Preparing For A Pandemic Spending Dynamics And Panic Buying During The Covid 19 First Wave
Summary
Institute for Fiscal Studies working paper 20/34, Preparing for a pandemic: Spending dynamics and panic buying during the COVID-19 first wave, by Martin O'Connell, Áureo de Paula and Kate Smith. The paper uses Kantar FMCG Purchase Panel scanner data for a balanced panel of 17,093 households to study purchases in the run-up to the UK's national lockdown on March 23, 2020. It reports spending spikes for storable products, with demand increases concentrated in 30 of 138 product categories such as soap, soup, canned goods and dried pasta. The authors find the spike was driven mainly by more households than usual buying these products, with larger increases for higher socioeconomic status households, and conclude that per-transaction unit limits would likely have had only a modest effect. Appendix figures chart spending by socioeconomic group.
Summary drafted by a model from the document's text below and checked by script against that text before publication. It is a navigation aid, not a reading of what the document proves. Where AI is used
Cite this document
“Preparing For A Pandemic Spending Dynamics And Panic Buying During The Covid 19 First Wave.”
Original at doi.org. Archived at Pandemic Darlings, https://pandemicdarlings.com/source-documents/preparing-for-a-pandemic-spending-dynamics-and-panic-buying-during-the-2838ab4421/
, SHA-256 fb2b557dbf76d608aa3fa8914028ae47da599347a8448d168cfd984cf7b92fa6.Full text
lnstitute for Fiscal Studies
Martin O'Connell
Áureo de Paula
Kate Smith
Preparing for a
Working paper
pandemic:
spending
dynamics and
panic buying
during the
COVID-19 first
20/34 wave
Preparing for a pandemic: Spending dynamics and
panic buying during the COVID-19 first wave
Martin O’Connell, Áureo de Paula and Kate Smith∗
October, 2020
Abstract
We study consumer spending dynamics during the first infection wave of the COVID-19
pandemic using household scanner data covering fast-moving consumer goods in the United
Kingdom. We document a large spike in spending for storable products, such as food staples
and household supplies, in the days before lockdown. Demand increases were concentrated
in 30 of 138 product categories, e.g. soap, soup, canned goods and dried pasta. Households
in all socioeconomic groups exhibit unusually high demand pre-lockdown, but there is a clear
gradient, with the largest demand spikes for wealthier households. Although stories of people
purchasing extreme amounts received a lot of attention, higher aggregate demand was mainly
driven by more households than usual choosing to buy storable products, with only small
increases in average quantities bought on a given trip. Temporary limits on the number of
units per transaction, introduced following the demand spike, are therefore unlikely to lead
to the avoidance of stock-outs. Given rapidly increasing case numbers in the ongoing second
wave, and the spectre of further national lockdowns, our work provides timely evidence for
preparing for a future demand spike.
Keywords: panic buying, hoarding, coronavirus
JEL classification: D12, D18, I30
Acknowledgements: The authors would like to gratefully acknowledge financial support
from the Nuffield Foundation under grant number WEL /FR-000022585 and the Economic
and Social Research Council (ESRC) under grant number ES/V003968/1, under the Centre
for the Microeconomic Analysis of Public Policy (CPP), grant number ES/M010147/1 and
under ESRC Centre for Microdata Methods and Practice (CeMMAP), grant number RES-
589-28-0001. Data supplied by Kantar FMCG Purchase Panel. The use of Kantar FMCG
Purchase Panel data in this work does not imply the endorsement of Kantar FMCG Purchase
Panel in relation to the interpretation or analysis of the data. All errors and omissions remain
the responsibility of the authors. Correspondence: martin o@ifs.org.uk, a.paula@ucl.ac.uk,
kate s@ifs.org.uk
∗
Institute for Fiscal Studies and University College London.
1 Introduction
Across the world countries have responded to the COVID-19 pandemic with the
introduction of restrictions on mobility, travel and social contact. In the run-up to
the adoption of these measures there were widespread reports of people engaging
in panic buying, with resulting shortages raising concerns about the sustainability
of the food system and people’s ability to access the products they need.1 Google
search data indicates sharp increases in searches for terms such as “panic buying”
and for necessities such as “toilet paper” across many countries (Keane and Neal
(2020)), and several papers that use financial transaction data to track the effects
of the pandemic on consumer spending have highlighted a pre-restriction spike in
grocery spending.2 A limitation of these data is that they are not able to tell us the
types of products that were hoarded, nor the changes in purchasing behavior that
led to demand spikes, and we therefore know little about the nature and extent of
panic buying in the early phase of the pandemic. In this paper, we are, to the best of
our knowledge, the first to use household level scanner data to document purchase
dynamics during the pandemic. Using data for the United Kingdom we show that
there were substantial spikes in demand for storable products, and this was primarily
driven by many more households than usual choosing to buy these products in the
run-up to lockdown, with only small increases in the average quantities bought per
transaction.
Rapidly rising COVID cases and the expectation of curbs on movement in the
early stage of the pandemic provided a rationale for consumers to build up their
stocks as a precautionary measure against, for instance, the need to quarantine or
significant supply disruptions. In addition, knowledge that other consumers were in
a similar position provided an additional incentive to purchase before stock ran out.
In the UK, unusually high demand for essentials led to calls from senior politicians
for people to behave responsibly,3 to supermarkets imposing limits on the quantity
of individual products that consumers were permitted to buy per visit, and to
the UK competition authority relaxing competition rules.4 Resurgent COVID case
numbers and the prospect of more stay-at-home orders mean that further bouts of
hoarding are a distinct possibility. For policy to adequately respond to this it is
important we understand the nature of hoarding prior to the first lockdown.
1
For instance, see USA Today in the US and Financial Times in the UK.
2
For example, see Baker et al. (2020a) and Cox et al. (2020) for the US, and Hacioglu et al.
(2020) for the UK.
3
On March 21, the UK Secretary of State for Environment, Food and Rural Affairs urged
consumers to “Be responsible when you shop and think of others” – see Reuters.
4
See UK Government.
1
We document purchase dynamics using household level scanner data, covering
fast-moving consumer goods. We use data on the purchases of a nationally repre-
sentative panel of 17,000 households. Each participant records all purchases they
make and bring into the home at the UPC (barcode) level. These data have a num-
ber of key advantages for measuring hoarding behavior over other real-time datasets
covering consumer spending. First, we observe the disaggregate products that con-
sumers purchase. This allows us to document precisely which product categories
drove the surge in demand. Second, the data contain information on quantities and
number of packs bought (as well as expenditure). Therefore, we are able to track
purchase incidence, pack size and number of units at the daily frequency. This
enables us assess the bite of limits placed on the number of units that consumers
are allowed to buy at any one time. Third, the long purchase history for households
allows us to compare their behavior during the pandemic with normal times. The
data cover the subset of consumer goods (roughly groceries and household supplies)
among which reports of panic buying was concentrated. Using this dataset we es-
tablish four sets of results regarding consumer purchase dynamics on the run-up to
the UK’s national lockdown on March 23, 2020.
