Pandemic Darlings The pandemic economy, in original documents
Home Source documents Tracking the Impact of COVID-19 on Economic Inequality

Tracking the Impact of COVID-19 on Economic Inequality

Summary

Tracking the impact of COVID-19 on economic inequality at high frequency is a research article published in PLOS ONE on March 31, 2021, by authors from Caixabank Research and Universitat Pompeu Fabra (UPF) in Barcelona. The article proposes using bank account data to measure wage inequality monthly, applying it to records from CaixaBank covering over three million residents of Spain. The authors report that, absent government intervention, inequality would have increased by almost 30% in just one month, driven primarily by job losses and wage cuts among low-wage earners. They find that extended unemployment insurance and furlough benefits were generally effective at mitigating the increase, though less so among young people and foreign-born workers. The article describes its sample restrictions and closes with a list of references.

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

PLOS ONE

                                                         RESEARCH ARTICLE

                                                         Tracking the impact of COVID-19 on economic
                                                         inequality at high frequency
                                                         Oriol Aspachs1☯, Ruben Durante2,3,4,5☯, Alberto Graziano1☯, Josep Mestres1☯,
                                                         Marta Reynal-Querol2,3,4,5☯, Jose G. Montalvo ID2,4,5☯*
                                                         1 Caixabank Research, Caixabank, Barcelona, Catalonia, Spain, 2 Department of Economics and Business,
                                                         Universitat Pompeu Fabra (UPF), Barcelona, Catalonia, Spain, 3 ICREA, Barcelona, Catalonia, Spain,
                                                         4 Institute for Political Economy and Governance (IPEG), Barcelona, Catalonia, Spain, 5 Barcelona Graduate
                                                         School of Economics (BGSE), Barcelona, Catalonia, Spain
a1111111111
                                                         ☯ These authors contributed equally to this work.
a1111111111                                              * jose.garcia-montalvo@upf.edu
a1111111111
a1111111111
a1111111111
                                                         Abstract
                                                         Pandemics have historically had a significant impact on economic inequality. However, offi-
                                                         cial inequality statistics are only available at low frequency and with considerable delay,
    OPEN ACCESS                                          which challenges policymakers in their objective to mitigate inequality and fine-tune public
Citation: Aspachs O, Durante R, Graziano A,              policies. We show that using data from bank records it is possible to measure economic
Mestres J, Reynal-Querol M, Montalvo JG (2021)           inequality at high frequency. The approach proposed in this paper allows measuring, timely
Tracking the impact of COVID-19 on economic
                                                         and accurately, the impact on inequality of fast-unfolding crises, like the COVID-19 pan-
inequality at high frequency. PLoS ONE 16(3):
e0249121. https://doi.org/10.1371/journal.               demic. Applying this approach to data from a representative sample of over three million res-
pone.0249121                                             idents of Spain we find that, absent government intervention, inequality would have
Editor: Shihe Fu, Xiamen University, CHINA               increased by almost 30% in just one month. The granularity of the data allows analyzing
                                                         with great detail the sources of the increases in inequality. In the Spanish case we find that it
Received: September 18, 2020
                                                         is primarily driven by job losses and wage cuts experienced by low-wage earners. Govern-
Accepted: March 11, 2021
                                                         ment support, in particular extended unemployment insurance and benefits for furloughed
Published: March 31, 2021                                workers, were generally effective at mitigating the increase in inequality, though less so
Copyright: © 2021 Aspachs et al. This is an open         among young people and foreign-born workers. Therefore, our approach provides knowl-
access article distributed under the terms of the        edge on the evolution of inequality at high frequency, the effectiveness of public policies in
Creative Commons Attribution License, which
                                                         mitigating the increase of inequality and the subgroups of the population most affected by
permits unrestricted use, distribution, and
reproduction in any medium, provided the original        the changes in inequality. This information is fundamental to fine-tune public policies on the
author and source are credited.                          wake of a fast-moving pandemic like the COVID-19.
Data Availability Statement: Data cannot be
shared publicly because they are owned by a third-
party commercial bank (Caixabank), and there are
legal restrictions to their use. The Legal Services of
the bank accepted the use of the microdata only to
researchers belonging to their Research Unit.
Therefore, the researchers of the team that did not      Introduction
belong to Caixabank Research could not access the        The COVID-19 pandemic has had a massive impact on economic activity around the globe.
microdata. They contributed with the
                                                         To tackle the economic consequences of the pandemic, most governments have used a combi-
conceptualization of the research, the writing of
code, the proposal of different empirical exercises
                                                         nation of family income support and credit facilities for firms. In particular, expanded unem-
and the writing of the manuscript. Data can be           ployment insurance and furlough schemes have been adopted to stabilize the income of the
made available by Caixabank Research (contact via        workers, and contain the impact of the crisis on consumption and economic inequality. The



