BIS Working Papers

No 1014
A shot in the arm:
stimulus packages and
firm performance during
Covid-19
by Deniz Igan, Ali Mirzaei and Tomoe Moore

Monetary and Economic Department

May 2022

JEL classification: G01, G14, G28, E65

Keywords: Economic stimulus, pandemic-prone,

COVID-19, policy effectiveness
BIS Working Papers are written by members of the Monetary and Economic
Department of the Bank for International Settlements, and from time to time by other
economists, and are published by the Bank. The papers are on subjects of topical
interest and are technical in character. The views expressed in them are those of their
authors and not necessarily the views of the BIS.

This publication is available on the BIS website (www.bis.org).

© Bank for International Settlements 2022. All rights reserved. Brief excerpts may be
reproduced or translated provided the source is stated.

ISSN 1020-0959 (print)

ISSN 1682-7678 (online)
A Shot in the Arm: Stimulus Packages and Firm Performance during
COVID-19†

by

Deniz Igana*^, Ali Mirzaeib and Tomoe Moorec

a
Monetary and Economic Department, Bank for International Settlements,
Basel, Switzerland, Tel: + 41 612808640, deniz.igan@bis.org
*
CEPR Centre for Economic Policy Research,
33 Great Sutton Street, London, EC1V 0DX, UK

b
Finance Department, School of Business Administration,
American University of Sharjah, PO Box 26666,
Sharjah, United Arab Emirates, Tel: + 97 165154645, amirzaei@aus.edu

c
Department of Economics and Finance, Brunel University London,
Uxbridge, Middlesex, UB8 3PH, UK,
Tel: + 44 1895274000, tomoe.moore@brunel.ac.uk

†
Awais Khuhro and Manzoor Gill provided valuable research assistance. We are grateful to
Jana Bricco, Federico Diez, Christian Ebeke, Pierre Guerin, Kristine Hankins, Sole Martinez
Peria, Manasa Patnam, and two anonymous referees for useful comments.

^ The views expressed here are those of the authors and not necessarily the views of the BIS.

1
A Shot in the Arm: Stimulus Packages and Firm Performance during
COVID-19

Abstract

We use firm-level data to provide some early evidence on the effectiveness of COVID-19
economic policy packages. Our empirical strategy relies on the varying degree of
vulnerability to the pandemic across industries. We find a robust association of fiscal
stimulus with changes in firm performance indicators (as measured by sales-to-assets ratio,
profit margin, interest coverage ratio as well as probability of default) in pandemic-prone
sectors. We also observe marginal effects of monetary policy on the sales-to-assets ratio
and of foreign exchange intervention on the interest coverage ratio in the hardest-hit firms.
These results broadly survive a battery of exercises to address endogeneity. Additionally,
we show that firms with a better financial position are more likely to take advantage of the
stimulus packages to withstand the pandemic shock. Overall, these provide preliminary
evidence suggesting that policy interventions have bought time for the hardest-hit
industries, by supporting turnover and improving liquidity.

JEL Classification Numbers: G01, G14, G28, E65

Keywords: Economic stimulus, pandemic-prone, COVID-19, policy effectiveness

 I. INTRODUCTION

COVID-19 prompted authorization and implementation of large economic policy packages

around the world, understandably so since a crisis like no other necessitated a response like no
other. These packages involved a combination of fiscal, monetary, financial, and capital-
account policies. An important question for academics and policymakers alike is how effective
these measures have been, especially by helping those sectors most in need.

In this paper, we use firm-level data to provide some answers to this question. Our
empirical strategy relies on the varying degree of vulnerability to the pandemic across
industries. Firms operating in sectors that rely more on face-to-face interactions when
producing goods or providing services are contact-intensive, and thus have a larger portion of
jobs that cannot be done at home. As a result, they are more vulnerable to non-pharmaceutical
interventions (such as social distancing or lockdown measures) that aim to stop or slow the
spread of virus. With the same token, economic policy support would aim to target these worst-
hit industries. Contrary to standard economic crises, stimulating real activity in a crisis like
COVID-19 is not only more challenging – given the complex nature of the shock combining
supply, demand and uncertainty factors – but could also be undesirable, in particular for the
contact-intensive sectors as this would go against the needed public health containment
measures. That said, economic policies would try to curb the Keynesian feedback loop
triggered by the abrupt and substantial loss of income in firms due to the shock, i.e. to minimize
spillovers and dislocation costs associated with business failures as well to ensure that liquidity
is sufficient enough to avoid unnecessary bankruptcies. One yardstick of success then is
whether policy actions have given more of a lift to these sectors relative to others, especially
with respect to supporting firms’ liquidity and capital.

To measure how prone different firms are to non-pharmaceutical interventions, we rely

on a proxy, namely, “distancing” measures that have been recently developed (Kóren and Petö
2020; Dingel and Neiman 2020; Hensvik et al. 2020) and utilized also by other researchers
(e.g., Pagano et al. 2020; Laeven 2020). These measures capture the degree to which jobs
require customer contact, teamwork, etc. at the sectoral level, as the share of workers in contact-
intensive occupations. We first confirm that firms in sectors with higher distancing indices
performed worse than the others in the same country, and especially so when the pandemic hit
to their country was more severe (as captured by the stringency of the lockdown measures,
which is highly correlated with the reported number of COVID-19 cases and deaths).

3
 We then examine whether performance metrics (efficiency, profitability, liquidity,
survival) in firms operating in more pandemic-prone sectors have fared better during the first
year of the pandemic (2020), if they are located in countries that deployed more comprehensive
economic stimulus packages (covering fiscal, monetary and foreign exchange). In other words,
if economic policies during the COVID-19 crisis portray an effective action in response to the
pandemic, we would expect this to be reflected in relatively better performance by firms that
are more pandemic-prone compared to those that are less so. Our main specification, thus,
focuses on the cross-sectional differences in firm performance depending on how sensitive to
distancing a sector is, controlling for sector and country fixed effects as well as firm
observables such as size, age and cash flow.

We find a robust positive association of fiscal stimulus with efficiency and profitability
(proxied by asset turnover, that is, the change in the sales-to-assets ratio and profit margin,
respectively) in pandemic-prone sectors: sales and profitability in firms that are more sensitive
to distancing have grown faster when the fiscal stimulus is larger. Furthermore, we observe
positive effects of fiscal packages on firm liquidity and survival (as measured by interest
coverage ratio and probability of default): interest coverage ratio increased while probability
of default decreased disproportionately more in pandemic-prone sectors.

Economically, moving from a country at the 10th percentile of the distribution of fiscal
stimulus (for example, Sri Lanka) to a country at the 90th percentile (for example, Germany),
the change in sales-to-asset ratio of firms in more pandemic-prone sectors is about 2 percent
more than their less pandemic-sensitive counterparts from 2019 to 2020. This is consistent with
Laeven and Valencia (2013), who report that fiscal policy disproportionately boosted the
growth of firms that were more dependent on external financing in the context of the global
financial crisis. Aghion et al. (2009) also find that counter-cyclical fiscal policy supported the
growth of manufacturing industries across 17 OECD countries over the period 1980–2005.

Additionally, we find that monetary stimulus is marginally associated with an

improvement in the sales-to-assets ratio. Prior to the COVID-19 outbreak, monetary policy
stance in major economies was already accommodative, raising questions about central banks’
ability to confront the next shock (Gagnon and Collins 2019). It appears that further easing has
proved to be still effective in improving revenues for firms that were hit hardest by the
pandemic. In this respect, the monetary policy transmission mechanism seems to have
remained functional during the pandemic, as opposed to the case of the global financial crisis

4
when banks were capital constrained and the lending channel was substantially weakened
(Laeven and Valencia 2013).

By contrast, we do not find a robust significant relationship between monetary policy

easing and the other firm performance indicators such as liquidity and probability of default.
This is in line with the argument that monetary policy may not be particularly well-suited to
deal with the implications of COVID-19 because of unsuitability of monetary policy in
addressing supply-side shocks and the difficulty to target stimulus to specific sectors that are
affected first and foremost by non-pharmaceutical interventions (Chen et al. 2020).

Foreign exchange interventions appear to arrest the decline of interest coverage ratio
during the pandemic for the hardest-hit (although this finding is not as robust as those on fiscal
and monetary policy measures). One plausible explanation for this finding may be that liquidity
in pandemic-prone sectors such as recreation services and tourism is highly responsive to
changes in the value of the domestic currency against foreign currencies.

Our findings are robust to a battery of checks, including different strategies to address
endogeneity issues and using alternative measures of distancing. We also verify that the results
remain broadly the same when we remove certain sectors or industries from the sample.
Additional analysis suggests that stimulus packages are generally more effective in larger firms
and firms entering the crisis with better liquidity, profitability and capital positions. The latter
finding provides some comfort that policy interventions in response to this entirely exogenous
shock may not have been distortive.

Our paper is linked to two strands of the literature. Firstly, it relates to those studies
investigating the effect of a crisis on corporate performance. Many recent additions to this
strand examine the real impact of the 2008–09 global financial crisis (see, among others,
Duchin et al. 2010; González 2015; Demirgüç-Kunt et al. 2020), adding to studies that look
more broadly at banking crises and sudden stops. Given that the COVID-19 crisis is still
unfolding, researchers have so far mostly examined the impact of the pandemic on stock market
performance (e.g., Alfaro et al. 2020; Fahlenbrach et al. 2020; Remelli and Wagner 2020).
Rather closely related to our analysis, Pagano et al. (2020) find that the impact of COVID-19
on stock performance was more severe for firms that operate in sectors that are more vulnerable
to social distancing. Ding et al. (2020) report that the adverse impact of the pandemic on stock
returns is more pronounced for those firms that have more anti-takeover devices, lower social
and corporate responsibility scores, and that depend more heavily on global supply chains.

5
Papanikolaou and Schmidt (2020) reveal that expected revenue growth of those sectors in
which a higher fraction of the workforce is not able to work remotely declined significantly
during the COVID-19 pandemic. Glover et al. (2020) document that the impact of the COVID-
19 pandemic is more serious amongst younger generations working in vulnerable sectors. Our
analysis adds to these studies by providing early evidence that balance sheet and income
indicators also show that the pandemic has taken a heavier toll on firms operating in sectors
that are more sensitive to distancing.

Secondly, we contribute to the literature that assesses the effectiveness of government

policy measures during a crisis (see, for instance in the context of the global financial crisis,
Laeven and Valencia 2013; Norden et al. 2013). By focusing on the differences across sectors,
we also build on studies investigating the channel through which the real effect of a crisis
materializes. See, for example, Claessens et al. (2012), Chaney et al. (2012), Chodorow-Reich
(2014) and Giroud and Mueller (2017) with regard to the global financial crisis, and Laeven
(2020) and Leibovici et al. (2020) with regard to COVID-19. Our study differs from these
papers since it focuses on the effectiveness of government economic policies during the
COVID-19 pandemic, rather than the transmission of the shock itself, by testing whether firms
in pandemic-prone sectors performed disproportionately better if they are domiciled in
countries with more comprehensive or larger stimulus packages. Closely related to our
analysis, Demirgüç-Kunt et al. (2021) examine the impact of policy measures taken in response
to the COVID-19 pandemic but only on the performance of the banking sector. They find that,
while policy interventions in the form of liquidity support, borrower assistance and monetary
easing, in general, mitigate the adverse impact of the crisis, this is not the case for all banks,
nor in all circumstances.

The rest of the paper is organized as follows. Section II summarizes the potential
channels through which policy stimulus could help firm performance in the hardest-hit sectors.
Section III lays out the methodology and the data. Section IV presents the findings. Section V
concludes.

II. POTENTIAL TRANSMISSION CHANNELS

A. Fiscal Stimulus

Fiscal stimulus packages implemented during the pandemic aimed to support businesses and
households at a time economic activity was intentionally curtailed to slow the spread of the

6
virus and allay the burden on public health systems. Specific measures included tax cuts, cash
handouts and social welfare payments on the demand side and tax relief measures and
guarantees for access to credit on the supply side (Padhan and Prabheesh 2021).

There are various ways such measures could help firms. Firstly, corporate tax breaks
could lessen the decline of profitability. Tax payment deferral is the common type of measure,
in particular, in less developed countries (OECD 2000a). Yet, this has a limited benefit to the
pandemic-prone sectors, as they have hardly generated profits and rather suffered from losses
during the crisis. In this instance, alternative measures such as loss carry-back tax provisions
can be more effective (OECD 2000a; Makin and Layton 2021). This allows firms to claim the
losses against taxable profits in previous years, which potentially reduces the losses incurred
during the COVID crisis. Such provisions have been introduced in some countries for the 2020
tax year.

Secondly, temporary increases in thresholds for low-value asset write-offs and

depreciation allowances could mitigate the decline of investment, since they effectively reduce
the tax liability of firms. The benefit should be felt across all sectors. However, if the contact-
intensive sectors have to alter their business structure in order to survive the pandemic and if
this requires investment, then this support should be more advantageous to these sectors. For
instance, restaurants may adapt their services away from in-person dining and towards
takeaway and delivery of food, or redesign the layout of the premises to maintain distance
among customers. Such changes necessitate new investment and could be supported by
investment incentives through temporary changes in the tax code. They would help maintain
sales and profitability.

Thirdly, direct government subsidies such as furlough schemes curb the massive
employment loss due to lockdowns. Many countries have helped the hardest-hit sectors retain
their workers by providing income support to employees whose working hours have been
curtailed or who have been temporarily laid off (OECD 2020b). The scheme enables firms to
maintain the contract with them and to preserve workers’ talent and experience. It also deters
the decline of production side of the firms, since firms are able to quickly resume operations
when the lockdown is eased.

Fourthly, many heavily affected businesses have been experiencing a sharp decline in
liquidity. To deal with the liquidity shortage, the most common instrument among developed
countries has been loan guarantee schemes, where the government guarantees all or part of the

7
bank loans granted to eligible businesses (OECD 2020a). Other measures have included
interest-free loans and cash grants. These measures are typically able to target or prioritize
those businesses adversely affected by the pandemic, alleviating cash flow difficulties,
enabling firms pay suppliers or creditors and, hence, avoid default or bankruptcy.

Finally, subsidies to consumers for consumption of certain goods and services could
also help the suppliers of such goods and services. This can target the hardest-hit sectors, for
example, some governments provided subsidies for eating out or domestic travel.

In general, the delivery speed of stimulus should be a key consideration. For instance,
countries may find it timelier to provide loan guarantees, business grants or wage subsidies
rather than tax measures. The effect of the latter is only felt at the end of the tax year. In order
to achieve prompt delivery, fiscal aid may also be provided broadly across all sectors rather
than targeting certain sectors, but then this is subject to taxation of regular profit. This would
imply that adversely affected firms are able to keep the full amount of support by documenting
the hit to their profits, while the firms whose economic circumstances have been affected the
least would return some of the support via the tax system (Mankiw 2020; Marron 2020).

There is, however, potentially an unintended side effect of fiscal stimulus. Higher
public debt fueled by the pandemic may harm business and household confidence, creating
uncertainty about how public debt would be repaid (OECD 2020a). To the extent that firms
perceive higher public debt to imply higher corporate taxes in the future, it would be reflected
as a negative repercussion on the firms’ performance. Note also that wage subsidy programs
implemented in some countries may prove to be an innovative yet extremely costly way of
sustaining business activities and employment, accelerating government debt. Besides, this
support may simply delay the inevitable re-deployment of labor away from unviable firms and
may not bring about a particular benefit to the vulnerable firms.

B. Other Policy Stimulus

With other stimulus actions such as monetary policy and foreign exchange intervention, unlike
fiscal stimulus, the channels of transmission are not as clear. This is partly because it is difficult
to target or prioritize specific sectors or firms that have been bruised by the pandemic.
Nevertheless, there is some scope for these measures to alleviate the adverse effect of COVID
on these sectors.

During the pandemic period, most economies have experienced exchange rate volatility
and often intervened in the foreign exchange market. Vulnerable firms engaged in tourism or

8
international trade may disproportionately benefit from such intervention, mitigating a decline
of profits and strengthening ability to meet debt obligations.

