2021 FIRM-LEVEL HETEROGENEITY IN THE IMPACT OF THE COVID-19 PANDEMIC - Documentos de Trabajo
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
FIRM-LEVEL HETEROGENEITY IN THE
IMPACT OF THE COVID-19 PANDEMIC
2021
Documentos de Trabajo
N.º 2120
Alejandro Fernández-Cerezo, Beatriz González,
Mario Izquierdo and Enrique Moral-BenitoFIRM-LEVEL HETEROGENEITY IN THE IMPACT OF THE COVID-19 PANDEMIC
FIRM-LEVEL HETEROGENEITY IN THE IMPACT OF THE COVID-19 PANDEMIC (*) Alejandro Fernández-Cerezo, Beatriz González, Mario Izquierdo and Enrique Moral-Benito BANCO DE ESPAÑA (*) Banco de España. Corresponding emails: alejandrofernandez@bde.es (Alejandro Fernández-Cerezo), beatrizgonzalez@bde.es (Beatriz González), mizquierdo@bde.es (Mario Izquierdo) and enrique.moral@bde.es (Enrique Moral-Benito). We would like to thank Manuel Ortega and Joaquin Rivero as well as the Central Balance Sheet Data Office Division for their invaluable help undertaking the survey. We are also grateful to the Information System Department for their work preparing the questionnaire and processing firms’ responses. Finally, we would like to thank Samuel Hurtado and José Luis Herrera for their help in the processing of the data. For useful comments, we thank an anonymous referee and Federico Curci. Documentos de Trabajo. N.º 2120 May 2021
The Working Paper Series seeks to disseminate original research in economics and finance. All papers have been anonymously refereed. By publishing these papers, the Banco de España aims to contribute to economic analysis and, in particular, to knowledge of the Spanish economy and its international environment. The opinions and analyses in the Working Paper Series are the responsibility of the authors and, therefore, do not necessarily coincide with those of the Banco de España or the Eurosystem. The Banco de España disseminates its main reports and most of its publications via the Internet at the following website: http://www.bde.es. Reproduction for educational and non-commercial purposes is permitted provided that the source is acknowledged. © BANCO DE ESPAÑA, Madrid, 2021 ISSN: 1579-8666 (on line)
Abstract This paper explores the heterogeneity across firms within each sector and region in the impact of and response to the COVID-19 shock. It relies on a survey conducted by Banco de España to 4,004 companies in November 2020 matched to very rich balance- sheet information on firm characteristics. According to our results, the impact of the COVID-19 shock was larger in the case of small, young and less productive firms located in urban areas within each sector-region pair. Moreover, these firms resorted relatively more to public-guaranteed loans, tax deferrals, and furlough schemes (ERTEs). More indebted companies, which were not hit relatively harder by the shock, also perceived public-guaranteed loans as very useful. Firms consider that uncertainty represents a key hindrance to the recovery, but observable characteristics do not explain the variation in the perception of uncertainty once the impact of the shock is accounted for. Finally, we use the announcement of the Pfizer vaccine on November 9th 2020 as a natural experiment to provide evidence that the vaccine announcement increased significantly firms’ subjective recovery expectations. Keywords: COVID-19, firms, sales, employment, uncertainty. JEL classification: D22, L20, L25.
Resumen Este artículo explora la heterogeneidad del impacto y la respuesta a la crisis del COVID-19 de las empresas españolas dentro de una misma región y sector. Los datos utilizados se basan en una encuesta realizada por el Banco de España a 4.004 empresas en noviembre de 2020, que cruzamos con información de los estados financieros de las empresas procedente de la Central de Balances del Banco de España. Los resultados muestran que la facturación y el empleo descendieron más en las empresas pequeñas, jóvenes y menos productivas localizadas en áreas urbanas. En el caso del empleo, una mayor ratio de temporalidad se encuentra asociada a mayores descensos de la ocupación. Además, estas empresas perciben los avales ICO, el aplazamiento de impuestos y los ERTE como herramientas útiles para mitigar los efectos de la actual crisis sanitaria. Las empresas más endeudadas también percibieron como muy útiles los avales ICO, pese a no haber sufrido relativamente más por esta perturbación. Las empresas consideran que la incertidumbre es uno de los principales factores que ha condicionado de forma negativa su actividad, pero las características observadas no pueden explicar la variación en la percepción de la incertidumbre una vez que se controla por la caída de las ventas. Finalmente, usamos el anuncio de la efectividad de la vacuna de Pfizer realizado el 9 de noviembre de 2020 como un experimento natural para mostrar que dicha noticia incrementó de forma significativa las expectativas subjetivas de recuperación de las empresas. Palabras clave: COVID-19, empresas, ventas, empleo, incertidumbre. Códigos JEL: D22, L20, L25.
1 Introduction
The global spread of the COVID-19 is having a significant human toll and repre-
sents an unprecedented shock for the economy, pushing most economies into recession.
One of the most salient features of the virus-induced economic crisis is the asymmetry
1
along Introduction
several dimensions. Although a pandemic represents a text-book example of an
ex-ante exogenous and symmetric shock, the actions taken by agents and policymakers
The global spread of the COVID-19 is having a significant human toll and repre-
have resulted in very different economic effects across sectors and regions/countries.1 This
sents an unprecedented shock for the economy, pushing most economies into recession.
is so because the scope of social-distancing measures depend on the social interaction in-
One of the most salient features of the virus-induced economic crisis is the asymmetry
tensity by sector of activity as well as the severity of the pandemic by region. While this
along several dimensions. Although a pandemic represents a text-book example of an
heterogeneity is well-documented, much less is known about the asymmetric effects across
ex-ante exogenous and symmetric shock, the actions taken by agents and policymakers
firms within each sector and region. The purpose of this paper is to shed light on this
have resulted in very different economic effects across sectors and regions/countries.1 This
issue. In particular, we investigate the heterogeneity of several aspects of the COVID-19
is so because the scope of social-distancing measures depend on the social interaction in-
shock across firms: its impact on sales and employment, the firms’ responses to mitigate
tensity by sector of activity as well as the severity of the pandemic by region. While this
the shock, their use of available policy instruments, and the main factors hindering firms’
heterogeneity is well-documented, much less is known about the asymmetric effects across
activity during the pandemic.
firms within each sector and region. The purpose of this paper is to shed light on this
To this purpose, we use more than 4,000 responses to a new firm-level survey
issue. In particular, we investigate the heterogeneity of several aspects of the COVID-19
launched by Banco de España, the so-called EBAE (Encuesta Banco de España sobre
shock across firms: its impact on sales and employment, the firms’ responses to mitigate
Actividad Empresarial in Spanish). A unique feature of this survey is that we can use the
the shock, their use of available policy instruments, and the main factors hindering firms’
firm identifier to match it to Central de Balances, a firm level data set that contains cash
activity during the pandemic.