PLOS ONE | https://doi.org/10.1371/journal.pone.0249121 March 31, 2021                                                                                        1 / 14
PLOS ONE                                                                              Tracking the impact of COVID-19 on economic inequality at high frequency



oaspachs@caixabank.com) to professional                concern is that a surge in inequality may erode social cohesion and spur support for populist
researchers who meet the criteria to access            or even undemocratic views.
confidential data. Researchers interested in
                                                           Yet, how appropriate and effective these policies are remains unclear, mainly due to a lack
obtaining access to the data are required to submit
a written application to Caixabank Research with a     of reliable indicators allowing to track economic activity at a fine temporal resolution. Indeed,
detailed research proposal consisting in a research    most official statistics on inequality are available only at yearly frequency and often with long
question and motivation, information on the            delays. This limits the ability of policymakers to rapidly adjust their responses in the effort to
researcher CV, and a detailed explanation of the       “flatten the recession curve” [1] after flattening the infection curve.
data needed, and the aggregation criteria to protect
                                                           The COVID-19 has pushed new international initiatives to track economic activity in real
the anonymity of the registers. The authors will
provide assistance to any researcher willing to
                                                       time [2–6]. Researchers analyze the impact of economic stimulus packages to mitigate the
analyze the data for replication purposes.             effect of the COVID-19 epidemic on economic activity using high-frequency administrative
                                                       data. Two examples are the effect on aggregate employment of the Paycheck Protection Pro-
Funding: JGM & MRQ: ECO2017-82696P, Spanish
Ministry of Science and Innovation (http://www.
                                                       gram of the US [7] or the effect on consumption of the stimulus checks sent by the US Admin-
ciencia.gob.es/); CEX 2019-000915S Severo Ochoa        istration [8] using the data from financial aggregation and service apps [9–12].
Program for Centers of Excellence (http://www.             One characteristic aspect of pandemics is their impact on inequality [13, 14]. However, offi-
ciencia.gob.es/). The Research Department of           cial inequality measures are calculated with long lags and low frequency. In the context of a
CaixaBank provided support in the form of salaries     fast-moving pandemic it is important to have a high-frequency measure of inequality to evalu-
for authors OA, AG and, JM. The specific roles of
                                                       ate the mitigating effect of policy measures. This is particularly important in countries, like
these authors are articulated in the ‘author
contributions’ section. JGM, MRQ and RD                Spain, that suffered very intensively the financial crisis of 2008 and that have experienced an
acknowledge the financial support of the "Ayudas       important increase in inequality since then. This process increased the support for populist
Fundación BBVA a Equipos de Investigación            parties, which in 2008 were not represented in the parliament and in 2020 accounted for
Cientı́fica SARS-CoV-2 y COVID-19.” The funders        32.8% of the representatives in Congress. It is interesting to notice that inequality increased
had no role in the study design, data collection and
                                                       significantly from 2008 to 2012 but the process of growing political representation of populist
analysis, decision to publish, or preparation of the
manuscript.
                                                       parties happens mostly after 2013, even though inequality was decreasing since 2013. This
                                                       seems to imply that there may be a threshold level of inequality that, once overcome, can trig-
Competing interests: OA, AG, and JM are
                                                       ger a set of popular grievances that persist over time, generating increasing support for popu-
employees of the Research Department of
CaixaBank. There are no patents, products in
                                                       list parties. Therefore, a further increase in inequality, even in the short run, could imply
development or marketed products to declare. This      reaching a level of inequality above the threshold that triggers future tension and political
does not alter our adherence to PLOS ONE policies      unrest. It could also ignite a process of increasing support for populist parties that could easily
on sharing data and materials.                         produce a significant deterioration of the institutional stability. Ultimately, this could have a
                                                       long run effect on economic performance.
                                                           This paper uses bank account data and proposes a methodology to track the impact of gov-
                                                       ernment policies on inequality immediately after they are taken. Inequality is a multifaceted
                                                       object and can concern dimensions as different as income, wealth, education etc. Our analysis
                                                       focuses on wage inequality which, in countries with a high proportion of wage-earners, is a
                                                       very precise indicator of overall income inequality (as we document for Spain). We do not
                                                       look at wealth inequality mainly because, using information from just one financial institution,
                                                       there is a high risk of not gauging a complete picture of the financial holdings of an individual.
                                                       Bank account data have many advantages to study the effect of policy responses to the
                                                       COVID-19 pandemic. They provide timely and reliable information on wages and govern-
                                                       ment benefits. Being able to use very granular data, and to construct a high-frequency measure
                                                       of inequality, allows to tailor policies to contain the increase of inequality in general, and by
                                                       subgroups of the population classified by income level, gender, age, and county of birth.
                                                           Recent research has also used bank account data to study the evolution of different macro-
                                                       magnitudes at very high frequency and, in particular, the effects of the pandemic on consump-
                                                       tion [15–17]. Our contribution to this literature is threefold. First, and opposite to many
                                                       papers in this literature [9, 10], our sample is very representative of the population of Spanish
                                                       wage-earners. As we show in next section, the distribution by gender and age are almost iden-
                                                       tical to the figures reported by the National Statistical Office. Second, and in contrast with a
                                                       large part of the literature that uses banks accounts data, we are not analyzing the evolution of



PLOS ONE | https://doi.org/10.1371/journal.pone.0249121 March 31, 2021                                                                                  2 / 14
PLOS ONE                                                                  Tracking the impact of COVID-19 on economic inequality at high frequency