Expansionary monetary policy may mitigate the effects of COVID-19 on the hardest-
hit sectors if firms in these sectors come under pressure from a tightening of credit conditions.
For instance, a fall in interest rates may enable vulnerable sectors to ease liquidity concerns
and reduce the probability of default.

III. METHODOLOGY AND DATA

A. Empirical Strategy

Our main empirical strategy is to examine whether firms in industries that are more pandemic-
prone (that is, industries in which a significant share of employment is affected by social
distancing) perform disproportionately better during the COVID-19 outbreak, if they happen
to be located in countries that have larger government stimulus packages. In other words, if
policy measures are to help firms during the pandemic, with regard to their real performance,
then one would expect them to have a larger effect on sectors that are more vulnerable to social
distancing measures. This inference can be empirically tested by estimating an econometric
model in which the effect of government policies on firms is allowed to differ depending on
how pandemic-prone is the industry to which the firm belongs. 1 Thus, our model specification
is given by:

∆𝑦𝑦𝑖𝑖𝑖𝑖,𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 = 𝜗𝜗𝑗𝑗 + 𝜗𝜗𝑐𝑐 + ∅. 𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝑗𝑗 × 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑐𝑐 + 𝜏𝜏. 𝑋𝑋𝑖𝑖𝑖𝑖,𝑃𝑃𝑃𝑃𝑃𝑃 + 𝜀𝜀𝑖𝑖𝑖𝑖,𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 (1)

where 𝑖𝑖 stands for firm, 𝑗𝑗 for sector, and 𝑐𝑐 for country. This is a cross-sectional regression
where ∆𝑦𝑦𝑖𝑖𝑖𝑖,𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 is the measure of change in performance indicators for firm 𝑖𝑖 in country 𝑐𝑐
between 2020—the latest data available—and 2019—a year prior. Following Claessens et al.
(2012), we use the changes in firm-level performance. Given that COVID-19 began to spread
in many countries and was declared a pandemic in 2020Q1, the pattern of change in
performance indicators is deemed to be due to the pandemic.

We employ four response variables to test the impact of government policy

interventions. Specifically, we use, in respective regressions, (i) change in the asset turnover
ratio [∆(𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆)], as measured by sales to total assets ratio; (ii) change in profit margin

1
This approach is an augmentation of the literature that examines the relationship between government
intervention and firm performance during a (financial) crisis (see, for example, Norden et al. 2013 and Laeven
and Valencia 2013).

9
[∆(𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃)], as measured by the net profit to total revenue ratio; (iii) change in the interest
coverage ratio [∆(𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼)], as measured by earnings before interest and tax divided by interest
expenses; and (iv) change in the probability of default [∆(𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃)], as measured by the default
risk of publicly listed firms by quantitatively analyzing numerous covariates (see Section III.B
for the detail of this data).

Following Claessens et al. (2012), we employ the changes in sales to asset ratio and
profit margin to investigate the impact of economic stimulus packages on firm efficiency and
profitability. In addition, crises have severe effects on firms’ financial health in two aspects
(Carletti et al. 2020): draining cash generation and liquidity that is necessary for functioning of
firms and evaporating capital. Since during the public health crisis firms find it difficult to
generate cash and thus could be expected to default on some obligations, we use the interest
coverage ratio to determine whether policy measures help a company to pay interest on its
outstanding debt. Also, following Ganganis et al. (2020) and Igan et al. (2022), we use an
indicator for the probability that a firm will continue operations, which is the probability of
default. It captures the likelihood of a default over a particular time horizon, reflecting not only
the borrower's characteristics but also the economic environment. Overall, the first three
response variables intend, in the main, to gauge whether government policies help firms in
maintaining their cash flow and, hence, improving liquidity, and the fourth variable aims to
capture the impact on firms’ survival.

𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑐𝑐 is a vector of variables that represent the economic stimulus package in country
c. We employ three policy variables as follows: (i) cumulative fiscal stimulus expressed in
percentage of GDP, (ii) cumulative change in the monetary policy rate expressed in basis
points, and (iii) interventions in foreign exchange markets (0 for no intervention, 1 for
intervention). All policy measures are computed over the period January 31st, 2020 (week 1)
to December 4th, 2020 (week 43). We investigate the change in the performance of firms over
the period 2019–2020 in response to government polices during the period from January to
December 2020. We believe that this period represents the most important initial stage of the
spread of the crisis, when countries declared the bulk of their policy packages. This is also the
period of the collapse of international trade due to non-pharmaceutical public health
interventions such as full (or partial) lockdowns and turmoil in financial markets as
expectations were quickly revised to take the impact of the pandemic fallout on the global
economy into account.

10
 𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝑗𝑗 is industry j’s degree of sensitivity to a pandemic, computed as the share
of industry employment affected by social distancing at the three-digit NAICS level (created
by Kóren and Petö 2020, and also used by Laeven 2020 and Pagano et al. 2020; we describe
this proxy further in the following subsection).

𝑋𝑋𝑖𝑖𝑖𝑖,𝑃𝑃𝑃𝑃𝑃𝑃 is a vector of firm-level explanatory variables, computed as of 2019. Note that,

because of the pure cross-sectional nature of our empirical strategy, we enter all firm-level
control variables as pre-determined (as do Laeven and Valencia 2013). We first consider the
following five variables: (i) size (𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆), measured as the natural log of total assets; (ii) age
(𝐴𝐴𝐴𝐴𝐴𝐴), calculated by subtracting the firm’s incorporation year from 2020; (iii) cash holdings
(𝐶𝐶𝐶𝐶𝐶𝐶ℎ𝐴𝐴), computed as the ratio of cash and cash equivalents to total assets; (iv) investment in
R&D (𝑅𝑅𝑅𝑅_𝐴𝐴), measured by R&D investment to total assets ratio; and (v) a dummy for private
firms (𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃). These controls are informed by the literature on the determinants of firm
performance. Small firms tend to perform worse than their larger counterparts during a crisis
(Gandhi and Lusting 2015). Younger firms face more constraints (Beck et al. 2006; D’Souza
et al. 2017). Firms with larger cash holdings tend to be more resilient during a crisis while firms
with better growth potential tend to invest more in R&D (Bates et al. 2009). Finally, privately-
held firms may be different from their listed counterparts along the dimensions we investigate.
For instance, Hall et al. (2014) document that public companies hold less cash given their
greater access to capital markets as compared to privately-held firms. In addition to these five
variables, we also include lags of the following variables as additional regressors: (vi) 𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆,
to control for efficiency in generating revenue for a given level of assets; (vii) 𝑅𝑅𝑅𝑅𝑅𝑅, to control
for pre-crisis differences in levels of profitability; (viii) 𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼, to control for ability to cover
current interest payments with available earnings; and (ix) 𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸, to control for leverage given
that more highly leveraged firms may face difficulty raising funds during a crisis (Giroud and
Mueller 2019). Overall, all these nine firm-level control variables are rather common in the
literature (e.g., Burns et al. 2017; Barbiero et al. 2020; Demirgüç-Kunt et al. 2020).

The main variable of interest is the interaction term 𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝑗𝑗 × 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑐𝑐 . The
coefficient ∅ measures the difference between performance in pandemic-prone sectors in
countries with high and low economic stimulus packages. A positive and significant point
estimate of ∅ indicates that the vulnerable industries in countries with higher levels of
government economic response do not suffer as much from the pandemic (we expect a negative
∅ for the probability of default).

11
 𝜗𝜗𝑗𝑗 refers to a vector of sectoral dummies (at three-digit NAICS level) to control for
sector-specific factors that could affect cross-sector performance differentials. 𝜗𝜗𝑐𝑐 are country
dummies that account for time-invariant country-specific features that might drive cross-
country differences in firm activity, such as the institutional environment. This set of fixed
effects absorbs all observable and unobservable time-invariant variations across sectors and
countries. Also, they subsume the direct level effects of social distancing and economic
policies, namely the 𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷 and 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃 variables in Eq. (1). By including this set of fixed
effects, our identification is obtained by looking at the differential performance of two
otherwise identical firms operating in more and less pandemic-prone sectors.

Eq. (1) is estimated with ordinary least squares (OLS). Residuals from OLS estimations
may be correlated across countries, resulting in biased standard errors. Thus, following
Demirgüç-Kunt et al. (2020), we cluster standard errors at the country level. An advantage of
our empirical strategy is that it incorporates information about heterogeneity across countries
in initiating and implementing economic stimulus packages.

One concern is that Eq. (1) is subject to the problem of endogeneity. Firstly, any
association between government policies and firm performance may be attributable to omitted
variables. Or, it could be that the effect of a particular policy is attributed to another because
of their simultaneous implementation. Secondly, firm performance during a crisis may affect
policy responses to the crisis, indicating the possibility of reverse causality.

Our empirical setup provides some leeway in alleviating these two endogeneity issues.
By including all policy measures at once, we reduce the issue of simultaneity while our use of
sector and country fixed effects mitigates the issue of omitted variable bias. In addition, we
control for other potential channels through which policy measures may affect firm
performance so as to gain more confidence that distancing remains a relevant channel for policy
measures to influence firm performance. Also noteworthy is that endogeneity may even play
against the chances to reject the null hypothesis: countries that are affected more by COVID-
19 — that is, where pandemic-prone sectors are large and fare particularly bad — may be more
likely to deploy large policy packages, giving rise to a negative correlation (which is the
opposite of what we find). Nevertheless, we admit that the issue of endogeneity may continue
to exist, hence we address this point by conducting several exercises in Section IV.B. 2

2
Arguably, our use of firm-level data, with distancing measured at the sectoral level, also introduces some degree
of separation. While it is plausible that policies are more likely to be enacted where pandemic-prone sectors make
up a larger portion of the economy, it is unlikely that government policy responds only to the performance of a

12
 B. Data Sources

Applying the empirical strategy laid out in Section III.A requires measures of firm
performance, sectoral pandemic sensitivity, and economic policy actions. This subsection
describes the process of compiling these data.

Firm Performance

Firm-level data come from the ORBIS database by Bureau Van Dijk, which provides
information on balance sheets and income statements for more than 40 million listed and
private companies from more than 100 countries worldwide. As one of the most comprehensive
databases of firm-level information, it has been increasingly used in academic research (e.g.
Frijns et al. 2016; Baumohl et al. 2019; Demirgüç-Kunt et al. 2020; Barbiero et al. 2020;
Cathcart et al. 2020).

We obtain data for 2019 and the latest year available, 2020 (at the time of conducting
this analysis). This enables us to calculate the change in firm performance during the COVID-
19 pandemic. We initially select all firms that belong to the nonfinancial corporate sector,
excluding financial firms (identified as firms with NAICS2017 code of 52). We drop firms in
sectors with no data on the distancing variable, which are “management of companies and
corporations,” “public administration,” and “unclassified establishments” (NAICS2017 codes
of 55, 92 and 99, respectively). Also, countries with no data on policy measures are excluded.
In addition, we drop offshore financial centers. Following Demirgüç-Kunt et al. (2020), we
further restrict our sample to countries with a minimum of 20 firms (with available information
for 2020). Last but not least, we drop the United States since data on U.S. sectors are used to
construct the distancing variable to avoid any mechanical endogeneity between this variable
and firm performance. 3 Given our interest in evaluating the effectiveness of policy measures
in improving firm performance during the COVID-19 crisis, we focus our baseline analysis on
firms that are present in both before and during the crisis. We thus clean the dataset further by

particular firm in a pandemic-prone sector. Indeed, the correlation between policies and average distancing of
firms in a given country is at most 7 percent (between distancing proxy and fiscal policy variable; for other
policies, the correlation is less than 3 percent).

3
In order to establish the benchmark of an industry’s pandemic sensitivity, Kóren and Petö (2020) use U.S. data.
Hence, one may argue that this proxy could be endogenous to the performance of U.S. firms. Therefore, following
other studies that apply the Rajan and Zingales (1998) approach such as Igan and Mirzaei (2020), we drop U.S.
firms from all regressions. Yet, for completeness, we check the robustness of the findings and present the results
obtained with U.S. firms. See Section IV.B.

13
excluding all firms with no data available on sales as well as on our main firm-level control
variables. This means we focus on the effects of policy measures on the intensive margin only.

As a result, 28,915 firms from 80 countries survive the filtering criteria. 4 Note that,
after imposing such additional criteria, we end up with one country (Mongolia) with less than
20 firms. We confirm the robustness of our results to excluding this country. The number of
firms in our dataset varies by country. On average, each country has about 361 firms with
available data. We reduce the influence of outliers by winsorizing all dependent variables at
the 1st and 99th percentiles.

Following Ganganis et al. (2020) and Igan et al. (2022), the data on probability of
default measure is from the Credit Research Initiative (CRI) of the National University of
Singapore. The probability of default estimates the default risk of publicly listed firms by
quantitatively analyzing numerous covariates that cover market-based and accounting-based
firm-specific attributes, as well as macro-financial factors (Duan et al. 2012). We use a
prediction horizon of 1 month. Note that the data on probability of default is not available for
all 80 countries and/or 28,915 firms. We have data only for 10,023 firms.

Pandemic Sensitivity

We now define our index for a sector’s relative sensitivity to social distancing (𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷).
As already stated, the impact of the COVID-19 shock on firms is likely to vary by the sector in
which they operate. More specifically, some sectors—for example, accommodation and food
services—are more vulnerable to pandemic-induced social distancing measures.

Kóren and Petö (2020) estimate each sector’s contact intensity, using pre-pandemic
data from the Occupational Information Network (O*NET) survey. Specifically, they use
information for 809 occupations from the 2010 Standard Occupational Classification System
to compute, for each NAICS three-digit code, the share of workers whose job requires a high
level of three occupational characteristics: customer contact, teamwork and physical presence. 5
They end up reporting two proxies. The first one is a measure of “communication” intensity

4
We acknowledge that the sample of firms we study is biased toward larger firms as almost all firms (about 95
percent) reporting 2020 data are listed firms. Thus, we are conservative when interpreting our results, as we cannot
analyze the overall effect of policy measures on the performance of small and medium-sized enterprises during
COVID-19.

5
Some industries will have high scores in all three dimensions while others may have high scores only in one.
For instance, most manufacturing requires physical presence but not necessarily face-to-face customer contact.

14
that incorporates teamwork-intensive and customer-facing activities. The second proxy,
“overall” incorporates the physical presence dimension to the first.

In our baseline, we use communication intensity as the metric for distancing. This
arguably captures the nature of non-pharmaceutical interventions put in place in response to
COVID-19 due to the fact that shelter-in-place or stay-at-home orders were gradually lifted
allowing industries that primarily rely on physical presence (e.g., construction, factories) to get
back to work whilst travel restrictions, bans on public gatherings and specific business closures
(e.g., gyms and restaurants) remained. We confirm the robustness of our results to the use of
“overall” index. We assume that distancing is an intrinsic characteristic of a sector and, thus,
indices derived using U.S. data can be used for the same sector across all countries.

Policy Measures

The data source for policy response is the IMF’s Policy Tracker. 6 Launched right around the
time COVID-19 was declared a pandemic by the WHO on March 11, 2020, this tracker relies
on responses by individual country teams to a survey designed by IMF staff. The teams are
asked to fill the survey on a weekly basis in order to capture any new announcements and
changes to previously implemented measures. The survey seeks responses on all policy actions
taken by the authorities in a country covering fiscal, monetary, external and financial policies.
The survey asks about not only whether a fiscal or monetary policy action has been taken but
also the size of the intervention. For external and financial policies, information is gathered in
a categorical manner with 0 denoting no action and +1 an intervention for foreign exchange
market. 7

In our analysis, we use three policy measures: (i) 𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹, cumulative fiscal stimulus
expressed in percentage of GDP, (ii) 𝑀𝑀𝑀𝑀_𝐵𝐵𝐵𝐵, cumulative change in the monetary policy rate
expressed in basis points, and (iii) 𝐹𝐹𝐹𝐹𝐹𝐹, the interventions in foreign exchange markets. All
policy measures are computed over the period January 31, 2020 (week 1) to December 4, 2020
(week 43) to overlap with the firm-level data we have.