flow and balance sheet information of the quasi-universe of Spanish firms. Therefore, we
To this purpose, we use more than 4,000 responses to a new firm-level survey
can investigate the impact of and response to the shock on the basis of the responses to
launched by Banco de España, the so-called EBAE (Encuesta Banco de España sobre
the survey and depending on firms’ ex-ante characteristics. There are several advantages
Actividad Empresarial in Spanish). A unique feature of this survey is that we can use the
of this matched data. First, some of the key variables for our analysis can only obtained
firm identifier to match it to Central de Balances, a firm level data set that contains cash
in surveys such as the EBAE, as they are not observed in administrative data. This is
flow and balance sheet information of the quasi-universe of Spanish firms. Therefore, we
the case, for instance, of timely information on business activity at the firm level, partic-
can investigate the impact of and response to the shock on the basis of the responses to
ularly for SMEs, information about firms’ expectations about future developments, their
the survey and depending on firms’ ex-ante characteristics. There are several advantages
evaluation of various policy instruments and their perception of the degree of uncertainty.
of this matched data. First, some of the key variables for our analysis can only obtained
Second, matching this survey data with balance sheet data allows us to exploit (exoge-
in surveys such as the EBAE, as they are not observed in administrative data. This is
nous) pre-crisis differences in a large set of firms’ characteristics, arguably with a higher
the case, for instance, of timely information on business activity at the firm level, partic-
degree of accuracy, and some of which would be hard to elicit from survey data, such as
ularly for SMEs, information about firms’ expectations about future developments, their
Total
1 Factor Productivity (TFP).
Conceptually, the COVID-19 shock involves simultaneous disruptions to both supply and demand.
evaluation
On theIn supply
ofside,
various
some
policy instruments and their perception of thespread
degree of uncertainty.
particular, weworkplaces and businesses are shut variation
exploit within-sector-region down to halt sothe of the virus. comes
that identification On the
Second, matching this survey data with balance sheet data allows us to exploit (exoge-
demand side, households are less willing to leave their homes, either because of mobility restrictions or
from
the feardifferences across which
of getting infected, firms depresses
operating in the same
consumption. sector the
Moreover, andfallthe same region.
in demand could beArmed
further
nous) pre-crisis
exacerbated differences
by theand
increase in a large setresulting
in unemployment of firms’ characteristics, arguably with a higher
with this data identification strategy, we from aim the
to supply
answershocks
fourhighlighted above,
main questions: which
(i)
degree of accuracy, and some of which would be hard to elicit from survey data, such as
represent the so-called Keynesian supply shocks in Guerrieri et al. (2020).
what the heterogeneous impact of the COVID-19 shock on firms’ turnover is; (ii) what
1
Conceptually,
firms’ responses the COVID-19
to this shock shock involves
are; (iii) simultaneous
which disruptions
policy measures aretodeemed
both supply
moreand demand.
useful by
On the supply side, some workplaces and businesses 1are shut down to halt the spread of the virus. On the
firms forside,
demand sustaining their
households activity;
are less willing (iv) which
to leave theirfactors
homes,affect
either firms’
becauseactivity therestrictions
of mobility most, with or
the fear of getting
a special focus oninfected,
firms’which depresses
recovery consumption.
expectations andMoreover, the fall in demand could be further
uncertainty.
exacerbated by the increase in unemployment resulting from the supply shocks highlighted above, which
Ourthefirst
represent set ofKeynesian
so-called results indicates thatin the
supply shocks COVID-19
Guerrieri shock hit harder small, young
et al. (2020).
and less productive firms within each sector and region. As a consequence, many firms
7
BANCO DE ESPAÑA DOCUMENTO DE TRABAJO N.º 2120
needed to adjust their employment, both in the
1 extensive margin (firing or hiring) and/or
in the intensive margin (temporary reduction in the staff thanks to the use of furlough
schemes - ERTEs). While firms with a larger share of temporary workers decreased moreIn particular, we exploit within-sector-region variation so that identification comes
from differences across firms operating in the same sector and the same region. Armed
with this data and identification strategy, we aim to answer four main questions: (i)
what the heterogeneous impact of the COVID-19 shock on firms’ turnover is; (ii) what
firms’ responses to this shock are; (iii) which policy measures are deemed more useful by
firms for sustaining their activity; (iv) which factors affect firms’ activity the most, with
a special focus on firms’ recovery expectations and uncertainty.
Our first set of results indicates that the COVID-19 shock hit harder small, young
and less productive firms within each sector and region. As a consequence, many firms
needed to adjust their employment, both in the extensive margin (firing or hiring) and/or
in the intensive margin (temporary reduction in the staff thanks to the use of furlough
schemes - ERTEs). While firms with a larger share of temporary workers decreased more
their staff, firms that are larger, more productive and with more savings were able to
better sustain employment.
The second set of results refers to the ways firms have adjusted to the shock: restor-
ing to e-commerce, reducing investment, introducing teleworking, or firing workers (ex-
tensive margin of employment). Reduction in investment was the margin most used by
firms (38%), followed by the implementation of working from home schemes (32%), the
introduction of e-commerce (22%) and firing workers (18%). Working from home was
useful for urban, large and young firms, with high share of intangible assets and a large
share of permanent workers in their staff. E-commerce and the reduction of investment
was more useful for less productive firms. Finally, firing was more used by firms with a
large share of temporary workers. While firing of workers is not that widespread, effective
employment used decreased significantly, as explained in the previous paragraph. This
suggests that most of the adjustment in employment was done via the intensive margin
(furlough schemes- ERTEs), which are explored further below.
The third set of results explores the role of the COVID-19 policy measures in sus-
taining firms’ activity. Public guaranteed loans (ICOs) was the most useful measure, with
nearly 40% of the surveyed firms reporting this policy measure had been important for
sustaining activity. Furlough schemes (ERTEs) were important for 29% of the respon-
dents, and tax deferrals and renegotiation of rental payments were deemed useful by 24%
and 21% of the respondents, respectively. Turning to firm-level heterogeneity, smaller,
less productive, younger, and more indebted firms resorted more intensively to public
guaranteed loans (ICOs) and tax deferrals, while medium-sized and less productive firms
resorted more intensively to furlough schemes (ERTEs).
The fourth set of results shows which 2are the main factors affecting firms’ activity.
Pandemic and political uncertainty take the lead (80% and 77%), followed by the evolution
of demand (48%), unpaid receivables (34%) and competition pressures (33%), problems in
access to financing (17%), disruptions in supply chains (13%) and availability of workers
(10%). Due to the prominent role of uncertainty in this pandemic, we dig deeper into the
heterogeneity of this uncertainty across firms. However, once we account for the size of
the shock, observable firm characteristics cannot explain differences in the perception of
uncertainty.
Finally, we make use of the announcement of the Pfizer vaccine effectiveness on
BANCO DE ESPAÑA 8 DOCUMENTO DE TRABAJO N.º 2120
November 9th 2020, right in the middle of the survey period, as a natural experiment
to compare the recovery expectations of firms that responded to the survey before and
after the announcement, and we find that this announcement improved significantly theirheterogeneity of this uncertainty across firms. However, once we account for the size of
the shock, observable firm characteristics cannot explain differences in the perception of
uncertainty.