First, we show that there were large spikes in spending on storable products in
the four weeks preceding lockdown. Spending on staples (such as canned goods,
pasta, rice and grains) rose sharply at the end of February, peaking on March 14 at
over 80% of the January–February daily average. A similar pattern is evident for
non-food household supplies (such as soap, cold treatments, and toilet tissue), with
demand peaking at over 70% of January–February levels on March 14. Spending
on the remaining set of fast-moving consumer goods (discretionary calories and
perishable foods) increased much more gradually into March, and continued to
rise as the UK entered the lockdown period. We show that the spike in spending
in the run-up to lockdown on staples and household supplies was driven by most
households increasing their demands rather than by a small number of extreme
purchasers.
Second, we identify which product categories were the primary drivers of these
increases in spending and the extent to which quantity spikes at the category level
were driven by changes in purchase frequency (extensive margin), or in the average
quantity bought, conditional on purchasing (intensive margin). 30, out of a total
of 138, product categories experienced an increase in demand of more than 25%;
these categories account for 70% of the increased spending on household supplies
and 53% of the increase for staples. Across these categories, on average, 70% of
the demand increase was due to increased purchase frequency, and the remainder
2
was due to households buying larger amounts, conditional on purchasing. The
increase in purchase frequency was relatively more important for those categories
that experienced the largest increases in demand. We show that the increase at the
extensive margin was mainly driven by more households than normal choosing to
buy these categories once over the “hoarding period” preceding the lockdown in late
March, rather than an increase in the number of times purchasing households bought
them. We also show that, although there was a small increase in the number of trips
that households made to the store over this period, this increase was not enough to
explain the higher purchase frequency for storable categories. Instead, households
were more likely to buy storable products when they did visit a store. This suggests
that further spates of hoarding could still occur, even though households are now
making fewer trips than they did prior to the pandemic.5
Third, we show that across socioeconomic groups, there was a sharp increase in
the quantity purchased of all storable categories, underlining that hoarding was a
widespread phenomenon. However, the average increase was substantially bigger for
higher socioeconomic status households: those in the top group increased purchases
by 55% across the affected categories, compared with 30% for the bottom group.
We show that this pattern is entirely driven by differences along the extensive
margin, with higher SES households increasing their probability of buying by more
than lower SES households. Wealthier households were therefore able to build up
larger precautionary stocks than less well-off households. Cox et al. (2020) find
that, although high income households in the US experienced a greater decline
in spending following the introduction of restrictions, this was entirely driven by
differences in non-essential spending. Our analysis sheds light on the changes in
purchasing behavior that drove the spike in spending on essential groceries prior to
the introduction of restrictions that Cox et al. (2020) (and others) have documented.
Fourth, we consider the likely impact of limiting the number of units of each
product that households are permitted to purchase on a visit to the store. These
limits were imposed by supermarkets following the demand spike just prior to the
UK’s lockdown. We show that it is unlikely that they would have prevented the
large spikes in demand seen in many categories. Had a limit of 2 units been in place
for the four weeks up to lockdown, and had households made no adjustment to their
behavior to circumvent the constraint, the average increase in quantity in the top
30 categories would have been 34% rather than 44%. Among the eight categories
that experienced spikes of more than 50%, the average quantity increase would have
5
We show that the number of shopping trips per four-week period dropped by 15%, relative
to the pre-pandemic period, following the beginning of lockdown.
3
been 55%, rather than 66%. This suggests that re-introducing limits may have only
a modest impact on preventing a further round of hoarding and shortages.
We build on and contribute to a fast growing literature that uses a range of real-
time datasets, including financial transaction, survey and publicly available data,
to document the impact of the pandemic on economic activity (e.g., Alexander and
Karger (2020), Andersen et al. (2020), Baker et al. (2020b), Bounie et al. (2020),
Carvalho et al. (2020), Chen et al. (2020), Chetty et al. (2020), Chronopoulos et al.
(2020), Coibion et al. (2020)). A contribution of this literature is to show how
aggregate expenditure, as well as spending in broad sectors of the economy, has
evolved over the pandemic, and how this varies across income groups. We comple-
ment this work by studying purchase dynamics in a sector of the economy that was
particularly hard hit by bouts of hoarding in the first phase of the pandemic. We
exploit granular, household-level scanner data to describe the changes in behavior
that drove the large spike in grocery spending prior to lockdown. This allows us
to provide valuable evidence on the nature of hoarding, and, notably, that it was
primarily driven by more people than usual buying a number of key categories,
rather than a small number of shoppers buying extreme amounts.
We also contribute to the wider literature that studies panic buying in other
settings, for example, in response to weather shocks and natural disasters. Con-
sistent with our results, Hori and Iwamoto (2014) find that the hoarding following
the 2011 earthquake in Tohoku, Japan, was primarily due to an increase in the
share of people buying. Hansman et al. (2020) study hoarding behavior during the
2008 Global Rice Crisis, highlighting the role of sticky prices as a motivating fac-
tor for households to stock-up. Like us, they also find that hoarding behavior was
more prevalent among richer households. Croson et al. (2014) study the motiva-
tions for hoarding in an experimental study of “the beer distribution game”, which
removes the incentives to hoard due to price expectations, demand uncertainty and
horizontal competitive effects. They find that, despite there existing no rational
motivations to hoard, there is considerable over-ordering; their results are robust
to excluding a number of outliers that reflect drastic increases in ordering. A key
finding of our analysis is that a lot of people buying a little bit more can lead to
substantial increases in aggregate demand and subsequent shortages.
The paper is structured as follows. In Section 2 we outline the dataset and in
Section 3 we document spending dynamics across broad sets of products. In Section
4 we explore how widespread hoarding was across disaggregate product categories
and households. In Section 5 we discuss some lessons for policy. A final section
concludes and we present additional tables and figures in an Online Appendix.
4
2 Data and setting
2.1 Dataset
We use household level scanner data collected by the market research firm Kan-
tar’s FMCG Purchase Panel. The data cover purchases of fast-moving consumer
goods (FMCG), which include all food and drinks (including alcohol), as well as
toiletries, cleaning products and pet foods, that are brought into the home by a
representative sample of households living in Great Britain (i.e., the UK exclud-
ing Northern Ireland).6 Participating households record purchases at the UPC (or
barcode) level using handheld scanners. For each transaction we observe quantity,
number of units, expenditure, price paid, store and UPC characteristics.
We use data covering the period January 1, 2019 to August 9, 2020. Our sample
is a balanced panel of 17,093 households who record their purchases over this period.
In Table A.1 in the Online Appendix we show that the sample is comparable along
key demographic dimensions with the UK’s main consumer spending survey. A
significant strength of our data is that they contain information on households’
purchases for a significant period prior to the pandemic and we are therefore able
to compare their behavior during the pandemic to normal times.
2.2 Timeline
The first case of COVID-19 was recorded in the UK on January 30. There was an
acceleration in case numbers across the globe during February. In Europe, the first
“lockdowns” (or “stay-at-home” orders) were introduced in the Lombardy region
in Italy on February 23, and were extended to the rest of the country on March 9.