The intensity of economic stimulus packages varies considerably (see Section III.C for
more details). By including countries with no or less strong policy responses as well as those

6
Available at https://www.imf.org/en/Topics/imf-and-covid19/Policy-Responses-to-COVID-19.

7
The survey also gathers information on capital flow management measures and broader financial market policy
measures such as loan forbearance and debt moratoria. We do not include these in our analysis given that there is
not enough variation across countries to tease out any differential effects.

15
with proactive intervention in our dataset, we reduce concerns that the results may be driven
by selection bias.

One shortcoming of this database is the lack of granularity as to the exact measures
implemented, in particular, under the fiscal stimulus packages. While a more in-depth analysis
would be desirable, it would be better conducted in a set of relatively homogeneous country
sample (if not, within a single country). We leave this for future research. That said, we do
confirm the robustness of our findings using policy measures from alternative source (see
Section IV.C)

Other Variables

Additional data are retrieved from standard databases such as the World Bank’s World
Development Indicators (WDI).

Appendix Table A1 details the construction of the main variables that we employ in the
analysis. Note also that Appendix Table A2 shows the number of firms by country.

C. Descriptive Statistics

Insert Table 1 (Panels A and B) around here

Table 1 shows the summary statistics of change in the asset turnover ratio [∆(𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆)], change
in profit margin [∆(𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃)], change in the interest coverage ratio [∆(𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼) and change in
the probability of default [∆(𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃)] over the period from 2019 to 2020. The relation between
firm performance and Distancing at the sector level is presented in Panel A. Retail Trade
(NAICS 44-45) and Health Care and Social Assistance (NAICS 62) have the highest share of
communication-intensive jobs, exceeding or around 60 percent. This is followed by Art,
Entertainment, and Recreation (NAICS 71) and Accommodation and Food Services (NAICS
72) at around 40 to 44 percent. These two sectors suffered from the largest decline in sales to
asset by around 20 to 24 percent and also in profit margin, dropping by around 18 to 24 percent.
We observe the same pattern for the other two indicators of firm performance, change in
interest coverage ratio and probability of default, for these two sectors. This heterogeneity
across sectors is important to understand the effect of the pandemic and associated policy
measures. The summary statistics for the main variables used in the regression analysis are
shown in Table 1, Panel B. The sectors are classified as more pandemic-prone (greater than
cross-country median) and less pandemic-prone (less than median) for the four dependent
variables capturing firm performance. Amongst others, the mean values clearly indicate lower

16
sales and profit margin, negative interest coverage and higher probability of default for sectors
that are more vulnerable.

Insert Figures 1a and 1b around here

Figure 1a presents the cumulative fiscal stimulus from January to December 2020 for
individual countries. Three advanced economies (Italy, Germany and Japan) top the chart
spending more than 30 percent of GDP. The cumulative change in monetary policy from
January to December 2020 by country is shown in Figure 1b. Emerging market economies
appear to utilize monetary policy more. These patterns are in line with perceived policy space:
ability to run budget deficits and increase in public debt being limited in developing countries,
whereas interest rates being already historically low and close to the effective lower bound in
advanced economies.

IV. EMPIRICAL FINDINGS

Using data for a maximum of 28,915 non-financial firms in 80 countries, we examine how the
COVID-19 crisis has propagated across industries and how policy actions have helped alleviate
the impact on firm performance.

Insert Table 2 around here

Before presenting our baseline results, we first show that the adverse impact of the
COVID-19 outbreak on firm performance is indeed more pronounced on pandemic-prone
sectors. This is to validate our main hypothesis that, if economic stimulus packages are
effective, then pandemic-sensitive sectors benefit more. Applying a form of Eq. (1), Table 2
shows the impact of the pandemic. Column 1 displays a negative significant sign on the
coefficient of 𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷 when the dependent variable is ∆(𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆). It implies that a decline
in the sales to be more noticeable for those sectors that are intrinsically more sensitive to social
distancing. This result suggests that there is indeed a significant channel captured by
𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷, consistent with Kóren and Petö (2020). The change in
in profit margin, ∆(𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃), in column 2 is also negative and significant. It is a plausible result
in that firms in pandemic-prone sectors are unable to generate profit, whilst experiencing a
drop in sales, possibly being forced to cut profit margins in order to survive the pandemic. We
observe a negative significant sign on the coefficients of ∆(𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼) (change in the ratio of
earnings to interest expenses) in column 3 and a positive sign on the coefficient of probability
of default, ∆(𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃) in column 4. The crisis has not only drained liquidity, but has also
increased the probability of default in vulnerable firms.

17
 In columns 5–8, 𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝑗𝑗 is interacted with 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶_𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑐𝑐 , that is, the country-
level severity of the lockdown measures in response to the pandemic. This is a composite
measure based on nine response indicators including school closures, workplace closures, and
travel bans, rescaled to a value from 0 to 100 with the score 100 being the strictest (Hale et al.
2020). Although the reported results suggest that there is still the same sign in all cases, they
are less significant both statistically and in terms of magnitude when compared with those in
columns 1–4. 8 The implication is that more weight is placed on the vulnerability of the specific
sectors, rather than the exposure to the pandemic at the country level when explaining firm
performance. Indeed, this is a reasonable outcome: sectors such as tourism and airlines are
severely affected, whereas others such as information technology markedly benefit from social
distancing. This phenomenon is common across countries.

As a separate exercise, we first run a regression of policy measures without interacting

them with 𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷 and then with the interactions. This is to check the overall impact of
policy measures on firm performance. As shown in Table A3 in the Appendix, almost all
coefficients on stimulus variables are insignificant at conventional levels except for one in
column 4. But when interacted with 𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷, we find significant coefficients in many cases
for pandemic-prone sectors.

Collectively, the results in Table 2 and Table A3 suggest that pandemic-prone sectors
were affected more severely by the COVID-19 pandemic, and that economic stimulus packages
have disproportionately benefited these sectors. However, we cannot make any conclusions yet
because these specifications do not employ a full set of fixed effects (though
𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶_𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑐𝑐 and several other country characteristics are included as controls). We now
turn to our baseline results using the full version of Eq. (1).

A. Baseline Results

Having observed the negative performance of pandemic-prone sectors in Table 2, we explore

whether these industries are the ones that benefited more from economic stimulus measures.
We specifically examine whether the stimulus measures alleviate the severity of the
pandemic’s impact on firm performance by interacting the policy measures with the distancing
proxy. In Table 3, all policy measures are simultaneously included.

8
Note that, in the specifications in columns 5–8, we include sector fixed effects. Hence, the coefficient on
distancing itself may be absorbed in the fixed effects.

18
 Insert Table 3 around here

Overall, we find evidence of positive impact of fiscal policy during the pandemic for
vulnerable sectors, and indeed, the various types of fiscal stimulus appear to be working as
intended by policy makers. 9 Fiscal policy is statistically significant in all four cases (columns
1-4) at the 5% or 1% levels, improving the sales, profit margin and liquidity position (proxied
by interest coverage) and, at the same time, decreasing the probability of insolvency for
vulnerable sectors, holding other policy variables constant. This is consistent with the findings
in Aghion et al. (2009), Claessens et al. (2012), and Laeven and Valencia (2013).

Monetary policy easing appears to have been effective in supporting sales revenue,
though at the marginal significance level of 10% (column 1). Note that the various robustness
tests conducted in the subsequent subsections reveal a clearer effect of monetary policy on firm
performance. In this respect, the functioning of monetary policy transmission appears to be, at
least, preserved during the pandemic. This is in contrast with the case of the global financial
crisis in 2008, when bank balance sheet constraints substantially weakened monetary policy
transmission (Van den Heuvel 2009), and attests to the importance of advances in prudential
regulation and deleveraging that allowed banks face the pandemic shock in much better shape
than they did the subprime mortgage shock.

Foreign exchange intervention seems to have mitigated the decline of interest coverage
(the ratio of earnings to interest expenses) during the pandemic (column 3). One of the possible
explanations of this outcome may be due to the fact that the earnings of pandemic-prone sectors
such as tourism are receptive to changes in the value of the domestic currency against foreign
currencies. By limiting excessive volatility in the exchange rate, FXI may have protected
earnings and kept interest expenses in check (especially if part of the debt is denominated in
foreign currency). In this respect, the positive sign on the coefficient of profit margin
(∆(𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃)) is consistent, though it is insignificant.

Note that sales and interest coverage seem to be more responsive to policy measures,
whereas the performance indicators of profit margin and the probability of default respond only
to fiscal measures. This may be explained as follows: during the pandemic, there is little scope

9
As a separate exercise, we run a regression of policy measures with two indicators of real activity as dependent
variables: change in ‘the number of employees’ and change in ‘value added’. The result is shown in Appendix
Table A4, where we find a highly significant effect of fiscal stimulus for both cases, supporting the baseline result
in Table 3. Given the limited sample size in this instance, we leave further exploration of real effects to the future
when more data become available.

19
for raising margins for those vulnerable firms, culminating in smaller response to other policy
stimulus than to specific fiscal instruments such as tax payment deferral or loss carry-back tax
provisions that may have a direct impact on the profit margin. Similarly, default is not an
unlikely outcome for vulnerable firms, in particular, for those with high levels of debt that were
accumulated before the pandemic. Fiscal measures such as loan guarantee schemes, interest-
free loans or cash grants may have effectively mitigated the risk of default. 10

The effect of fiscal stimulus is economically meaningful. Focusing on the sales-to-

assets ratio, the estimation results in column 1 of Table 3 suggest that a firm from an industry
at the 90th percentile of distancing would have change in sales-to-asset that is 2.05 percentage
points higher than a firm from a sector that is at the 10th percentile of distancing, if it were
located in a country that is at the 90th percentile compared to a country at the 10th percentile of
fiscal stimulus. Similarly, the estimation results in column 4 of Table 3 suggest that, relative to
less pandemic-prone sectors (in the 10th percentile level), the probability of default in more
pandemic-sensitive sectors (in the 90th percentile level) is around 0.68 percent less in a country
that launched significant fiscal stimulus packages (in the 90th percentile) than in a country with
a limited adoption of economic support in the form of fiscal stimulus (in the 10th percentile).
Note that both differentials in changes in sales-to-asset ratio and the probability of default are
substantial, compared to the average rates of change in 𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆 and 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃 (which are –8.26
percent and 0.22 percent, respectively).

B. Endogeneity Issues

The key challenges for our identification strategy are the two conventional endogeneity
problems – omitted variable bias and reverse causality. In this section, we present several
exercises to address these issues.
Omitted variable bias
The established positive correlation between stimulus packages and firm performance
is in line with our hypothesis, which suggests that government policies during the pandemic
have provided life support for hardest-hit firms. Yet, the stimulus packages may have been
automatically picking up the effects of some country-level and/or sectoral-level omitted
variables that are also likely to affect firm performance. In fact, latent omitted variables that

10
Due to differences in number of observations across our four dependent variables, it is worth confirming the
validity of the findings in the restricted sample (that is, only the firms with non-missing observations for all
response variables). Results reported in Appendix Table A5 columns 1-4 indicate that our main findings remain
broadly unchanged, except for the case where profit margin is the dependent variable.

20
are correlated with both firm activities and stimulus policies may raise the issue of endogeneity
(Hyytinen and Toivanen, 2005). For instance, countries more open to trade and cross-border
capital flows may launch more economic support given their larger exposure to the negative
shocks and the broader lift to the economy cushions firms’ performance, and this may, in turn,
lead to a spurious positive association between stimulus packages and firm performance. To
address the omitted variable bias, we employ two approaches to evaluate the significance of
such variables. Firstly, we control for observable characteristics – especially at the
country/industry level – that may affect firms’ performance. Secondly, we make selection on
these observable factors to determine the likelihood that our estimates are being driven by
unobserved heterogeneity across countries/sectors.

Following the existing literature, the first approach involves including a set of country-
and sector-level control variables. We rely on two sets of characteristics. The first set includes
those that are related to firm activities through country characteristics (Other country
characteristics). These include five variables classified into four groups of a) pandemic
resilience, b) channels of transmission, c) bank stability, and d) macroeconomic stability (see,
for example, Claessens et al., 2010; Martin and Nagler, 2020; Igan et al., 2022).

a) Pandemic resilience: We consider the vulnerability of a country to the pandemic by

utilizing the variable of private health expenditure per capita (𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝐻). When people in
a country are able to obtain health services without suffering financial hardship, the
country can reasonably be expected to be more resilient to a pandemic such as COVID-
19. Data are collected from the World Health Organisation as reported by the World
Bank.
b) Channels of transmission: Previous research has highlighted the role of real and
financial channels through which a crisis can spread across countries. While arguably
less applicable to the case of COVID-19 given the different nature of the shock, these
channels may still matter in the transmission of the economic effects. For instance,
given restrictions imposed on movement across and within borders, supply can be
disrupted and countries that are more connected to global value chains may feel the
effects more profoundly. We consider two variables: (i) foreign direct investment
(𝐹𝐹𝐹𝐹𝐹𝐹), as a proxy for financial interconnectedness, and (ii) total exports and imports in
% of GDP (𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇), as a proxy for a country’s economic integration with the rest of the
world. Data are retrieved from the World Bank.

21
 c) Bank stability: The health of bank balance sheets could be an amplifier of the economic
shocks. In order to capture bank health, we include the ratio of non-performing loans to
total loans (𝑁𝑁𝑁𝑁𝑁𝑁). Data come from the World Bank.
d) Macroeconomic stability: We capture the general macroeconomic stability of a country
by including inflation (𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼). Data are collected from the WDI. 11
These country-specific features are interacted with Distancing. Estimates of equation (1)
controlling for these controls are reported in Table 4A.

Insert Tables 4A, 4B, 4C and 4D around here

We find almost identical estimated coefficients of the interaction terms between

distancing and policy variables in terms of direction, significance and magnitude. This
highlights that pandemic-prone firms disproportionately benefit from fiscal stimulus in firms’
performance, in general. Expansionary monetary policy continues to raise sales, and mitigates
the risk of default. The latter result is new relative to the baseline in Table 3 and a plausible
outcome as monetary policy easing can relieve debt service pressures. As before, the liquidity
position in vulnerable sectors improves when there is government intervention in the foreign
exchange market.

Other channels of propagation emphasized in the literature involve liquidity constraints

and sensitivity to consumer demand in non-financial firms. 12 Hence, as a second set of
characteristics, we consider the effect of these two sectoral characteristics, which may interact
particularly with financial policy measures. It follows that we control for sectoral
characteristics by interacting the variables of external financial dependence (𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹) and
demand sensitivity (𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷) of individual sectors, respectively, with the stimulus variables.
We use the Rajan and Zingales (1998) index for external finance dependence and an index of

11
One of the important country-level variables that may affect firm performance is the debt-to-GDP ratio. The
ability of a country in diminishing the adverse impact of a crisis on corporate activity and risk may be related to
its level of debt (Martin and Nagler, 2020). However, Benmelech and Tzur-Ilan (2020) show that debt-to-GDP is
positively related to fiscal spending during COVID-19. Indeed, we face a high correlation with more than 0.8
between the sovereign debt ratio and the size of the fiscal response to the crisis. The correlation further increases
to 0.9 when interacted with distancing. There are other likely control variables one could consider, such as share
of population aged 65+ and number of hospital beds. However, due to the problem of multicollinearity as indicated
by a large variance inflation factor (VIF), we have to be selective and exclude all these potential control variables
from the baseline model.

12
See, for instance, Tong and Wei (2008), who examine how the subprime crisis spilled over to the real economy
and find that these two channels indeed explain the negative impact on stock prices during the global financial
crisis.

22
sensitivity to demand shocks based on the stock price response to the September 11 shock, as
computed by Tong and Wei (2008). We examine the extent to which policy measures affect
firm performance through these two potential channels.

