more likely to give away their grant to others (14
versus 6 percent). Surprisingly, again there is virtually no difference in the fraction that
report mostly spending.
While the simple breakdowns reported in Figure 2 are informative, they mask correlations
across households in income, age, financial damage, and other variables such as education,
religion, religiosity, and having children. In the next subsection we use a multinomial re-
gression model to analyze the determinants of grant use in a framework that simultaneously
controls for these variables.
15
3.5 Determinants of grant use
Table 4 reports the average marginal effects from a multivariate multinomial logistic regres-
sion (robust standard errors are in parentheses). The dependent variable is reported grant
use, coded as in Table 2’s column 4; however the eight “Other” responses that we could not
recode are assigned to the “Don’t know” category, yielding seven dependent-variable cate-
gories. The independent variables are those reported in the table and, in addition, geographic
region (nine regions) and household size (with a top category of seven or more).
To deal with multiple hypothesis testing, the discussion in the rest of this section focuses
on (all) “statistically significant” results, using a false discovery rate (FDR) threshold of 10
percent (Benjamini and Hochberg, 1995). Rather than controlling the Type I error rate for
each coefficient in isolation, the FDR threshold ensures that the null hypothesis is true (in
expectation) for at most 10 percent of the entire set of coefficients we refer to as significant
throughout the discussion. We look at the impact of each of our independent variables on
the probabilities of choosing among the possible ways to use the grant.
Income. Perhaps surprisingly, even in a regression framework with a long list of controls,
the percent of subjects who mostly spent the grant is fairly stable across the income cate-
gories. However, this (non)result in fact replicates past studies.21 It has been interpreted as
casting doubt on the conventional wisdom—originating from canonical theory and the liter-
ature on consumption smoothing—that the MPC is higher at lower incomes. Higher income
households are more likely to donate compared to the base category, by 10–11 percentage
points (pp)—more than doubling the baseline share of 6 percent from column 4 of Table 2
for the 17,001-20,000 and 20,001+ NIS categories. The highest income category is also less
likely to pay down debt and more likely to save (18 and 13 pp, respectively).
21
Shapiro and Slemrod (2003) use three questions to assess liquidity constraints, Shapiro and Slemrod
(2009) use income, Sahm, Shapiro and Slemrod (2010) use income and stock ownership, and Leigh (2012)
uses income. None of them finds a significant relationship with the share of mostly spenders. Parker and
Souleles (2019) and Coibion, Gorodnichenko and Weber (2020) use several measures and find mixed results.
As they discuss, their mixed results, as well as differences across studies, could result from measurement
issues, including those related to survey methodology, e.g., differences in questions, samples, and the response
behavior of high- versus low-income households.
16
Age. Households in the 48–59 age group are more likely to pay down debt and less likely
to save (20 and 14 pp, respectively). These are both economically large impacts given the
baseline of 37 and 13 percent from Table 2. Households in the 60+ age category are also more
likely to pay down debt and less likely to save (18 and 15 pp, respectively). Finally, while
not statistically significant, the large point estimates suggest that these latter households
are roughly twice as likely to donate the grant or give it to family or friends compared to
any other age category.
Education. Education does not show much explanatory power across the various cat-
egories of grant use. With a single exception, no coefficient is statistically significant even
by old t-test standards. The exception is that respondents with at least a Bachelor’s degree
are more likely to respond “No effect” (p-value = 0.01). While narrowly missing our FDR
threshold, the 2.7 pp estimate is economically significant: it means a doubling of the baseline
rate (of 2.7 percent).
Religion and children. Jewish households are nearly eight pp more likely to save.
Relative to secular respondents, traditionally observant ones are less likely to save and report
“No effect,” and more likely to spend (the only statistically significant predictor of “Mostly
spend”), while ultra-orthodox ones are less likely to save, give away the grant, and report
“No effect.” Families with children under the age of 18 are 14 pp likelier to pay down debt.
Financial damage and liquidity constraints. Households that report being finan-
cially damaged or liquidity constrained are more likely to pay down debt (12 and 14 pp,
respectively). Moreover, liquidity constrained households are seven pp less likely to save.
4 Discussion
Depending upon the underlying assumptions, we estimate a range of MPCs from 28 to 51
percent in the near term and up to 58 percent in the longer run. We cautiously conclude that
the government was modestly successful in achieving its goal of stimulating consumption, and
17
perhaps more than modestly when we take into account delayed spending. We also maintain
that these grants serve a larger purpose than simply stimulating aggregate demand. Paying
down debt, especially by financially distressed households, is, in our view, an important
outcome. It could, in addition, have future expansionary effects on aggregate demand.