By March 3, when the UK Government first published its strategy for responding
to the pandemic, the “Coronavirus action plan” (DHSC, 2020), there were a total
of 49 confirmed cases in the UK. A rapid increase in case numbers resulted in
the government introducing a nationwide lockdown on March 23. The lockdown
entailed closure of all non-essential businesses; however, businesses specializing in
the sale of fast-moving consumer goods, such as supermarkets, convenience stores
and liquor stores were permitted to remain open. On May 11, England moved into
the “Stay Alert” phase, with the government no longer encouraging people to stay
at home. From this point forward, lockdown restrictions were gradually lifted.
6
In prior years the scanner data account for approximately 40% of household expenditures on
goods, and 15% of total household expenditures on both goods and services (see Jaravel (2019)).
5
3 Purchase dynamics
In this section we document purchase dynamics over the pandemic for four broad
groups of products that together comprise all fast-moving consumer goods. For
each group, Figure 3.1 shows the evolution of daily expenditure between January 1
and August 9 in 2019 and 2020. The two vertical red lines denote the day in 2020
that the Coronavirus action plan was published and the beginning of lockdown.
Figure 3.1: Aggregate spending
(a) Staples (b) Household supplies
1.75
1.75
Average daily spending normalized Average daily spending normalized
by mean value over Jan-Feb by mean value over Jan-Feb
.75 1 1.25 1.5 .75 1 1.25 1.5
.5 .5
0 20 40 60 80 100 120 140 160 180 200 220 0 20 40 60 80 100 120 140 160 180 200 220
Weeks since start of year Weeks since start of year
(c) Discretionary calories (d) Perishables
1.75
Average daily spending normalized
Average daily spending normalized
by mean value over Jan-Feb
by mean value over Jan-Feb
1 1.25 1.5 1.75 .75 1 1.25 1.5
.75
.5 .5
0 20 40 60 80 100 120 140 160 180 200 220 0 20 40 60 80 100 120 140 160 180 200 220
Weeks since start of year Weeks since start of year
Notes: Each panel shows total daily expenditure on staples, household supplies, discretionary
calories, and perishables. See Table A.3 in the Online Appendix for a list of the product categories
in each grouping. Solid lines show smoothed daily expenditure, dotted lines show non-smoothed
daily expenditure after day of the week and holiday effects are removed. In each case the line
is normalized by the mean value over January and February. The vertical red lines indicate the
announcement of UK’s Coronavirus action plan on March 3, and the beginning of lockdown on
March 23.
Panel (a) focuses on “staples”. These products are storable and include canned
goods, rice and grains (see Table A.3 in the Online Appendix for a full list of product
categories). This product group accounts for 16% of total fast-moving consumer
good expenditure in 2019. The graph shows that spending evolved similarly in 2019
6
and 2020 up until the end of February. At this point spending in 2020 rose sharply,
peaking on March 14 at over 80% the January to February daily average. Spending
then fell back to close to normal levels at the beginning of lockdown. After lockdown
began, spending rose gradually before settling at an elevated level compared to the
beginning of the year and to 2019, but considerably below the peak in March.
Panel (b) focuses on “household supplies”, which cover all non-food and drink
fast-moving consumer goods – such as soap, cleaning products, and toiletries – and
comprise 14% of spending in 2019. The evolution of daily expenditure follows a
similar pattern to staples up until the beginning of lockdown: expenditure is very
similar in 2019 and 2020 until the end of February, at which point spending in 2020
rises sharply, peaking at over 70% above the January to February daily average on
March 14. Unlike staples, expenditure on household supplies in 2020 was similar to
2019 from the beginning of lockdown onwards.
Panels (c) and (d) show the path of expenditure on “discretionary calories” –
e.g., alcohol, desserts, confectionery and soft drinks – and “perishables” – i.e., fresh
food such as fruit, vegetables, meat and dairy. These account for 27% and 42%,
respectively, of fast-moving consumer good spending in 2019. For both of these
groups of products the spike in daily expenditure in March is smaller than the
increase in spending during lockdown.
Figure 3.1 makes clear there was a large spike in spending on staples and house-
hold supplies in the run-up to lockdown. Following the spike, at the onset of
lockdown, spending on household supplies returned to a level similar to in 2019.
Spending on staples dipped at the beginning of lockdown, though remained at a
level higher than the same time in 2019, before moderately rising during lockdown.
Spending on discretionary calories and perishables were also higher during lock-
down than the same time in 2019. This higher spending is predominantly due to
increased demand during lockdown as households worked from home and switched
away from shut-down restaurants. In addition, a spike in inflation for fast-moving
consumer goods in the first week of lockdown, driven by fewer promotions (doc-
umented in Jaravel and O’Connell (2020b)), contributes toward higher spending
during the lockdown period.7 However, the increases in spending on staples and
household supplies during lockdown are dwarfed by the March spikes.
7
Jaravel and O’Connell (2020b) show that month-to-month inflation for fast-moving consumer
goods in the first month of lockdown was 2.4%. Prices then gradually declined towards pre-COVID
levels (see Jaravel and O’Connell (2020a)). The increase in spending on fast-moving consumer
goods from the beginning of lockdown to the end of our data (a period when households were
urged to work from home) compared to 2019 was 11%.
7
Press reports over the four-week period prior to lockdown suggest that a small
number of greedy consumers were leading to stock outs.8 In Figure 3.2 we show how
the distribution of household spending on staples and household supplies changed
in the run-up to lockdown. This allows us to assess whether it was the case that
a small number of extreme purchasers drove the spike in aggregate spending on
staples and household supplies.
Figure 3.2: Distribution of real spending in four weeks up until March 23
(a) Staples (b) Household supplies
.5 .5
.4 .4
Density Density
.3 .3
.2 .2
.1 .1
0 0
0 2 4 6 8 0 2 4 6 8 10
Daily real expenditure (£) over 4 weeks up to March 23 Daily real expenditure (£) over 4 weeks up to March 23
Notes: For each household we compute their average daily real expenditure on staples and household
supplies over the four week period ending March 22 in both 2019 and 2020. Each panel shows the
distribution in 2019 and 2020 for staples (panel (a)) and household supplies (panel (b)). See
Table A.3 in the Online Appendix for a list of the product categories in each grouping. Daily real
expenditure is constructed holding UPC prices fixed at their average level over the four-week period
ending March 22 in 2019.
Panel (a) of Figure 3.2 focus on staples. It shows the distribution of average
daily real expenditure on staples in the four week period up until March 23 across
households in both 2019 and 2020.9 For each distribution we hold UPC prices at
their 2019 level, so that the graph is not influenced by price inflation. The graph
shows that there was a rightward shift in the distribution in 2020 relative to 2019.
Panel (b), which focuses on household supplies, shows a similar rightward shift in the
distribution of real expenditure. Together they point towards a moderate increase
in demand for staples and household supplies by many households, rather than a
small number of extreme purchasers, driving the spike in aggregate spending. This
does not rule out that extreme purchasing was important at the more disaggregate
8
See, for instance, The Sun report on March 15 speculating over the introduction of food
rationing if greedy shoppers ignore warnings to stop panic buying.
9
Since 2020 was a leap year, the four weeks cover 23 February to 22 March in 2019 and 24
February to 22 March in 2020.
8
level of product categories. We turn to the analysis of purchase dynamics at the
product category level in the next section.