Table 4B presents the results. 13 With consistently high statistical significance at the 1
or 5% level, we continue to find that fiscal policy gives a boost to firm performance in sectors
hit hard by the pandemic. Contrary to previous studies (e.g. Aghion et al. 2009; Laeven and
Valencia 2013), we do not find much significant effect of fiscal stimulus on firms that are more
dependent on external finance and firms that are more demand sensitive. A plausible
interpretation of this result is that, in the context of the COVID-19 shock on firm performance
and survival, the main channel through which fiscal stimulus helped is the alleviation of the
impact of non-pharmaceutical interventions. The impact of monetary policy on sales growth of
pandemic-prone firms remains significant, though still marginally at the 10% level (see the
coefficient on 𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷 × 𝑀𝑀𝑀𝑀_𝐵𝐵𝐵𝐵 in column 1). There appears to be a similar link to profit
margins (see column 2). The impact of FXI on ∆(𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼) is still attributable to differences
across sectors in terms of how vulnerable they are to distancing rather than other sectoral
characteristics (column 3). Note that there seems to be a marginally significant negative impact
of FXI on the probability of default (column 4). On balance, the key findings for the interaction
terms between distancing and policy variables appear to be consistent with the baseline results
in Table 3.

In Table 4C, we further estimate the model by controlling for interactions of both
country and sector characteristics simultaneously. The results are close to those in the baseline
reported in Table 3 in terms of the sign, magnitude and significance of the coefficients (or even
better with more significant coefficients).

While the above control variables provide a reasonable amount of country- and sectoral
specific information, they may not entirely account for all relevant factors and thus, the
likelihood of some omitted variable bias continues to be present. In fact, while we control for
observable factors, our result may still be biased due to unobservable variables that may be
correlated with government stimulus packages and subsequently with firm performance.

13
Note that since 𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹 and 𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷 are at the SIC 3-digit level whilst 𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷 is at the NAICS 3-digit
level, NAICS 4-digit-level fixed effects are utilized in order to cover both SIC 3-digit levels and NAICS 3-digit
levels in Table 4B. In Appendix Table A5 Columns 5-8, we use 𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹 at the NAICS 3-digit level from Bilir et
al. (2019), and also impose sector fixed effects at the 3-digit level to be consistent with the specifications in Table
3. The main results are supportive to those in Table 4B.

23
Henceforth, following Altonji et al. (2005), as a second approach, we measure the relative
significance of omitted variable bias by testing to what extent the coefficients of interest are
altered by the inclusion of more regressors (unobservable factors). This approach quantifies
how much greater the influence of unobservable factors would be required to be, relative to
observables, to completely explain away the positive association between the economic
stimulus packages and the performance of firms that operate in pandemic-prone sectors. If such
influence is substantial, inclusion of more controls (i.e., unobservable factors) would reduce
the estimated effect even further. If it is trivial, we can be more assured in proposing a causal
interpretation to the estimated relationship. We utilize the method proposed by Oster (2019),
who argues that one should scale the coefficient movements by the observed increase in 𝑅𝑅 2 as
the measurement of the change. In this set-up, we need to have two types of regressions:
restricted regression (the one with a restricted set of control variables – those in baseline Table
3) and full regression (the one with a full set of country and sector controls – those in Table
4C).

Table 4D reports the coefficients of the interaction term between government stimulus
packages and distancing (those that are significant in Table 3), along with the associated R2
obtained by estimating Eq. (1) in a restricted version and in a full model. We find that the full
model increases the magnitude of the coefficient, while R2 increases from about 8% to 16%,
depending on the proxy used for stimulus policy. This result indicates that, holding other
factors constant, unobservable factors generally bias our coefficient toward zero (similar to the
case in Claessens et al. 2021). Therefore, the estimated effects are likely to be conservative,
resulting in negative figures for Oster delta as presented in columns 9-12, except for the one
case for ∆(𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼) that is just above the threshold unity.

Reverse causality
Besides omitted variables, reverse causality could be another challenge. Receiving government
economic support may be endogenous to a firm’s activities. Even if the COVID-19 shock is
exogenous, the reaction of policy makers may not be random (Demirgüç-Kunt et al. 2021). For
instance, companies, especially large ones, that were adversely affected by the COVID-19
pandemic and its associated lockdown measures could be more likely to be supported by the
government.

To deal with reverse causality concerns, we apply three different strategies. Firstly, we
drop the top 3 pandemic-prone industries in each country from our sample. The underlying

24
idea is that the most vulnerable sectors in a country may be the ones to influence government
policies. Secondly, we exclude large firms (firms with revenue greater than USD 5 billion)
from the dataset. If activities of firms may determine the degree of government intervention,
then this would be more likely to be the case with large influential firms. By contrast, one may
expect that smaller firms are more vulnerable to COVID-19 and, thus, may benefit more from
government policies, rather than the other way round. Finally, we remove countries with a high
share of pandemic-prone sectors to the GDP. This is because the reverse causality effect should
be in tandem with the size of vulnerable industries relative to the overall size of the economy
(Levintal, 2013). In other words, one would expect a larger reverse causality bias in countries
where the pandemic-prone sectors constitute a significant portion of GDP. We measure this
share as 𝑆𝑆ℎ𝑎𝑎𝑎𝑎𝑎𝑎𝑐𝑐 = ∑𝑛𝑛𝑗𝑗=1 𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝑗𝑗 × 𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑗𝑗 / 𝐺𝐺𝐺𝐺𝐺𝐺𝑐𝑐 where 𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉 is value added of sector
𝑗𝑗 computed as the sum of earnings before taxes, depreciation and labor expense (Laeven and
Valencia, 2013). 𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑗𝑗 / 𝐺𝐺𝐺𝐺𝐺𝐺𝑐𝑐 is measured for year 2019. We then remove countries in the
75th percentile of 𝑆𝑆ℎ𝑎𝑎𝑎𝑎𝑎𝑎.

Insert Table 5 around here

The results are reported in Table 5 and resonate with those in the baseline reported in
Table 3. Removing the top 3 pandemic-prone sectors (columns 1-4), largest firms (columns 5-
8) or countries with a high share of pandemic-prone sectors (columns 9-12) do not alter the
findings and, actually, in some cases deliver larger and more statistically significant
coefficients.

To address any remaining endogeneity issue, our final attempt is to utilize the shock
components for each policy as instruments. Specially, we follow Biljanovska et al (2021) to
construct the shocks for year 2020 as residuals of the following regressions:

• Fiscal policy: the shock for fiscal policy is the residuals obtained from regressing the
primary balance on its lag, output gap and a dummy for positive output gap.
• Monetary policy: the shock for monetary policy is the residuals obtained from
regressing the policy rate on real output gap, inflation, one quarter lagged policy rate
and the log difference of the real effective exchange rate (in the case of emerging
economies).
• FXI: the shock for foreign exchange intervention is the residuals obtained from
regressing the FXI (as calculated by Adler et al. 2021) on the change in the real effective
exchange rate, change in portfolio flows, inflation, change in credit-to-GDP, the VIX,

25
 change in commodity price index, interest rate differential to the US policy rate, foreign
reserves and a dummy variable for floating exchange rate regime.
The regressions are run by country for fiscal and monetary policies. For FXI, they are run
separately for emerging market economies and advanced economies. Additional information
is available in Biljanovska (2021). We instrument our variable of interest (that is,
𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝑗𝑗 × 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑐𝑐 ) with the estimated policy shocks interacted with distancing (that is,
𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝑗𝑗 × 𝑆𝑆ℎ𝑜𝑜𝑜𝑜𝑜𝑜𝑐𝑐 ). Due to data availability, we do not have estimates of the FXI shock
for Japan and some EU countries, although FXI shocks in these cases can arguably be assumed
to be zero. Thus, we choose to report the IV results for two cases: (i) FXI is not included and
(ii) FXI is included (and instrumented).

In addition to policy shocks, we also use a proxy for the quality of the institutional
environment to further orthogonalize the policy response. The structure and timely launch of
an economic stimulus package when a country faces an external shock could depend on its
overall institutional quality. Hence, we complement the shock data with data on institutional
quality retrieved from the World Bank’s World Governance Indicators Database, as of 2020,
the so-called KKZ institution. It measures different dimensions of governance, which include
government effectiveness, political stability, regulatory quality, rule of law, voice and
accountability, and control of corruption. Again, we interact this variable with the sectoral
distancing proxy.

Insert Table 6 around here

We report the IV estimates in Table 6, Panel A for fiscal and monetary stimuli and Panel B for
all three policies. The F-test of the excluded instruments rejects the null hypothesis of weak
instruments, and Hansen’s J-test does not reject the null hypothesis that the over-identifying
restrictions are valid. This indicates that our two instruments, policy shocks and institutional
quality, are valid. We find that fiscal stimulus relates positively and statistically significantly
to the performance of pandemic-prone sectors, with a magnitude which is even larger than that
reported in Table 3 for the OLS case. We find no impact on the profit margin when FXI is
included. Overall, the IV estimator confirms our baseline results.

To sum up, in this section, we have conducted a number of exercises to verify that the
association we unveiled in Table 3 between stimulus packages and firm performance during
COVID-19 can reasonably be considered to not suffer from omitted variable bias and reverse

26
causality. 14 Yet, the above strategies may not entirely address such problems. Thus, we conduct
several additional robustness tests in the next subsection.

C. Robustness Tests

Several robustness tests are conducted in order to ascertain the baseline results in Table 3. First,
an alternative proxy for pandemic sensitivity is examined in Table 7A. Recall that, in the
previous estimations, the proxy for Distancing is based on communication-intensive jobs. In
this table, we employ Kóren and Petö (2020)’s overall index that incorporates not only
communication intensity but also the need for physical presence. Note the interpretation of the
Kóren-Petö index is as follows: a higher share of jobs that cannot be done at home indicates
higher sensitivity to non-pharmaceutical interventions, which is captured by a higher level of
reliance on face-to-face communication and physical presence. The results appear to be
somewhat mixed, however, the effectiveness of fiscal policy remains as the main driver of
differences in firm performance. With the alternative measure of Distancing, profit margin and
interest coverage become more responsive to expansionary monetary policy as compared with
the baseline results.

Insert Tables 7A, 7B, 7C and 7D around here

Another concern could be the composition of our dataset. Since firms in our dataset
may be unequally distributed across countries, one issue could be that the results are driven by
the unbalanced nature of the dataset. To alleviate such a concern, we re-estimate our baseline
regressions in Table 7B using weighted least squares where observations are weighted by the
inverse of the number of firm observations in each country (following Laeven and Valencia,
2013), which allows the covariance matrix of errors to be different from an identity matrix (in
columns 1–4). In a related exercise, we remove ‘critical’ or ‘essential’ sectors in columns 5–8.
The concern in this case is that some sectors may be mandated to stay open even under full
lockdown and receive (direct) policy support in order to do so. Such sectors include: utilities,
infrastructure, food and agriculture, critical manufacturing, healthcare and public health
services, security and emergency services. We drop 36 three-digit industries as informed by
Papanikolaou and Schmidt (2020). In both cases, the main findings are qualitatively unaltered.
Fiscal stimulus, in general, help hardest-hit sectors weather the COVID-19 shock. The effect

14
Recall that we excluded the United States in conducting the empirical analysis due to the concern of reverse
causality. The estimates obtained when we include the US data are in Appendix Table A5 Columns 9-12. The
results are almost identical to those in Table 3.

27
of monetary policy to profit margin becomes more pronounced. The boost provided by foreign
exchange intervention to interest coverage remains robust. In fact, the coefficient on
𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷 × 𝐹𝐹𝐹𝐹𝐹𝐹 is highly significant in probability of default (column 8) by excluding 36
critical sectors, though it is still marginally significant with weighted least squares (column 4).
This hints that FXI could be more instrumental than revealed in the baseline results.

It is possible that omitted variable bias could stem not only from country and sector
characteristics (studied in Table 4), but also from differences at the firm level. Note that we
already control for a battery of firm characteristics in the baseline, Table 7C extends the set of
controls to include overhead costs and cash flow, both deflated by total assets. Tobin’s Q
(defined as the market value of common stocks plus the book value of total liabilities divided
by the book value of total assets) is also specified in order to control for the market value of
firms. Our sample includes both non-listed and listed firms, hence columns 1–4 are for all firms
without Tobin’s Q and columns 5–8 are for listed firms with Tobin’s Q. The resonance of the
baseline results in Table 3 is also apparent in Table 7C: fiscal policy is an effective tool to
mitigate the decline in overall firm performance, whereas monetary policy seems to support
revenue and foreign exchange intervention to improve interest coverage. An interesting
observation is that, either listed or unlisted, hardest-hit firms disproportionately gain from
stimulus packages implemented during the pandemic.

There are many databases that have been tracking what governments across the world
are doing in response to the COVID-19 pandemic. They vary in terms of coverage in terms of
country, frequency, and type of policy measures. There is often a trade-off between how many
countries such databases can cover and how granular they can go in categorizing policy
measures. While our dataset on stimulus packages from IMF policy tracker used in the baseline
analysis is comprehensive, we, as a final sensitivity test, also check the robustness of our
findings to two alternative sources: (i) ESCAP Policy Responses to COVID-19 in Asia and the
Pacific Database, (ii) Yale COVID-19 Financial Response Tracker (CFRT) by the Yale
Program on Financial Stability Database. 15 The advantage of these alternative sources over
some others is that the data are reported for most countries in our main dataset. Since our
analysis so far mainly produced robust results for fiscal stimulus policy, we focus only on this

15
We do not utilize other stimulus policy databases because either their coverage is limited to a specific region
and/or a small number of countries or they do not include sufficient quantitative information. For instance, the
OECD provides a database with qualitative information on COVID policy measures in a narrative format but
transforming this information into numerical values is not straightforward given lack of direct comparability.

28
policy dimension. The results reported in Table 7D with alternative stimulus policy data appear
to affirm, generally, the effectiveness of fiscal policies during the pandemic, notably firms in
pandemic-prone sectors performed better if they are located in countries that adopted larger
fiscal stimulus packages.

D. Additional Analyses

In this subsection, we investigate the relationship between pre-COVID firm size and basic
financial conditions and their performance during the crisis. More specifically, we test whether
pre-crisis firm characteristics influence the link between government policies and firm
performance during the COVID-19 crisis.

Insert Tables 8A, 8B, 8C and 8D around here

We focus on four characteristics commonly-studied in the literature (Giroud and

Mueller 2017). The first one is size, and the other three relate to liquidity constraints and
leverage. Because of the adverse impact of the COVID-19 crisis on revenues and free cash
flow, one may expect that smaller firms and firms with less cash, more leverage and less
profitability to be more vulnerable and, thus, more favorably affected by government economic
stimulus policies. In this vein, Ding et al. (2020) find that firms entering the COVID-19 crisis
with a better position in terms of cash holdings, leverage and profitability performed relatively
better during the crisis, with respect to stock prices. Fahlenbrach et al. (2020) find that financial
flexibility (proxied, for example, by cash holdings) is one of the factors explaining why some
firms performed better during the COVID-19 crisis. Laeven (2020) finds that large firms and
firms with cash buffers are better able to absorb the pandemic shock. Carletti et al. (2020) also
report that distress in terms of book value of equity is more frequent for small and medium-
sized enterprises and for firms with high pre-COVID-19 leverage, using a sample of 80,972
Italian firms. Although the nature of the initial shock is very different, these findings are in line
with those reported in studies looking at the 2008 financial crisis (e.g., Duchin et al. 2010 find
that the decline in investment is greatest for firms that have low cash reserves and are
financially more constrained). Thus, we re-estimate the baseline specifications in Table 3, by
considering the role of pre-crisis firm conditions with respect to size (Table 8A), cash holdings
(Table 8B), profitability (Table 8C) and leverage (Table 8D). Columns 1–4 in all four tables
include only the firms that are below the median values and columns 5–8 above the median
values, respectively.

29
 Larger pandemic-prone firms predominantly benefit more in all cases except for profit
margin from fiscal stimulus (see the significant coefficient on the interaction for 𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹 in
column 5, 7 and 8 of Table 8A). On the other hand, smaller firms benefit more from foreign
exchange intervention (see the significant coefficient on the interaction for FXI in columns 3
and 4). Expansionary monetary policy seems to exert a preferable effect regardless of the firm
size by easing the cost of borrowing which is raising profit in smaller firms (column 2), whereas
increasing sales in larger firms (column 5).

