Chowdhury Dixon 2024 Oer Measuring Inflation During The Pandemic Hindsight

Summary

Measuring Inflation During the Pandemic with the Benefit of Hindsight, an Open Economies Review article by Aftab Chowdhury and Huw Dixon, accepted in 2024. Using household expenditure data from national accounts to build a true inflation rate with a Fisher index, the authors find that during the first wave of the COVID-19 pandemic the official inflation rate overestimated true inflation in 23 of 33 OECD countries and underestimated it in 10. Their results match those of Cavallo and Reinsdorf for countries where true inflation exceeded the official rate, but differ for the others. They attribute the gap to expenditure weights and urge caution in using credit-card spending data when spending behaviour changes sharply.

Summary drafted by a model from the document's text below and checked by script against that text before publication. It is a navigation aid, not a reading of what the document proves. Where AI is used

Full text

Text recognized from a scanned document; may contain errors. Pages 9, 14, 18, and 22–24 of 26 are not shown because the text could not be recognized reliably.

Open Economies Review
https://doi.org/10.1007/s11079-024-09776-3

ORIGINAL RESEARCH

®

Check for
updates

Measuring Inflation During the Pandemic with the Benefit
of Hindsight

Aftab Chowdhury'?® - Huw Dixon!

Accepted: 10 June 2024
© The Author(s) 2024

Abstract

This study has adopted the actual household expenditure data from the national
accounts to construct a true inflation rate (using the Fisher index) and found that the
official inflation rate in the 33 OECD countries was an overestimate of true infla-
tion for 23 and underestimate in 10 countries in the first wave of the COVID-19
pandemic. The result obtained for the countries where true inflation was higher than
the official rate in this study matches the results obtained by Cavallo (Inflation with
covid consumption baskets, 2020) and Reinsdorf (COVID-19 and the CPI: Is infla-
tion underestimated?, 2020). However, a significant difference has been detected
for the countries where the official inflation exceeds the true measure in this study.
The core reason behind the discrepancies is in the use of appropriate expenditure
weights. This suggests caution in using credit-card based expenditure data when
spending behaviour has changed dramatically.

Keywords Coronavirus (COVID-19) - Pandemic - Inflation measurement -
Household Final Consumption Expenditure (HHFCE) - Consumer Price Index (CPI)

JEL Classification C43 - E01 - E31 - 110

1 Introduction

The first documented case of Covid-19 was found in December 2019, and within
three months the World Health Organization (WHO) declared Covid-19 as a Pan-
demic. Economies around the world experienced significant losses due to the dis-
ruption in global supply chains, temporary and permanent closures of businesses,

Db Huw Dixon
DixonH @cardiff.ac.uk

Aftab Chowdhury
chowdhuryam @cardiff.ac.uk

Cardiff Business School, Cardiff University, Aberconway Building, Colum Dr,
Cardiff CF10 3EU, UK

Department of Banking and Insurance, University of Dhaka, Dhaka, Bangladesh

Published online: 29 June 2024 va Springer
A. Chowdhury, H. Dixon

temporary closures/limits of the production plants, restriction on international travel
etc. It had a massive effect on the spending patterns within economies.

As was realized early on (see for example Dixon 2020b and Tenreyro 2020), this
raised specific challenges for the measurement of CPI inflation, since the standard
methods of constructing inflation statistics were not designed for periods when there
were rapid and large changes in expenditure patterns. There were several attempts to
use real time data such as credit card expenditures to try to understand exactly how
expenditure patterns were evolving and how this might distort the official inflation
figures. See for example Cavallo (2020) Chronopoulos et al. (2020), Dixon (2020a),
Jaravel and O’Connell (2020b) in the UK; Chetty et al. (2020), Bachas et al. (2020),
Dunn et al. (2020), Baker et al. (2020) in the US; Andersen et al. (2022) in the Den-
mark; Carvalho et al. (2021) in Spain; Seiler (2020) in Switzerland. As national
account information was not available immediately, all the above-mentioned stud-
ies had to rely on the alternative sources of data, such as debit and credit card data,
scanner data etc. Cavallo (2020) used US and Spanish card-based expenditure data,
whilst Reinsdorf (2020) used US and Canada card-based expenditure data to meas-
ure the inflation in the Covid-19 pandemic across a range of countries. Both the
studies found significant biases in the headline inflation rate during pandemic.’ One
of the strong assumptions of those papers was the use of US and Canada’s expendi-
ture share changes to other countries.” There might be no other options available for
real-time measurement as the national account’s information often has a considera-
ble lag. However, since we now have the national accounts data for a most countries,
we can look at the problem with the data from each country.*

The official headline inflation of most countries uses the “Lowe” price index
which is a Laspeyres-type or fixed based weight index. Moreover, the advice offered
by the international agencies* was to keep the pre-pandemic weights. Where goods
and services were unavailable due to lockdowns the advice given was to impute the
missing prices on the basis of the prices that were available.°

As is well known, using a base weighted Laspeyres index will tend to overstate
inflation (since consumers are likely to consume less of items that become more
expensive), whilst a current weighted Paasche index will tend to understate infla-
tion. The true inflation will be between these two, and an alternative “ideal index”
is the Fisher Index, which is the geometric average of the Laspeyres and Paasche
indices (Fisher 1921). In this paper, we construct a Fisher index to evaluate the true
inflation during the pandemic and compare to the official inflation figures with this

' Cavallo (2020) found that out of 17 countries 10 countries have higher inflation than that of the official
CPI. Reinsdorf (2020) found that out of 83 economies where official CPI is underestimated in 65 coun-
tries during Feb-May 2020.

? Cavallo (2020) used Spanish card-based expenditure data to estimate the weight for the European
countries in his sample.