4 How widespread was hoarding?
In this section we explore how widespread hoarding was, both across disaggregate
product categories and households. We document the size of quantity spikes across
product categories, the relative importance of changes in the probability of pur-
chasing and quantity conditional on buying, and provide evidence on the extent
to which the size of households’ demand increase in one category correlates with
their increase in purchases in other categories. We then document heterogeneity in
demand spikes across households from different socioeconomic groups.
4.1 Which product categories drove the demand spike?
We use the granular nature of our data to provide evidence on the product cate-
gories that experienced the largest spikes in demand, and the extent to which this
contributed to the overall spikes in spending shown in the previous section. We
study purchase dynamics across 138 product categories; see Table A.3 in the On-
line Appendix for a list. For each category, we calculate the percentage increase in
average daily quantity purchased over the four-week period running up to the start
of lockdown on March 23, relative to the same period in 2019.
Figure 4.1(a) shows the distribution of these percent changes across categories.
30 categories, all of which belong to the broader product groupings of staples or
household supplies, experienced an increase in quantity bought of more than 25%
(see panel (b) for the list of these). The quantity of soap purchased more than
doubled over this period relative to the previous year. Other categories that expe-
rienced significant spikes in purchases include soup (75%), cold treatments (64%),
rice and noodles (54%), and dried pasta (49%). Given the “just-in-time” nature
of UK grocery supply chains (Garnett et al. (2020)), it is not surprising that there
were widespread stories of shortages in these categories. The 30 categories that ex-
perienced an increase in quantity of more than 25% jointly account for 70% of the
overall increase in spending on household supplies, and 53% of the overall increase
in spending on staples over this period.
9
Figure 4.1: Quantity increases across categories in four weeks up until March 23
(a) All categories
% increase in quantity purchased
in 2020 relative to 2019
-25 0 25 50 75 100
-50
0 20 40 60 80 100 120 140
Ranked category
Staples Household supplies
Discretionary Perishables
(b) Top 30
% increase in quantity purchased
in 2020 relative to 2019
0 20 40 60 80 100
W thn E
as ic
hi In
ng g
Vi Cl ne
ta a dC
e an U red
p ie
Pr nt
od s
uc
m ni V
in ng eg T
s
Pr an Ac ta a
ep d ce ble e e ts
ar Su ss s
ed pp or
Pe lem ies
as e
Am b
Bl ien Fe nne an
ea t C m d s C &B nts
a
ch o inin Fis e
es ok e h
&L ing Ca
av S re
Ki ato uc a
tc ry es
he C
C Ba To rs
an k w n ln
D ne ed els
ry d B
Pu Pu ea
ls dd n
es in
To Tin Ce s
m ne rea
at d l
o + g
P Fr
P ro uit
O ack du
ra e ct
lA tS s
To nal oup
Am bi Hou ile ge
t T sic
is s
su es
en s
t R eh Dr Flou
o
ic ld P r
e+ C a y
Sv le sta
C an ne C N er ry an
d an oo s
Pa ne dl e
C sta d M s
ol P e
d ro at
Tr d
Fa ea uct
ci tm s
a
Am l T ent
bi issu
en e s
tS s
ou
So app
Extensive margin Intensive margin Covariance
Notes: For each category we compute the mean quantity purchased (across household-dates) over
the four-week period ending March 22 in 2019 and 2020. The bars show the percentage change
between these two periods. Panel (a) shows changes for all categories. See Table A.3 in the Online
Appendix for a list. Panel (b) zooms in on those with an increase of more than 25%, decomposing
the increase into changes due to the extensive margin, intensive margin and a covariance term.
10
Extensive versus intensive margin changes
Was the spike in category demands driven by households purchasing more often, or
buying larger quantities, conditional on purchasing? To answer this, we decompose
the change in quantity purchased in the four week period running up to March 23
in 2020 relative to the same period in 2019. Let Qjy denote average daily quantity
purchased of category j in year y = {2019, 2020} over this four week period; this is
given by:
1 XX
Qjy = qijt
N i t∈P
y
where qijt denotes the total quantity of j purchased by household i on date t, Py
indexes the set of 28 dates over the four week period in year y, and N denotes
the number of households in the sample multiplied by the number of days (28)
+
= i t∈Py 1{qijt > 0} denote the number of household-
P P
in the period. Let Njy
days on which category j was bought. We can re-write Qjy = Ejy × Qcjy , where
N+
Ejy = Njy is the fraction of household-days on which category j was bought and
Qcjy = N1+ i t∈Py qijt is the average quantity bought, conditional on the category
P P
jy
being purchased. Defining ∆X = X2020 − X2019 , we can then write:
∆Qj = Qcj2019 × ∆Ej + Ej2019 × ∆Qcj + ∆Ej × ∆Qcj .
| {z } | {z } | {z }
extensive margin intensive margin covariance
Figure 4.1(b) shows the contribution that each of these components makes to
the demand spike in the 30 categories that experience overall increases in demand
of more than 25%. We report these numbers in Table A.4 of the Online Appendix.
Two things are evident from the figure.
First, the extensive margin contributes more towards the demand spike than
the intensive margin: in 27 of the 30 categories, the extensive margin accounts for
at least 50% of the spike. On average across the categories, an increase in the
fraction of household-days on which the category was bough accounts for 70% of
the spike in demand. Second, the increase in quantity attributable to the intensive
margin is roughly similar across categories – for most categories the increase in
quantity, conditional on buying ranges between 5-15%. Therefore, the extensive
margin change was relatively more important in driving demand spikes of those
categories that experienced the largest overall increases.
11
Did households make multiple purchases or shop more often?
An increase in the fraction of household-days on which the category was purchased
was the primary driver of the large quantity spikes in key storable categories. To
what extent was this driven by some households making more multiple purchases
of this category (on different days), or more households than usual choosing to
buy the category over the March period? Relatedly, was the increase driven by
households visiting the store more often in general, or being more likely to buy
storable foods and supplies, conditional on visiting the store? To answer these
questions, we conduct two further decompositions.
First, we decompose the increase in the fraction of household-days on which
the category was purchased into the change due to an increase in the fraction of
households buying the category at least once over the period, and the part due
to a change in the number of times these households bought the category over
+,hh
= i 1{maxt∈Py qijt > 0} denote the number of
P
the four-week period. Let Njy
hh jy N +,hh
households that buy category j at least once over period y, Ejy = N/28 denote the
mult N+
corresponding fraction of households, and let Ejy = N +,hhjy×28 denote, for those
jy
households that buy at least once, the average fraction of days on which they buy.
mult hh
We can then write Ejy = Ejy × Ejy and decompose as follows:
mult
∆Ej = Ej2019 × ∆Ejhh + Ej2019
hh
× ∆Ejmult + ∆Ejhh × ∆Ejmult .
| {z } | {z } | {z }
new shoppers multiple purchases covariance
Table A.4 in the Online Appendix summarizes the results for the top 30 cat-
egories. It shows that the extensive margin change was primarily driven by an
increase in the number of unique households buying the category, which accounted
for 76% of the extensive margin response, on average. Thus the biggest driver of
the demand spikes was the fact that more households than usual chose to buy these
categories over this period, rather than some households repeatedly buying them.