Finally, we make use of the announcement of the Pfizer vaccine effectiveness on
November 9th 2020, right in the middle of the survey period, as a natural experiment
to compare the recovery expectations of firms that responded to the survey before and
after the announcement, and we find that this announcement improved significantly their
prospects of recovery. This finding puts forward evidence that during a pandemic firms
take into account
developing medical
countries, and developments
show that thewhen forming shock
COVID-19 their expectations about economic
has had persistent negative
recovery.on sales, but the response of employment has been mostly along the intensive
impact
The Bartik
margin. remainder of the
et al. paper
(2020) useis survey
organized
data as follows. A brief
for the US overview
to show thatofthe
the pandemic
literature
closes this
brought introduction.
a significant Section 2ofpresents
proportion closures,thejob
survey
cuts details and thefinancial
and a fragile balancesituation
sheet data.
of
SectionBennedsen
firms. 3 presents the heterogeneity
et al. (2020) use in the impact
a large surveyofon
thesmall,
shockmedium
on turnover,
and and
largethe reac-
Danish
tion ofand
firms firms’
findemployment.
evidence thatSection 4.1 shows
firms using the heterogeneity
government of the
aid were also firm
those levelmost
in the responses
need,
to the shocks.
suggesting thatSection
support4.2measures
exploreswere
the heterogeneity in the useemployment.
effective in preserving of the policyBloom
instruments
et al.
aimed at
(2021) usemitigating the negative
a panel survey consequences
of 2,500 SMEs in theof US,
the COVID-19 shock.
and document Section
smallest 5 looks
offline at
firms
the main factors
experienced salesconditioning
drops of over firms’
40%activity, with
compared to subsection 5.1 focusing
less than 10% for the on the impact
largest online
of uncertainty
firms. Humphrieson et
firms’ activity.
al. (2020) use Section 6 shows
survey data causalthe
to assess evidence
impact ofof the impact
targeted of the
COVID-
vaccine
19 policies on SMEs.2 on
announcement firms’
Our recovery
paper differsexpectations. Sectionin7that
from these papers concludes.
we can match the
survey responses to very rich balance sheet data of firms, providing us with a variety
of firm-level heterogeneity
Literature review dimensions to look at. Bloom et al. (2020) also match their
survey data to Amadeus database to discern the impact of the COVID-19 shock on TFP.
The paper contributes to the flourishing literature studying the impact of COVID-19
Our paper differs from theirs in that our sample is more representative, including small
on businesses. Although studies examining the impact of previous pandemics on business
and very small firms, and that we focus on the heterogeneous impact of the shock using
activity are quite limited and typically focus on macroeconomic indicators (see Turner &
a broader set of heterogeneity measures, such as age, debt, cash holdings, etc. This is
Akinremi (2020) for a review), a rapidly growing literature on the economic consequences
also an important difference with respect to other readily available data sources, such as
of COVID-19 and government response is emerging since the outbreak of the pandemic.
Chetty et al. (2020), that analyze heterogeneity at group levels (area, industry, income
The closest papers to ours are those using survey data to understand the impact of
level, business size), but are not suitable for exploring heterogeneity in other dimensions,
the COVID-19 shock on firms. Apedo-Amah et al. (2020) perform a survey focused on
such as productivity or indebtedness.
developing countries, and show that the COVID-19 shock has had persistent negative
This paper is also related to the literature dealing with the impact of the COVID-19
impact on sales, but the response of employment has been mostly along the intensive
shock on subjective perceptions and uncertainty.3 Altig et al. (2020) use several macroe-
margin. Bartik et al. (2020) use survey data for the US to show that the pandemic
conomic uncertainty indicators for the US and UK to show a huge uncertainty increase in
brought a significant proportion of closures, job cuts and a fragile financial situation of
reaction to the pandemic, but with different peak amplitudes and time paths in these two
firms. Bennedsen et al. (2020) use a large survey on small, medium and large Danish
countries. One of their uncertainty measures also relies on subjective uncertainty mea-
firms and find evidence that firms using government aid were also those in the most need,
sures computed from business expectation surveys, which shows that sales uncertainty
suggesting that support measures were effective in preserving employment. Bloom et al.
rose by more than 100%. Furthermore, Barrero & Bloom (2020) argue this huge increase
(2021) use a panel survey of 2,500 SMEs in the US, and document smallest offline firms
in uncertainty might be slowing the subsequent recovery and reducing the impact of policy
experienced sales drops of over 40% compared to less than 10% for the largest online
measures taken. Buchheim et al. (2020) show with a panel of German firms that firms that
firms. Humphries et al. (2020) use survey data to assess the impact of targeted COVID-
perceived higher uncertainty, proxied by the perception of shutdown lasting longer, were
19 policies on SMEs.2 Our paper differs from these papers in that we can match the
more likely to implement strong measures like layoffs or canceling investments. Our pa-
survey responses to very rich balance sheet data of firms, providing us with a variety
per contributes to this strand of literature by showing that, once the shock is accounted
of firm-level heterogeneity dimensions to look at. Bloom et al. (2020) also match their
2
There is a growing number of papers using survey data to assess the impact of COVID-19 of firms,
survey data to Amadeus database to discern the impact of the COVID-19 shock on TFP.
which are not listed here for the sake of brevity.
Our paper differs from theirs in that our sample is more representative, including small
and 9very small firms, and that we focus on the heterogeneous impact of the shock using
BANCO DE ESPAÑA DOCUMENTO DE TRABAJO N.º 2120
4
a broader set of heterogeneity measures, such as age, debt, cash holdings, etc. This is
also an important difference with respect to other readily available data sources, such asfirms. Humphries et al. (2020) use survey data to assess the impact of targeted COVID-
19 policies on SMEs.2 Our paper differs from these papers in that we can match the
survey responses to very rich balance sheet data of firms, providing us with a variety
of firm-level heterogeneity dimensions to look at. Bloom et al. (2020) also match their
survey data to Amadeus database to discern the impact of the COVID-19 shock on TFP.
Our paper differs from theirs in that our sample is more representative, including small
and very small firms, and that we focus on the heterogeneous impact of the shock using
a broader set of heterogeneity measures, such as age, debt, cash holdings, etc. This is
also an important difference with respect to other readily available data sources, such as
Chetty et al. (2020), that analyze heterogeneity at group levels (area, industry, income
level, business size), but are not suitable for exploring heterogeneity in other dimensions,
such as productivity or indebtedness.