3 We review a detail some of the main papers published using real time data in Appendix B.

4 Advice from Eurostat to European Union Countries, UNCE, IMF, US Bureau of Labor Statistics.

> Diewert and Fox (2020) argued that the implicit price of unavailable goods was in fact very high (the
price that would have driven demand to zero in normal times), so that the imputation method recom-
mended would likely understate the real inflation.

va Springer
Measuring Inflation During the Pandemic with the Benefit of...

measure. Most statistical agencies have now published the actual expenditure shares
for the period 2020-2021 which we can use to answer the question of how the offi-
cial measures published in 2020-21 compare with the ex-post true inflation. Real
time indicators used at the time were mostly based on a limited range of consumer
expenditure (mostly using credit card data,° as in Cavallo 2020, Carvalho et al 2021,
Reinsdorf 2020, and Sieler 2020) and the national statistical agencies have since
published figures that are much more comprehensive and cover the whole range of
consumer expenditure. We will also compare the actual data with some examples of
the contemporary real time data. Hindsight is indeed an advantage when it comes to
tracking what was going on in 2020-2021.

When we use the actual household expenditure data from the national accounts of
the 33 OECD countries, that the official inflation figures were too high in 23 coun-
tries (including Belgium, Italy, Norway, and the UK) and too low in 10 (including
the US, Australia, and Iceland).’

This paper is structured as follows. Section 2 is about the measurement issues
of the official inflation rate during the pandemic. Section 3 explains the methodol-
ogy and the data. Section 4 has measured the impact on the US and the UK infla-
tion rates. Section 5 illustrates the impact on the other OECD countries’ inflation
rates and the differences between this study’s outcome with some earlier studies.
Finally, we conclude the paper by mentioning the importance of appropriate data set
in measuring inflation.

2 Measurement Issues

During COVID-19 Pandemic, the nationwide lockdown has produced significant chal-
lenges to the inflation measurement by the national statistical offices (NSOs) as there
were large and sudden changes in the consumers’ expenditure patterns. The standard
method of measuring inflation measure does not consider the possibility of sudden
large changes in the expenditure patterns, but rather a gradual adjustment with most
NSOs updating the expenditure weights used in the official CPI annually, with lagged
expenditure data from previous calendar years (Jaravel and O’Connell 2020b). There-
fore, sudden large changes in the expenditure shares across different sectors, products
and outlets may all create biases in the official headline inflation rate.

There are of course other issues that made measurement more challenging during
the pandemic. The physical collection of prices by agents was not possible, so col-
lection moved online. Many goods and services became unavailable during the lock-
down, and hence there were no prices to be collected. Some items were completely

© The main exception being Jaravel and O’Connell (2020b), which used store scanner data. This is best
for capturing grocery expenditure and in particular food and beverage prices.

7 However, the results were a little mixed for 23 out of 33 countries: the bias during the 2020 Q2 was the
opposite to the whole period 2020-2021. See Table 3 below.

va Springer
A. Chowdhury, H. Dixon

unavailable to buy, and even when items were available, consumers were less inter-
ested in purchasing the items during lockdown.

How did the UK calculate the official price index during the pandemic? If
the item was unavailable, or with limited availability (with the sample of price
quotes being less than 20% of its usual size), the Office for National Statistics
(ONS) decided to impute the prices, based on the method that best reflects the
price behavior of the missing item. For unavailable items,*® the ONS followed the
Eurostat guidelines to impute using one of three methods: firstly, imputing from
the index immediately above it in the classification structure; secondly, imputing
based in the price movement of a similar item; thirdly, carrying forward prices
from earlier months. If the unavailable item is seasonal, the ONS will use aggre-
gate annual growth of all available non-imputed items. However, if the unavailable
item is not seasonal, the ONS will use aggregate monthly growth of all available
non-imputed items.

Dixon (2020b) argued that the imputation of unavailable prices is equivalent to
altering expenditure shares, since if we use the available prices to impute the una-
vailable prices, we are effectively increasing the weight of the available prices in
the price index. Moreover, Diewert and Fox (2020) also presented that the advice
from International Monetary Fund, Eurostat and the United Nations to the NSOs to
implement carry-forward methodology for the items that are missing, will lead to
a downward bias in estimating Cost of Living Index (COLD and an upward bias in
estimating the changes in the real consumption. As the headline inflation rate calcu-
lated by different NSOs is heavily biased due to the treatment of missing prices and
use of out-of-date expenditure weights, Diewert and Fox (2020) has suggested to
produce a new analytic CPI whenever new consumer information for the Covid-19
Pandemic periods becomes available. Therefore, we would like to say that this is key
motivation for doing this study to calculate inflation rate during the Covid-19 Pan-
demic period with benefit of new consumer expenditure information.

3 Data and Methodology
3.1 Methodology

Generally, CPI uses a “Lowe” price index which is a Laspeyres-type or fixed
based weight index. The Lowe price index measures the proportional change
between reference period and the current t in the total value of a specific basket of
goods and services.