These are hard to measure.
Table 3 shows that over all grant episodes studied, the fraction of those who mostly
pay down debt is remarkably consistent: it is around 50 percent, over time and economic
crisis, with the single exception of Leigh (2012), who estimates 35 percent for Australia’s
2009 episode. The heterogeneity across studies is seen primarily in the fraction who mostly
spend and the fraction who mostly save. We offer one possible insight into this observation.
As Coibion, Gorodnichenko and Weber (2020) note, one explanation for the low level of
spending during the 2020 stimulus grant episode in the US is that the grants were relatively
large compared to previous episodes and relative to median monthly household income in
the US.22 The larger these one-time payments get, diminishing returns induces individuals to
consume smaller fractions of their temporarily higher income. Coibion, Gorodnichenko and
Weber (2020) find 15 and 33 percent, respectively, of mostly spenders and mostly savers;
in comparison, we find roughly twice the share of spenders and half the share of savers
in Israel’s COVID-19 episode. One potential explanation is a mental accounting argument
along the lines of Shefrin and Thaler (1988) and Feldman (2010). It maintains that larger
payments may enter the “wealth” mental account—and therefore generally face a higher
marginal propensity to save—and smaller payments may enter the “consumption” mental
account—where they are meant to be spent.
There is however an important difference in the Israeli context compared to other grant
episodes: the Israeli grant was, for all practical purposes, universal (excluding the top few
percent of the wealthiest taxpayers). The government had effectively decided that speed of
22
Median household income in the US was $68,700 in 2019, so a $3,400 payment for a family of four is
equal to nearly 60 percent of median monthly income. In comparison, the payment in Israel was nearly 18
percent of median monthly income in 2018.
18
delivery was more important than more careful targeting. Furthermore, the government may
have lacked the information and the tools necessary to conduct effective targeting.
With this universality came political backlash. However, our results suggest that some
of the protest behavior itself may have become part of the solution. While the biggest
disadvantage of universal transfers is that many people who “don’t need” the money get
it, our results suggest that many of them redirected it towards people they believed to
be of higher need. To give a concrete example, an online crowdfunding campaign called
“Move it Forward” collected nearly 18.5 million NIS from 21,600 donors—about 850 NIS per
donation, on average. The express purpose of this and similar campaigns was to encourage
people to donate their grants to those more needy (see Figure 3). Thus, while not a typical
argument in such a situation, it may well be possible under some cultural, social, and political
circumstances to send money quickly to everybody, trusting to some extent that people will
generously move it along to where it is most needed. Individuals may have better relevant
information than the government, especially when speed is of the essence and given the
asymmetric-information limitations governments inherently face.
Our results suggest that many recipients of the grant took matters into their own hands,
effectively creating a decentralized grant-allocation mechanism. Fourteen percent of house-
holds claimed to have given their grants away, a response that is not typically studied in
previous grant episodes in the US.23 In particular, higher income households were more likely
to mostly donate their grants, while 28 percent of those above age 60 either mostly donated
their grant or gave it to family or friends. These actions of the older population are in line
with the theory of Ricardian equivalence if they expect that future generations will face
higher taxation as a result of the grant—again, an argument that was frequently made in
Israel at the time.
23
We do not suggest that donations are unique to Israel. They were made particularly salient during the
2020 episode by the vocal opposition to the universality of the grant (and, more generally, to the “emergency”
national-unity government at the time). In addition, as discussed above, because “Donate” and “Help a friend
or family member” were not explicitly offered as a response option in the surveys administered in the US,
we are unable to offer a clean comparison. We encourage future iterations of such surveys to offer a more
complete picture of the available options for using direct government grants.
19
Finally, we have noted what we call “protest” responses—those who claim “No/negligible
effect.” We find it remarkable that income is not a statistically significant determinant of
such responses; it suggests against interpreting them simply as reflecting the fact that a fixed-
size grant matters less for high-income individuals. Short of burning the money, respondents
must have done something with the grant but choose instead to provide a response that, to
us, seems to suggest that they disliked this particular fiscal policy. The demographics that
most strongly predict such protest are being a secular Jew with higher education—a group
that was less likely to support the government at the time.24 This finding is an important
reminder that when certain groups are strongly opposed to a particular fiscal policy, that
itself may hinder the policy’s goals—but it may also diminish the policy effects as measured in
surveys. In such cases, alternative to surveys—which, as we noted, are becoming increasingly
viable—may be particularly important.