                                           expenditure but the changes in the distribution of wages over time. Finally, the papers that
                                           have data on expenditure and income, like [16], deal with the issue of the sensitivity of con-
                                           sumption to income and do not consider the evolution of inequality, which is our basic
                                           objective.
                                               We study empirically the evolution of inequality, before and after considering government
                                           support, comparing the period before the lockdown with the lockdown stage. We apply this
                                           methodology to data from a large Spanish bank. Spain is one of the countries most affected by
                                           the pandemic not only in terms of the number of people infected, but also regarding the eco-
                                           nomic impact. The comparison of the situation before and after the activation of the new poli-
                                           cies of income support allows analyzing the effect of government interventions in the
                                           mitigation of inequality.
                                               Using these bank account data, and our research design, we find that the largest impact of
                                           COVID-19 on inequality is transmitted through the movement of the distribution of salary
                                           changes among low wage earners. Second, we also find that most of the increase of inequality
                                           in the period after the beginning of the pandemic is mitigated by the action of the new
                                           extended unemployment benefits and furlough schemes activated by the government. There
                                           are no other changes in other government benefits during the period of analysis. We provide
                                           further details on the public income measures to support workers in the next section. Third we
                                           show that the policy response could not fully mitigate the large increase in inequality among
                                           young people and foreign-born individuals.


                                           Materials and methods
                                           We study the effect of COVID-19 on inequality using bank account data from CaixaBank, the
                                           second largest Spanish bank. Caixabank had monthly records on more than 3 million wage
                                           earners in 2020, and accounted for 27.1% of the wages, salaries and benefits deposited monthly
                                           in the Spanish financial sector. In Spain, differently from other countries like the US, the pay-
                                           ment of the salaries or benefits using checks is a very rare event. Almost all the payments of sal-
                                           aries and benefits use direct deposits on bank accounts.
                                               The wages and government benefits recorded by CaixaBank provide a large, precise and
                                           granular data source. Banks’ administrative data avoid most of the problems of surveys: there
                                           are no measurement errors or imperfect recollection mistakes, and they are obtained with
                                           short delays compared to surveys. For instance, the CaixaBank data provides the universe of
                                           wages through June 15, 2020 while the latest official measure of wage inequality in Spain, pro-
                                           duced by the National Institute of Statistics, was published at the end of June of 2020, but
                                           referred to the situation in 2018.
                                               The granularity of CaixaBank data allows also calculating inequality for subgroups of the
                                           population. Unlike other financial institutions, such as digital banks and personal finance
                                           management software, CaixaBank collects demographic information directly (gender, age,
                                           province, country of birth). We also provide a methodology to calculate monthly Gini indices
                                           and Lorentz curves, before and after accounting for public benefits, to analyze if the schemes
                                           to support workers temporarily out of the labor market are being effective at containing
                                           inequality.
                                               The raw data are the wages and salaries deposited monthly at CaixaBank, and they present
                                           some challenges in order to construct wage inequality measures. We restrict our sample to
                                           accounts with either only one account holder or with multiple account co-holders but only
                                           one employer paying-in wages. This way, we ensure that payrolls or transfers recorded corre-
                                           spond to only one individual and avoid recording multiple payrolls or transfers from multiple
                                           account holders. In addition, we exclude from the sample those individuals who died during



PLOS ONE | https://doi.org/10.1371/journal.pone.0249121 March 31, 2021                                                                      3 / 14
PLOS ONE                                                                             Tracking the impact of COVID-19 on economic inequality at high frequency



                                                     our period of study or who did not use the bank account for their usual financial transactions
                                                     during the period. Finally, to ensure some stability on the sample of individuals studied, we
                                                     require observing either wages or government benefits during two months (that is, in Decem-
                                                     ber 2019 and in January 2020) prior to the beginning of the period of study (February 2020).
                                                     The S1 File explains all the details of our methodology to select the data.
                                                         Our reference sample includes individuals aged 16-64 who received either wages or unem-
                                                     ployment benefits in December of 2019 and January of 2020. We follow those individuals in
                                                     the months starting in February 2020. Since our main source of data is related with holding a
                                                     bank account it is important to start analyzing the level of financial inclusion in Spain. The
                                                     data of the Global Findex, the index of financial inclusion of the World Bank, shows that
                                                     97.6% of Spanish people over 15 years old holds a bank account when the average in high
                                                     income countries is 93.7%.
                                                         We exclude the self-employed from our sample since it is difficult to calculate their net
                                                     monthly income from bank account data: they receive payments from many different sources,
                                                     and it is complicated to calculate expenses associated with their business. However, it is impor-
                                                     tant to note that the proportion of wage earners among the Spanish working population was
                                                     84.4% in the first quarter of 2020 (Labor Force Survey of Spain, EPA). The relevance of wages
                                                     as the main source of income can also be seen in the similarity of the inequality measures
                                                     using income or gross wages. For instance, for the last period for which both measures are
                                                     available, income inequality in Spain, measured by the Gini index, was 0.345 while wage
                                                     inequality was 0.343.
                                                         Since most of the individuals in the sample are workers, to analyze its representativeness in
                                                     terms of the distribution of wages we compare our data with the data of the latest Spanish
                                                     National Statistical Office’s Wage Survey (Encuesta de Estructura Salarial, EES). For this pur-
                                                     pose we consider the individuals in our sample who were working in February of 2020. First,
                                                     we compare the distribution of individuals by gender and age with other sources. Table 1 sum-
                                                     marizes the comparisons. In general, samples from digital banks and financial aggregation ser-
                                                     vices have more young males than the general population. This is not the case with large and
                                                     diversified traditional banks like our data source, CaixaBank. Table 1 shows that the gender
                                                     and age distribution of our data is very similar to the working population.