In the UK and most NSOs, the expenditure weights come from previous years
and are used to weight the price relatives across months within the current calendar

8 ONS has identified 92 items as unavailable for April Index, where 90 items for CPIH (around 16.3% of CPIH
basket), 90 items for CPI (around 20.2% of CPI basket) and 89 items for RPI (around 17.7% of RPI basket).

va Springer
Measuring Inflation During the Pandemic with the Benefit of...

year. For example, in 2020 the expenditure weights used by the ONS were based
on the household final consumption expenditure (HHFCE) data from 2018,” whilst
the price relatives were between each month and January of 2020. Similarly, U.S.
Bureau of Labor Statistics (BLS) calculate headline CPI (All Urban Consumers CPI-
U) inflation rate in 2020 based on the expenditure share (relative importance) from
2017-2018 Consumer Expenditure Surveys (CE).'° Before the outbreak of Covid-19
Pandemic, there is nothing significantly wrong with estimation method used by the
several NSOs as the consumer expenditure patterns doesn’t change dramatically in
the most recent years where the current year inflation rate is calculated based on the
two-year lag period of the expenditure share. However, the consumer expenditure
has changed significantly in the Pandemic period as a result of lockdown measures
and related changes in behaviour.'' Diewert and Fox (2020) suggest using the Fisher
Price Index because the substitution bias in Laspeyres (too high) and Paasche (too
low) or any fixed basket index will be large during the period of exceptional changes
in the consumer expenditure pattern. More studies are also advocating the use of
various superlative price indices such as the Fisher and Térnqvist Index during pan-
demic period, including Fox et al. (2022), Jaravel and O’Connell (2020a), Kantur
et al. (2021), and Alvarez and Lein (2020).

We will briefly revisit the standard formula of the simple fixed basket indexes
and superlative price index.

Laspeyres price index:

pt = i=1 Fj 7i =-y Rw (1)

=l

Paasche price index:
-1

I= Sa = | Dee (2)

Fisher price index:

° The major sources of data used to construct HHFCE estimates are the Retail Sales Inquiry (RSI) and
the Living Costs and Food Survey (LCF). In LCF, annual and quarterly household sample size are 5000
and 1200 households respectively. While RSI is a monthly survey of 5,000 retail businesses including all
900 ‘large retailers’ and 4100 ‘small and medium retailers’ on a sampled basis.

'0 The U.S. Census Bureau collects CE data for BLS. The CE consists of two separate surveys, the Inter-
view Survey and the Diary Survey. BLS designs the Interview Survey to collect data on large and/or
recurring expenditures, such as rent, utilities, or insurance. On the other hand, BLS designed the Diary
Survey to collect data on frequently purchased items, such as food and beverage expenditures at home
and in eating places; housekeeping supplies and services; nonprescription drugs; and most personal care
products and services. The sample size of Interview Survey and Diary Survey are 20,000 and 11,000
independent surveys completed annually respectively.

‘| The literature on the changes in the expenditure patterns during the Pandemic is reviewed in the
online Appendix B.

va Springer
A. Chowdhury, H. Dixon

Pe = x YS (3)

where, /°'= Index value for period t based on period 0.
p'= Price level of item i at period t
Pp, = Price level of item i at period 0 (base period)
q, = Quantity on item i at period t
q; = Quantity on item i at period 0 (base period)

R,” = —, is the price relative
0 peg
w= y. >» is the expenditure share of item i at base period 0
PG: . . . .
wi = Sa is the expenditure share of item i at period t

The Laspeyres price index is thus the weighted arithmetic mean of price relatives.
One of the notable strengths of the Laspeyres price index is CPI can be produced
on a timely basis without knowing the current period quantity information (Diewert
2001). On the other hand, the Paasche price index is a weighted harmonic mean
of price relatives which is based on the current period’s quantities and expenditure
shares. The Paasche price index can only be produced with a delay, as it requires
current expenditure weights which are generated by the NSOs with a lag. However,
both the indexes are easy to calculate and understand. From an economic perspec-
tive, the Laspeyres tends to overstate inflation, since it assumes the base quantities/
expenditure-shares do not respond to price changes, where as the Paasche index
understates inflation since it fixes the current quantities/expenditure-shares. The
Fisher index takes the geometric mean on the two, and will be closer to the true
value. Furthermore, the Fisher index is considered as a superlative or ideal price
index because it passes a wide battery of desirable axiomatic tests (see Diewert
2001, 2021 for a detailed analysis of the indices in relation to a range of axiomatic
tests). Crucially, the Fisher price index satisfies the important time-reversal test
(which derives from the theory of rational choice), which means that if the prices
and quantities in the two periods being compared are interchanged the resulting
price index is the reciprocal of the original price index. Unlike the Fisher Index,
both the Laspeyres and Paasche price index fail this fundamental test.

An alternative and popular index is the Térnqvist price index, which also passes
the time reversal test. As Dumagan (2002) has shown analytically, the Fisher and
ToOrnqvist indices are generally numerically approximate. As a check, we also
applied the Tornqvist index to the US data and found the results very similar to the
Fisher Index (see Appendix D).

We now have the consumer expenditure information for the Pandemic period
(the benefit of hindsight), so we are able to use the Fisher method to measure
the “true” inflation rate during the pandemic, which we call the Pandemic infla-
tion rate, and compare this with the official rate published by the NSO. The
key difference between official inflation and the Pandemic inflation rates are
that the Official inflation rate used a Laspeyres (or Lowe) price index usually
based on one or two-year old consumer expenditure information,!* whereas the

2 For example, in the UK, the 2020 CPI figures were based on 2018 expenditure data, whilst the 2021
CPI data was based on 2020 expenditures and 2022 CPI on 2021 expenditures.

va Springer
Measuring Inflation During the Pandemic with the Benefit of...

Pandemic inflation rate uses both current and base expenditures for each period.
While calculating Pandemic price index, we have used the weights for 2019 as
our pre-pandemic base, needed for constructing the Laspeyres part of the Fisher
index. For the Paasche part of the Fisher index, we will be using the “current
month” expenditure weights based on the frequency of expenditure share data in
each country.

Our method is the “direct” Fisher price index approach (see Australian Bureau
of Statistics (2022)). The “chained” Fisher index is obtained by comparing
each “current” period with using the previous period as the “base”: this gives a
sequence of two-monthly indices which can then be joined together by chaining.
We also calculate the chained Fisher index measure for the US data in Sect. 4
to check whether choosing this methodology yields different results. Chaining
of course has its own problems: one-off outliers in the time series can become
permanently embedded in the index (chain drift), which is a potential issue even
without the pandemic due to the seasonality of expenditure data. However, we
found that the chained Fisher was little different to the direct Fisher index when
we compare them using the US data.