5 Conclusion
In response to the COVID-19 pandemic that hit the Israeli economy—along with many other
economies around the globe—the Israeli government disbursed one-time universal grants of
750 NIS (about $220) per adult and 500 NIS (about $150) per child to resident citizens.
The time from the government’s first announcement of its intention to provide the grants
to their deposit in bank accounts was barely three weeks. The stated goal of the grants
was to stimulate the domestic economy. There was significant public backlash against the
grants, largely due to their essential universality (i.e., their delivery also to those who did not
necessarily need them) and to their being commonly perceived among certain constituencies
as simply adding to a view that Israel’s fiscal policies at the time were haphazard and
24
Survey data from the Israel National Election Studies (2019-2020) administered around the September
2019 elections show that 50 percent of secular Jewish respondents ranked their support for the Prime Minister
at 2 or below, on a scale from 0 (rejection) to 10 (support)—compared with only 13 percent of non-secular
Jewish respondents. The data also show a lack of support among higher-educated individuals (with at least
a B.A. degree), albeit less dramatic. The sharpest difference is seen at the top of the scale: 19 and 7 percent
of lower- and higher-educated respondents, respectively, gave the Prime Minister a full 10.
20
inadequate to address the needs of those who had been suffering from financial damage due
to the pandemic.
Using data collected from an internet survey in the weeks after the grants had been
disbursed, and our preferred classification of responses (column 5, Table 2), we find that
42 percent of households reported using the grants to mostly pay down debt, 26 percent to
mostly spend it, 15 percent intended to mostly save the grant, 14 percent reported to have
either donated the money to a third party or given it to family or friends and 3 percent
claimed no effect. We find these latter two categories particularly interesting. We speculate
that they may suggest, first, that under some circumstances, individuals may be able to target
payments to those who are truly needy more efficiently and quickly than the government; and
second, that estimating the true impact of such grants with a survey may be problematic
when respondents are politically opposed to the policy at hand. Future research could
investigate both conjectures, e.g., by using other, newly available data to supplement survey
data. In addition, future surveys could explicitly invite respondents to express a wider range
of possible outcomes than those suggested by narrowly interpreting economic theory.
References
Baker, Scott, R.A. Farrokhnia, Steffen Meyer, Michaela Pagel, and Constantine
Yannelis. 2020. “Income, Liquidity, and the Consumption Response to the 2020 Economic
Stimulus Payments.” NBER WP 27097.
Benjamini, Yoav, and Yosef Hochberg. 1995. “Controlling the False Discovery Rate:
A Practical and Powerful Approach to Multiple Testing.” Journal of the Royal Statistical
Society Series B, 57(1): 289–300.
Central Bureau of Statistics, CBS. 2018. “Ownership of Financial Assets by Households:
Findings from the Longitudinal Survey 2018.”
Coibion, Oliver, Yuriy Gorodnichenko, and Michael Weber. 2020. “How did U.S.
Consumers Use their Stimulus Payments?” NBER WP 27693.
Feldman, Naomi. 2010. “Mental Accounting Effects of Income Tax Shifting.” Review of
Economics and Statistics, 92: 70–86.
21
Guerrieri, Veronica, Guido Lorenzoni, Ludwig Straub, and Iván Werning. 2020.
“Macroeconomic Implications of COVID-19: Can Negative Supply Shocks Cause Demand
Shortages?” NBER WP 26918.
Heffetz, Ori, and Daniel B. Reeves. 2019. “Difficulty to Reach Respondents and Non-
response Bias: Evidence from Large Government Surveys.” Review of Economics and
Statistics, 101: 176–191.
Israel National Election Studies. 2019-2020. “Election Studies Survey.” https://www.
tau.ac.il/~ines/elections.html.
Leigh, Andrew. 2012. “How Much Did the 2009 Australian Fiscal Stimulus Boost Demand?
Evidence from Household-Reported Spending Effects.” The B.E. Journal of Macroeco-
nomics: Contributions, 12.
Misra, Kanishka, Vishal Singh, and Qianyun Zhang. 2020. “Impact of the CARES
Act Stimulus Payments on Consumption.” Working Paper.
Parker, Jonathan A., and Nicholas S. Souleles. 2019. “Reported Effects versus
Revealed-Preference Estimates: Evidence from the Propensity to Spend Tax Rebates.”
American Economic Review-Insights, 1(3): 273–290.