This paper is also related to the literature dealing with the impact of the COVID-19
shock on subjective perceptions and uncertainty. Altig et al. (2020) use several macroe-
conomic uncertainty indicators for the US and UK to show a huge uncertainty increase in
reaction to the pandemic, but with different peak amplitudes and time paths in these two
countries. One of their uncertainty measures also relies on subjective uncertainty mea-
sures computed from business expectation surveys, which shows that sales uncertainty
rose by more than 100%. Furthermore, Barrero & Bloom (2020) argue this huge increase
in uncertainty might be slowing the subsequent recovery and reducing the impact of policy
measures taken. Buchheim et al. (2020) show with a panel of German firms that firms that
perceived higher uncertainty, proxied by the perception of shutdown lasting longer, were
more likely to implement strong measures like layoffs or canceling investments. Our pa-
per contributes to this strand of literature by showing that, once the shock is accounted
for,2 There
observable characteristics
is a growing cannot
number of papers explain
using survey any differences
data to in perceived
assess the impact uncertainty.
of COVID-19 of firms,
which are not listed
Furthermore, here for the
we provide sake of
causal brevity. of the impact of the vaccine announcement on
evidence
subjective firm level recovery expectations. On this front, our results complement those
of Heap et al. (2021), who find that the vaccine
4 announcement decreased the trust in
government and elected politicians.
Finally, this paper is also related to the literature about the impact of COVID-19 on
firms at different margins: incumbent’s firm behaviour (Brotherhood & Jerbashian, 2020,
Barrero et al., 2020 ); liquidity needs, credit constraints and solvency of firms (Schivardi
& Guido, 2020, Balduzzi et al., 2020, Greenwald et al., 2020, Blanco et al., 2020), the
efficiency of policies implemented to mitigate the COVID-19 shock (Gonzalez-Uribe &
Wang, 2020, Goodhart et al., 2020, Zoller-Rydzek & Keller, 2020), the entry decision
(Sedlacek & Sterk, 2020, Albert et al., 2020), among many others.
2 Data
2.1 Survey details
The survey was launched by Banco de España in November 2020, the so-called
EBAE (Encuesta Banco de España sobre Actividad Empresarial ), with the purpose of
monitoring Spanish non-financial corporations’ activity in real time. The participation
10
of companies
BANCO DE ESPAÑA
is voluntary and responses were collected through a questionnaire sent to
DOCUMENTO DE TRABAJO N.º 2120
firms by e-mail. The survey was carried during the fortnight between the 4th and 19th of
November 2020. The survey was submitted to a sample of 12,940 Spanish non-financialThe survey was launched by Banco de España in November 2020, the so-called
EBAE (Encuesta Banco de España sobre Actividad Empresarial ), with the purpose of
monitoring Spanish non-financial corporations’ activity in real time. The participation
of companies is voluntary and responses were collected through a questionnaire sent to
for, observable
firms by e-mail. characteristics
The survey wascannot
carriedexplain any fortnight
during the differences in perceived
between the 4th uncertainty.
and 19th of
Furthermore,
November we The
2020. provide causal
survey wasevidence of the
submitted to aimpact
sampleofofthe vaccine
12,940 announcement
Spanish non-financial on
subjective firmand
corporations, level
werecovery
received expectations. On this front,
4,004 valid responses, which our results complement
represents a response ratethoseof
Questionnaire
of HeapFigure
30,9%. et al. 10
(2021), who find
of Appendix A.1that thethe
shows vaccine announcement
responses 3
received bydecreased
day. Theretheis trust in
a slight
government and elected
over-representation
The survey politicians.
of some
included a sectors
total of(e.g. manufacturing)
8 questions, and
split into twolarge firms
parts. (see firms
First, Appendix
were
Finally, this paper is also related to the literature about the impact of COVID-19
A.1). about their views on the current and future evolution of business turnover, employ-
asked on
firms Survey
ment, at
anddifferent
input margins:
responses
and output incumbent’s
aggregated
prices,atasthefirm
well behaviour
sectoral
the effect (Brotherhood
andofregional
some & Jerbashian,
level capture
general factors well 2020,
the recent
affecting their
Barrero
activity, et
developments al.,
such as 2020
in the ); liquidity
the demand needs,
Spanish changes, credit
economy.access constraints
For instance,
to external evenand solvency
at a high
financing, of firms
degree
supply (Schivardi
of disaggre-
disruptions or
& Guido,
gation
economic 2020,
regional Balduzzi
at uncertainty, et al.,
or industry
among 2020,A
level,
others. Greenwald
survey
second setetofon
figures al., 2020, Blanco
employment
questions et calibrate
aimedgrowth
to al., are
2020), the
highly
more
efficiency the
correlated
precisely of policies
with thoseofimplemented
impact ofand
other to mitigate
the sources
response (see the 9COVID-19
Figure
to the crisis.shock
in Appendix
COVID-19 InA.1).(Gonzalez-Uribe
It is worth
particular, &
high-
companies
Wang,
lighting
were 2020,
that
asked Goodhart
both
about the et al.,
thedegree
survey 2020,
ofand thisZoller-Rydzek
uptake paper
of focus
support & the
on Keller,
measures, how2020),
intensive thenormal
margin,
far from entry decision
namely, the
activity
(Sedlacek 4
they&were,
performance
levels Sterk, 2020,
howAlbert
of surviving
and long itetwill
firms. al., take
2020),them
among many
to get others.
back to normal. Appendix A.2
contains
3
the full questionnaire,
The distribution andthe
of firms that received in survey,
each ofthethe following
distribution subsections
of firms we detail
that responded the
the survey,
and its comparison to aggregate data can be found on Appendix A.1.
Questionnaire
questions
2 4 Data used.
Indeed, non-responses to the survey may reflect companies that have closed permanently as a result
The survey included a total of 8 questions, split into two parts. First, firms were
2.2 Balance
2.1 about
Survey sheet data: Central 5 de Balances Integrada
asked their details
views on the current and future evolution of business turnover, employ-
ment,Data
and input
The and
wasoutput
on firm-level
survey prices,
responses
launched asBanco
byto well the
de effect
the survey of some
is combined
España general factors
2020, affecting
with information
in November the on their
firms’
so-called
activity, such as
characteristics
EBAE (Encuesta the
that isdemand
Banco changes,
available
de España access
at a yearly
sobre to external
frequency
Actividad from financing,
the Central
Empresarial ), supply disruptions
Balance
with the Sheet or
Data
purpose of
economic
(CBI,
monitoring uncertainty,
Central
Spanish among
de Balances others.
Integrada),
non-financial Awhich
secondis set
corporations’ of questions
sourced
activity from
in aimed
realfirms’
time. to calibrate
voluntary
The more
responses
participation
precisely
to
of Banco the impact
de España
companies of and and
Central
is voluntary the response
Balance towere
Sheet
responses the COVID-19
Data crisis. surveys
Office (CBSO)
collected through In
a particular,
and thecompanies
questionnaire Spanish
sent to
were
firms asked
Mercantile about thedata;
Registry
by e-mail. The degreethe
survey of ultimate
was uptake
carriedofduring
support
sources ofmeasures,
the the datahow
fortnight are far from
therefore
between normal
theand
the 4th activity
Banco
19th de
of
levels
Españathey
November were,
and2020. and survey
howMercantile
the Spanish
The long
wasit submitted
willand
takeProperty
them to Registrars’
get back
to a sample toAssociation.
of 12,940normal.