Table 1. Check our data versus labor surveys.
                                           Our sample (CBK)                        EES                         EPA4T19                        EPA1T20
N                                               3,028,204                        209,473                       �200,000                       �200,000
    Gender
Male                                                0.54                           0.52                           0.52                           0.52
Female                                              0.46                           0.48                           0.48                           0.48
    Age
� 19                                                0.01                           0.00                           0.01                           0.01
20-29                                               0.18                           0.12                           0.15                           0.14
30-39                                               0.25                           0.31                           0.25                           0.24
40-49                                               0.28                           0.30                           0.30                           0.30
50-59                                               0.21                           0.21                           0.23                           0.24
� 60                                                0.07                           0.05                           0.06                           0.06

Notes—EES stands for Encuesta de Estructura Salarial (Spanish Wages Survey); EPA4T19 refers to the sample ofemployees in the Spanish Labor Survey (EPA) in the last
quarter of 2019; EPA1T2020 refers to the sample of employees in theSpanish Labor Survey in the first quarter of 2020.

https://doi.org/10.1371/journal.pone.0249121.t001




PLOS ONE | https://doi.org/10.1371/journal.pone.0249121 March 31, 2021                                                                                       4 / 14
PLOS ONE                                                                         Tracking the impact of COVID-19 on economic inequality at high frequency




                                           Fig 1. Distribution of monthly net wages: Our sample (CABK) versus the sample of the official wage survey (EES).
                                           https://doi.org/10.1371/journal.pone.0249121.g001




                                               In our sample, 54% of the individuals are male. This compares satisfactorily with the 52% of
                                           males in the sample of the last official survey (EES). In order to compare with more recent esti-
                                           mates, columns 3 and 4 include the proportions of males among employees in the Labor Force
                                           Survey of the last quarter of 2019 and first quarter of 2020. February of 2020 is between the last
                                           quarter of 2019 and the first quarter of 2020. The proportion of males is identical to the one in
                                           the EES and very close to the one in our sample. With respect to age, we also find that the pro-
                                           portions of workers in each age bracked in our sample are very similar to those reported in the
                                           EES and the EPA.
                                               Fig 1 shows the distribution of the monthly wages of our sample compared with the distri-
                                           bution of monthly net salaries in the EES. The wages received by workers in their bank
                                           accounts are net of payroll taxes. In order to compare our data with the EES we have calculated
                                           the distribution of net salaries transforming the gross salaries of the EES into net salaries by
                                           subtracting social insurance payments and taxes withheld. The S1 File includes a detailed
                                           explanation of this transformation. Since there is a time difference between the last EES avail-
                                           able and our data we have adjusted the wages by moving the whole distribution by the increase
                                           in the average wage since the last available EES. We can see that the histogram of the net wages
                                           of our sample is very well adjusted by the density estimation of the adjusted distribution of net
                                           salaries in the official wage survey. Both distributions are remarkably similar. The similarity of
                                           the distribution of wages, and also the characteristics of the workforce, confirms the represen-
                                           tativeness of our sample.
                                               Since the distributions are so similar it is not surprising to see that the quantile ratios used
                                           regularly to describe inequality are very similar in both distributions as shown in S1 Table in
                                           S1 File.




PLOS ONE | https://doi.org/10.1371/journal.pone.0249121 March 31, 2021                                                                               5 / 14
PLOS ONE                                                                 Tracking the impact of COVID-19 on economic inequality at high frequency



                                           Government support schemes for workers
                                           The public policy response to mitigate the impact of the COVID-19 crisis in Spain has been
                                           large, as in most developed countries. For a detailed description of the economic impact of
                                           COVID-19 on the Spanish economy and the public policy reaction see [18]. The Spanish gov-
                                           ernment has deployed income and liquidity support measures that are expected to reach 3.7%
                                           of GDP in discretionary measures and around 15.6% of GDP in off-budget measures [19].
                                              Income measures to support workers have consisted mostly in the deployment of a furlough
                                           scheme (“Expediente de Regulación Temporal de Empleo”, or ERTEs) that was scarcely used
                                           until then. This scheme consists in a temporary job suspension (or a reduction in working
                                           hours) that avoids dismissals while maintaining the employment relationship. The Spanish
                                           government facilitated the use of this ERTE scheme due to COVID-19 (considering coronavi-
                                           rus as a force majeure, etc.) and increased coverage to all workers affected by a temporary job
                                           suspension. In addition, the benefits received did not reduce future unemployment benefit
                                           entitlements.
                                              In addition to the job retention scheme, the government facilitated and extended the cover-
                                           age of unemployment benefits. Regular unemployment benefits require a minimum of 360
                                           days of contract employment in the previous 6 years and its duration is proportional to the
                                           amount of time worked (up to 18 months). Due to the pandemic, however, special unemploy-
                                           ment subsidies were created for those who exhausted their unemployment benefits.
                                              Those workers affected by job retention schemes and by unemployment received unem-
                                           ployment benefit transfers, which normally amounts to 70% of their social security contribu-
                                           tion base. The schemes ensured an income stream during the duration of the contract
                                           suspension or unemployment, although of a lower amount than the regular salary.
                                              Public transfers programs partially compensated wage losses for those workers that received
                                           them. However, despite the increase in coverage not all affected workers were entitled or had
                                           the same degree of coverage. In particular, those workers already unemployed before the pan-
                                           demic or in temporary contracts that expired might not have had the right to unemployment
                                           benefits or only to reduced amounts. In addition, many of the beneficiaries experienced several
                                           months of delay before actually receiving their unemployment benefits in their bank accounts.
                                           All these developments might have affected the effectiveness of the government support to
                                           reduce inequality. This is of particular relevance in a country as Spain, which suffers from a
                                           very high labour market duality. In particular, subgroups of the population like young and for-
                                           eign-born individuals are most likely to be in temporary contracts, and thus more heavily
                                           affected.