Sahm, Claudia, Matthew D. Shapiro, and Joel Slemrod. 2010. “Household Response
to the 2008 Tax Rebates: Survey Evidence and Aggregate Implications.” Tax Policy and
the Economy, , (24): 69–110.
Sahm, Claudia, Matthew D. Shapiro, and Joel Slemrod. 2012. “Check in the Mail
or More in the Paycheck: Does the Effectiveness of Fiscal Stimulus Depend on How It Is
Delivered?” American Economic Journal: Economic Policy, 4(3): 216–50.
Shapiro, Matthew D., and Joel Slemrod. 2003. “Did the 2001 Tax Rebate Stimulate
Spending? Evidence from Taxpayer Surveys.” Tax Policy and the Economy, , (17): 83–109.
Shapiro, Matthew D., and Joel Slemrod. 2009. “Did the 2008 Tax Rebates Stimulate
Spending?” American Economic Review, 99(2): 374–379.
Shefrin, Hersh, and Richard Thaler. 1988. “The Behavioral Life Cycle Hypothesis.”
Economic Inquiry, 26: 609–643.
22
Table 1: Summary Statistics: Percent of Households
Survey Sample National Census
Male 49 50
Age
18-29 25 26
30-39 23 20
40-49 16 17
50-59 17 14
60+ 19 23
Religion
Jewish 80 75
Muslim 14 18
Christian 3 2
Druze 2 2
Other 1 4
Religiousity
Secular 47 43
Traditional 37 33
Religious 11 17
Ultra-orthodox 5 7
Income (NIS)
less than 5,000 14
5,001-8,000 17
8,001-11,000 17
11,001-14,000 15
14,001-17,000 13
17,001-20,000 12
20,001+ 12
50th perc. (˜14,000 in 2018) 63 50
80th perc. (˜20,000 in 2018) 88 80
Education
HS degree and below 26 56
BA student 7 3
Bachelor’s degree + 49 30
Other (e.g., trade school) 19 11
No. in household (avg) 3.0 (1.6) 3.3
Households with at least one child:
Under 18 40
Over 18 36
Children
1-2 73 37
3-4 23 43
5+ 4 20
= 1 if the household’s income has declined due
Financially damaged 61 to Covid-19, 0 otherwise. (Appendix: Survey
question 5)
= 1 if the household reports having falling behind
Liquidity constrained 22
on bills (Appendix: Survey question 8)
Household grant (avg NIS) 1,774 (1,017) household self-reported grant
Notes: Source: Central Bureau of Statistics. Children: Number of children is conditional upon having at least one child living
at home under the age of 17 (Census) or 18 (survey). Religiosity: in census, based upon Jewish and Arab population only.
Standard deviations in parentheses for averages (household size and grant amount); all other statistics are in percentages. N =
998.
23
Table 2: Responses to 2020 Universal-Transfer Survey
Recoded
Raw data Recoded + Delayed
(1) (2) (3) (4) (5) (6) (7) (8) (9)
A: Responses to Main Survey Question (# = Frequency, % = Percent)
# % % # % % # % %
Mostly spend 201 25 29 228 26 31 325 37 45
Mostly pay off debt 358 45 52 367 42 51 315 36 44
Mostly save 126 16 18 129 15 18 84 10 12
Mostly donate 59 7 61 7 61 7
Mostly help friend or family 60 7 60 7 60 7
No/negligible effect 0 0 27 3 27 3
Other 76 8 8
Don’t know 118 118 118
N 998 804 685 998 872 724 998 872 724
B: Calibrating Marginal Propensity to Consume (MPC; Percent)
Input from A:
Share mostly spend 20 25 29 23 26 31 33 37 45
Output: average MPC by method
Shapiro and Slemrod (2003) 28 33 37 31 35 39 40 44 48
Parker and Souleles (2019) 36 39 41 38 39 42 43 46 50
Average MPC using US estimates
Coibion et al. (2020)a 50 51 58
Notes: Percentages are rounded to the nearest whole number. Marginal propensity to consume (MPC),
or to spend, out of grant: Shapiro-Slemrod calibration is based on the Parker-Souleles implementation of
Shapiro-Slemrod (see their Table 4 and Appendix A, with χ = 1); Parker-Souleles calibration is based on
their “Alternative calibration” (see their Table 4 and Appendix A); Coibion et al. calibration is based on
their MPC estimates for mostly spenders, savers, and debt payers (see their Table 3).
a
This method uses two additional inputs from A: share mostly pay off debt and share mostly save.