Firms with operations in a pandemic-prone industry and a low level of cash holdings
appear to have been protected from insolvency by the fiscal stimulus, given the significant
coefficient on the interaction for 𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹 in column 4 of Table 8B. The benefit from the fiscal
intervention tends to extend when it comes to meeting their debt obligations (column 3). For
those firms with more cash holdings, the result appears to echo that for larger firms in Table
8A: both fiscal and monetary stimuli are, in general, suitable to their needs with highly
significant coefficients found on the interaction terms for FisStim (column 5, 7 and 8) and
MP_BP (column 5).

By distinguishing between higher and lower profitability amongst vulnerable firms in

Table 8C, the effectiveness is not well determined with only 5 coefficients on the interaction
terms are significant. Yet, interestingly, firms in the higher profitability group seem to benefit
from fiscal and, to a lesser degree, from monetary stimuli more than their lower-profitability
counterparts do (see column 5 and 7, where the interaction terms are highly significant). A
somewhat speculative interpretation could be that fiscal and monetary policy stimuli provide a
lifeline to even to those that were not doing well before the pandemic hit, but only firms of a
certain level of profitability are then able to lever this lifeline to weather the blow to their
profitability.

Looking at the results in Table 8D, there is clear distinction between higher debt-
holding companies (columns 1-4) and lower debt-holding companies (columns 5-8).
Predominantly, economic policy packages favor the latter, in particular both fiscal and
monetary policy have shown to be operative to all types of firm performance, ranging from
sales revenue, profit margin, interest coverage and probability of default. On the other hand,
low leveraged firms appear to only receive the benefit of an improved interest coverage ratio
from foreign exchange intervention (column 3) and of a decline in probability of default from
fiscal stimulus (column 4). These results seem to support the argument in that firms with a

30
better financial position are more likely to take advantage of the stimulus packages to withstand
the pandemic shock.

V. CONCLUSION

In this paper, we use firm-level data to provide some early evidence on the effectiveness of
COVID-19 economic policy packages. Our empirical strategy relies on the varying degree of
vulnerability to the pandemic across industries. If policy actions have been targeted enough,
they would give a lift to pandemic-prone sectors.

After confirming that firms in sectors with higher distancing indices performed worse
than the others in the same country, we find a robust positive association of fiscal stimulus with
growth in the sales-to-assets ratio, profit margin, and interest coverage ratio, and negative
association with probability of default in pandemic-prone sectors. Put differently, firms that
are more sensitive to distancing have performed better when the fiscal stimulus is larger. There
is also some evidence that monetary stimulus has been associated with improved sales and
foreign exchange intervention with increased interest coverage ratio for the hardest-hit firms.
Overall though, the evidence indicates that fiscal stimulus packages are more effective than
other policies during the COVID-19 pandemic. Thus, this early evidence seems to suggest that
policy interventions have bought time for the hardest-hit industries, by supporting sales and
improving liquidity.

31
References

Aghion, P., Hémous, D., Kharroubi, E., 2014. Cyclical fiscal policy, credit constraints, and
industry growth, Journal of Monetary Economics 62, 41-58.

Alfaro, L., Chari, A., Greenland, A.N., Schott, P.K., 2020. Aggregate and Firm-Level Stock
Returns during Pandemics, in Real Time, NBER Working Paper No. 26950.

Altonji, J.G., Todd E. E., and Christopher R. T., 2005. Selection on Observed and Unobserved
Variables: Assessing the Effectiveness of Catholic Schools. Journal of Political
Economy, 113(1): 151–84.

Barbiero, F., Popov, A., Wolski, M., 2020. Debt overhang, global growth opportunities, and
investment, Journal of Banking & Finance 120.

Bates, T.W., Kahle, K.M., Stulz, R.M., 2009. Why Do U.S. Firms Hold So Much More Cash
than They Used to? Journal of Finance 64(5), 1985–2021.

Baumohl, E., Iwasaki, I., Kočenda, E., 2019. Institutions and determinants of firm survival in
European emerging markets. Journal of Corporate Finance 60, 431–453.

Beck, T., Demirgüç-Kunt, A., Laeven, L., Maksimovic, V., 2006. The determinants of
financing obstacles. Journal of International Money and Finance 25(6), 932–952.

Benmelech, E., and Tzur-Ilan, N., 2020. The determinants of fiscal and monetary policies
during the Covid-19 crisis, NBER Working Papers 27461.

Bilir, L.K., Chor, D., Manova, K., 2019. Host-country financial development and multinational
activity. European Economic Review 115, 192–220.

Biljanovska, N., Chen, S., Igan, D., Martinez Peria, M.S., 2021. Exploring the Interactions of
Macroprudential Regulation with Other Policies, forthcoming IMF Working Paper.

Carletti, E., Oliviero, T., Pagano, M., Pelizzon, L., Subrahmanyam, M.G., 2020. The COVID-
19 shock and equity shortfall: Firm-level evidence from Italy, Review of Corporate
Finance Studies 9, 534–568.

Cathcart, L., Dufour, A., Rossi, L., Varotto, S., 2020. The differential impact of leverage on
the default risk of small and large firms, Journal of Corporate Finance 60.

Chaney, T., Sraer, D., and Thesmar, D., 2012. The Collateral Channel: How Real Estate Shocks
Affect Corporate Investment, American Economic Review 102(6), 2381–2409.

Chen, S., Igan, D., Pierri, N., Presbitero, A.F., 2020. Tracking the Economic Impact of COVID-
19 and Mitigation Policies in Europe and the United States, CEPR Covid Economics
36.

Chodorow-Reich, G., 2014. The Employment Effects of Credit Market Disruptions: Firm-
Level Evidence from the 2008–09 Financial Crisis, Quarterly Journal of Economics
129(1), 1–59.

32
Claessens, S., Dell’Ariccia, G., Igan, D., Laeven, L., 2010. Cross-country experiences and
policy implications from the global financial crisis, Economic Policy 25(62), 267–
293.
Claessens, S., Tong, H., Wei, S. J., 2012. From the financial crisis to the real economy: Using
firm-level data to identify transmission channels. Journal of International Economics
88(2), 375–387.

Claessens, S., Cornelli, G., Gambacorta, L. et al., 2021. Do macroprudential policies affect
non-bank financial intermediation? BIS Working Papers 27.

Demirgüç-Kunt, A, Pedraza, A., Ruiz-Ortega, C., 2021. Banking sector performance during
the COVID-19 Crisis, Journal of Banking and Finance 133.

Demirgüç-Kunt, A., Peria, M.S.M., Tressel, T. 2020. The global financial crisis and the capital
structure of firms: Was the impact more severe among SMEs and nonlisted firms?
Journal of Corporate Finance 60.

Ding, W., Levine, R., Lin, C., 2020. Corporate immunity to the COVID-19 pandemic, NBER
Working Paper No. 27055.

Dingel, J.I., Neiman, B., 2020. How many jobs can be done at home? Journal of Public
Economics 189, 1–8.

D'Souza, J., Megginson, W.L., Ullah, B., Wei, Z., 2017. Growth and growth obstacles in
transition economies: Privatized versus de novo private firms, Journal of Corporate
Finance 42, 422–438.

Duan, J.C., Sun, J., Wang, T., 2012. Multiperiod corporate default prediction –a forward
intensity approach, Journal of Econometrics 170 (1), 191–209.

Duchin, R., Ozbas, O., Sensoy, B., 2010. Costly external finance, corporate investment, and
the subprime mortgage credit crisis, Journal of Financial Economics 97, 418–435.

Fahlenbrach, R., Rageth, K., Stulz, R.M., 2020. How valuable is financial flexibility when
revenue stops? Evidence from the COVID-19 crisis, NBER Working Papers 27106.

Frijns, B., Dodd, O., Cimerova, H., 2016. The impact of cultural diversity in corporate boards
on firm performance, Journal of Corporate Finance 41, 521–541.

Gaganis, C., Lozano-Vivas, A., Paradimitri, P., Pasiouras, F., 2020. Macroprudential policies,
corporate governance and bank risk: cross-country evidence. Journal of Economic
Behavior and Organization 169, 126–142.

Gagnon, J.E., Collins, C.G., 2019. Are Central Banks Out of Ammunition to Fight a Recession?
Not Quite, Peterson Institute for International Economics Policy Brief 19–18.

Gandhi, P., Lustig, H., 2015. Size anomalies in U.S. bank stock returns. Journal of Finance 70,
733–768.

33
Giroud, X., Mueller, H.M., 2017. Firm leverage, Consumer Demand, and Employment Losses
during the Great Recession, Quarterly Journal of Economics 132, 271–316.

Giroud, X., Mueller, H.M., 2019. Firms’ Internal Networks and Local Economic Shocks,
American Economic Review 109(10), 3617–49.

Glover, A., Heathcote, J., Krueger, D., Ríos-Rull, J.V., 2020. Health versus Wealth: On the
Distributional Effects of Controlling a Pandemic, mimeo, University of Pennsylvania.

González, V.M., 2015. The financial crisis and corporate debt maturity: the role of banking
structure, Journal of Corporate Finance 35, 310–328.

Hale, T., Webster, S., Petherick, A., Phillips, T., Kira, B., 2020. Oxford Covid-19 government
response tracker, Blavatnik School of Government.

Hall, T., Mateus, C., Mateus, I.B., 2014. What determines cash holdings at privately held and
publicly traded firms? Evidence from 20 emerging markets, International Review of
Financial Analysis 33(3), 104–116.

Hensvik, L., Le Barbanchon, T., Rathelot, R., 2020. Which jobs are done from home? Evidence
from the American Time Use Survey, CEPR Discussion Paper No. 14611.

Hyytinen, A., Toivanen, O., 2005. Do financial constraints hold back innovation and growth?
Evidence on the role of public policy. Research Policy 34 (9), 1385–1403.

Igan, D., Mirzaei, A., 2020. Does going tough on banks make the going get tough? Bank
liquidity regulations, capital requirements, and sectoral activity. Journal of Economic
Behavior & Organization 177, 688–726.

Igan, D., Mirzaei, A., Moore, T., 2022. Does macroprudential policy alleviate the adverse
impact of COVID-19 on the resilience of banks? Journal of Banking and Finance, (in
press).

Kóren, M., Petö, R., 2020. Business disruptions from social distancing, PLoS ONE 15(9):
e0239113.

Laeven, L., 2020. Pandemics, Intermediate Goods, and Corporate Valuation. CEPR Discussion
Paper No. 15022.

Laeven, R., Valencia, F., 2013. The real effects of financial sector interventions during crises,
Journal of Money, Credit and Banking 45 (1), 147–177.
Leibovici, F., Santacreu, A.M., Famiglietti, M., 2020. How the Impact of Social Distancing
Ripples through the Economy, St. Louis Fed On the Economy Blog, April 7.

Levintal, O., 2013. The real effects of banking shocks: evidence from OECD countries. Journal
of International Money and Finance 32, 556–578.

Makin, A.J., Layton, A., 2021. The global fiscal response to COVID-19: Risks and
repercussions, Economic Analysis and Policy 69, 340-349.

34
Mankiw, G., 2020. A Proposal for Social Insurance During the Pandemic,
http://gregmankiw.blogspot.com/2020/03/a-proposal-for-social-insurance-during.html

Marron, D., 2020. If We Give Everybody Cash To Boost The Coronavirus Economy, Let’s
Tax It | Tax Policy Center, https://www.taxpolicycenter.org/taxvox/if-we-give-
everybody-cash-boostcoronavirus-economy-lets-tax-it

Martin, T., Nagler, F., 2020. Sovereign Debt and Equity Returns in the Face of Disaster
(September 15, 2020). Available at
SSRN: https://ssrn.com/abstract=3572839 or http://dx.doi.org/10.2139/ssrn.3572839

Norden, L., Roosenboom, P., Wang, T., 2013. The impact of government intervention in banks
on corporate borrowers’ stock returns, Journal of Financial and Quantitative Analysis
48 (5), 1635–1662.
OECD, 2020a. Tax and Fiscal Policy in Response to the Coronavirus Crisis: Strengthening
Confidence and Resilience, https://www.oecd.org/coronavirus/policy-responses/tax-
and-fiscal-policy-in-response-to-the-coronavirus-crisis-strengthening-confidence-and-
resilience-60f640a8/.

OECD, 2020b. Supporting people and companies to deal with the COVID-19 virus: Options
for an immediate employment and social-policy response,
https://read.oecdilibrary.org/view/?ref=119_119686-
962r78x4do&title=Supporting_people_and_companies_to_deal_with_the_Covid-
19_virus.

Oster, E., 2019. Unobservable Selection and Coefficient Stability: Theory and Evidence,
Journal of Business & Economic Statistics 37(2), 187–204.

Padhan, R., Prabheesh, K.P., 2021. The economics of COVID-19 pandemic: A survey,
Economic Analysis and Policy 70, 220-237.

Pagano, M., Wagner, C., Zechner, J., 2020. Disaster Resilience and Asset Prices, CEPR
Discussion Paper No. 14773.

Papanikolaou, D., Schmidt, L.D.W., 2020. Working Remotely and the Supply-side Impact of
COVID-19, NBER Working Paper No. 27330.

Rajan, R., Zingales, L., 1998. Financial dependence and growth. American Economic Review
88(3), 559–586.

Ramelli, S., Wagner A.F., 2020. Feverish Stock Price Reactions to COVID-19, CEPR
Discussion Paper No. 14511.

Tong, H., Wei, S. J., 2008. Real effects of the subprime mortgage crisis: Is it a demand or a
finance shock? IMF Working Paper No. 08/186.

Van den Heuvel, S., 2009. The bank capital channel of monetary policy, mimeo, Federal
Reserve Board.