                                           Results
                                           Large effect of the shutdown on pre-benefits inequality mostly due to low
                                           wage earners
                                           To analyze the role of government benefits on inequality our analysis considers two scenarios:
                                           pre- and post-government benefits. In the pre-benefits scenario, we consider monthly wages
                                           before taking into account the benefits. The post-benefits scenario also considers unemploy-
                                           ment insurance benefits, subsidies and furlough schemes.
                                              Fig 2 shows the distribution of changes in pre-benefit wages between February and April
                                           2020 (i.e., before vs. during the lockdown), represented by the solid lines. The x-axis reports
                                           the percentage change in wages experienced between the two months, while the y-axis reports
                                           the share of account holders in each category. The dashed lines represent the distribution for
                                           the same months of 2019, i.e., prior to the pandemic. The top left panel reports the distribution



PLOS ONE | https://doi.org/10.1371/journal.pone.0249121 March 31, 2021                                                                     6 / 14
PLOS ONE                                                                          Tracking the impact of COVID-19 on economic inequality at high frequency




                                       Fig 2. Changes in payments between April and February by level of wages in the reference period. Pre-benefits scenario.
                                       Comparing 2020 and 2019.
                                       https://doi.org/10.1371/journal.pone.0249121.g002

                                           for the entire sample; the other panels report the distribution for each of five wage brackets
                                           (measured as of February of 2020): i) the interval between 900 to 1,000 euros, which includes
                                           the 25th percentile of the wage distribution; the interval between 1,200 to 1,300 euros, which
                                           includes the median; the interval between 1,700 to 1,800 euros, which includes the 75th



PLOS ONE | https://doi.org/10.1371/journal.pone.0249121 March 31, 2021                                                                                  7 / 14
PLOS ONE                                                                  Tracking the impact of COVID-19 on economic inequality at high frequency



                                           percentile; the interval between 2,900 and 3,000 euros, which includes the 95 percentile, and
                                           the interval between 4,700 and 4,800, which represents the top 1%.
                                               Several interesting facts emerge from Fig 2. First, in 2020 the probability mass of the no-
                                           change interval is about half than in 2019. Compared to 2019, in 2020 a sizeable portion has
                                           moved to the no-income category. Furthermore, and most interestingly, in 2020 the probabil-
                                           ity of shifting to the no-income category is higher for individuals in the lower wage brackets.
                                               Another noticeable aspect is that a substantial share of the highest wage earners experience
                                           a drastic wage reduction in April relative to February. This seasonal pattern, observed both in
                                           2019 and 2020, is due to the payment of bonuses which occurs in February, and reaches over
                                           30% for the top earners in our sample (i.e., the top 0.01%, not shown in the figure). There is no
                                           evidence of an analogous pattern for low wage earners.
                                               S1 Fig in S1 File shows the difference in payments received by account holders between
                                           April and February after accounting for extended unemployment insurance and other bene-
                                           fits. Compared to the pre-benefits wages depicted in Fig 2, the shift to the no-income category
                                           is much less pronounced. There is still a large wage reduction for high earners who are largely
                                           unaffected by government transfers.
                                               S2 Fig in S1 File compares in the same graph all the levels of initial wages, before and after
                                           government benefits, to facilitate the comparisons.
                                               To account for seasonality Fig 3 shows the difference in the proportion of changes in sala-
                                           ries between April and March of 2020 net of the the difference between the same two months
                                           in 2019. The S1 File shows the precise transformation to deal with seasonality. When seasonal-
                                           ity is controlled for, the effect of the February bonuses for high wage earners disappears. Inter-
                                           estingly, the effect of the pandemic on pre-transfer earnings is very different for low and high
                                           wages. For wages below 1,300 euros the lower mass in the no-change brackets is associated
                                           with a corresponding shift to no-income category. The importance of the decline of employ-
                                           ment for the lowest-income workers is common to other countries like the US [20]. For wages
                                           above 1,700 euros, instead, the lower mass in the no-change brackets is associated with a
                                           higher share of individuals experiencing small wage cuts. S3 Fig in S1 File shows the changes
                                           for all wage categories in the same figure.
                                               To summarize the evolution of inequality we compute the Gini index. The S1 File presents
                                           a discussion of its calculation. Fig 4 Panel (a) depicts the evolution of the Gini index between
                                           February and May for 2020 and 2019, respectively. Both the pre- and post-benefits curves are
                                           basically parallel until April 2020, when the pre-benefits Gini index increases considerably
                                           while the post-benefit one only moderately. In May 2020 the pre-benefits Gini index remains
                                           very high, while the post-benefits index returns to the pre-pandemic level. From February
                                           until April of 2020 the pre-benefits Gini index increased close to 0.11 points. This implies a
                                           25% increase in just two months. To evaluate the statistical significance of this large movement
                                           in the Gini index we can calculate the confidence intervals around our estimate. There are
                                           basically two possible procedures: using a Jakknife or a WLS estimator [21]. The S1 File
                                           describes the calculation of the standard error of the Gini index using a WLS estimator. As
                                           expected, given our large sample size, the standard error is very low (0.0002). This implies that
                                           the increase of 0.11 points observed in the Gini index between February and April of 2020 is
                                           highly statistically significant (well over the level of significance of 1%). Since the confidence
                                           intervals are tiny they cannot be visualized in the figures.
                                               To confirm the robustness of the documented pattern to alternative measures of inequality,
                                           in Fig 4 Panel (b) we show the evolution of the Theil index, an inequality measure related to
                                           the concept of entropy and to Shannon’s index. The S1 File discusses the computation of this
                                           index. The Theil index shows a pattern very similar to the Gini index: a sizeable increase in