24
Table 3: Spending Propensities Across Countries, Episodes, and Studies
Study Country & Mostly Mostly M. pay Estimated
Dataset episode spend save off debt N MPC
Shapiro and Slemrod (2003)
MSCa Sep-Oct 2001 US 2001 21.8 32.0 46.2 1,444 33–36*
MSC Mar-Apr 2002 US 2001 24.9 27.1 48.0 1,002 34–37
HARb Sep-Oct 2001 Hypothetical 16.6 36.5 46.9 752
Shapiro and Slemrod (2009)
MSC Feb-June 2008 US 2008 19.9 31.8 48.2 2,245 33 (32–35*)
Sahm, Shapiro and Slemrod (2010)
MSC Feb-Dec 2008 US 2008 20 28 52 3,417
MSC Nov-Dec 2008 US 2008 22 23 55 990 33 (30–35)
Leigh (2012)
ANUc 2009 Australia 2009 40.5 24.0 35.5 805 41–42
CEXd 2008 US 2008 31.2 18.2 50.7 5,192 37–39
SIPPe 2008 US 2008 28.9 17.6 53.6 54,402 36–38
Parker and Souleles (2019)
CEX 2008 US 2008 32 17 51 4,076
Shapiro-Slemrod 40
Alternative calibration 42
Reported spending in NCP 58
NCPf 2008 US 2008 19 27 54 19,149
Shapiro-Slemrod 27
Alternative calibration 36
Reported spending in NCP 50
Coibion, Gorodnichenko and Weber (2020)
NCP June 2020 US 2020 15 33 52 11,793
Reported amounts (shares) [42] [27] [31] 42
The present paper
Aug 2020 (Table 2 column 6) Israel 2020 31 18 51 724
Shapiro-Slemrod 39
Alternative calibration 42
Coibion et al. calibration 51
Notes: With exception to Coibion et al. (2020) and where indicated under Parker and Souleles (2019)
and this present paper, all estimates use Shapiro-Slemrod calibration. Parker and Souleles (2019) use NCP
scanner data to calculate aggregate MPC, and combine it with survey answers (Reported spending in NCP
rows). Coibion et al. (2020) did not use NCP scanner data; in addition to answering what they mostly
used their rebate for, their survey respondents gave quantitative answers (reported dollar amounts) to how
much of the rebate they used for spending, saving and for paying off debt. The respective row—Reported
amounts (shares)—reports these as mean shares in brackets; they are not comparable with other entries in
their columns. The row’s estimated MPC is the share reported as spending. The present paper’s alternative
calibration uses that in Parker-Souleles, and its Coibion et al. calibration uses mean share expenditures from
Coibion et al. to calibrate the qualitative answers.
Notice that the data presented in Shapiro and Slemrod (2009) is a subset of the data in Sahm, Shapiro and
Slemrod (2010).
Estimates with * are given in Leigh (2012).
a
Michigan Surveys of Consumers
b
How America Responds Survey (University of Michigan Survey Research Center)
c
Australian National University, survey conducted by Social Research Center in Melbourne
d
Consumer Expenditure Survey: https://www.bls.gov/cex/taxrebate.htm
e
The Survey of Income and Program Participation
f
Nielsen Consumer Panel
25
Table 4: Determinants of Grants Use (Multinomial Logit–Avg. Marginal Effects)
Spend Pay debt Save Donate Family No Effect DK/Ot.