Finally, we will compare our Pandemic inflation rate with Cavallo (2020) and
Reinsdorf (2020) Covid inflation rates in detail. The methodology and consumer
expenditure information used in both the studies are different from this study.
Cavallo (2020) uses a chained index and Reinsdorf (2020) uses the Paasche
index. They also use different expenditure data. We are able to show that what
drives the differences in our estimates is the expenditure share data used.

3.2 Data

In this study, we use (depending on availability) the annual and quarterly/monthly
household final consumption expenditure (HHFCE) data from the national accounts
of 33 OECD countries to directly measure the changes in spending pattern during
the pandemic. In the UK, the HHFCE data is contained in Consumer Trends pub-
lished by the Office for National Statistics (ONS). Statistics Canada and Statistics
Korea also published quarterly consumer expenditure data for the Canada and the
South Korea, which we also converted into monthly data as in the UK (see Appen-
dix E). For the US, we have collected monthly personal consumption expenditure
(PCE) data from the U.S. Bureau of Economic Analysis (BEA), which is proxied by
the consumer expenditure survey (CEX) from the U.S. Bureau of Labor Statistics
(BLS) due to the significant error in collecting data directly from consumers dur-
ing the pandemic period (Curtin 2022). PCE collected data from several statistical
reports, such as reports come from various government agencies, administrative and
regulatory agencies, and other private organizations. The HHFCE data for the rest
of the countries in the list are collected mostly from Eurostat, OECD Stat, and other
statistical agencies, including Australian Bureau of Statistics, Statistics of Japan. In
Appendix-A, we will find the full list of countries along with the frequency of the
data and collecting institution.

va Springer
A. Chowdhury, H. Dixon

4 Impact on the US and the UK Inflation

In this section we will look in detail at the US and UK experience to illustrate the main
issues involved in measuring the effect of the pandemic on inflation. Each country has
a slightly different set up for its CPI and indeed how it dealt with the pandemic.

4.1 Impact on US Inflation

The first COVID-19 case was found in the US on the 20th of January 2020. Since then,
more than one-million people have died and the US remains the top country in the world
in terms of total cases and deaths. Table | gives us information regarding the change in
the expenditure pattern by consumers during the period of the Pandemic. The expendi-
ture pattern changed dramatically in the first wave, i.e., from March to May 2020.

On 13th March 2020, the president of the United States (POTUS) declares the
Covid-19 outbreak as a national emergency (Department of Defense, n.d.). The
declaration significantly affected consumer behaviour in March 2020 as people
started to stocking food and beverage at home which is reflected in Table 1. In addi-
tion, U.S. State Department issues Global Level 3 Health Advisory: Do Not Travel
on the 15th of March 2020 which is reflected in the expenditure share in transporta-
tion. As the consumer started to cook at home, the expenditure share for the food
away from home has significantly declined from March 2020. Moreover, Centers for
Disease Control and Prevention (CDC) recommends that travelers defer all cruise
travel worldwide and avoid all nonessential travel to China (Department of Defense,
n.d.) that significantly reflect in the consumer expenditure share in recreation (see
Table 1). On August 6, 2020, the U.S. State Department in coordination with the
CDC lifted the Global Level 4 Health Advisory and returning to its previous sys-
tem of country-specific levels of travel advice. That announcement has significantly
affected people’s movement along with increase in the expenditure share in the third
quarter of 2020 in compared to the second quarter in several categories, such as
Food and beverages purchased for off-premises consumption; Clothing, footwear,
and related services; Transportation; Recreation; Food services and accommoda-
tions (see Table 1). In December, U.S. Food and Drug Administration has started
to issue authorization to emergency use for Pfizer and Moderna’s Covid-19 vaccine
(Department of Defense, n.d.). This declaration has brought mass people’s confi-
dence as well as in their consumption behaviour.

Figure | depicts the official inflation rate and the Pandemic inflation rate, as well as
the corresponding indices. Significant differences have been detected in the US infla-
tion during the Pandemic period in these two measures. In 2020 the Pandemic inflation
rate was higher, reflecting large price changes and the increased consumer expenditure in
food and non-alcoholic beverage, alcoholic beverage, housing, and household services. In
2021, the situation was reversed. This is partly a result of the base effect (higher pandemic
prices in 2020 tend to reduce inflation implied by prices in 2021), but also the differences
in expenditure shares (the Fisher price index price levels are below the official levels).

va Springer
A. Chowdhury, H. Dixon

112.000 8.00

110.000 7.00

108.000 6.00

106.000 3 5.00

104.000 ~ 3 3 4.00
3 3 4

102.000 > ? 3 3.00
3 ? 3

100.000 4 Fo 4 2.00

98.000 x 3 4 1.00

96.000 : a 0.00

wens sans Official Price Index ‘¥ 9? Pandemic Price Index

Official inflation rate (annual) —©<©« Pandemic inflation rate (annual)

Fig. 1. US Official and Pandemic Inflation Source: Author calculation

The U.S. Bureau of Economic Analysis also calculates an alternative price index
known as the Personal Consumption Expenditure (PCE) Price Index, which is cal-
culated using the Fisher index.!* This is the measure primarily used by the FED
in formulating its policy. Curtin (2022) has also confirmed that during pandemic
the data quality in the CE Survey by the BLS is poor in comparison to the PCE
by BEA. In Fig. 2 we show the PCE inflation measure in addition to the Official
and Pandemic inflation rates. Both PCE and Pandemic inflation rates has used the
Fisher price index and the identical consumer expenditure share, but in this study,

8.00
7.00
6.00
5.00
4.00 aan
3.00 7
1.00 -
0.00

Official Inflation rate (annual) PCE Inflation rate (annual) == «= Pandemic inflation rate (annual)

Fig. 2. US Official, PCE and COVID Inflation Source: Author calculation

‘3 The relative weights used in the PCE index are derived from business surveys—for example, the Cen-
sus Bureau’s annual and monthly retail trade surveys, the Service Annual Survey, and the Quarterly Ser-
vices Survey.

va Springer
Measuring Inflation During the Pandemic with the Benefit of...