35
Table 1. Summary statistics

Panel A: Change in firm performance and distancing by sector

Change in firm performance
NAICS NAICS Sub
(2d) (3d) sector Sector Obs ∆(SaleA) ∆(ProfM) ∆(IntrC) ∆(ProbD) Distancing

11 113-115 3 Agriculture, Forestry, Fishing and Hunting 178 -0.042 -0.020 -3.054 -0.011 0.147
21 211-213 3 Mining, Quarrying, and Oil and Gas Extraction 1012 -0.060 -0.030 -2.146 0.004 0.196
22 221 1 Utilities 1013 -0.044 -0.003 1.548 0.002 0.200
23 236-238 3 Construction 1022 -0.065 -0.021 -1.966 0.002 0.164
31-33 311-339 20 Manufacturing 14504 -0.073 -0.006 1.188 0.002 0.100
42 423-425 3 Wholesale Trade 2039 -0.108 -0.009 -3.900 0.002 0.154
44-45 441-454 12 Retail Trade 978 -0.152 -0.020 1.016 0.007 0.642
48-49 481-493 9 Transportation and Warehousing 1038 -0.106 -0.050 1.427 0.003 0.134
51 511-519 6 Information 1683 -0.071 -0.022 -3.334 0.002 0.146
53 531-533 3 Real Estate and Rental and Leasing 1293 -0.029 -0.088 -0.952 0.005 0.216
54 541 1 Professional, Scientific, and Technical Services 2016 -0.091 -0.009 -2.038 -0.0001 0.120
56 561-562 2 Administrative and Support and Waste Management … 901 -0.150 -0.042 -6.801 0.008 0.264
61 611 1 Educational Services 126 -0.149 -0.056 -7.885 0.0004 0.190
62 621-624 4 Health Care and Social Assistance 298 -0.084 -0.018 -3.995 0.005 0.596
71 711-713 3 Arts, Entertainment, and Recreation 234 -0.200 -0.179 -14.364 0.003 0.405
72 721-722 2 Accommodation and Food Services 458 -0.239 -0.238 -17.783 0.011 0.440
81 811-813 3 Other Services (except Public Administration) 122 -0.119 -0.064 -11.221 0.005 0.351

36
Panel B: Summary statistics of main variables
Variable Obs Mean Std p25 Median p75 Min Max

Change in firm performance (∆y ic,COVID )

∆(SaleA) 28915 -0.08 0.25 -0.16 -0.05 0.02 -1.15 0.68
sectors more pandemic prone 13694 -0.10 0.27 -0.17 -0.05 0.01 -1.15 0.68
sectors less pandemic prone 15221 -0.07 0.22 -0.15 -0.05 0.03 -1.15 0.68

∆(ProfM) 26993 -0.02 0.17 -0.05 -0.001 0.03 -0.76 0.56

sectors more pandemic prone 12484 -0.04 0.20 -0.07 -0.01 0.03 -0.76 0.56
sectors less pandemic prone 14509 -0.01 0.15 -0.04 -0.01 0.04 -0.76 0.56

∆(IntrC) 27845 -0.73 63.94 -4.55 0.02 4.62 -327.34 292.72

sectors more pandemic prone 13099 -3.29 61.04 -5.59 -0.33 3.17 -327.34 292.72
sectors less pandemic prone 14746 1.54 66.33 -3.76 0.39 5.92 -327.34 292.72

∆(ProbD) 10023 0.002 0.04 -0.002 0.001 0.01 -0.16 0.18

sectors more pandemic prone 4748 0.004 0.03 -0.001 0.001 0.01 -0.16 0.18
sectors less pandemic prone 5275 0.001 0.04 -0.003 0.0003 0.01 -0.16 0.18

Pandemic-prone j
Distancing 79 0.16 0.13 0.09 0.11 0.16 0.04 0.9

Policy c
FisStim (% of GDP) 80 11.62 10.29 6.1 6.1 14.6 0 40.9
MP_BP (-1*basis pint/100) 80 0.71 1.09 0.15 0.3 1.15 -3 10
FXI 80 0.24 0.43 0 0 0 0 1

Controls (X ic,Pre )
Size (log) 28915 11.64 2.25 9.98 11.5 13.13 1.66 20.12
Age (log) 28915 3.17 0.77 2.71 3.09 3.66 0 5.9
CashA 28915 0.13 0.14 0.03 0.09 0.18 0 1
RD_A 28915 0.02 0.04 0 0 0.02 -0.18 1.27
Private (dummy) 28915 0.05 0.22 0 0 0 0 1
SaleA 28915 0.87 0.85 0.41 0.71 1.09 -0.62 30.13
ROA (%) 28915 2.70 13.62 0.31 3.84 8.31 -99.64 97.34
IntrC 28915 38.09 117.42 0.81 4.49 19.89 -99.83 999.84
EqitA 28915 0.49 0.37 0.35 0.52 0.68 -27.29 1

37
Table 2. Social distancing and firm performance during COVID-19
This table reports the results estimating ∆𝑦𝑦𝑖𝑖𝑖𝑖,𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 = 𝜗𝜗𝑐𝑐 + ∅. 𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝑗𝑗 + 𝜏𝜏. 𝑋𝑋𝑖𝑖𝑖𝑖,𝑃𝑃𝑃𝑃𝑃𝑃 + 𝜀𝜀𝑖𝑖𝑖𝑖,𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 and ∆𝑦𝑦𝑖𝑖𝑖𝑖,𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 = 𝜗𝜗𝑗𝑗 + 𝜗𝜗𝑐𝑐 + ∅. 𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝑗𝑗 × 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶_𝑆𝑆𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑐𝑐 + 𝜏𝜏. 𝑋𝑋𝑖𝑖𝑖𝑖,𝑃𝑃𝑃𝑃𝑃𝑃 + 𝜀𝜀𝑖𝑖𝑖𝑖,𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 where
𝑖𝑖 stands for firm, 𝑗𝑗 for sector, and 𝑐𝑐 for country. ∆𝑦𝑦𝑖𝑖𝑖𝑖,𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 is the change in performance ratios for firm 𝑖𝑖 in country 𝑐𝑐 between 2020 and 2019. We use, alternatively, change in asset turnover ratio
[∆(𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆)], change in profit margin [∆(𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃)], change in interest coverage ratio [∆(𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼)], and change in probability of default [∆(𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃)]. 𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝑗𝑗 is industry j’s degree of sensitivity
to a pandemic from Kóren and Petö (2020). 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶_𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑐𝑐 is a proxy for severity of COVID-19 in country c, using the Oxford stringency index. 𝑋𝑋𝑖𝑖𝑖𝑖,𝑃𝑃𝑃𝑃𝑃𝑃 is a vector of firm-level explanatory
variables, computed as of 2019. We include sector fixed effects (𝜗𝜗𝑗𝑗 ) in Columns 5-8 at the three-digit NAICS level as well as country fixed effects (𝜗𝜗𝑐𝑐 ) in all regressions. See Appendix, Table A1
for detailed definition of variables. Regressions are estimated using OLS. The statistical inferences are based on clustered standard errors at the country level (associated t-values reported in
parentheses). ***, **, and * denote statistical significance at the 1%, 5%, and 10% levels, respectively.

Without interaction Interacted with COVID-19 severity

∆(SaleA) ∆(ProfM) ∆(IntrC) ∆(ProbD) ∆(SaleA) ∆(ProfM) ∆(IntrC) ∆(ProbD)

(1) (2) (3) (4) (5) (6) (7) (8)

Distancingj -0.122*** -0.124*** -1.190*** 0.014*

(-6.499) (-13.098) (-8.142) (1.900)
Distancingj x Covid_Severity c -0.002* -0.002*** -0.015 0.001***
(-1.771) (-2.901) (-1.427) (3.334)

Controls ic,pre
(Size, Age, CashA, RD_A, Private,
√ √ √ √ √ √ √ √
SaleA, ROA, IntrC, EqitA)

Constant 0.019 -0.039*** 0.365** -0.016* 0.011 -0.061 -0.293 -0.025***

(0.814) (-3.009) (2.561) (-1.871) (0.470) (-1.327) (-0.681) (-2.744)

Sector FEs N N N N Y Y Y Y
Country FEs Y Y Y Y Y Y Y Y

# Countries 80 80 80 80 80 80 80 80
# Sectors 79 79 79 79 79 79 79 79
N 28,915 26,993 27,841 10,023 28,860 26,941 27,791 10,023
Adj. R 2 0.154 0.096 0.080 0.024 0.175 0.142 0.093 0.035

38
Table 3. Social distancing and firm performance during COVID-19: Baseline results
This table reports the results estimating ∆𝑦𝑦𝑖𝑖𝑖𝑖,𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 = 𝜗𝜗𝑗𝑗 + 𝜗𝜗𝑐𝑐 + ∅. 𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝑗𝑗 × 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑐𝑐 + 𝜏𝜏. 𝑋𝑋𝑖𝑖𝑖𝑖,𝑃𝑃𝑃𝑃𝑃𝑃 + 𝜀𝜀𝑖𝑖𝑖𝑖,𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 where 𝑖𝑖
stands for firm, 𝑗𝑗 for sector, and 𝑐𝑐 for country. ∆𝑦𝑦𝑖𝑖𝑖𝑖,𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 is the change in performance ratios for firm 𝑖𝑖 in country 𝑐𝑐 between
2020 and 2019. We use, alternatively, change in asset turnover ratio [∆(𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆)], change in profit margin [∆(𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃)], change
in interest coverage ratio [∆(𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼)], and change in probability of default [∆(𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃)]. 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑐𝑐 is a vector of variables
represent government stimulus packages in country c. 𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝑗𝑗 is industry j’s degree of sensitivity to a pandemic from
Kóren and Petö (2020). 𝑋𝑋𝑖𝑖𝑖𝑖,𝑃𝑃𝑃𝑃𝑃𝑃 is a vector of firm-level explanatory variables, computed as of 2019. We include sector fixed
effects (𝜗𝜗𝑗𝑗 ) at the three-digit NAICS level as well as country fixed effects (𝜗𝜗𝑐𝑐 ) in all regressions. See Appendix, Table A1 for
detailed definition of variables. Regressions are estimated using OLS. The statistical inferences are based on clustered standard
errors at the country level (associated t-values reported in parentheses). ***, **, and * denote statistical significance at the 1%,
5%, and 10% levels, respectively.

∆(SaleA) ∆(ProfM) ∆(IntrC) ∆(ProbD)

(1) (2) (3) (4)

Distancingj x FisStimc 0.003** 0.001** 0.030*** -0.001**

(2.369) (2.321) (2.795) (-2.500)
Distancingj x MP_BPc 0.031* 0.023 0.169 -0.004
(1.922) (1.649) (1.186) (-0.813)
Distancingj x FXIc -0.011 0.026 0.736** -0.010
(-0.255) (0.978) (2.043) (-1.518)
Controls ic,pre
(Size, Age, CashA, RD_A, Private,
√ √ √ √
SaleA, ROA, IntrC, EqitA)

Constant 0.008 -0.067 -0.443 -0.022**

(0.303) (-1.446) (-0.967) (-2.493)

Sector FEs Y Y Y Y
Country FEs Y Y Y Y

# Countries 80 80 80 80
# Sectors 79 79 79 79
N 28,915 26,993 27,841 10,023
Adj. R 2 0.175 0.142 0.093 0.035

39
Table 4. Addressing omitted variable bias
This table reports the results estimating ∆𝑦𝑦𝑖𝑖𝑖𝑖,𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 = 𝜗𝜗𝑗𝑗 + 𝜗𝜗𝑐𝑐 + ∅. 𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝑗𝑗 × 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑐𝑐 + 𝜏𝜏. 𝑋𝑋𝑖𝑖𝑖𝑖,𝑃𝑃𝑃𝑃𝑃𝑃 + ∇. 𝑍𝑍𝑖𝑖𝑖𝑖𝑖𝑖,𝑃𝑃𝑃𝑃𝑃𝑃 +
𝜀𝜀𝑖𝑖𝑖𝑖,𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 where 𝑖𝑖 stands for firm, 𝑗𝑗 for sector, and 𝑐𝑐 for country. ∆𝑦𝑦𝑖𝑖𝑖𝑖,𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 is the change in performance ratios for firm 𝑖𝑖 in
country 𝑐𝑐 between 2020 and 2019. We use, alternatively, change in asset turnover ratio [∆(𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆)], change in profit margin
[∆(𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃)], change in interest coverage ratio [∆(𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼)], and change in probability of default [∆(𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃)]. 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑐𝑐 is a
vector of variables represent government stimulus packages in country c. 𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝑗𝑗 is industry j’s degree of sensitivity to
a pandemic from Kóren and Petö (2020). 𝑋𝑋𝑖𝑖𝑖𝑖,𝑃𝑃𝑃𝑃𝑃𝑃 is a vector of firm-level explanatory variables, computed as of 2019. 𝑍𝑍𝑖𝑖𝑖𝑖𝑖𝑖,𝑃𝑃𝑃𝑃𝑃𝑃
is a vector of country-specific or sector-specific (interacted with Distancing or policy variables) new control variables. We
include sector fixed effects (𝜗𝜗𝑗𝑗 ) at the three-digit NAICS level (or at the four-digit in Panel B) as well as country fixed effects
(𝜗𝜗𝑐𝑐 ) in all regressions. See Appendix, Table A1 for detailed definition of variables. Regressions are estimated using OLS. The
statistical inferences are based on clustered standard errors at the country level (associated t-values reported in parentheses).
***, **, and * denote statistical significance at the 1%, 5%, and 10% levels, respectively.

4A: Controlling for other (pre-crisis) country characteristics

∆(SaleA) ∆(ProfM) ∆(IntrC) ∆(ProbD)

(1) (2) (3) (4)

Distancingj x FisStimc 0.004** 0.002** 0.032*** -0.001**

(2.128) (2.176) (3.642) (-2.182)
Distancingj x MP_BPc 0.037*** 0.019 0.182 -0.008**
(2.753) (1.245) (1.424) (-2.170)
Distancingj x FXIc -0.023 0.025 0.639* -0.009
(-0.484) (0.797) (1.723) (-1.003)
Other country characteristics
Distancingj x HealEc 0.018 0.003 0.256 -0.007
(0.612) (0.106) (1.028) (-1.279)
Distancingj x FDIc -0.003** -0.004 -0.037* -0.000
(-2.458) (-1.471) (-1.980) (-0.768)
Distancingj x Tradec 0.001** 0.000 0.004** -0.000
(2.050) (1.359) (2.074) (-0.669)
Distancingj x NPLc 0.003 -0.001 0.001 -0.003*
(1.243) (-0.324) (0.075) (-1.851)
Distancingj x Inflationc 0.005 0.004 0.088 0.006**
(1.225) (1.176) (1.002) (2.545)

Controls ic,pre
(Size, Age, CashA, RD_A, Private,
√ √ √ √
SaleA, ROA, IntrC, EqitA)

Constant -0.050 -0.071 -1.090** -0.020**

(-1.610) (-1.525) (-2.387) (-2.172)

Sector FEs Y Y Y Y
Country FEs Y Y Y Y

# Countries 80 80 80 80
# Sectors 79 79 79 79
N 28,502 26,603 27,452 10,020
Adj. R 2 0.176 0.141 0.093 0.035

40
4B: Controlling for other sector characteristics (other possible channels)

∆(SaleA) ∆(ProfM) ∆(IntrC) ∆(ProbD)

(1) (2) (3) (4)

Distancingj x FisStimc 0.003*** 0.002*** 0.029** -0.001**

(3.827) (2.902) (2.177) (-2.517)
Distancingj x MP_BPc 0.033* 0.024* 0.232 -0.005
(1.947) (1.849) (1.658) (-0.987)
Distancingj x FXIc 0.002 0.034 0.791** -0.013*
(0.049) (1.295) (2.147) (-1.717)
Other sector characteristics
FinDepj x FisStimc -0.000 -0.000** -0.001* -0.000
(-0.381) (-2.570) (-1.683) (-0.870)
FinDepj x MP_BPc 0.002 -0.000 -0.002 -0.000
(0.914) (-0.522) (-0.180) (-0.834)
FinDepj x FXIc 0.002 0.001 0.043 -0.000
(0.580) (0.348) (1.507) (-0.100)
DemSenj x FisStimc -0.001 0.002 -0.022* 0.000
(-0.774) (1.516) (-1.786) (0.720)
DemSenj x MP_BPc -0.022 0.008 -0.229 0.002
(-1.042) (0.421) (-1.260) (0.341)
DemSenj x FXIc -0.097** -0.012 -0.114 0.005
(-2.545) (-0.298) (-0.247) (0.442)

Controls ic,pre
(Size, Age, CashA, RD_A, Private,
√ √ √ √
SaleA, ROA, IntrC, EqitA)

Constant -0.060* -0.183*** -1.065*** 0.019

(-1.767) (-6.585) (-4.419) (1.031)

Sector FEs (4-digit level) Y Y Y Y

Country FEs Y Y Y Y

# Countries 80 80 80 80
# Sectors 79 79 79 79
N 25,954 24,237 24,999 9,042
Adj. R 2 0.174 0.156 0.099 0.032

41
4C: Controlling for other country and sector characteristics

∆(SaleA) ∆(ProfM) ∆(IntrC) ∆(ProbD)

(1) (2) (3) (4)

Distancingj x FisStimc 0.004*** 0.002** 0.023** -0.001**

(3.116) (2.144) (2.026) (-2.485)
Distancingj x MP_BPc 0.035** 0.020 0.246* -0.013***
(2.403) (1.288) (1.670) (-3.078)
Distancingj x FXIc 0.001 0.042 0.894** -0.013
(0.015) (1.337) (2.157) (-1.317)
Other country characteristics
Distancingj x HealEc -0.000 -0.000** -0.001 -0.000
(-0.445) (-2.582) (-1.523) (-0.881)
Distancingj x FDIc 0.002 -0.001 -0.001 -0.000
(0.901) (-0.982) (-0.098) (-0.816)
Distancingj x Tradec 0.002 0.001 0.032 -0.000
(0.658) (0.320) (1.074) (-0.132)
Distancingj x NPLc -0.001 0.002 -0.023* 0.000
(-0.784) (1.576) (-1.850) (0.703)
Distancingj x Inflationc -0.014 0.008 -0.166 0.002
(-0.619) (0.353) (-0.834) (0.306)
Other sector characteristics -0.098** -0.028 -0.138 0.004
FinDepj x FisStimc (-2.613) (-0.686) (-0.291) (0.369)