PLOS ONE | https://doi.org/10.1371/journal.pone.0249121 March 31, 2021                                                                      8 / 14
PLOS ONE                                                                          Tracking the impact of COVID-19 on economic inequality at high frequency




                                       Fig 3. Diff-in-diffs in payments for each level of salaries in the reference month. April vs February—2020 vs 2019.
                                       https://doi.org/10.1371/journal.pone.0249121.g003


                                           March for both the pre- and post-benefits distribution which persists in April for the pre-bene-
                                           fit measure but not for the post-benefit one.
                                               Panels (c) and (d) of Fig 4 show the changes in the pre- and post-benefits Lorenz curves
                                           respectively for every month between February and May 2020. It is apparent that, for the pre-



PLOS ONE | https://doi.org/10.1371/journal.pone.0249121 March 31, 2021                                                                                       9 / 14
PLOS ONE                                                                               Tracking the impact of COVID-19 on economic inequality at high frequency




            Fig 4. Inequality measures. (a) Gini index (b) Theil index (α = 1) (c) Lorentz curve: Pre-benefits, 2020 (d) Lorentz curve: Post-benefits, 2020.
            https://doi.org/10.1371/journal.pone.0249121.g004


                                                 benefits curve, the downward movement accelerates in April and stabilizes in May, while, for
                                                 the post-benefit curve, the evolution is smoother.


                                                 Within group inequality post benefits has increased among young and
                                                 foreign-born people
                                                 Given the granularity of the data we can also analyze the evolution of inequality within differ-
                                                 ent subgroups of the population, differentiating by gender, age, and country of origin. Panel
                                                 (a) in Fig 5 shows that there are not major differences in within inequality of males and females
                                                 before the shock. The magnitude of the increase in the Gini index after the beginning of the
                                                 pandemic is similar across genders before public transfers, but slightly higher for females in
                                                 the post-benefits case.



PLOS ONE | https://doi.org/10.1371/journal.pone.0249121 March 31, 2021                                                                                         10 / 14
PLOS ONE                                                                         Tracking the impact of COVID-19 on economic inequality at high frequency




                                           Fig 5. Evolution of the Gini index by gender, age and country of origin. (a) By gender. (b) By age group. (c) By place
                                           of birth.
                                           https://doi.org/10.1371/journal.pone.0249121.g005

                                              Panel (b) of Fig 5 shows the evolution of inequality for different age groups. For the youn-
                                           gest cohort (i.e., 16 to 29 years old), there is a considerable increase in the Gini index for pre-
                                           transfer earnings. The other groups also experience an increase in inequality though much
                                           smaller than for the young. The spike in the Gini index for the young is mitigated when



PLOS ONE | https://doi.org/10.1371/journal.pone.0249121 March 31, 2021                                                                                    11 / 14
PLOS ONE                                                                  Tracking the impact of COVID-19 on economic inequality at high frequency