8.00
7.00

6.00

5.00 7
4.00 7
3.00 -

2.00 ~ ae se ?
1.00 we

0.00

Pandemic inflation rate- annual, fisher price index-chained
= © Pandemic inflation rate- annual, fisher price index-direct

Official Inflation rate-annual

Fig. 3 Fisher Price Index US: Chain Vs Direct Source: Author calculation

we calculate Pandemic price index as the weighted average of the PCE share and the
Consumer Price Index for All Urban Consumers (CPI-U). That’s why PCE inflation
rate is closer to the Pandemic inflation rate. In 2020, the PCE followed the Official
figure more closely (both below the pandemic measure), whilst in 2021 the PCE fol-
lows the Pandemic measure more closely (both below the official measure).

4.1.1 Pandemic inflation and Cavallo (2020) Covid inflation

Figure 4 highlights the differences between the Pandemic inflation rate and the
Covid inflation rate calculated in this study and the Cavallo (2020) respectively.
Before comparing these two approaches of calculating annual inflation rate of US
during the Pandemic period (2020 to 2021), it is worth to mention the methodo-
logical differences between these two measurements.

7.00
6.00

5.00 ,
4.00

3.00 3
2.00 ~ /
1.00 ~

0.00

= © Pandemic inflation rate (annual) COVID inflation rate (annual) - Cavallo, 2020

Fig.4 Pandemic inflation rate (annual) and COVID inflation rate (annual)- Source: Author calculation

va Springer
A. Chowdhury, H. Dixon

First, Cavallo (2020) used a chained price index to calculate Covid infla-
tion rate. With chaining, prices in each month are compared to the previous
month using the expenditure weights in the two consecutive months, where in
our approach we use the current month and base month weights (the “direct”
method). Our method is more comparable to the official methodology, which
uses fixed weights within the year. In Fig. 3 we compare the annual Pandemic
inflation using both the direct and chained method to calculate the price index for
the US, and they are very close (it appears chain drift was not an issue). There-
fore, the difference between the inflation measurement during the Pandemic in
this study and Cavallo (2020) isn’t driven by the use of chained or direct Fisher
methods.

Second, expenditure share/weights used in this paper are updated from Per-
sonal Consumption Expenditure (PCE) by The U.S. Bureau of Economic Analy-
sis (BEA), whereas Cavallo (2020) updated the weight from data collected by
Opportunity Insights (based at Harvard University). PCE collected data from sev-
eral statistical reports, such as reports come from various government agencies,
administrative and regulatory agencies, and other private organizations. On the
other hand, Opportunity Insights (OD collected data from credit and debit card
spending collected by Affinity Solutions Inc. Therefore, whilst neither dataset
wasn’t collected from consumer-based sample like Consumer Expenditure Survey
(CE) by the U.S. Bureau of Labor Statistic (BLS).

The key reason that our measure differs from Cavallo is the changes in weights
for different categories of products in the CPI basket during pandemic (see
Appendix C for a full breakdown). The source of the lower inflation rate esti-
mated by Cavallo from March’20 to July’20 are:

1. The changes in Cavallo weights in the health category are more, being -13%,
-39%, -19% in March, April and May 2020 as compared to January 2020. The
comparative changes in weights from PCE where only -3%, -10% and -4%...

2. For recreation activity the Cavallo expenditure weight fell significantly across
all months in 2020 compared to January. For example, changes in the Cavallo
(2020) weights were -25%, -60%, -48%, -32%, and -34% for March, April, May,
June and July 2020 respectively. The corresponding changes in PCE weights were
only -6%, -12%, -8%, -2%, and -3%.

3. The expenditure share changes in the food away from home (food services) in
Cavallo were also very large -50% and -37% respectively in April and May 2020.
The corresponding changes in PCE weights are -41% and -32%.

Lower weights in these sectors (and higher ones in other sectors) combined
with the price changes led to the differences in the inflation rate Fig. 4.

The pandemic inflation rate calculated in this study from November 2020 to
December 2021 was lower than Cavallo.

The main reasons behind these discrepancies in these two approaches again
reflects differences in the changes in expenditure shares relative to January 2020:

va Springer
Measuring Inflation During the Pandemic with the Benefit of...

1. In food, the weight changes in Cavallo (2020) are consistently higher than the
PCE weight changes. For example, the weight changes in Cavallo (2020) are
32% and 34% in January and February 2021 compared to 14% and 13% in PCE
expenditure weights. Since food had significantly high inflation this raised the
Cavallo measure more than our estimate.

2. For apparel, there is a base effect as negative annual inflation rate prevails till
March’21 and returned to positive until the end of 2021. The weight changes in
Cavallo uare higher than the PCE weight changes from October’20 to December’21.
Again, higher inflation rate along with the higher weight led higher inflation rate
calculated in Cavallo (2020).

3. Similar to apparel, a base effect was also seen in transportation category where
official inflation rate increased significantly from March 2021. The weight change
in Cavallo was negative but significantly increased from January 2021 as com-
pared to PCE. This also lead to higher covid inflation rate for Cavallo in com-
parison to this study’s pandemic inflation rate.