SpanishAppendix
This isA.2an
non-financial
contains the full
administrative
corporations, and questionnaire,
database and valid
that contains
we received 4,004 in each of the
information on following
responses, firms’ subsections
which financial we detail
statements
represents ratethe
(required
a response of
questions
by law to
30,9%. used.
be submitted
Figure to the A.1
10 of Appendix commercial registry)
shows the as received
responses well as on day.3income
bytheir There corporate
is a slight
tax returns. The data
over-representation coverssectors
of some around(e.g.
90%manufacturing)
of firms in the and
non-financial market
large firms economy
(see Appendix
2.2
for Balance
all size
A.1). sheet
categories, data:
including bothCentral de number
turnover and Balances Integrada
of employees. The correlation
between micro-aggregated
Survey employment
responses aggregated at the and output
sectoral growth and
and regional levelthe National
capture Accounts
well the recent
Data on firm-level responses to the survey is combined with information on firms’
counterparts
developmentsisinabove 0.90 (seeeconomy.
the Spanish Almunia For et al. (2018) for
instance, evenmore
at adetails). The approach
high degree of disaggre-of
characteristics that is available at a yearly frequency from the Central Balance Sheet Data
matching
gation at CBI and survey
regional data enables
or industry level, us to explore
survey theonrole
figures of firms’ characteristics,
employment which
growth are highly
(CBI, Central de Balances Integrada), which is sourced from firms’ voluntary responses
would be difficult
correlated to replicate
with those of otherinsources
a survey, along
(see Figureseveral
9 in dimensions
Appendix A.1).observedIt isinworth
the survey,
high-
to Banco de España Central Balance Sheet Data Office (CBSO) surveys and the Spanish
such as the
lighting thatimpact
both theof COVID-19
survey andon thisturnover and employment,
paper focus on the intensive sincemargin,
this information
namely, the is
Mercantile Registry data; the ultimate sources of the data are therefore the Banco de
only availableofinsurviving
performance firms.4 registers with a sizable lag, firms’ perception of policies
administrative
España and the Spanish Mercantile and Property Registrars’ Association. This is an
to tackle
3 the crisisofand
The distribution firmsthe
thatdegree
receivedofthe
uncertainty about future
survey, the distribution developments.
of firms that responded After the
the survey,
administrative database that contains information on firms’ financial statements (required
and its comparison
match EBAE-CBI, to aggregate
we end up data can 3,584
with be found on Appendix A.1.
observations.
by law
4 to be
Indeed, submittedtotothe
non-responses the commercial
survey may reflectregistry)
companiesasthat
wellhave
as on their
closed income corporate
permanently as a result
of
taxthereturns.
COVID-19 shock.
The data With respect
covers to the 90%
around extensive margininand
of firms theinnon-financial
line with our findings
marketbelow for the
economy
intensive margin, Social Security records shows that small firms have been hit harder by the COVID-19
for all the
shock: sizenumber
categories, including
for firms with less both turnover
than 49 5 and
employees numberbyof-3.3%
decreased employees. Thethe
in 2020 while correlation
fall in the
case of larger firms (>500 emp.) was only -1.4%.
between micro-aggregated employment and output growth and the National Accounts
counterparts is above 0.90 (see Almunia et al. (2018) for more details). The approach of
11
BANCO DE ESPAÑA DOCUMENTO DE TRABAJO N.º 2120
6
matching CBI and survey data enables us to explore the role of firms’ characteristics, which
would be difficult to replicate in a survey, along several dimensions observed in the survey,
such as the impact of COVID-19 on turnover and employment, since this information is(CBI, Central de Balances Integrada), which is sourced from firms’ voluntary responses
to Banco de España Central Balance Sheet Data Office (CBSO) surveys and the Spanish
Mercantile Registry data; the ultimate sources of the data are therefore the Banco de
España and the Spanish Mercantile and Property Registrars’ Association. This is an
administrative database that contains information on firms’ financial statements (required
by law to be submitted to the commercial registry) as well as on their income corporate
tax returns. The data covers around 90% of firms in the non-financial market economy
for all size categories, including both turnover and number of employees. The correlation
between micro-aggregated employment and output growth and the National Accounts
counterparts is above 0.90 (see Almunia et al. (2018) for more details). The approach of
matching CBI and survey data enables us to explore the role of firms’ characteristics, which
would be difficult to replicate in a survey, along several dimensions observed in the survey,
such as the impact of COVID-19 on turnover and employment, since this information is
only available in administrative registers with a sizable lag, firms’ perception of policies
to tackle the crisis and the degree of uncertainty about future developments. After the
match EBAE-CBI, we end up with 3,584 observations.
For each firm,
of the COVID-19 shock.among other to
With respect variables, we observe
the extensive theinfirm’s
margin and sector
line with of activity
our findings (4-digit
below for the
intensiveRev.
NACE margin, Social Security
2 code), locationrecords shows
(5-digit zipthat small
code), firms havematerial
turnover, been hit expenditures,
harder by the COVID-19
number
shock: the number for firms with less than 49 employees decreased by -3.3% in 2020 while the fall in the
of
caseemployees, share
of larger firms (>500of emp.)
temporary employees,
was only -1.4%. age, debt ratio (interest-bearing borrowed
funds to interest-bearing liabilities), share of intangible assets, the ratio of cash to total
assets, and total fixed assets. Moreover, from
6 these variables we compute a measure of
total factor productivity (TFP) for each firm.
2.3 A first glimpse at the data
Table 1 presents some descriptive statistics at the sector level for the main variables
of interest in our analysis. In particular, it provides figures on two sets of variables:
responses to the survey (columns 1 and 2), and firm characteristics (columns 3-10). From
Panel A, it is worth emphasizing the wide heterogeneity across sectors in both the average
impact of the shock and the average firm characteristics. For instance, the average firm
in manufacturing experienced a fall of -12.66% in turnover against the -45.53% fall for
the average firm operating in the hospitality sector (-5.49% and -34.97% for employment,
respectively). Also, the average firm in manufacturing is 33% more productive than that
of hospitality (log TFP 1.28 versus 0.95), 8 years older, more rural (26% hospitality firms
are out of cities against 44% manufacturing firms), less indebted, holds less cash and are
much larger in terms of employees.
While heterogeneity across sectors is well-known, Panel B of Table 1 documents a
more interesting and potentially more important source of heterogeneity, that is, hetero-
geneity across firms within the same sector, which is the main focus of the paper. In
particular, it uncovers huge variation across firms within each sector as measured by the
interquartile range (IQR) given by the difference between the 75th and the 25th per-
centiles. For example, the TFP difference between the 75th percentile and the 25th in
the administrative services sector is even larger than the difference between the average
firm in manufacturing and hospitality from the table above: 0.54 against 0.33 (0.33=1.28-
0.95).