(1) (2) (3) (4) (5) (6) (7)
Income (NIS), Base = ≤ 5K
5–8K −0.021 0.004 0.016 0.004 0.011 −0.013 0.000
(0.054) (0.057) (0.032) (0.015) (0.032) (0.018) (0.037)
8–11K −0.097 0.065 0.011 0.030 0.014 −0.013 −0.010
(0.055) (0.060) (0.034) (0.20) (0.032) (0.020) (0.039)
11–14K −0.087 0.070 0.053 0.038 −0.026 −0.011 −0.059
(0.057) (0.061) (0.037) (0.021) (0.030) (0.025) (0.038)
14–17K −0.022 −0.147 0.064 0.058 −0.006 0.000 0.053
(0.064) (0.063) (0.042) (0.025) (0.036) (0.022) (0.047)
17–20K −0.065 −0.078 0.048 0.105 −0.002 0.009 −0.017
(0.066) (0.068) (0.043) (0.036) (0.037) (0.023) (0.046)
20K+ −0.062 −0.179 0.127 0.112 0.024 0.008 −0.028
(0.067) (0.065) (0.047) (0.039) (0.045) (0.026) (0.045)
Age, Base = 18–28 years
29–35 0.068 0.066 −0.068 0.005 −0.040 0.035 −0.068
(0.044) (0.044) (0.042) (0.021) (0.018) (0.022) (0.037)
36–47 0.100 −0.004 −0.100 0.023 −0.011 0.064 −0.072
(0.046) (0.044) (0.043) (0.025) (0.023) (0.031) (0.039)
48–59 0.009 0.204 −0.138 −0.013 −0.001 0.015 −0.076
(0.048) (0.053) (0.049) (0.021) (0.026) (0.018) (0.039)
60+ −0.057 0.178 −0.150 0.072 0.056 0.005 −0.104
(0.047) (0.056) (0.047) (0.036) (0.034) (0.013) (0.042)
Education, Base = ≤ HS graduate
Non−academic trade −0.085 0.045 0.018 −0.024 0.016 −0.002 0.031
(0.043) (0.046) (0.032) (0.027) (0.021) (0.009) (0.033)
Student −0.039 0.083 0.017 −0.007 −0.012 0.009 −0.050
(0.059) (0.066) (0.046) (0.039) (0.028) (0.019) (0.034)
Bachelors + −0.035 0.012 0.004 −0.037 0.023 0.027 0.006
(0.037 ) (0.038) (0.026) (0.023) (0.019) (0.010) (0.026)
Jewish 0.035 −0.013 0.075 −0.015 −0.032 0.022 −0.073
(0.0.4) (0.046) (0.024) (0.032) (0.028) (0.010) (0.043)
Religiousity, Base = Secular
Traditional 0.114 −0.005 −0.063 0.004 −0.008 −0.028 −0.012
(0.030) (0.033) (0.023) (0.017) (0.016) (0.011) (0.022)
Religious 0.022 0.026 −0.067 −0.001 0.005 −0.031 0.045
(0.044) (0.053) (0.032) (0.024) (0.029) (0.014) (0.038)
Ultra−Orthodox 0.146 −0.038 −0.012 −0.062 −0.066 −0.042 0.074
(0.074) (0.066) (0.058) (0.010) (0.011) (0.009) (0.056)
Kids over 18 0.038 −0.097 0.025 −0.040 0.014 0.045 0.014
(0.043) (0.045) (0.044) (0.024) (0.022) (0.034) (0.035)
Kids under 18 −0.013 0.142 −0.002 −0.045 −0.020 −0.016 −0.045
(0.046) (0.051) (0.036) (0.027) (0.025) (0.013) (0.033)
Financial damage 0.022 0.122 −0.037 −0.022 −0.026 −0.010 −0.049
(0.030) (0.032) (0.023) (0.016) (0.018) (0.012) (0.025)
Liquidity constraints −0.027 0.138 −0.073 0.000 −0.032 0.013 −0.019
(0.035) (0.042) (0.024) (0.021) (0.016) (0.020) (0.028)
Notes: All columns include binary indicators for geographic region and number of members in the household. Robust standard
errors in parentheses. N = 998.
26
Figure 1: Categories of Grant Use
(a) How the grant is spent (conditional upon spending)
(b) Debts paid (conditional upon paying down debt)
Notes: Fraction of respondents who report (a) spending on (N = 228) or (b) paying down debt on (N =
367). Total sums exceed 100 percent because respondents could choose as many categories as applied to
them.
27
Figure 2: Heterogeneity in grant use
Income Age
18 - 28
Above
Median 29 - 35
36 - 47
Below
Median 48 - 59
60+
0 20 40 60 80 100 0 20 40 60 80 100
Percent Percent
Financial damage Liquidity constraints
No No
Yes Yes
0 20 40 60 80 100 0 20 40 60 80 100
Percent Percent
Notes: See Table 1 for variable definitions. N = 998.
28
Figure 3: “Move It Forward” Campaign
Notes: https://giveback.co.il (Accessed on October 11, 2020.) Translation in top-right corner is provided
for the reader (not part of original webpage).
29
A Appendix: English Translation of Online Survey
[Survey starts on next page.]
30
A Grant to Every Citizen
During or after August and due to the Coronavirus crisis, most Israeli citizens received or will receive A
Grant to Every Citizen:
• NIS 750 to every citizen over the age of 18 who earns less than NIS 650,000 a year.
• An additional NIS 500 to parents for each child under the age of 18, up to four children, and an
additional NIS 300 for the fifth child and above.
• Enlarged grants for recipients of government benefits.