FinDepj x MP_BPc 0.022 0.014 0.435 -0.009

(0.668) (0.511) (1.641) (-1.660)
FinDepj x FXIc -0.004*** -0.004 -0.042 -0.000
(-2.844) (-1.151) (-1.494) (-0.577)
DemSenj x FisStimc 0.001* 0.000 0.003 -0.000
(1.815) (0.982) (1.228) (-0.420)
DemSenj x MP_BPc 0.004 -0.002 -0.008 -0.003
(1.500) (-0.644) (-0.408) (-1.474)
DemSenj x FXIc 0.001 0.003 0.027 0.008***
(0.291) (0.822) (0.322) (3.045)

Controls ic,pre
(Size, Age, CashA, RD_A, Private,
√ √ √ √
SaleA, ROA, IntrC, EqitA)

Constant -0.145** -0.199*** -2.160*** 0.028

(-2.541) (-4.552) (-4.890) (1.482)

Sector FEs (4-digit level) Y Y Y Y

Country FEs Y Y Y Y

# Countries 80 80 80 80
# Sectors 79 79 79 79
N 25,581 23,884 24,648 9,039
Adj. R 2 0.174 0.155 0.098 0.033

42
4D: Coefficient stability - test for omitted variable bias.
This table reports the results of the coefficient stability test of Oster (2019). ∅ is the coefficient of the policy variable, the one that is statistically significant in Table 3, along
with the associated R-squared, obtained by estimating Eq. (1) in a restricted version (omitting all country-level and industry-level control variables) and in a full model (as
presented in Table 4C). The Oster Delta statistic represents the degree of selection on unobserved variables relative to that on observed variables, where we set 𝑅𝑅𝑚𝑚𝑚𝑚𝑚𝑚 = 1.3 ∗
𝑅𝑅𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓 . Note that 𝑅𝑅𝑚𝑚𝑚𝑚𝑚𝑚 is described as the R-squared for a speculative regression that contains unobserved confounders.

from restricted model (Table 3) from full model (Table 4C) Oster Delta

∆(SaleA) ∆(ProfM) ∆(IntrC) ∆(ProbD) ∆(SaleA) ∆(ProfM) ∆(IntrC) ∆(ProbD) ∆(SaleA) ∆(ProfM) ∆(IntrC) ∆(ProbD)

(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)

Distancingj x FisStimc 0.0034 0.0014 0.0296 -0.0008 0.0035 0.0017 0.0227 -0.0006 -2.54 -2.16 1.01 -6.88

Distancingj x MP_BPc 0.0314 0.0348 -0.67

Distancingj x FXIc 0.7362 0.8935 -1.74

R2 0.180 0.147 0.099 0.047 0.184 0.166 0.109 0.031

43
Table 5. Addressing reverse causality
This table reports the results estimating ∆𝑦𝑦𝑖𝑖𝑖𝑖,𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 = 𝜗𝜗𝑗𝑗 + 𝜗𝜗𝑐𝑐 + ∅. 𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝑗𝑗 × 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑐𝑐 + 𝜏𝜏. 𝑋𝑋𝑖𝑖𝑖𝑖,𝑃𝑃𝑃𝑃𝑃𝑃 + 𝜀𝜀𝑖𝑖𝑖𝑖,𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 where 𝑖𝑖 stands for firm, 𝑗𝑗 for sector, and 𝑐𝑐 for country. ∆𝑦𝑦𝑖𝑖𝑖𝑖,𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 is the change
in performance ratios for firm 𝑖𝑖 in country 𝑐𝑐 between 2020 and 2019. We use, alternatively, change in asset turnover ratio [∆(𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆)], change in profit margin [∆(𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃)], change in interest
coverage ratio [∆(𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼)], and change in probability of default [∆(𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃)]. 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑐𝑐 is a vector of variables represent government stimulus packages in country c. 𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝑗𝑗 is industry j’s
degree of sensitivity to a pandemic from Kóren and Petö (2020). 𝑋𝑋𝑖𝑖𝑖𝑖,𝑃𝑃𝑃𝑃𝑃𝑃 is a vector of firm-level explanatory variables, computed as of 2019. We include sector fixed effects (𝜗𝜗𝑗𝑗 ) at the three-digit
NAICS level as well as country fixed effects (𝜗𝜗𝑐𝑐 ) in all regressions. See Appendix, Table A1 for detailed definition of variables. Regressions are estimated using OLS. The statistical inferences
are based on clustered standard errors at the country level (associated t-values reported in parentheses). ***, **, and * denote statistical significance at the 1%, 5%, and 10% levels, respectively.

Removing countries with a high share of

Removing top 3 pandemic-prone sectors Removing large firms
pandemic-prone sectors to GDP

∆(SaleA) ∆(ProfM) ∆(IntrC) ∆(ProbD) ∆(SaleA) ∆(ProfM) ∆(IntrC) ∆(ProbD) ∆(SaleA) ∆(ProfM) ∆(IntrC) ∆(ProbD)

(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)

Distancingj x FisStimc 0.003* 0.002** 0.027*** -0.001** 0.003** 0.001** 0.031*** -0.001** 0.004** 0.001** 0.035*** -0.001***
(1.794) (2.561) (2.692) (-2.517) (2.400) (2.318) (2.978) (-2.513) (2.571) (2.025) (4.473) (-5.276)
Distancingj x MP_BPc 0.033* 0.028* 0.193 -0.001 0.032* 0.023 0.171 -0.004 0.035* 0.029** 0.122 -0.007
(1.761) (1.981) (1.391) (-0.279) (1.986) (1.655) (1.200) (-0.826) (1.810) (2.064) (0.864) (-1.098)
Distancingj x FXIc -0.025 0.028 0.794** -0.016** -0.010 0.027 0.749** -0.010 -0.001 0.019 1.302*** -0.015*
(-0.572) (1.000) (2.010) (-2.100) (-0.243) (0.995) (2.076) (-1.549) (-0.014) (0.636) (4.523) (-1.786)

Controls ic,pre
(Size, Age, CashA, RD_A, Private,
√ √ √ √ √ √ √ √ √ √ √ √
SaleA, ROA, IntrC, EqitA)

Constant 0.011 -0.066 -0.451 -0.022** 0.006 -0.067 -0.434 -0.022** 0.021 -0.067 -0.629 -0.004
(0.391) (-1.415) (-0.989) (-2.570) (0.214) (-1.442) (-0.950) (-2.572) (0.719) (-1.428) (-1.398) (-0.445)

Sector FEs Y Y Y Y Y Y Y Y Y Y Y Y
Country FEs Y Y Y Y Y Y Y Y Y Y Y Y

# Countries 80 80 80 80 80 80 80 80 63 63 63 63
# Sectors 76 76 76 76 79 79 79 79 79 79 79 79
N 28,775 26,858 27,709 9,950 28,825 26,904 27,751 9,980 24,246 22,867 23,308 7,632
Adj. R 2 0.173 0.142 0.093 0.035 0.175 0.142 0.093 0.035 0.169 0.135 0.088 0.039

44
Table 6. IV strategy
This table reports the results estimating ∆𝑦𝑦𝑖𝑖𝑖𝑖,𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 = 𝜗𝜗𝑗𝑗 + 𝜗𝜗𝑐𝑐 + ∅. 𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝑗𝑗 × 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑐𝑐 + 𝜏𝜏. 𝑋𝑋𝑖𝑖𝑖𝑖,𝑃𝑃𝑃𝑃𝑃𝑃 + 𝜀𝜀𝑖𝑖𝑖𝑖,𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 where 𝑖𝑖 stands for firm, 𝑗𝑗 for sector, and 𝑐𝑐 for country. ∆𝑦𝑦𝑖𝑖𝑖𝑖,𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 is the change
in performance ratios for firm 𝑖𝑖 in country 𝑐𝑐 between 2020 and 2019. We use, alternatively, change in asset turnover ratio [∆(𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆)], change in profit margin [∆(𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃)], change in interest
coverage ratio [∆(𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼)], and change in probability of default [∆𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃)]. 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑐𝑐 is a vector of variables represent government stimulus packages in country c. 𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝑗𝑗 is industry j’s
degree of sensitivity to a pandemic from Kóren and Petö (2020). 𝑋𝑋𝑖𝑖𝑖𝑖,𝑃𝑃𝑃𝑃𝑃𝑃 is a vector of firm-level explanatory variables, computed as of 2019. We include sector fixed effects (𝜗𝜗𝑗𝑗 ) at the three-digit
NAICS level as well as country fixed effects (𝜗𝜗𝑐𝑐 ) in all regressions. See Appendix, Table A1 for detailed definition of variables. Regressions are estimated using IV approach where
instrument variables are the policy shocks and a proxy for institutional quality. The statistical inferences are based on clustered standard errors at the country level (associated t-values
reported in parentheses). ***, **, and * denote statistical significance at the 1%, 5%, and 10% levels, respectively.

∆(SaleA) ∆(ProfM) ∆(IntrC) ∆(ProbD) ∆(SaleA) ∆(ProfM) ∆(IntrC) ∆(ProbD)

(1) (2) (3) (4) (5) (6) (7) (8)

Distancingj x FisStimc 0.005*** 0.001* 0.040** -0.001*** 0.005*** 0.001 0.040** -0.001***
(3.351) (1.647) (2.278) (-5.002) (3.215) (0.776) (2.235) (-4.358)
Distancingj x MP_BPc 0.058** 0.014 0.210 -0.024* 0.070 0.051 0.379 -0.026*
(2.443) (0.604) (0.704) (-1.737) (1.313) (1.215) (0.891) (-1.807)
Distancingj x FXIc -0.034 -0.096 -0.368 0.008
(-0.265) (-0.952) (-0.481) (0.519)
Controls ic,pre
(Size, Age, CashA, RD_A, Private,
√ √ √ √ √ √ √ √
SaleA, ROA, IntrC, EqitA)

Constant 0.019 -0.075 -0.381 -0.008 0.026 -0.053 -0.355 -0.008

(0.539) (-1.327) (-0.599) (-1.000) (0.560) (-0.918) (-0.556) (-0.928)
Sector FEs Y Y Y Y Y Y Y Y
Country FEs Y Y Y Y Y Y Y Y

# Countries 45 45 46 47 44 44 44 44
# Sectors 79 79 79 79 79 79 79 79
N 24,876 23,223 23,968 8,886 24,182 22,609 23,289 8,863
Adj. R 2 0.171 0.138 0.091 0.034 0.165 0.136 0.089 0.034

F-value (first stage)

FisStim 8.12*** 8.63*** 7.80*** 14.08*** 6.02*** 6.46*** 5.76*** 14.44***
MP_BP 3.31** 3.21** 3.31** 4.61*** 3.20** 3.19** 3.16** 3.57**
FXI 6.37*** 6.29*** 6.38*** 4.14***
Instruments relevance (LM χ2) 7.84** 7.51** 7.82** 3.32 6.83** 6.50** 6.77** 3.51
J-statistics (p-value) 0.97 0.15 0.40 0.57 0.54 0.27 0.66 0.73

45
Table 7. Other sensitivity tests
This table reports the results estimating ∆𝑦𝑦𝑖𝑖𝑖𝑖,𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 = 𝜗𝜗𝑗𝑗 + 𝜗𝜗𝑐𝑐 + ∅. 𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝑗𝑗 × 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑐𝑐 + 𝜏𝜏. 𝑋𝑋𝑖𝑖𝑖𝑖,𝑃𝑃𝑃𝑃𝑃𝑃 + 𝜀𝜀𝑖𝑖𝑖𝑖,𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 where 𝑖𝑖
stands for firm, 𝑗𝑗 for sector, and 𝑐𝑐 for country. ∆𝑦𝑦𝑖𝑖𝑖𝑖,𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 is the change in performance ratios for firm 𝑖𝑖 in country 𝑐𝑐 between
2020 and 2019. We use, alternatively, change in asset turnover ratio [∆(𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆)], change in profit margin [∆(𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃)], change
in interest coverage ratio [∆(𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼)], and change in probability of default [∆(𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃)]. 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑐𝑐 is a vector of variables
represent government stimulus packages in country c. 𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝑗𝑗 is industry j’s degree of sensitivity to a pandemic (labeled
“overall”) from Kóren and Petö (2020). 𝑋𝑋𝑖𝑖𝑖𝑖,𝑃𝑃𝑃𝑃𝑃𝑃 is a vector of firm-level explanatory variables, computed as of 2019. We include
sector fixed effects (𝜗𝜗𝑗𝑗 ) at the three-digit NAICS level as well as country fixed effects (𝜗𝜗𝑐𝑐 ) in all regressions. See Appendix,
Table A1 for detailed definition of variables. Regressions are estimated using OLS. The statistical inferences are based on
clustered standard errors at the country level (associated t-values reported in parentheses). ***, **, and * denote statistical
significance at the 1%, 5%, and 10% levels, respectively.

7A: Robust to alternative distancing proxy

∆(SaleA) ∆(ProfM) ∆(IntrC) ∆(ProbD)

(1) (2) (3) (4)

Distancingj x FisStimc 0.002 0.001*** 0.029*** -0.001*

(1.546) (2.813) (2.737) (-1.728)
Distancingj x MP_BPc 0.023** 0.013** 0.190* 0.000
(2.428) (2.072) (1.897) (0.046)
Distancingj x FXIc -0.013 0.012 0.392 -0.007
(-0.498) (0.767) (1.553) (-0.926)

Controls ic,pre
(Size, Age, CashA, RD_A, Private,
√ √ √ √
SaleA, ROA, IntrC, EqitA)

Constant 0.006 -0.071 -0.485 -0.021**

(0.231) (-1.499) (-1.032) (-2.298)

Sector FEs Y Y Y Y
Country FEs Y Y Y Y

# Countries 80 80 80 80
# Sectors 79 79 79 79
N 28,915 26,993 27,841 10,023
Adj. R 2 0.175 0.142 0.093 0.034

46
7B: Robust to WLS estimation and to excluding critical sectors

Weighted Least Square (WLS) Excluding 36 critical sectors

∆(SaleA) ∆(ProfM) ∆(IntrC) ∆(ProbD) ∆(SaleA) ∆(ProfM) ∆(IntrC) ∆(ProbD)

(1) (2) (3) (4) (5) (6) (7) (8)

Distancingj x FisStimc 0.003*** 0.001** 0.025** -0.001*** 0.003** 0.001* 0.0001 -0.001***
(3.043) (2.058) (2.387) (-2.984) (2.330) (1.939) (0.016) (-3.096)
Distancingj x MP_BPc 0.022** 0.024** 0.077 -0.002 0.049** 0.029* 0.168 -0.006
(1.976) (2.513) (0.724) (-0.568) (2.184) (1.676) (0.845) (-0.975)
Distancingj x FXIc -0.016 0.024 0.613** -0.013* -0.003 0.030 0.652* -0.022***
(-0.586) (1.162) (2.126) (-1.738) (-0.053) (0.853) (1.689) (-2.782)
Controls ic,pre
(Size, Age, CashA, RD_A, Private,
√ √ √ √ √ √ √ √
SaleA, ROA, IntrC, EqitA)

Constant -0.022 -0.065*** 0.022 -0.016 -0.022 -0.034 -0.150 -0.027***

(-0.971) (-3.578) (0.089) (-0.441) (-0.704) (-0.671) (-0.304) (-3.023)

Sector FEs Y Y Y Y Y Y Y Y
Country FEs Y Y Y Y Y Y Y Y

# Countries 80 80 80 80 80 80 80 80
# Sectors 79 79 79 79 43 43 43 43
N 28,915 26,993 27,841 10,023 21,343 19,924 20,533 7,358
Adj. R 2 0.179 0.138 0.099 0.046 0.189 0.157 0.099 0.039

47
7C: Robust to other (pre-crisis) firm characteristics

Controlling for firm efficiency and investment opportunities Controlling for firm growth opportunities

∆(SaleA) ∆(ProfM) ∆(IntrC) ∆(ProbD) ∆(SaleA) ∆(ProfM) ∆(IntrC) ∆(ProbD)