                                           considering the distribution of post-benefit earnings. Yet, the level of post-benefits inequality
                                           for this group is still remarkable, both in absolute and in relative terms. Such increase is argu-
                                           ably related to the fact that young workers account for a high proportion of temporary jobs in
                                           low wage occupations.
                                               Panel (c) of Fig 5 shows the evolution of the Gini index separately for foreign-born individ-
                                           uals and for natives. As of January 1st 2020, foreign-born individuals represented 14.77% of
                                           the total Spanish resident population. Looking at the distribution of pre-benefits earnings, it is
                                           clear that inequality increases much more among foreign born than among natives. Such
                                           increase is less pronounced when looking at the post-benefits distribution, though, in this case
                                           as well, the Gini index for foreign born is significantly higher than for natives.
                                               Interesting differences emerge when dividing foreign-born individuals by the per capita
                                           GDP level of the country of origin. For example, as shown in S4 Fig in S1 File, while post-bene-
                                           fits inequality decreases over time for both natives and foreign-born from high-income coun-
                                           tries, it remains high for foreign-born from low-income countries.
                                               The disproportionate increase in post-benefit inequality among poorer migrants attests to
                                           their vulnerability in times of crisis as their social welfare net is thinner. Foreign born workers
                                           from low income countries tend to have occupations with low salaries, and a high proportion
                                           of temporary jobs. In many cases they work without a formal contract which means that they
                                           cannot prove they were working before the pandemic and, therefore, they cannot get the bene-
                                           fits that other workers get. On the other hand expatriates from high income countries still
                                           enjoy a high salary.
                                               Finally, inequality increases more in regions that rely heavily on tourism (e.g., Balearic and
                                           Canary Islands) than in other parts of the country (S5 Fig in S1 File). This is not surprising
                                           since the touristic sector is characterized by a high proportion of low wage workers who, as
                                           shown above, are the ones most affected by the job losses and wage cuts caused by the
                                           pandemic.


                                           Discussion
                                           The financial crisis of 2008 generated a large increase in inequality in many countries. When
                                           some countries were still trying to recover from the financial crisis a new shock, the COVID-
                                           19, has hit the economy. Recent research shows that social distancing laws are not responsible
                                           for the economic harm [15] and the responses to emergency declarations are strongly differen-
                                           tiated by income [22]. In this paper we show that the economic impact is also very heteroge-
                                           neous by income level which, in turn, is reflected in large increases in inequality before
                                           governments policy response.
                                               Our findings contribute to a recent literature on the measurement of economic indicators
                                           in real-time, or at very high frequency. Most of the economic research on the impact of
                                           COVID-19 has concentrated on its effect on consumption [6, 8, 15–17]. We present evidence
                                           on the impact of COVID-19 on economic inequality. Our findings show that, before account-
                                           ing for extended unemployment insurance and furlough benefits, the economic impact of the
                                           pandemic caused a large increase of inequality. After considering public benefits the effect of
                                           the crisis on inequality is mitigated. We show how bank account data of a representative finan-
                                           cial institution can be used to track inequality and monitor the effect of economic polity on its
                                           evolution. In contrast with some previous research that uses data on personal finance websites
                                           and bank accounts, our data replicates very precisely the distribution of the population of
                                           wage earners.
                                               We present evidence that shows a very heterogeneous impact of the pandemic on inequality
                                           by income level, age and country of birth of the individuals. Our methodology could be applied



PLOS ONE | https://doi.org/10.1371/journal.pone.0249121 March 31, 2021                                                                     12 / 14
PLOS ONE                                                                      Tracking the impact of COVID-19 on economic inequality at high frequency



                                           to many other countries that have introduced income-support schemes similar to the ones
                                           considered in Spain (furlough benefits and extended unemployment insurance). Tracking, at
                                           high frequency, the effect of policy responses on inequality allows tuning the policy instru-
                                           ments to mitigate inequality, targeting the groups that contribute the most to the increase of
                                           inequality.


                                           Supporting information
                                           S1 File.
                                           (PDF)


                                           Acknowledgments
                                           We want to thank Miguel Angel Barcia for his helpful suggestions. Daniele Alimonti provided
                                           excellent research assistance.


                                           Author Contributions
                                           Conceptualization: Oriol Aspachs, Ruben Durante, Josep Mestres, Jose G. Montalvo.
                                           Data curation: Alberto Graziano, Josep Mestres.
                                           Formal analysis: Alberto Graziano, Jose G. Montalvo.
                                           Funding acquisition: Ruben Durante, Marta Reynal-Querol, Jose G. Montalvo.
                                           Investigation: Ruben Durante, Josep Mestres, Marta Reynal-Querol.
                                           Methodology: Marta Reynal-Querol, Jose G. Montalvo.
                                           Project administration: Oriol Aspachs, Jose G. Montalvo.
                                           Software: Alberto Graziano, Josep Mestres, Jose G. Montalvo.
                                           Supervision: Oriol Aspachs.
                                           Writing – original draft: Oriol Aspachs, Ruben Durante, Josep Mestres, Jose G. Montalvo.