For the purpose of researching and evaluating the effectiveness of the grant program, we would like
to know if the grant funds helped you and how you used them.
The survey is written in masculine for convenience but is intended for all genders.
This survey is anonymous.
As part of the study, we ask of you to answer on behalf of your household. That is, if you live and
manage your finances in partnership with a spouse, and / or have children under the age of 18, please
answer on behalf of all of you together. If you live alone or live with others but manage your finances
separately from them, please answer on behalf of yourself only.
For this survey children over the age of 18 are not considered part of your household.
1. Have you and / or a member of your household received this grant?
o Yes
o No
o Don’t know
1.1. To those who answered “No”
Are you and / or someone in your household eligible for the grant and are expecting to receive it?
o I and / or someone in my household is eligible, and I believe that we will receive the grant
o I and / or someone in my household is eligible, but I do not believe that we will receive the
grant [GO TO QUESTION 5]
o I and / or someone in my household is eligible, but I do not know if we will receive the grant
o I do not know if I and / or someone in my household is eligible to receive the grant
o No one in my household is eligible [GO TO QUESTION 5]
1.2. To those who answered “I and/or … is eligible, but I do not know if we will receive the grant”
or “I do not know if I and/or … is eligible to receive the grant”
Regardless of your eligibility, do you think you will receive the grant?
o Yes
o No [GO TO QUESTION 5]
31
1.3. To those who answered “Don’t know”
A. Why don’t you know?
o Usually someone else in the household takes care of financial matters
o I have not checked my bank account recently
o I did not pay attention to whether the grant was received or not
o I do not know how to check whether my household received the grant
o Other [with textbox for open response]
B. Although you do not know for sure, do you think you and / or someone in your household
received the grant?
o I think that we received the grant
o I think that we did not receive the grant, but we will receive it in the future
o I think that we did not receive the grant, and we will not receive it in the future [GO TO
QUESTION 5]
2. To the best of your knowledge, how much money did you / your household receive or are going to
receive from the grant?
NIS __________________ [Open response, numbers only]
3. How soon, in your estimation, will you / your household use or did use most (over half) of the grant
money?
o Even before actually receiving the grant
o Within two weeks of receiving the grant
o Within a month of receiving the grant
o Within three months of receiving the grant
o Within more than three months of receiving the grant
o Only after the Coronavirus crisis is over
4. Compare the current situation in which you / your household received or will receive the grant to a
situation in which no grant existed.
What do you think is the grant’s primary effect on you and your household?
o Increased spending
o Increased saving
o Reduced debt
o Giving to charity
o Giving to friends or family
o Don’t know
o Other [with open textbox for explanation]
4.1. To those who answered “Increased saving”
32
Do you / your household plan to use the money saved, including the grant money, by the end of the
year?
o We plan to use the savings by the end of the year, including the grant money added to the
savings
o We plan NOT to use the savings by the end of the year, including the grant money added to
the savings
4.2. To those who answered “Reduced debt"
A. As a result of reducing your debt with the assistance of the grant money - do you think that you
will spend more money later this year compared to what you would have spent without the grant?
o Yes, we will spend more money because of the grant we received
o No, the grant we received will not cause us to spend more money
B. What is the nature of the debt that has been paid or will be paid with the assistance of the grant
funds?
o Debt that has accumulated due to the Coronavirus crisis
o Debt that is not related to the Coronavirus crisis
o Both debt that has accumulated due to the Coronavirus crisis and debt that is not related to
the Coronavirus crisis
C. We would be happy if you could elaborate, what debts were paid with the assistance of the grant
money? [Multiple choices can be selected]
o Negative bank account balance
o Bills (electricity, water, gas, etc…)
o Mortgage
o Rent
o Property tax
o Previous purchases being paid in monthly installments
o Other loan
o Other debt
4.3. To those who answered “Increased spending” in question 4, or “We plan to use the savings by
the end of the year…” in question 4.1 or “Yes, we will spend more money…” in question 4.2.A
A. You indicated that you / your household spent or will spend more money due to receiving the
grant. In what way?