12 This indicates that, while the average manufacturing firm is 33% more productive
BANCO DE ESPAÑA DOCUMENTO DE TRABAJO N.º 2120
that the average hospitality firm, the 75th percentile firm in administrative services is 54%
more productive than the 25th percentile firm in the same sector. In terms of age, theinterquartile range (IQR) given by the difference between the 75th and the 25th per-
centiles. For example, the TFP difference between the 75th percentile and the 25th in
the administrative services sector is even larger than the difference between the average
firm in manufacturing and hospitality from the table above: 0.54 against 0.33 (0.33=1.28-
0.95). This indicates that, while the average manufacturing firm is 33% more productive
that the average hospitality firm, the 75th percentile firm in administrative services is 54%
more productive than the 25th percentile firm in the same sector. In terms of age, the
manufacturing-hospitality average gap is 9 years, while the 75th-25th gap within admin.
services is 17 years. Regarding the cash holdings of firms, and hence the starting buffer
against the shock, there is also substantial heterogeneity, with the 75th percentile of firms
in thevs
(0.36 IT0.03).
services
5
Weholding
show ina Table
share that
9 on is 10 timesBlarger
Appendix than that
that these of thestatistics
summary 25th percentile
remain
5
(0.36 vs 0.03).similar
qualitatively We show
if weinuse
Table 9 on Appendix
weights B that
to match the these summary
sector-size statistics
distribution. remain
In light of
qualitatively
these figures, similar
it seemsif crucial
we use to
weights
bettertounderstand
match the the
sector-size distribution.
heterogeneity In light of
of the COVID-19
7
these figures,
shock and the itresponses
seems crucial
across to better
firms understand
within the something
each sector, heterogeneity of the COVID-19
we investigate further
shock Table 1: Summary statistics by sector
in the and the responses
remaining across firms within each sector, something we investigate further
of the paper.
in the
PANELremaining
A of the paper.
Averages ∆ ∆ log Age Rural Temp. Intangible Debt Size Cash
Turnover Emp TFP Workers capital Ratio (Emp) ratio
3 Manufacturing
The impact -12.66
of
-5.49
the
1.28
COVID-19
29.03 0.44 0.12
shock
0.09
across
0.31
firms
142.34 0.12
3 Construction
The impact -12.23 of
-5.20 the
1.18 COVID-19
21.86 0.27 0.30 shock
0.16 across
0.28 firms
39.58 0.15
Trade -15.55
In order to assess the-7.34impact
0.84 of25.67
COVID-19 0.26 on0.13businesses,
0.13 we 0.30
rely on63.37 0.15 6
question
Transport -16.34 -8.34 1.99 24.15 0.29 0.20 0.11 0.35 119.90 0.14
of theInsurvey,
order to
Hospitality assess
which the
-45.53reads as
-34.97impact
0.95 of20.76
follows: COVID-19
‘How0.26 on0.29businesses,
are your firms’0.10 we 0.36
turnover rely
andon question
0.17 6
employment
38.02
IT services -11.33 -3.29 1.51 18.57 0.08 0.16 0.42 0.22 78.38 0.22
of the
in Real survey,compared
the estate
4Q20 which
-10.27reads
to the assame
-3.23 follows:
1.11period‘How
23.37 are
last0.14 your
year? firms’that
’. Note
0.05 turnover
0.05 while and 6.03
employment
0.26turnover is0.12
more
Prof. services -10.03 -5.01 1.70 19.34 0.13 0.12 0.25 0.24 45.68 0.24
in the 4Q20
informative
Admin. compared
about
services the to
-16.84 theof same
size-11.92 the1.77 period
shock the
17.98 lastfirm
year?
0.16 ’. Notethe
received,
0.22 that while
0.26change turnover
in
0.31 254.79 is0.23
employment more
is
Other services -32.23 -19.42 1.30 19.36 0.21 0.24 0.19 0.28 50.50 0.24
informative
more aboutabout
informative the size
theofreaction
the shock the firm
of23.71
the firm to
received,
the the change
shock. in employment
The question is
specifically
Total -16.10 -8.58 1.25 0.27 0.17 0.15 0.29 85.32 0.16
more
asked informative about
about the total the reaction
change of the firm
in employment used,tothat
theis,shock. The the
including question specifically
extensive margin
Obs 3,523 3,457 3,161 3,584 3,584 3,160 3,584 3,584 3,584 3,582
asked about(hiring
adjustment the total
or change in employment
firing), and used,
the intensive thatadjustment
margin is, including the extensive
(workers margin
on temporary
PANEL B
adjustment (hiring
leave through or firing),
furlough schemesand- the intensive
ERTEs). margin
There wereadjustment
ten possible(workers
answersonexpressed
temporary
in
IQRs ∆ ∆ log Age Rural Temp. Intangible Debt Size Cash
leave through
intervals, furlough
depending onschemes
Turnover the -TFP
ERTEs).
Emppercentage Theredecrease/increase.
change were
Workersten possible
capital answers expressed
ratio in
The distribution
Ratio (Emp) of
intervals,
responses depending
Manufacturing inonFigure
is shown22.50 the
7.50percentage
1. 0.20 change
17.00
A first look1.00atdecrease/increase.
the0.17distribution
0.04 ofThe
the distribution
0.45 65.62
reported 0.16 of
year-
Construction 22.50 2.50 0.36 16.00 1.00 0.50 0.03 0.48 31.06 0.21
responses
on-year is shown
Trade turnover in Figure
changes
22.50 1. 0.17
(Panel
7.50 A first
A) 18.00look
reveals at the
that
1.00 the distribution
0.18 firmsofdeclared
bulk of0.06 the reported
0.53 year-
a negative
36.00 0.19
Transport 22.50 7.50 0.39 16.00 1.00 0.26 0.02 0.56 43.00 0.16
on-year
impact turnover
of
Hospitality changes
COVID-19 (Panel
20.00 (63%),
52.50 whileA)24%
0.23 reveals thatnothe
report
17.00 1.00 0.30bulk of
change. In firms
0.01 declared
contrast,
0.61 Panel aB
31.80 negative
shows
0.20
IT services 22.50 2.50 0.52 14.00 0.00 0.21 0.96 0.40 63.00 0.33
impact
that only
Real of 38%
estateCOVID-19 12.50 (63%),
of respondents while
0.71 24%
0.00 report report
having
14.00 0.00 no change.
decreased In
0.00contrast,
their employment,
0.00 Panel
0.42 and 54%
2.05 B report
shows
0.14
Prof. services 22.50 2.50 0.36 14.00 0.00 0.15 0.39 0.41 29.88 0.35
that onlyservices
no Admin.
change. 38%
6 of respondents
These patterns
40.00 report
suggest
12.50 having
0.54 that decreased
firms
17.00 have been
0.00 their
0.32 ableemployment,
to absorb
0.50 and
part
0.53 of54%
47.76 report
the shock,
0.31
no Other
since
services
change. 6
These52.50
their employment
40.00
patterns 0.50
suggest
decreased
16.00
that
less firms
than
0.00
have
their
0.30
been
sales. able 0.14 0.53
to absorb part45.00 0.32
of the shock,
Total 22.50 7.50 0.64 17.00 1.00 0.24 0.07 0.51 42.25 0.21
since To
their employment
analyze decreased
the type less than
of firms most their sales.
impacted by the COVID-19 shock, we investigate
Obs 3,523 3,457 3,161 3,584 3,584 3,160 3,584 3,584 3,584 3,582
whichTo analyze
firm the type (productivity,
characteristics of firms most impacted
age, shareby of the COVID-19
temporary shock,
workers, we investigate
intangible assets
Source:
which EBAE
firm survey and Central
characteristics de Balances Integrada.