Second, we decompose the increase in the fraction of household-days on which
the category was purchased, into the fraction due to households making more trips
to the store to buy anything (shopping frequency) versus households being more
likely to buy the category, conditional on visiting the store (purchase incidence).
Let Ny+,trips = i t∈Py 1{maxj qijt > 0} denote the number of household-days on
P P
which any category was purchased over the period y, i.e., the number of shopping
+
incid Njy
trips undertaken by households in the sample, Ejy = N +,trips denote the fraction of
y
+,trips
trip N
trips on which the category was purchased, and Ejy = yN denote the fraction
of households-days on which a shopping trip was undertaken. We can then write
12
incid trip
Ejy = Ejy × Ejy , and decompose this into:
incid
∆Ej = Ej2019 × ∆Ejtrips + Ej2019
trips
× ∆Ejincid + ∆Ejtrips × ∆Ejincid .
| {z } | {z } | {z }
shopping frequency purchase incidence covariance
Table A.4 presents the results of this decomposition. We find that the extensive
margin change is almost entirely driven by increases in the purchase incidence for
these categories, rather than households visiting a store more frequently. Increased
probability of buying, conditional on visiting a store, accounts for, on average, 90%
of the extensive margin change for the categories that saw the largest spikes in
demand. The period of hoarding of key storable categories was primarily driven
by more households than normal choosing to buy these categories once, and not by
increased shopping frequency, nor households buying larger quantities conditional
on purchasing.
This is relevant when we consider how changes in shopping patterns may affect
the propensity for future bouts of hoarding. In the Online Appendix we show how
the average number of shopping trips made by households varies over 2019–20 (see
Figure A.1). We find that there was a small increase in number of shopping trips
during the four-week period leading up to lockdown (March, 2020), followed by a
10-15% decline in shopping frequency once lockdown restrictions were introduced.
The fact that the quantity spikes were not driven by increased shopping frequency
suggests that further spates of hoarding could still occur, even if households continue
to shop less frequently than they were before the pandemic.
Cross-category correlation
To what extent were household level spikes in demand correlated across categories?
To answer this, we compute household level changes in average daily quantity pur-
chased in the four weeks running up to March 23 in 2020 relative to the same period
in 2019. For each of the 30 categories that exhibited the biggest demand spike we
take the pairwise correlation in these changes across households. 434 of the 435
pairwise correlations are positive. The median correlation coefficient is 0.1. The
categories with the largest correlation coefficients (above 0.25) are: “dry pasta”
and “ambient cooking sauces”; “canned pasta products” and “baked beans”; and
“tomato products” and “dry pasta”. Overall, there is a consistent but modest cor-
relation in household level demand changes across categories. If a household raised
their demand for one category it is likely that they also raised demand across others,
13
but the predictive power of changes in demand for one category for changes in a
specific second category is low.10
4.2 Heterogeneity by socioeconomic group
A common concern raised by the media and policymakers in the run-up to lock-
down was that some vulnerable households may be failing to access the products
they need. In this section we explore differences in purchase dynamics across so-
cioeconomic groups. We use the social grade of the household, which is based on
the occupation of the head of the household, and is a good proxy for the household’s
permanent income. Table A.2 in the Online Appendix lists the five socioeconomic
groups and the share of households in each. “AB” is the highest group, consist-
ing of households with a head who is occupied in a managerial, administrative or
professional role; “E” is the lowest group, consisting of those households with non-
working heads e.g., state pensioners, casual and lowest grade workers, unemployed
with only state benefit income.
Figure A.2 of the Online Appendix shows how spending on staples, household
supplies, discretionary calories and perishables evolved over January to August 2020
for each socioeconomic group. All five exhibit a substantial increase in their spend-
ing on staples and household supplies during the four-week period prior to lockdown.
However, there is a clear gradient in the size of these spikes: for staples, over the
four weeks up to March 23, daily spending was 31% above its prior 2020 average
for AB households and 18% higher for E households. The increase in spending on
household supplies was 27% for AB households and 18% for E households. This
gradient in spending changes persists into the lockdown period, where spending is
considerably below the March peak but elevated compared to before the pandemic.
These patterns are also evident at the product category level. In Figure 4.2
we show how the average increase in quantity in the four weeks in the run-up to
lockdown relative to the same period in 2019, calculated across the 30 categories
with the largest overall demand spikes, varies across the socioeconomic groups.
It also shows the contribution made to these increases by changes in the number
of household-days on which the category was bought (the extensive margin) and
quantity conditional on buying (the intensive margin). On average, AB households
increased the quantity bought of these categories by 55%, compared with 30% for
households in the lowest group. This gradient is almost entirely driven by changes
along the extensive margin; changes in conditional quantity are similar across the
10
Note, if instead we compare February 24 - March 22, 2020 with the four week period January
27 - February 23, 2020 the pattern is very similar.
14
socioeconomic groups. The greater propensity of higher socioeconomic households
to hoard was therefore driven by the fact they increased the probability of buying
these categories by more than other households.
Figure 4.2: Average change in product categories, by socioeconomic group
% increase in 2020 relative to 2019
20 40 60
0
AB (top) C1 C2 D E (bottom)
% change in:
Mean quantity
Fraction of hh-days on which category was bought
Mean quantity, conditional on buying
Notes: The figure shows the unweighted mean percentage change in quantity, fraction of household-
days on which the category was bought, and quantity conditional on buying across the top 30
categories shown in Figure 4.1(b) for each socioeconomic group.
5 Lessons for policy
During the run-up to the national lockdown there were reports of shortages in many
stores. These led to calls for policy intervention to tackle the shortages. As the
second wave of infections grows, these calls are being renewed.
In the final few days before lockdown supermarkets introduced limits on the
number of units of a product households could buy per transaction.11 By the time
that quantity limits came into effect, demand was already returning to normal levels.
A resurgence of case numbers and the threat of a second nationwide lockdown has
already led some supermarkets to reintroduce quantity limits in September 2020.
We show above that much of the demand spike in March was driven by more
11
For instance, on March 19, Tesco, the largest supermarket chain, introduced a limit that
people could purchase no more than three of any item pre visit. This followed the adoption of
similar restrictions by Sainsbury’s (another large chain) the day before.
15
households than usual choosing to buy storable food and supplies on at least one of
their shopping trips over this period, rather than buying much larger quantities.12
This suggests that quantity limits may not prevent the shortages induced by the
demand spikes.
In Figure 5.1 we graph the quantity spikes for the top 30 categories, as well as
what they would have been if a limit of 3 or 2 units per transaction was in place
over the whole four weeks running up to lockdown. This assumes that households
do not circumvent the limits by purchasing different UPCs in the same category, or
undertaking more regular grocery store visits. It therefore should be a viewed as an
upper bound for the impact of the quantity limits. The graph shows that the limits
are likely to have had only a moderate effect on reducing demand; even with the
limits there are still very large increases in demand for the majority of categories –
for instance, were the 2 pack limit place the average increase in quantity in the 30
categories would have been 34% rather than 44%.