As we can see, there are significantly different expenditures weights in Cavallo
(2020) as compared to the PCE weights. This probably reflects the partial coverage of
the predominantly credit card data underlying Cavallo, which excludes cash payments
and direct bank transfers which are more important is some sectors than others.

4.2 Impact on UK Inflation

The first nationwide lockdown was enforced on the 23rd of March 2020, and it con-
tinued until the relaxing of restrictions and social distancing rules on the 23rd of
June 2020. The subsequent lockdowns were the Second Lockdown (Sth November
2020 to 2nd December 2020) and Third Lockdown (6th January 2021 to 12th April
2021).'4 From the literature, we have already identified that the consumers’ expendi-
ture pattern has changed significantly during the lockdown period. Table 2 has well
demonstrated this concern with the help of actual household final consumption
expenditure during the COVID-19 pandemic period. Major changes took place dur-
ing the first lockdown. Among the 12 divisions in the COICOP category, significant
changes happened in the Food and non-alcoholic beverages; Alcoholic beverages
and tobacco; Housing, water, electricity, gas, and other fuels; Transport; Education;
Restaurants and hotels division. However, the expenditure patterns were moving to
their previous track after the end of the third lockdown in the UK.

Figure 5 shows how far the official headline inflation rate is from the pandemic
inflation rate (the Fisher index). The figure depicts both the price-index level (in
bars) and the Pandemic inflation rate. The two measures diverge in two periods: the
official inflation rate is slightly higher than the pandemic inflation in the first lock-
down period (i.e., second quarter of 2020) and lower after the third lockdown period
(i.e., second quarter of 2021) till at end of 2021. The major contributors accounting

'4 For details see Institute for Government Analysis, UK.

va Springer
Measuring Inflation During the Pandemic with the Benefit of...

109.00 6.00

108.00

107.00 5.00

106.00
4.00
105.00

104.00
3.00
103.00

102.00
2.00

101.00
100.00 1.00

99.00

98.00 0.00

antici Official Price Index ravuvuray Pandemic Price Index
Official inflation rate (annual) — =

Fig. 5 UK Official and COVID Inflation Source: Author calculation

for the differences in the second quarter of 2020 inflation were restaurant and hotels,
recreation and culture, housing and household services, food and non-alcoholic bev-
erage, clothing and footwear, and transport. As the official inflation used the ear-
lier expenditure weight, the higher inflation in the restaurant and hotels, recreation
and culture divisions have positively contributed to the overall inflation. However,
the increased expenditure in the food and non-alcoholic, housing and household
services divisions has also positively contributed to higher pandemic inflation but
remaining below the official headline inflation rate because of the sharp decline in
the expenditure on transport where inflation rate was largely negative.

In 2021, the ONS updated its expenditure weight significantly considering the
dramatic change in the expenditure behaviour in 2020. The major contributor to
the inflation in the second quarter of 2021 are housing and household services,
transport, recreation and culture, other goods, and services divisions. As the actual
expenditure pattern has started to return to the earlier track, the increase in transport,
restaurants and hotels expenditure weight in the pandemic inflation calculation is
the main factor behind the difference with the official inflation rate. However, the
official inflation rate and the pandemic inflation rate look quite similar after the
third quarter of 2021 because the increased inflation rate in the housing and house-
hold services and the transport division were cancelled out due to the expenditure
weight in these two different inflation rate calculations. The official expenditure
weight of the housing and household services is higher than the actual expenditure
and the vice-versa for the transport after the second quarter of 2021. These are the
main divisions where the change took place in the form of the inflation rate as well
as the expenditure weight in the two methods of calculating the inflation rate.

va Springer
A. Chowdhury, H. Dixon

The UK experience shows how the comparison of the official inflation figures
with the pandemic measure depends on a detailed comparison of often offsetting
effects across the expenditure categories with offsetting trends.

5 Impact on the other OECD Countries

The US and UK provide contrasting experiences of how the pandemic inflation
and official figures can differ. In the UK there is a small difference, with pandemic
inflation being slightly lower than the official figures, whilst there is a more sub-
stantial difference in the US with pandemic inflation being higher in 2020 and lower
in 2021. We now go on to explore the differences across the remaining OECD coun-
tries using the same methodology, as shown in Table 3. The third and fourth col-
umns give us the difference between the official and pandemic inflation: in the third
column we see the difference in 2020 Q2 and in the fourth column we see the differ-
ence over the two years 2020-2021.

Among the rest of the OECD countries, the official inflation rate is mostly higher
than the pandemic measure in the second quarter of 2020 and less over the whole
COVID period of 2020-21. A common pattern in those countries is that the bias
is greater in the second quarter of 2020 compared to the whole year as a result of
the large change in the expenditure patterns in that quarter. An identical pattern
observed in almost all the countries with a higher official rate: (4) pandemic weight
is higher where the price level is decreasing; (ii) official weight is higher where the
price level is increasing.

In two countries out of the sample (Iceland and Poland), the official inflation rate
was lower than the pandemic inflation in both the second quarter of 2020 and the
whole COVID period of 2020-21. Among these two, the changes in the expenditure
pattern were quite similar. In these countries, pandemic weights have increased in food
and non-alcoholic beverage, alcoholic beverage, housing and household services, fur-
niture, health, communication, and education. In addition, the price level has increased
in food and non-alcoholic beverage, alcoholic beverage, housing and household ser-
vices, furniture, health, recreation and culture, education, restaurant, and hotels; and
the price level has decreased in clothing and footwear, transport, and communication.
Again, a similar pattern has also been observed in those countries where the official
inflation rate is higher, they are: (i) pandemic weight is higher where the price level is
increasing; (ii) official weight is higher where the price level is decreasing.