o Mainly on the same things we would have spent money on regardless of the grant, but we
will now spend more money on them
o Mainly on things different from the things we would have spent money on without receiving
the grant
o Mainly on expenses unique to the Coronavirus crisis
B. What specifically did you / your household spend or spend more on after receiving the grant?
(Mark all correct answers) [Multiple choices can be selected]
33
o Food and non-alcoholic beverages
o Tobacco and alcoholic beverages
o Clothes and shoes
o Housing, electricity, fuel and gasoline
o Household equipment
o Health and medical services
o Transportation (public transportation, buying a vehicle, vehicle repairs, etc…)
o Communication (TV, internet or cell phone bills, etc…)
o Leisure and culture
o Education
o Restaurants, cafes and hotels
o Other
5. Because of the Coronavirus crisis, has your or your household’s income declined?
o Yes
o No
6. Because of the Coronavirus crisis, have you or someone in your household lost a job, been laid off,
been furloughed, or worked fewer hours? [Multiple choices can be selected]
o Yes, I or someone in my household lost a job or was laid off
o Yes, I or someone in my household was furloughed
o Yes, I or someone in my household owns a business or is self-employed and worked fewer hours
or lost income due to the crisis
o Yes, I or someone in my household is an employee and worked fewer hours or lost income due
to the crisis
o No
7. Do you and / or anyone in your household receive or have received unemployment benefits during
the Coronavirus crisis? (Mark all correct answers) [Multiple choices can be selected]
o Yes, starting from before the Coronavirus crisis
o Yes, during the beginning of the crisis
o Yes, I and / or someone in my household is currently receiving unemployment benefits
o No
8. Did you and / or your household fall behind on mortgage, rent or other payments due to the
Coronavirus crisis?
o Yes
o No
9. Did you and / or anyone in your household contract the Coronavirus?
o Yes
o No
34
[Demographic questions]
10. According to the definition of “household” in this survey, are there any other people included in your
household in addition to you? (You can select multiple choices) [Multiple choices can be selected]
o Spouse or significant other
o Children under the age of 18
o I am the only one included in my household
10.1 To those who answered “Children under the age of 18”
How many children under the age of 18 live in your household?
__________________ [Dropdown menu: 1…9, 10 or more]
11. Do you have children over the age of 18?
o Yes
o No
11.1 To those who answered “Yes”
A. How many children over the age of 18 do you have?
__________________ [Dropdown menu: 1…9, 10 or more]
B. Do your children over the age of 18 have their own bank account and manage their finances
separately from you?
o Yes
o No
C. Do you financially support your children that are above the age of 18? (By giving them a monthly
allowance, help with rent, living in your home, etc...)
o Yes
o No
o Some of my children yes, some of them no
D. How many of your children above the age of 18 live in your home?
__________________ [Dropdown menu: 0, 1…9, 10 or more]
12. What is your birth date?
__________________ [Dropdown menu: 2003 or later, 2002…1931, 1930 or earlier]
13. What is your gender?
o Man
o Woman
o Other
14. What is your area of residence?
35
o Northern Israel
o Haifa Metropolitan Area
o Tel Aviv Metropolitan Area
o The West Bank
o The Sharon Plains (Herzliya, Raanana, etc…)
o Jerusalem Metropolitan Area
o The Shfela (Beit Shemesh, Modi’in, Rehovot, etc…)
o South (Near Gaza Strip to Beer Sheva)
o South (South of Beer Sheva)
15. What is the highest level of education among the members of your household?
o High school or less (no diploma)
o High school Diploma
o Professional training (non-academic)
o Practical Engineering degree
o Undergraduate university student
o Bachelor’s degree
o Graduate degree
16. To which sector do you ascribe yourself?
o Jewish
o Arab – Muslim
o Arab – Christian
o Druze
o Other
17. How do you describe your level of religiosity?
o Secular
o Traditional
o Religious
o Ultra-Orthodox [only available to those who answered “Jewish” to question 16]
18. What is your marital status?
o Single
o Married
o Civil Union
o Divorced
o Widowed
19. To the best of your knowledge, what is the total monthly income level of your household (before
taxes, in NIS)?
o Up to 5,000
o 5,001 – 8,000
o 8,001 – 11,000
36
o 11,001 – 14,000
o 14,001 – 17,000
o 17,001 – 20,000
o 20,001 – 55,000
o Above 55,000
20. What is your native language?
o Hebrew
o Arabic
o English
o Russian
o Amharic
o French
o Other
37
File and source
- File
- a-grant-to-every-citizen-survey-evidence-of-the-impact-of-a-direct-government-payment-in-israel.pdf
- Size
- 596,925 bytes
- SHA-256
- bf89892284a8b0278c6859eb2829dc41cee599a78de43e1d8f4440c86efb1d88
- Our copy
- a-grant-to-every-citizen-survey-evidence-of-the-impact-of-a-direct-government-payment-in-israel.pdf
- Original
- www.nber.org