(productivity, age, share of temporary workers, intangible
share, indebtedness,
Notes: Panel cash and
A shows the averages ratio
Paneland
B thesize) correlate
interquantile rangeswith the asfall
(measured in activity
p75-p25 within the at theassets
industry). firm
The
share, indebtedness,
first column
level. In cash
of both tables corresponds
particular, our ratio andchange
to yearly size)incorrelate
turnover, andwith
of athethe
the second fall intoof
column activity
the yearly at thein firm
change
employment. Column 3, 4 and 5 baseline
correspond specification
to log TFP, age ofconsists
the firm, and regression
dummy variable either turnover
indicating the or
level. In
headquarters particular, our
are in rural areas. baseline
Column specification
6 shows consists
the share of temporary of a
workers. regression
Column 7, 8of
employment growth of firm i (yi ) on a vector of firm characteristics (Xi ) as well as different either
and 10 showturnover
the share ofor
intangible capital (intangible capital over total capital), the debt ratio (total debt over total assets), and the cash ratio
employment
configurations growth
(cash over total assets) of firm
of respectively.
sector i (yi ) on
Column
s (NACE a vector
9 shows
rev of firm
andbycharacteristics
size, measured
2-digit) (Xi ) as well
the number of employees.
region j (Autonomous as different
Communities)
configurations
fixed effects (γs,j of):sector s (NACE rev 2-digit) and region j (Autonomous Communities)
fixed
5
effects (γs,j ):
Note that the number of observations differs depending on which explanatory variable we look at. We
show5 Note
in Table
that 8theonnumber
Appendix B that the results
of observations differs are qualitatively
depending similar
on which if we keep
explanatory only the
variable weobservations
look at. We
for which
show we have
in Table 8 onall the regresors
Appendix B thatofthe
interest
results(2,715 observations).
are qualitatively similar if we keep only the observations
6
This we
for which ‘inaction
have allrange’ masks aofvast
the regresors heterogeneity
interest between sectors: while only 16.5% of firms in
(2,715 observations).
hospitality
6
This services claim
‘inaction range’ amasks
no change
a vastin heterogeneity
their employment withsectors:
between respect while
to theonly
previous
16.5%year, 82% in
of firms of
BANCO DE ESPAÑA 13 DOCUMENTO DE TRABAJO N.º 2120
the firms inservices
hospitality the realclaimstate asector have been
no change ableemployment
in their to sustain their
withemployment.
respect to theAlso, aroundyear,
previous 10%82%of the
of
respondents
the firms in ofthethe survey
real statehave 0 employees
sector have beenand most
able of themtheir
to sustain claimemployment.
not having changed their employment.
Also, around 10% of the
respondents of the survey have 0 employees and most8of them claim not having changed their employment.
9(0.36 vs 0.03).5 We show in Table 9 on Appendix B that these summary statistics remain
qualitatively similar if we use weights to match the sector-size distribution. In light of
these figures, it seems crucial to better understand the heterogeneity of the COVID-19
shock and the responses across firms within each sector, something we investigate further
in the remaining of the paper.
(0.36 vs 0.03).5 We show in Table 9 on Appendix B that these summary statistics remain
3 The similar
qualitatively impact if we ofuse the
weightsCOVID-19 shockdistribution.
to match the sector-size across firms In light of
(0.36 vs
these 0.03).5it We
figures, show
seems in Table
crucial 9 on Appendix
to better understandB that
the these summaryofstatistics
heterogeneity remain
the COVID-19
In order to assess the impact of COVID-19 on businesses, we rely on question 6
qualitatively
shock and thesimilar if we
responses use weights
across to match
firms within the sector-size
each sector, distribution.
something In light
we investigate of
further
of the survey, which reads as follows: ‘How are your firms’ turnover and employment
these
in the figures,
remainingit seems
of the crucial
paper. to better understand the heterogeneity of the COVID-19
in the 4Q20 compared to the same period last year? ’. Note that while turnover is more
shock and the responses across firms within each sector, something we investigate further
informative about the size of the shock the firm received, the change in employment is
in the remaining of the paper.
3
more The impact
informative about the ofreaction
the COVID-19
of the firm to the shock
shock. Theacross
question firms
specifically
asked about the total change in employment used, that is, including the extensive margin
In order to assess the impact of COVID-19 on businesses, we rely on question 6
3 The(hiring
adjustment impact of and
or firing), the the COVID-19 shock (workers
intensive margin adjustment acrossonfirms temporary
of the survey, which reads as follows: ‘How are your firms’ turnover and employment
leave through furlough schemes - ERTEs). There were ten possible answers expressed in
in theIn4Q20
ordercompared
to assesstothe theimpact of COVID-19
same period last year?on’. businesses,
Note that while we rely on question
turnover is more6
intervals, depending on the percentage change decrease/increase. The distribution of
of the survey,
informative which
about thereads
size ofas the
follows:
shock ‘Howthe firmare received,
your firms’ theturnover
change inand employment
employment is
responses is shown in Figure 1. A first look at the distribution of the reported year-
in theinformative
more 4Q20 compared abouttothe thereaction
same periodof thelastfirmyear? ’. Note
to the shock.that Thewhile turnover
question is more
specifically
on-year turnover changes (Panel A) reveals that the bulk of firms declared a negative
informative
asked about about
the totalthechange
size ofinthe shock the used,
employment firm received, the change
that is, including the in employment
extensive marginis
impact of COVID-19 (63%), while 24% report no change. In contrast, Panel B shows
more informative
adjustment (hiringabout the reaction
or firing), of the firm
and the intensive to the
margin shock. The
adjustment question
(workers on specifically
temporary
that only 38% of respondents report having decreased their employment, and 54% report
asked through
leave about the total change
furlough schemes in employment
- ERTEs). There used, were
that ten
is, including the extensive
possible answers margin
expressed in
no change.6 These patterns suggest that firms have been able to absorb part of the shock,
adjustmentdepending
intervals, (hiring or on firing),
the and the intensive
percentage changemargin adjustment (workers
decrease/increase. on temporary
The distribution of
since their employment decreased Figure less than their sales.