                                           References
                                            1.   Gourinchas PO. “Flattening the Pandemic and Recession Curves” in Mitigating the COVID Economic
                                                 Crisis: Act Fast and Do Whatever, Ed. Baldwin R., Weder di Mauro B. (Chapter 2, 31–40). CEPR
                                                 Press, London, UK), 2020.
                                            2.   Cicala Steve. Early economic impacts of covid-19 in europe: A view from the grid. Technical report, Uni-
                                                 versity of Chicago, 2020.
                                            3.   Bick Alexander and Blandin Adam. Real time labor market estimates during the 2020 coronavirus out-
                                                 break. Technical report, Arizona State University, 2020.
                                            4.   Chen Sophia, Igan Deniz, Pierri Nicola, and Presbitero Andrea F. Tracking the economic impact of
                                                 covid-19 and mitigation policies in europe and the united states. Technical report, IMF WP/20/125,
                                                 2020.
                                            5.   Chetty Raj, Friedman John N, Hendren Nathaniel, and Stepner Michael. Real-time economics: A new
                                                 platform to track the impacts of covid-19 on people, businesses, and communities using private sector
                                                 data. Technical report, NBER WP 27431, 2020.
                                            6.   Carvalho Vasco M, Hansen Stephen, Ortiz Alvaro, Garcia Juan Ramon, Rodrigo Tomasa, Mora Sevi
                                                 Rodriguez, et al. Tracking the covid-19 crisis with high-resolution transaction data. Technical report,
                                                 CEPR DP14642, 2020.
                                            7.   Autor D., Cho D., Crane L., Goldar M., Lutz B., Montes J., et al. An evaluation of the paycheck protection
                                                 program using administrative payroll microdata. Technical report, MIT WP, 2020.




PLOS ONE | https://doi.org/10.1371/journal.pone.0249121 March 31, 2021                                                                              13 / 14
PLOS ONE                                                                       Tracking the impact of COVID-19 on economic inequality at high frequency



                                            8.   Baker Scott R, Farrokhnia Robert A, Meyer Steffen, Pagel Michaela, and Yannelis Constantine. How
                                                 does household spending respond to an epidemic? consumption during the 2020 covid-19 pandemic.
                                                 Review of Asset Pricing Studies, 2020.
                                            9.   Gelman Michael, Kariv Shachar, Shapiro Matthew D, Silverman Dan, and Tadelis Steven. Harnessing
                                                 naturally occurring data to measure the response of spending to income. Science, 345(6193):212–215,
                                                 2014. https://doi.org/10.1126/science.1247727
                                           10.   Baker Scott R. Debt and the response to household income shocks: Validation and application of linked
                                                 financial account data. Journal of Political Economy, 126(4): 1504–1557, 2018. https://doi.org/10.1086/
                                                 698106
                                           11.   Olafsson Arna and Pagel Michaela. The liquid hand-to-mouth: Evidence from per- sonal finance man-
                                                 agement software. The Review of Financial Studies, 31(11):4398–4446, 2018. https://doi.org/10.1093/
                                                 rfs/hhy055
                                           12.   Kueng Lorenz. Excess sensitivity of high-income consumers. The Quarterly Journal of Economics,
                                                 133(4):1693–1751, 2018. https://doi.org/10.1093/qje/qjy014
                                           13.   Wade L. An unequal blow. Science (New York, NY), 368(6492):700, 2020. https://doi.org/10.1126/
                                                 science.368.6492.700
                                           14.   Scheidel Walter. The great leveler: Violence and the history of inequality from the stone age to the
                                                 twenty-first century. Princeton University Press, 2018.
                                           15.   Sheridan A., Andersen A., Hansen E., and Johannesen N. Social distancing laws cause only small
                                                 losses of economic activity during the covid-19 pandemic in scandi-navia. Proceeding of the National
                                                 Academy of Sciences, (August 2020):1–6, 2020.
                                           16.   Hacioglu S., Kanzig D., and Surico P. The distributional impact of the pandemic. CEPR Discussion
                                                 Paper 15101, 2020.
                                           17.   Bachas N., Ganong P., Noel P., Vara J., Wong A., Farrell D., et al. Initial impact of the pandemic on con-
                                                 sumer behavior: evidence from linked income, spending, and savings data. NBER Working Paper
                                                 27617, 2020.
                                           18.   Aspachs Oriol, Durante Ruben, Graziano Alberto, Garcia-Montalvo Jose, and Reynal-Querol Marta.
                                                 Real time inequality and the welfare state in motion: evidence from covid-19 in spain. 2020.
                                           19.   IMF. Spain 2020 article iv consultation. 2020.
                                           20.   Cajner Tomaz, Crane Leland D, Decker Ryan A, Grigsby John, Hamins-Puertolas Adrian, Hurst Erik,
                                                 et al. The us labor market during the beginning of the pan- demic recession (nber wp. Technical report,
                                                 National Bureau of Economic Research, 2020.
                                           21.   Giles D. Calculating the standard error for the gini coefficient: some further results. Oxford Bulletin of
                                                 Economics and Statistics, 66(3):425–433, 2004. https://doi.org/10.1111/j.1468-0084.2004.00086.x
                                           22.   Weill J., Stigler M., Deschenes O., and Springborn M. Social distancing responses to covid-19 emer-
                                                 gency declaration strongly differentiated by income. Proceeding of the National Academy of Sciences,
                                                 (July 2020):1–3, 2020.




PLOS ONE | https://doi.org/10.1371/journal.pone.0249121 March 31, 2021                                                                                 14 / 14

File and source

File
tracking-the-impact-of-covid-19-on-economic-inequality-at-high-frequency.pdf
Size
1,758,501 bytes
SHA-256
bc21ee1c163ecf0a12335ab32498d3aa6dee83f0cb4fc7e20c0a00e7de42bce5
Our copy
tracking-the-impact-of-covid-19-on-economic-inequality-at-high-frequency.pdf
Original
doi.org
Back to top