Figure 5.1: Change in quantity spikes under counterfactual quantity limits
% increase in quantity purchased
in 2020 relative to 2019
20 40 60 80 100
0 h
as nic
hi In
ng g
U red
p ie
Vi Cl neC
ta ea d
m ni Van Pr nt
od s
uc ts
in ng eg T
s
Pr n c ta a
a
ep d ce bleA
ar Su ss s e e
Am ed pp or
Pe lem ies
as e
C &B nts
b
Bl ien Fe nne an
ea t C m d s a
ch o inin Fis
es ok e h e
&L ing Ca
av S re
Ki ato uc
tc ry es
he C a
C Ba To s
an k w
D ne ed els
ry d B n lnr
Pu Pu ea
ls dd n
es in
To Tin Ce s + g
m ne rea
at d l
o P Fr
P ro uit
O ack du
Et Am
ra e ct
lA tS s
To nal oup
ile ge
t T sic
is s
bi ouH
en s
t R eh Dr Flo
ic ld P ur y su es
W C an o
e+ C a
Sv le sta
ne C N ner ry a
d an oo s
Pa ne dl e
C sta d M s
ol P e
d ro at
Tr d
Fa ea uct
ci tm s
a
Am l T ent
bi issu
en e s
tS s
ou
So app
Observed 3 pack limit 2 pack limit
Notes: The red bars show the percentage increase in quantity purchased for each category between
the four-week period ending March 22 in 2020 compared with the same period in 2020. The dark
and light blue bars show the analogous, counterfactual increases if households were limited from
buying no more than 3 and no more than 2 packs per transaction, respectively.
This begs the question, so what then, can be done about shortages arising from
hoarding? Another policy implemented by several supermarkets in March was to
12
There are exceptions. In September, 2020 the Daily Mail reported on someone who had just
purchased 400 tins and 700 nappies due to fear of rationing.
16
have dedicated shopping hours for the elderly and other vulnerable consumers. If
these were timed to coincide with deliveries of new stock, this may help to en-
sure that vulnerable consumers are able to procure the supplies that they need.
The differences across socioeconomic group documented above are indicative of the
potential importance of this: households in lower socioeconomic groups did not
increase their purchase frequency by nearly as much as richer households, and did
not offset this through buying larger quantities conditional on purchasing. Thus
wealthier households were able to build up larger precautionary stocks than less
well-off ones.
6 Summary and conclusions
In this paper, we provide new evidence on consumer purchase dynamics and hoard-
ing during the early phase of the COVID-19 pandemic. We show that a number
of storable categories experienced dramatic spikes in demand, and that this was
primarily driven by more consumers than usual buying products in these categories
on at least one of their shopping trips over this period, rather than increasing
the amount they bought on any particular trip. Unusually high demands were
widespread across households, although higher socioeconomic status households in-
creased the quantities they bought by more than lower socioeconomic households.
Understanding the motivations behind why people hoard is important in deter-
mining what, if any, is the appropriate policy response. For example, we show that
the category that saw the largest increase in demand was soap – given the focus
on the importance of hand washing to prevent the spread of the virus, this likely
reflects a shift in preferences that could be persistent. It is also important to recog-
nize that government policy can affect the degree of panic and subsequent hoarding.
An important avenue for future research is to exploit differences in the course of
the pandemic and resulting restrictions across different localities, to further unpack
the drivers of panic buying and hoarding behavior.
17
References
Alexander, D. and E. Karger (2020). Do stay-at-home orders cause people to stay
at home? Effects of stay-at-home orders on consumer behavior. Federal Reserve
Bank of Chicago.
Andersen, A. L., E. T. Hansen, N. Johannesen, and A. Sheridan (2020). Consumer
Responses to the COVID-19 Crisis: Evidence from Bank Account Transaction
Data. SSRN Electronic Journal .
Baker, S. R., R. A. Farrokhnia, S. Meyer, M. Pagel, and C. Yannelis (2020a). How
Does Household Spending Respond to an Epidemic? Consumption During the
2020 COVID-19 Pandemic. Review of Asset Pricing Studies forthcoming, 35.
Baker, S. R., R. A. Farrokhnia, S. Meyer, M. Pagel, and C. Yannelis (2020b). In-
come, Liquidity, and the Consumption Response to the 2020 Economic Stimulus
Payments. NBER Working Paper No. 27097.
Bounie, D., Y. Camara, and J. W. Galbraith (2020). Consumers’ Mobility, Expen-
diture and Online-Offline Substitution Response to COVID-19: Evidence from
French Transaction Data. SSRN Electronic Journal .
Carvalho, V. M., J. R. Garcia, S. Hansen, Á. Ortiz, T. Rodrigo, and J. V. R. Mora
(2020). Tracking the COVID-19 Crisis with High-Resolution Transaction Data.
Cambridge Working Papers in Economics 2030.
Chen, H., W. Qian, and Q. Wen (2020). The Impact of the COVID-19 Pandemic on
Consumption: Learning from High Frequency Transaction Data. SSRN Electronic
Journal .
Chetty, R., J. N. Friedman, N. Hendren, and M. Stepner (2020).
The Economic Impacts of COVID-19: Evidence from a New Public
Database Built from Private Sector Data. https://opportunityinsights.org/wp-
content/uploads/2020/05/tracker paper.pdf .
Chronopoulos, D. K., M. Lukas, and J. O. S. Wilson (2020). Consumer Spending
Responses to the COVID-19 Pandemic: An Assessment of Great Britain. SSRN
Electronic Journal .
Coibion, O., Y. Gorodnichenko, and M. Weber (2020). The cost of the COVID-19
crisis: Lockdowns, macroeconomic expectations, and consumer spending. NBER
Working Paper No. 27141.
Cox, N., P. Ganong, P. Noel, J. Vavra, A. Wong, D. Farrell, and F. Greig (2020).
Initial impacts of the pandemic on consumer behavior: Evidence from linked
income, spending and savings data. Brookings Papers on Economic Activity.
Croson, R., K. Donohue, E. Katok, and J. Sterman (2014). Order Stability in Supply
Chains: Coordination Risk and the Role of Coordination Stock. Production and
Operations Management 23 (2), 176–196.
Department of Health & Social Care (DHSC) (2020). Coronavirus (COVID-19)
action plan. https://www.gov.uk/government/publications/coronavirus-action-
plan/coronavirus-action-plan-a-guide-to-what-you-can-expect-across-the-uk.
Garnett, P., B. Doherty, and T. Heron (2020, June). Vulnerability of the United
Kingdom’s food supply chains exposed by COVID-19. Nature Food 1 (6), 315–318.
Hacioglu, S., D. Känzig, and P. Surico (2020). Consumption in the time of COVID-
19: Evidence from UK transaction data. CEPR Discussion Paper DP14733.
18
Hansman, C., H. Hong, Á. de Paula, and V. Singh (2020). A Sticky-Price View
of Hoarding. Technical Report w27051, National Bureau of Economic Research,
Cambridge, MA.
Hori, M. and K. Iwamoto (2014). The Run on Daily Foods and Goods After the
2011 Tohoku Earthquake. The Japanese Political Economy 40 (1), 69–113.
Jaravel, X. (2019, May). The Unequal Gains from Product Innovations: Evidence
from the U.S. Retail Sector*. The Quarterly Journal of Economics 134 (2), 715–
783.
Jaravel, X. and M. O’Connell (2020a). Grocery prices and promotions during the
COVID-19 pandemic. Technical Report IFS Briefing Note.