For most countries, there are mixed results, meaning the bias during 2020 Q2 was
the opposite to the whole period 2020-2021. In Norway, the official inflation rate
was significantly higher than the pandemic in 2020 and lower over the whole period.
The opposite scenario was seen the Czech Republic where the official inflation rate
was lower in the second quarter of 2020 and higher for the whole. Clearly, the differ-
ences can be very small (smaller than 0.1 pp in absolute terms) ranging to over | pp
in absolute terms in 2020 Q2 (Norway and the US). There is great heterogeneity
in how the pandemic impacted om the inflation experience in different countries,
reflecting different expenditure patterns, government policies and the behavior of
households and firms.

va Springer
Measuring Inflation During the Pandemic with the Benefit of...

(L) (9)

pringer

Qs
Measuring Inflation During the Pandemic with the Benefit of...

(L) (9)

pringer

Qs
A. Chowdhury, H. Dixon

In columns 5, 6 and 7 we show the earlier results obtained by Cavallo (2020)
and Reinsdorf (2020) using real time debit and credit card-based expenditure data.
Whereas this study is based on the actual data from the national accounts with the
benefit of hindsight, these studies were attempts to measure the effects as they were
happening. Both Cavallo and Reinsdorf used credit card data from one country and
applied it to several other countries: Cavallo used US card data, Reinsdorf Canadian
and US data. Even when applied to the US or Canadian data, the credit card data
can be misleading. For example, Reinsdorf (2020) and Cavallo (2020) use 37.12
and 35.80 respectively for the COICOP category 04 housing and household services
in Canada, but the actual data from Statistics Canada gives us 32.59 for this COI-
COP category. In the case of the UK, the official expenditure share for the COI-
COP category 01 food and non-alcoholic beverage in the second quarter of 2020 is
11.3 whereas the expenditure share for the similar category in Cavallo (2020) is esti-
mated as 29.5 in April 2020.'> The difference is quite high and the potential source
of bias in the inflation measurement. Whilst using card data provides real time infor-
mation, it can be significantly different to the real expenditure behavior since it only
captures a part of expenditure.

6 Conclusion

During the coronavirus pandemic, specifically the first wave (second quarter of
2020) of the pandemic, the official Consumer Price Index (CPI) inflation figure
wasn’t as informative as usual about the balance of supply and demand in the econ-
omy. The reason behind that is the conceptual challenges that have affected price
measurement during the lockdown period. The major change were the large
shifts in spending patterns due to the stringent lockdown policies adopted by the
government across the world, which changed the representative household con-
sumption basket.

The aim of this study is to identify whether the official inflation rate was higher
or lower than the true inflation rate (using the Fisher index) during the COVID-
19 pandemic. Adopting the actual household expenditure data from the national
accounts, this study found that the official inflation rate in the 33 OECD countries
was an overestimate of true inflation for 23 and underestimate in 10 countries in the

'S Cavallo (2020) used Spanish card-based expenditure data to estimate the weight for the European
countries in his sample.

va Springer
Measuring Inflation During the Pandemic with the Benefit of...

first wave of the COVID-19 pandemic. The result obtained for the underestimating
countries in this study has matched the result obtained by Cavallo (2020) and
Reinsdorf (2020). However, a significant difference has been detected for the coun-
tries where the official inflation rate was higher than the true inflation rate in this
study. The core reason behind the discrepancies in the use of appropriate expendi-
ture weight. Finally, this study suggests caution while adopting card-based expendi-
ture data in calculating any superlative price indexes for measuring general price
levels when spending behaviour has changed dramatically.

One of the limitations of this study is time aggregation. As Ivancic et al. (2011)
showed, time aggregation choices lead to the differences in price estimates of the
chained index for both the superlative index (though relatively lower) and the non-
superlative indices. As this study is based on the national account’s expenditure
share data, there is a possibility of potential time aggregation bias for the countries
which have generally published expenditure share data only on an annual or quar-
terly basis.

Supplementary Information The online version contains supplementary material available at https://doi.
org/10.1007/s11079-024-09776-3.

Acknowledgements We would like to thank the editor and referees for very helpful comments, along
with participants at the Cardiff ONS workshop in 2021, the 8th International Conference on Economics
of the Turkish Economic Association ICE-TEA 2022) and the 2nd Welsh Postgraduate Research Confer-
ence in Business, Management & Economics (WPGRC 2022) organized by the Cardiff Business School.
Faults remain our own.

Data Availability In this study, we recalculate the official CPI of 33 OECD countries based on the
true household expenditure share from the national account during the period of the COVID-19 Pan-
demic. The monthly official CPI/CPIH/HICP index and the Household Final Consumption Expenditure
(HHFCE) are the key data used in this study. All the data used in this study are publicly available and
collected from the following sources:

va Springer
Measuring Inflation During the Pandemic with the Benefit of...

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License,
which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long
as you give appropriate credit to the original author(s) and the source, provide a link to the Creative
Commons licence, and indicate if changes were made. The images or other third party material in this
article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line
to the material. If material is not included in the article’s Creative Commons licence and your intended
use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permis-
sion directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/
licenses/by/4.0/.