1: Distribution of responses
leave through
responses furlough
is shown in Figure Figure
schemes 1: first
A Distribution
1.- ERTEs). There
look of responses
at were
the ten possibleofanswers
distribution expressed
the reported in
year-
To analyze the type of firms
Figure
Figure most
1: impacted
Distribution
1: reveals
Distribution byof the COVID-19
responses
of responses shock, we investigate
intervals,
on-year depending
turnover on the
changes percentage
(Panel
Figure A) change
1: Distribution Panel
the B: Change
thatdecrease/increase.
bulk
of temporary
responsesof firmsin indeclared
Theemployment
distribution
a negativeof
which firmPanel A: Change
characteristics in sales
(productivity, age, share Panel
of B: Change workers, employment
intangible assets
responses
impact Panel
of is A:
COVID-19 Change
shown in (63%), in sales
Figure while
1. A 24% first report Panel
look atnothe
Panel B: Change
distribution
change. B: ofin
In contrast,
Change employment
the reported
Panel
inactivity
employment year-
B shows
Panel
Panel A:
share, indebtedness, Change
A: Change cashin sales
inratio
salesand size) correlatePanel withB: the fall inin
Change at the firm
employment
on-year
that onlyPanel
38%A:
turnover of Change
changesin(Panel
respondents sales
reportA)having
revealsdecreased
that the their
bulk employment,
of firms declared a negative
and 54% report
level. In particular, our baseline specification consists of a regression of either turnover or
6
impact
no change.of COVID-19
These patterns(63%), while that
suggest 24% firms
report no been
have change.able In contrast,
to absorb partPanel
of theB shock,
shows
employment growth of firm i (yi ) on a vector of firm characteristics (Xi ) as well as different
that only
since their38% of respondents
employment decreasedreportlesshaving decreased
than their sales.their employment, and 54% report
configurations of sector s (NACE rev 2-digit) and region j (Autonomous Communities)
6
no change.
To analyzeThese thepatterns
type of suggest
firms mostthatimpacted
firms have bybeen able to absorb
the COVID-19 shock,partweofinvestigate
the shock,
fixed effects (γs,j ):
since their
which employment decreased
firm characteristics less than
(productivity, age,their
sharesales.
of temporary workers, intangible assets
5
Note that the number of observations differs depending on which explanatory variable we look at. We
share, ToTable
analyze
show in indebtedness,
the type
cashBofratio
8 on Appendix
firmsand
that the
most impacted
size)
results
by with
correlate
are
thesimilar
qualitatively
COVID-19
the fall
if we in
shock,
keepactivity we investigate
only theat the firm
observations
which
level. firm
In
for which we characteristics
particular, our
have all the (productivity,
baseline
regresors age, share
specification
of interest (2,715 of temporary
consists workers,
of a regression
observations). intangible
of either assets
turnover or
6
This ‘inaction range’ masks a vast heterogeneity between sectors: while only 16.5% of firms in
share, indebtedness,
employment
hospitality growth
services ofcash
claim no ratio
afirm (yi ) and
ichange on size)
in atheir
vector correlate with the fall
of firm characteristics
employment with respect in
theiactivity
to (X as well at
)previous as the82%
firm
different
year, of
level.
the firmsIn inparticular,
configurations the real oursector
of sector
state baseline
have specification
s (NACE rev able
been 2-digit)consists of employment.
and their
to sustain region a regression of either
j (Autonomous
Also, turnover
Communities)
around or
10% of the
respondents
Source: EBAEof the survey have 0 employees and most of them claim not having changed their employment.
survey
employment
fixed
Source: effects growth
(γ
EBAE survey s,j ): of firm i (yi ) on a vector of firm characteristics (Xi ) as well as different
Notes: Panel A shows the reported year-on-year change in turnover. Panel B shows the reported year-on-year change in
Source: EBAE
EBAEA survey
configurations
Source:
Notes:
5 Panel
Note
employment,
Source: that
EBAE theof
survey
shows
taking
sector
the
number
into
survey ofs
reported
account
(NACEand
year-on-year
observations
hires/layoffs
rev 2-digit)
change
differs
workers
andPanel
in turnover. region
in9a furlough
depending on which
scheme
jexplanatory
B shows (Autonomous
(ERTE)
Communities)
the reported year-on-year
variable
change in
we look at. inWe
Notes: Panel
Notes: Panel taking
employment, A shows
A shows theaccount
the
into reported
reported year-on-year
year-on-year
hires/layoffs change
change
and in
in turnover.
workers turnover.
in a Panel
Panel
furlough B
B shows
schemeshows the
the reported
(ERTE) reported year-on-year
year-on-year change
change in
fixed Panel
show
Notes: effects
in Table (γs,jthe
8 on
A shows ): reportedhires/layoffs
Appendix B that the results are qualitatively similar if we keep only the observations
year-on-year change in turnover. Panel B shows the reported year-on-year change in
employment,
for which wetaking
employment, haveinto
taking all account
into account hires/layoffs
the regresors and
and workers
workers
of interest (2,715in
in a furlough
furlough scheme
scheme (ERTE)
aobservations). (ERTE)
employment,
6
5 taking into account hires/layoffs and workers in a furlough scheme (ERTE)
This ‘inaction
Note that range’ of
the number masks a vast heterogeneity
observations differs depending between
on whichsectors: while only
explanatory 16.5%
variable of firms
we look in
at. We
hospitality
show in Table services claim a noB change
8 on Appendix that theinresults
their are
employment withsimilar
qualitatively respect to the
if we keepprevious
only theyear, 82% of
observations
the firms in
for which wethehave real
allstate sector have
the regresors of y been
interest able to sustain
(2,715 their employment. Also, around 10% of (1)
observations).
i = α + β Xi + γs,j + i
the
6
respondents of the survey
This ‘inaction range’have masks yi =heterogeneity
0 employees
a vast α +most
and β Xofi + γs,jclaim
them
between+ sectors:
i not having changed
while only their of firms(1)
16.5%employment. in
hospitality services claim a no change yyiiin=
= α
α
their+
+ β
β Xi + γs,j + i
X
employment+ γ +
with respect to the previous year, 82% (1)
(1)
of
Table 2 shows the estimation ybeen results. The
i
+ βto Xsustain
i =results.
αable first
s,j
i + γs,j + four
i columns use sales growth (1)
iemployment.
as
the
BANCO DE ESPAÑA
Table
firms in 2
the shows
real state the
14 DOCUMENTO DE TRABAJO N.º 2120 estimation
sector have The
9ofThe first
theirfour columns use
Also, sales
around growth
10% of as
the
dependent
Tableofvariable,
Table
respondents 22the
shows
surveythe
shows whereas
the the last
haveestimation
0 employees
estimation andfour
mostcolumns
results.
results. Thethemfirstusefour
claim
first four employment
not columns
having
columns growth.
use
changed
use sales The four
growth
their employment.
sales growth as
as
dependent
Table variable, whereas
2 shows the the last
estimation four columns
results. The firstusefour
employment growth.
columns use The four
sales growth as
columns
dependent
dependentforvariable,
each dependent
variable, whereas
whereas variable
the
the last
lastdiffer
four
four in the FEuse
columns
columns configuration
use employment
employment considered.
growth.
growth. Our four
The
The pre-
four
columns for each dependent variable differ in
dependent variable, whereas the last four columns the FE
9 full FE configuration considered. Our
use employment growth. The four pre-
ferred
columns
columnsspecifications
for
for each is the one
each dependent
dependent that includes
variable
variable differ ina the
differ in set of
the FE sector-region
configuration
configuration dummies, Our
considered.
considered. and thus
Our pre-
pre-You can also read