Jaravel, X. and M. O’Connell (2020b). Real-time price indices: Inflation spike
and falling product variety during the Great Lockdown. Journal of Public Eco-
nomics 191, 104270.
Keane, M. and T. Neal (2020). Consumer panic in the COVID-19 pandemic. Journal
of Econometrics, forthcoming.
19
Appendix
For Online Publication
Preparing for a pandemic: Spending dynamics and panic
buying during the COVID-19 first wave
Martin O’Connell, Áureo de Paula and Kate Smith
1
A Additional tables and figures
Table A.1: Household demographics
Kantar LCFS
Region
England - North (%) 28.1 27.4
[27.4, 28.8] [26.1, 28.6]
England - Midlands (%) 17.9 19.1
[17.3, 18.5] [18.0, 20.2]
England - South and East (%) 44.8 44.4
[44.0, 45.6] [43.0, 45.8]
Scotland (%) 9.2 9.1
[8.8, 9.7] [8.3, 9.9]
Employment status of household head
Full time (%) 38.7 39.6
[38.0, 39.5] [38.2, 41.0]
Part time (%) 20.3 11.1
[19.6, 20.9] [10.2, 12.0]
Self-employed* (%) 7.9
[7.2, 8.7]
Unemployed (%) 1.6 2.4
[1.4, 1.8] [1.9, 2.8]
Retired or not working (%) 39.4 39.0
[38.6, 40.1] [37.6, 40.4]
Socioeconomic group
Highly skilled (%) 20.7 18.7
[20.0, 21.4] [17.2, 20.1]
Semi skilled (%) 59.9 59.4
[59.0, 60.8] [57.6, 61.2]
Unskilled (%) 19.4 21.9
[18.7, 20.1] [20.4, 23.4]
Notes: Table shows the share of households in the Kantar Worldpanel and Living Costs and Food
Survey (LCFS) in various demographic groups. Numbers are shown for the most recently available
data for the LCFS, which is 2014, and for the households in our 2019-20 sample from the Kantar
data. ∗ The self-employed are not distinguished from employees in the Kantar data. Socioeconomic
status is based on the occupation of the head of the household and is shown for the set of non-
retired households – this is not recorded for non-retired households in the LCFS. Highly skilled
corresponds to AB, semi-skilled to C1 and C2, and unskilled to DE listed in Table A.2. 95%
confidence intervals are shown below each share.
2
Table A.2: Socioeconomic groups
Socioeconomic
group Description % households
AB (top) Managerial, administrative or professional 21.0
C1 Supervisory or clerical and junior managerial, 40.0
administrative or professional
C2 Skilled manual workers 17.6
D Intermediate managerial, administrative or pro- 13.3
fessional
E (bottom) State pensioners, casual and lowest grade work- 8.0
ers, unemployed with state benefits only
Notes: Table shows the description of the socioeconomic groups that we use and the share of
households belonging to each group.
3
Table A.3: Product categories
Product category Share of spending
in 2019 (%)
Staples 16.16
Ambient Condiments 0.20
Ambient Cooking Sauces 0.68
Ambient Rice+Svry Noodles 0.57
Ambient Soup 0.30
Ambnt Salad Accompanimet 0.24
Baked Bean 0.36
Breakfast Cereals 1.59
Canned Fish 0.54
Canned Meat 0.26
Canned Pasta Products 0.09
Canned Puddings 0.03
Canned Vegetables 0.19
Cooking Oils 0.32
Dry Pasta 0.21
Dry Pulses+Cereal 0.16
Ethnic Ingredients 0.21
Flour 0.10
Food Drinks 0.17
Frozen Fish 0.95
Frozen Meat 1.25
Frozen Pizzas 0.57
Frozen Potato Products 0.83
Frozen Ready Meals 0.71
Frozen Savoury Bakery 0.25
Frozen Vegetables 0.78
Herbal Tea 0.11
Herbs+Spices 0.20
Instant Coffee 0.85
Instant Hot Snacks 0.23
Liquid+Grnd Coffee+Beans 0.47
Meat Extract 0.38
Other Frozen Foods 0.16
Packet Soup 0.10
Pickles Chutneys+Relish 0.09
Prepared Peas+Beans 0.15
Preserves and spreads 0.48
Sour+Speciality Pickles 0.12
Sweet+Savoury Mixes 0.10
Table Sauces 0.30
Tea 0.47
Tinned Fruit 0.16
Tomato Products 0.25
4
Table A.3 cont.
Product category Share of spending
in 2019 (%)
Household 14.04
supplies
Air Fresheners 0.31
Anti-Diarrhoeals 0.15
Bath+Shower Products 0.38
Batteries 0.24
Bin Liners 0.20
Bleaches+Lavatory Clnrs 0.26
Body Sprays 0.05
Cleaning Accessories 0.14
Cold Treatments 0.27
Cotton Wool 0.04
Dental Products 0.81
Deodorants 0.41
Electric Light Bulbs 0.04
Eye Care 0.04
Fabric Conditioners 0.40
Facial Tissues 0.25
Feminine Care 0.34
First Aid Dressings 0.03
Foot Preparations 0.07
Hair Colourants 0.14
Hair Conditioners 0.18
Hair Styling 0.07
Hairsprays 0.07
Household Cleaners 0.57
Household Food Wraps 0.22
Kitchen Towels 0.39
Machine Wash Products 0.92
Moist Wipes 0.13
Oral Analgesics 0.25
Other Healthcare 0.25
Other Household 0.17
Pet Food 2.67
Pot Pourri+Scented Candles+Oil 0.08
Shampoo 0.30
Shaving 0.22
Skincare 0.63
Soap 0.19
Sun Care 0.09
Toilet Tissues 1.28
Vitamins and Supplements 0.36
Washing Up Products 0.45
5
Table A.3 cont.
Product category Share of spending
in 2019 (%)
Discretionary 27.38
calories
Ambient Cakes+Pastries 1.54
Beer+Lager 1.81
Chilled Cakes 0.96
Chocolate Confectionery 2.61
Cider 0.50
Crisps 0.90
Fruit Juice 0.87
Home Baking 0.82
Ice Cream 1.40
Long Life Milk and Desserts 0.40
Milk Drinks 0.17
Mineral Water 0.45
Mixers 0.18
Nuts 0.62
Popcorn 0.10
Savoury Biscuits 0.72
Savoury Snacks 1.04
Soda 2.28
Spirits 2.64
Sugar Confectionery 0.83
Sweet Biscuits 2.02
Total Fruit Squash 0.54
Wine 3.99
Perishables 42.41
Butter and Fats 1.46
Chilled Burgers+Grills 0.29
Chilled Cooking Sauces 0.08
Chilled Deli 0.34
Chilled Pizza+Bases 0.51
Chilled Prepared Frt+Veg 0.92
Chilled Prepared Salad 0.48
Chilled Ready Meals 2.52
Cooked Meats 2.12
Cooked Poultry 0.51
Eggs 0.77
Fresh Bacon 1.15
Fresh Beef 2.01
Fresh Cream 0.35
Fresh Flavoured Meats 0.16
Fresh Lamb 0.46
Fresh Pasta 0.15
Fresh Pork 0.63
Fresh Poultry 2.19
Fresh Sausages 0.92
Fresh Soup 0.09
6
Table A.3 cont.