References

Alvarez SE, Lein SM (2020) Tracking inflation on a daily basis. SJES 156:1-13

Andersen AL, Hansen ET, Johannesen N, Sheridan A (2022) Consumer responses to the COVID-19 cri-
sis: Evidence from bank account transaction data. Scand J Econ 124(4):905-929

Australian Bureau of Statistics (2022) Price index theory, ABS, viewed 22 Feb 2024

Bachas N, Ganong P, Noel PJ, Vavra JS, Wong A, Farrell D, Greig FE (2020) Initial impacts of the
pandemic on consumer behavior: evidence from linked income, spending, and savings data. NBER
Work Pap Ser. https://doi.org/10.3386/w27617

Baker SR, Farrokhnia RA, Meyer S, Pagel M, Yannelis C (2020) How does household spending respond
to an epidemic? Consumption during the 2020 COVID-19 pandemic. RAPS 10(4):834—862

Balleer A, Link S, Menkhoff M, Zorn P (2022) Demand or supply? price adjustment heterogeneity dur-
ing the Covid-19 Pandemic. Int J Cent Bank forthcoming

Barclays (2020) UK Consumer spending report-August 2020. . PLC., B.B.U

Benchimol J, Caspi I, Levin Y, Benchimol J, Caspi I, Levin Y (2021) The COVID-19 Inflation weighting
in Israel. The Economists’ Voice. https://doi.org/10.1515/ev-202 1-0023

Bonam D, Smadu A, Bonam D, Smadu A (2021) The long-run effects of pandemics on inflation: Will
this time be different? Econ Lett 208:110065. https://doi.org/10.1016/j.econlet.2021.110065

Carvalho VM, Garcia JR, Hansen S, Ortiz A, Rodrigo T, Rodriguez Mora JV, Ruiz P (2021) Tracking the
COVID-19 crisis with high-resolution transaction data. R Soc Open Sci 8(8):210218. https://doi.
org/10.1098/rsos.210218

Cavallo A (2020) Inflation with covid consumption baskets. NBER Work Pap Ser. https://doi.org/10.
3386/w27352

Chetty R, Friedman JN, Hendren N, Stepner M, Team TOI (2020) How did COVID-19 and stabilization
policies affect spending and employment? A new real-time economic tracker based on private sector
data. NBER Work Pap Ser. https://doi.org/10.3386/w2743 1

Chronopoulos DK, Lukas M, Wilson JO (2020) Consumer Spending Responses to the COVID-19 Pan-
demic. An Assessment of Great Britain. Available at SSRN 3586723

Curtin S (2022) The Impact of COVID-19 on CE Survey Estimates for 2020. Consumer Expenditure Sur-
veys Program Report Series. U.S. Bureau of Labor Statistics (BLS). Available at. https://www.bls.
gov/cex/research_papers/pdf/curtin-2020-annual-COVID-impact.pdf

Department of Defense (n.d.) Coronavirus DOD Response Timeline. Retrieved from https://www.defense.
gov/Spotlights/Coronavirus-DOD-Response/Timeline/. Accessed: 19 Feb 2024

Diewert WE, Fox KJ (2020) Measuring Real Consumption and CPI Bias under Lockdown Conditions.
NBER Work Pap Ser. https://doi.org/10.3386/w27 144

Diewert WE (2001) The consumer price index and index number theory: a survey Vancouver. Depart-
ment of Economics, University of British Columbia, pp 1-104

Diewert, W.E. (2021), “The Axiomatic or Test Approach to Index Number Theory”, Chapter 3 in Con-
sumer Price Index Theory, Washington D.C.: International Monetary Fund. https://www.imf.org/en/
Data/Statistics/cpi-manual

Dixon H (2020a) The measurement of inflation during the lockdown: a trial calculation. NIESR Policy
Paper 016, April 2020

Dixon H (2020b) Will CPIH inflation understate inflation for April 2020? , NIESR Blog, 8" May 2020.

Dumagan JC (2002) Comparing the superlative Toérnqvist and Fisher ideal indexes. Econ Lett
76(2):25 1-258

va Springer
A. Chowdhury, H. Dixon

Dunn A, Hood K, Driessen A (2020) Measuring the effects of the COVID-19 pandemic on consumer
spending using card transaction data. US Bureau of Economic Analysis Working Paper WP2020-5

Fisher I (1921) The best form of index number. Q Publ Am Stat Assoc 17(133):533-551

Fox KJ, Levell P, O’Connell M (2022) Multilateral index number methods for Consumer Price Statistics
(No. ESCoE DP-2022-08). Economic Statistics Centre of Excellence (ESCoE)

Ivancic L, Diewert WE, Fox KJ (2011) Scanner data, time aggregation and the construction of price
indexes. J Econom 161(1):24-35

Jaravel X, O’Connell M (2020b) Inflation spike and falling product variety during the Great Lockdown.
IFS Working Paper W20/17

Jaravel X, O’Connell M (2020a) High-frequency changes in shopping behaviours, promotions and the
measurement of inflation: evidence from the great lockdown. Fisc Stud 41(3):733-755

Kantur Z, Ozcan G, Kantur Z, Ozcan G (2021) What pandemic inflation tells: Old habits die hard. Econ
Lett 204:109907. https://doi.org/10.1016/j.econlet.2021.109907

Reinsdorf M (2020) COVID-19 and the CPI: Is inflation underestimated? IMF Working Paper No. 2020/224

Seiler P (2020) Weighting bias and inflation in the time of COVID-19: evidence from Swiss transaction
data. SJES 156(1):13. https://doi.org/10.1186/s41937-020-00057-7

Tenreyro S (2020) Monetary policy during pandemics: inflation before, during and after Covid-19. Bank
of England Speech. Available at https://www.bankofengland.co.uk/speech/2020/silvana-tenreyro-
speech-monetary-policy-during-pandemics

Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps
and institutional affiliations.

va Springer


File and source

File

chowdhury-dixon-2024-oer-measuring-inflation-during-the-pandemic-hindsight.pdf

Size

1,630,932 bytes

SHA-256

bab55e4e0cc3e6cf8951d095e1dd0990e21c331de46d3b60b3a90b63aff39954

Our copy

chowdhury-dixon-2024-oer-measuring-inflation-during-the-pandemic-hindsight.pdf

Original

No public link identified.