Product category Share of spending
in 2019 (%)
Fruit 4.93
Morning Goods 1.70
Other Chilled Convenience 0.39
P/P Fresh Meat+Veg+Pastry 1.71
Shellfish 0.18
Total Bread 1.67
Total Cheese 2.85
Total Milk 2.79
Vegetables 5.21
Wet/Smoked Fish 0.92
Yoghurt 1.96
Notes: The table lists the 138 products that together comprise all fast-moving consumer goods,
grouped into staples, household supplies, discretionary calories and perishables. The final column
shows the share of total spending on fast-moving consumer goods in 2019 accounted for by each
category.
7
Table A.4: purchase dynamics, top 30 categories
(1) (2) (3) (4) (5) (6) (7) (8) (9)
% of overall % of extensive % of extensive
change due to: margin due to: margin due to:
∆Qjy
Category Qj2019 Ext. Int. Cov. No.hhs Mult. Trips Incid.
Soap 103.7 85.0 8.0 7.0 68.7 19.5 2.6 95.2
Ambient Soup 74.5 73.6 17.0 9.4 84.9 10.3 4.2 93.6
Facial Tissues 73.2 75.8 15.6 8.6 78.6 14.9 4.2 93.7
Cold Treatments 63.6 85.0 9.7 5.3 87.3 8.6 4.3 93.6
Canned Pasta Products 57.6 69.9 21.5 8.6 101.9 -1.4 5.7 92.1
Canned Meat 54.7 62.9 27.6 9.5 82.2 13.9 6.7 91.2
Ambient Rice+Svry Noodles 54.2 61.7 28.7 9.6 63.9 29.7 6.9 91.0
Household Cleaners 51.7 72.7 19.8 7.5 57.7 34.7 6.1 91.7
Dry Pasta 49.1 58.6 32.2 9.3 69.5 25.4 8.0 89.9
Flour 46.0 68.5 23.9 7.6 64.9 29.1 7.3 90.6
Toilet Tissues 44.9 59.5 32.0 8.5 58.3 36.1 8.6 89.3
Oral Analgesics 44.2 77.2 17.0 5.8 88.1 9.1 6.8 91.1
Packet Soup 43.8 103.2 -2.2 -1.0 87.8 8.7 5.1 92.7
Tomato Products 42.3 59.7 32.2 8.1 63.8 31.2 9.2 88.8
Tinned Fruit 42.1 66.5 26.2 7.3 94.8 4.1 8.3 89.7
Dry Pulses+Cereal 40.6 38.5 53.1 8.3 78.3 19.3 14.8 83.3
Canned Puddings 40.5 76.2 18.2 5.6 110.4 -7.7 7.5 90.4
Baked Bean 39.8 69.1 24.2 6.7 71.5 23.8 8.4 89.5
Kitchen Towels 37.4 85.4 11.1 3.5 77.7 17.9 7.2 90.7
Bleaches+Lavatory Clnrs 37.3 97.8 1.6 0.6 65.0 28.3 6.3 91.5
Ambient Cooking Sauces 34.7 49.5 43.1 7.4 52.9 43.1 13.4 84.6
Feminine Care 33.6 67.1 26.8 6.1 54.8 40.2 10.3 87.7
Canned Fish 32.6 64.5 29.4 6.2 77.8 19.1 11.0 87.0
Prepared Peas+Beans 31.9 62.7 31.1 6.2 88.9 9.4 11.6 86.4
Vitamins and Supplements 31.2 87.4 9.9 2.7 84.5 12.6 8.5 89.4
Cleaning Accessories 29.8 72.8 22.4 4.9 79.3 17.7 10.6 87.3
Canned Vegetables 29.8 54.4 39.2 6.4 86.4 11.9 14.3 83.8
Tea 29.0 72.3 22.9 4.8 64.6 31.2 11.0 87.0
Washing Up Products 27.4 86.6 10.8 2.6 69.7 26.0 9.7 88.2
Ethnic Ingredients 26.9 42.6 51.5 5.9 81.6 16.8 20.1 78.1
Notes: Column (2) shows the percentage change in average quantity purchased between the four-
week period up to March 22 and in the same period in 2019. Columns (3)-(5) decompose this
change into the share attributable to the extensive margin (2), the intensive margin (4) and a
covariance term (5). Columns (6)-(7) further decompose the change in the extensive margin into
that attributable to the change in the number of unique households that bought the category (6),
and the number of times each household bought the category, conditional on buying at least once
(7). Columns (8)-(9) conduct a separate decomposition, also on the extensive margin, but into the
increase attributable to households making more shopping trips (8), and probability of buying the
category, conditional on visiting the store (9).
8
Figure A.1: Average number of shopping trips per four-week period
% change in number of shopping trips
per 4-week period, relative to Jan 2019
-15 -10 -5 0 5
19 19 20 20 20
20 20 20 20 20
Ja
n Ju l n
Ja ar Ju l
M
Notes: For each household-four week period we calculate the number of days that they record buying
groceries. The markers show the average difference (relative to the first period) in the log of this
variable in each four-week period from January 2019 to July 2020. The red dashed line shows the
four-week period prior to the beginning of lockdown.
9
Figure A.2: Aggregate spending, by socioeconomic group
(a) Staples (b) Household supplies
2 2
Average daily spending normalized Average daily spending normalized
by mean over Jan-Feb 2020 by mean over Jan-Feb 2020
1 1.25 1.5 1.75 1 1.25 1.5 1.75
.75 .75
.5 .5
0 20 40 60 80 100 120 140 160 180 200 220 0 20 40 60 80 100 120 140 160 180 200 220
Weeks since start of 2020 Weeks since start of 2020
(c) Discretionary calories (d) Perishables
2
2
Average daily spending normalized Average daily spending normalized
by mean over Jan-Feb 2020 by mean over Jan-Feb 2020
1 1.25 1.5 1.75 1 1.25 1.5 1.75
.75 .75
.5 .5
0 20 40 60 80 100 120 140 160 180 200 220 0 20 40 60 80 100 120 140 160 180 200 220
Weeks since start of 2020 Weeks since start of 2020
Notes: Each panel shows total daily expenditure on staples, household supplies, discretionary
calories, and perishables, by socioeconomics group. See Table A.3 for a list of the product categories
in each grouping. Solid lines show smoothed daily expenditure, dotted lines show non-smoothed
daily expenditure after day of the week and holiday effects are removed. In each case the line
is normalized by the mean value over January and February. The vertical red lines indicate the
announcement of UK’s Coronavirus action plan on March 3, and the beginning of lockdown on
March 23.
10
File and source
- File
- preparing-for-a-pandemic-spending-dynamics-and-panic-buying-during-the-covid-19-first-wave.pdf
- Size
- 1,109,575 bytes
- SHA-256
- fb2b557dbf76d608aa3fa8914028ae47da599347a8448d168cfd984cf7b92fa6
- Our copy
- preparing-for-a-pandemic-spending-dynamics-and-panic-buying-during-the-covid-19-first-wave.pdf
- Original
- doi.org