(1) (2) (3) (4) (5) (6) (7) (8)

Distancingj x FisStimc 0.004** 0.001** 0.028*** -0.001** 0.004** 0.002** 0.023** -0.001**
(2.591) (2.039) (2.778) (-2.356) (2.309) (2.121) (2.307) (-2.585)
Distancingj x MP_BPc 0.031* 0.020 0.123 -0.004 0.030 0.024 0.039 -0.005
(1.835) (1.430) (0.857) (-0.762) (1.372) (1.582) (0.263) (-0.812)
Distancingj x FXIc -0.002 0.022 0.678* -0.010 0.013 0.022 0.674* -0.009
(-0.039) (0.841) (1.917) (-1.433) (0.286) (0.856) (1.779) (-1.425)
Other firm characteristics
(OverA, CashFlowA) Y Y Y Y Y Y Y Y
(Tobin's Q) N N N N Y Y Y Y

Controls ic,pre
(Size, Age, CashA, RD_A, Private,
√ √ √ √ √ √ √ √
SaleA, ROA, IntrC, EqitA)

Constant 0.043 -0.070 -0.439 -0.020** 0.025 -0.149** -0.777 -0.025**

(1.425) (-1.521) (-0.959) (-2.324) (0.656) (-2.629) (-1.378) (-2.618)

Sector FEs Y Y Y Y Y Y Y Y
Country FEs Y Y Y Y Y Y Y Y

# Countries 80 80 80 80 80 80 80 80
# Sectors 79 79 79 79 79 79 79 79
N 28,335 26,519 27,310 9,942 18,320 17,076 17,811 9,562
Adj. R 2 0.183 0.145 0.096 0.035 0.193 0.171 0.099 0.038

48
7D: Robust to alternative policy stimulus databases

Alternative policy stimulus databases: Alternative policy stimulus databases:

Policy response to COVID-19 in Asia and Pacific Yale dataset

∆(SaleA) ∆(ProfM) ∆(IntrC) ∆(ProbD) ∆(SaleA) ∆(ProfM) ∆(IntrC) ∆(ProbD)

(1) (2) (3) (4) (5) (6) (7) (8)

Distancingj x FisStimc 0.004*** 0.001* 0.024** -0.001*** 0.0003*** 0.00002 0.001 -0.00004*
(4.520) (1.827) (2.715) (-4.720) (4.080) (0.454) (0.716) (-1.733)

Controls ic,pre
(Size, Age, CashA, RD_A, Private,
√ √ √ √ √ √ √ √
SaleA, ROA, IntrC, EqitA)

Constant -0.061*** -0.084* -1.157*** -0.023** -0.001 -0.062 -0.341 -0.027***

(-3.187) (-1.829) (-3.384) (-2.919) (-0.031) (-1.378) (-0.776) (-2.898)

Sector FEs Y Y Y Y Y Y Y Y
Country FEs Y Y Y Y Y Y Y Y

# Countries 22 22 22 22 60 60 60 60
# Sectors 79 79 79 79 79 79 79 79
N 20,900 19,844 20,070 8,482 25,528 23,846 24,565 9,736
Adj. R 2 0.165 0.121 0.090 0.035 0.176 0.141 0.094 0.028

49
Table 8. Heterogeneity in firms’ size and financial positions entering the pandemic
This table reports the results estimating ∆𝑦𝑦𝑖𝑖𝑖𝑖,𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 = 𝜗𝜗𝑗𝑗 + 𝜗𝜗𝑐𝑐 + ∅. 𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝑗𝑗 × 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑐𝑐 + 𝜏𝜏. 𝑋𝑋𝑖𝑖𝑖𝑖,𝑃𝑃𝑃𝑃𝑃𝑃 + 𝜀𝜀𝑖𝑖𝑖𝑖,𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 where 𝑖𝑖 stands for firm, 𝑗𝑗 for sector, and 𝑐𝑐 for country. ∆𝑦𝑦𝑖𝑖𝑖𝑖,𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 is the change
in performance ratios for firm 𝑖𝑖 in country 𝑐𝑐 between 2020 and 2019. We use, alternatively, change in asset turnover ratio [∆(𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆)], change in profit margin [∆(𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃)], change in interest
coverage ratio [∆(𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼)], and change in probability of default [∆(𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃)]. Each panel displays the results obtained by running the regression in a subsample determined by the median value of
various pre-crisis financial variables. 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑐𝑐 is a vector of variables represent government stimulus packages in country c. 𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝑗𝑗 is industry j’s degree of sensitivity to a pandemic from
Kóren and Petö (2020). 𝑋𝑋𝑖𝑖𝑖𝑖,𝑃𝑃𝑃𝑃𝑃𝑃 is a vector of firm-level explanatory variables, computed as of 2019. We include sector fixed effects (𝜗𝜗𝑗𝑗 ) at the three-digit NAICS level as well as country fixed
effects (𝜗𝜗𝑐𝑐 ) in all regressions. See Appendix, Table A1 for detailed definition of variables. Regressions are estimated using OLS. The statistical inferences are based on clustered standard errors
at the country level (associated t-values reported in parentheses). ***, **, and * denote statistical significance at the 1%, 5%, and 10% levels, respectively.

8A: Size

Size (<Mdn.) Size (>Mdn.)

∆(SaleA) ∆(ProfM) ∆(IntrC) ∆(ProbD) ∆(SaleA) ∆(ProfM) ∆(IntrC) ∆(ProbD)

(1) (2) (3) (4) (5) (6) (7) (8)

Distancingj x FisStimc 0.002 0.002 0.026 -0.001 0.004*** 0.001 0.031*** -0.001***
(1.020) (1.446) (1.507) (-0.847) (3.070) (1.391) (3.190) (-3.379)
Distancingj x MP_BPc 0.026 0.028*** 0.108 -0.005 0.037** 0.023 0.253 -0.003
(1.031) (2.726) (0.642) (-0.688) (2.127) (1.024) (1.412) (-0.705)
Distancingj x FXIc -0.010 0.046 0.940** -0.021** -0.015 -0.002 0.418 0.003
(-0.174) (1.043) (1.996) (-2.040) (-0.346) (-0.037) (1.271) (0.321)

Controls ic,pre
(Size, Age, CashA, RD_A, Private,
√ √ √ √ √ √ √ √
SaleA, ROA, IntrC, EqitA)

Constant -0.036 -0.095 -0.995** -0.047*** 0.033 -0.046 0.120 0.005

(-0.577) (-1.352) (-2.203) (-4.019) (0.502) (-0.777) (0.159) (0.708)

Sector FEs Y Y Y Y Y Y Y Y
Country FEs Y Y Y Y Y Y Y Y

# Countries 80 80 80 80 80 80 80 80
# Sectors 79 79 79 79 79 79 79 79
N 14,440 13,237 13,630 2,361 14,475 13,756 14,211 7,662
Adj. R 2 0.155 0.122 0.096 0.080 0.231 0.183 0.091 0.045

50
8B: Cash holdings

CashA (<Mdn.) CashA (>Mdn.)

∆(SaleA) ∆(ProfM) ∆(IntrC) ∆(ProbD) ∆(SaleA) ∆(ProfM) ∆(IntrC) ∆(ProbD)

(1) (2) (3) (4) (5) (6) (7) (8)

Distancingj x FisStimc 0.003 0.001 0.021* -0.001** 0.004*** 0.001 0.044*** -0.001***
(1.016) (1.362) (1.937) (-2.165) (3.622) (1.514) (3.028) (-2.793)
Distancingj x MP_BPc 0.021 0.038** 0.178 -0.003 0.058** -0.016 0.035 -0.005
(1.077) (2.535) (1.053) (-0.523) (2.471) (-0.763) (0.136) (-1.056)
Distancingj x FXIc 0.009 -0.001 0.546 -0.018 -0.036 0.073* 0.892* -0.001
(0.217) (-0.035) (1.422) (-1.533) (-0.499) (1.966) (1.915) (-0.128)

Controls ic,pre
(Size, Age, CashA, RD_A, Private,
√ √ √ √ √ √ √ √
SaleA, ROA, IntrC, EqitA)

Constant 0.080* -0.103 -0.720 -0.037*** -0.155* 0.063 0.633* 0.005

(1.774) (-1.423) (-1.089) (-2.834) (-1.803) (0.866) (1.670) (0.284)

Sector FEs Y Y Y Y Y Y Y Y
Country FEs Y Y Y Y Y Y Y Y

# Countries 80 80 80 80 80 80 80 80
# Sectors 79 79 79 79 79 79 79 79
N 14,446 13,162 14,065 3,903 14,469 13,831 13,776 6,120
Adj. R 2 0.168 0.158 0.088 0.059 0.181 0.134 0.100 0.012

51
8C: Profitability

ROA (<Mdn.) ROA (>Mdn.)

∆(SaleA) ∆(ProfM) ∆(IntrC) ∆(ProbD) ∆(SaleA) ∆(ProfM) ∆(IntrC) ∆(ProbD)

(1) (2) (3) (4) (5) (6) (7) (8)

Distancingj x FisStimc 0.002 0.002 0.011 -0.001** 0.005*** 0.001 0.046*** -0.0002
(1.452) (1.582) (0.845) (-2.703) (2.666) (1.216) (3.536) (-1.399)
Distancingj x MP_BPc 0.031 0.027 0.010 -0.007 0.035 0.016 0.433** -0.001
(1.414) (1.462) (0.080) (-0.721) (1.198) (1.223) (2.171) (-0.673)
Distancingj x FXIc 0.010 0.014 0.661** -0.019 -0.032 0.046 0.700 -0.003
(0.240) (0.342) (2.389) (-1.475) (-0.526) (1.498) (1.248) (-0.611)

Controls ic,pre
(Size, Age, CashA, RD_A, Private,
√ √ √ √ √ √ √ √
SaleA, ROA, IntrC, EqitA)

Constant 0.044 -0.021 0.020 -0.035*** 0.046 -0.254** -1.763** -0.004

(1.250) (-0.399) (0.045) (-2.885) (1.229) (-2.553) (-2.045) (-0.634)

Sector FEs Y Y Y Y Y Y Y Y
Country FEs Y Y Y Y Y Y Y Y

# Countries 80 80 80 80 80 80 80 80
# Sectors 79 79 79 79 79 79 79 79
N 14,471 12,929 13,904 4,798 14,444 14,064 13,937 5,225
Adj. R 2 0.162 0.158 0.095 0.052 0.190 0.140 0.097 0.040

52
8D: Leverage

EqitA (<Mdn.) EqitA (>Mdn.)

∆(SaleA) ∆(ProfM) ∆(IntrC) ∆(ProbD) ∆(SaleA) ∆(ProfM) ∆(IntrC) ∆(ProbD)

(1) (2) (3) (4) (5) (6) (7) (8)

Distancingj x FisStimc 0.002 0.000 0.003 -0.001** 0.007*** 0.003*** 0.090*** -0.001***
(0.939) (0.313) (0.297) (-2.128) (8.007) (2.933) (4.885) (-3.037)
Distancingj x MP_BPc 0.028 0.018 0.020 0.003 0.054** 0.039** 0.526** -0.009**
(1.368) (1.333) (0.159) (0.718) (2.463) (2.040) (2.347) (-2.086)
Distancingj x FXIc -0.030 0.003 0.758** -0.016 0.012 0.031 0.431 -0.007*
(-0.676) (0.084) (2.521) (-1.214) (0.200) (1.061) (0.685) (-1.975)

Controls ic,pre
(Size, Age, CashA, RD_A, Private,
√ √ √ √ √ √ √ √
SaleA, ROA, IntrC, EqitA)

Constant -0.026 -0.091 -0.628 -0.035*** 0.056 -0.082 -0.708 0.004

(-0.594) (-1.389) (-1.114) (-4.121) (0.751) (-1.449) (-0.732) (0.622)

Sector FEs Y Y Y Y Y Y Y Y
Country FEs Y Y Y Y Y Y Y Y

# Countries 80 80 80 80 80 80 80 80
# Sectors 79 79 79 79 79 79 79 79
N 14,708 13,739 14,467 4,913 14,207 13,254 13,374 5,110
Adj. R 2 0.179 0.166 0.095 0.052 0.181 0.128 0.100 0.021

53
Appendix
Table A1. Variable definitions and sources
Variable Definition Source

Change in firm performance (∆y ic,COVID )

∆(SaleA) The change in a bank sale to asset ratio (SaleA) between 2019 and Bureau van Dijk,
2020, calculated as ∆(SaleA)=(SaleA20-SaleA19). SaleA is an asset ORBIS, and own
turnover ratio, which measures the efficiency of a company's assets to calculation.
generate revenue or sales.
∆(ProfM) The change in a bank profit margin (ProfM) between 2019 and 2020, "
calculated as ∆(ProfM)=(ProfM20-ProfM19). Profit margin is a measure
of profitability where it is calculated as the net profit as a share of the
revenue.

∆(IntrC) The change in a bank interest coverage ratio (IntrC) between 2019 and "
2020, calculated as ∆(IntrC)=(IntrC20-IntrC19). Interest coverage ratio is
a company's ability to meet its debt obligations. The interest coverage
ratio is calculated by dividing a company's earnings before interest and
taxes by its interest expense.

∆(ProbD) The change in indicator of probability of default (ProbD) between Dec. Credit Research
2019 and Dec. 2020. ProbD reflects the default risk of publicly listed Initiative – CRI,
firms by quantitatively analyzing numerous covariates that cover market- National University of
based and accounting-based firm-specific attributes, as well as macro- Singapore.
financial factors. We use a prediction for horizon of 1 month. Higher
figures denote higher risk.
Pandemic-prone j
Distancing Kóren and Petö' (2020) sectoral pandemic-prone proxy, using data from Kóren and Petö
O*NETZ. It represents share of worker whose job requires a high level of (2020)
teamwork and customer contact. Kóren and Petö document that US
industries are different when it comes to reliance on teamwork and
customer contact in their operations. This proxy suggests that firms in
economic sectors with a high degree of such pandemic-prone proxy are
particularly vulnerable to social distancing.
Policy c
FisStim Fiscal policy: Cumulative fiscal stimulus package (% of GDP) from IMF, and own
January to December 2020 (week 1 to week 43). calculation.
MP_BP Monetary policy: cumulative change in basis points from January to "
December 2020 times (-1) divided by 100.
FXI Foreign exchange intervention (0=No, 1=Yes): cumulative from January "
to December 2020.
Controls ic,pre
Size Natural logarithm of a firm total assets in 2019. Bureau van Dijk,
ORBIS, and own
calculation.
Age Firm age measured by logarithm of subtracting the firm's year of
incorporation from year 2020. "

CashA Firm cash assets to total assets ratio in 2019. "

RD_A Research and development expenditure divided by total assets in 2019.
"

Private (dummy) A dummy variable that takes value 1 if the firm is a private firm, and 0
otherwise. "

SaleA Firm sales to total assets ratio in 2019. "

ROA Return on assets, which is defined as profit before tax as a percentage
of average assets of a bank, in 2019. "

54
Table A1: Continued …
IntrC Interest coverage ratio is earnings before interest and taxes (EBIT) to
interest expenses ratio in 2019. It determines how easily a company "
can pay interest on its outstanding debt.

EqitA The ratio of shareholder fund (equity) to total assets of a firm in 2019. "
Other variables
Covid_Severity The country-level severity of the lockdown measures in response to the Hale et al. (2020).
pandemic. This is a composite measure of the scale of school closures,
workplace closures and travel bans based on the data on the 31st
December 2020. The indicator is normalised to be from 0 to 100 with the
score100 being the strictest.

HealE Current private expenditures on health per capita expressed in World Bank - WDI.
international dollars at purchasing power parity in year 2019.

FDI Foreign direct investment, which refers to direct investment equity flows "
in the reporting economy, as % of GDP in year 2019.
Trade Total exports and imports as % of GDP in year 2019. "

NPL The ratio of a country bank nonperforming loans to total gross loans in "
year 2019.
Inflation Inflation, measured by consumer price index, which is defined as the "
yearly change in the prices of a basket of goods and services in year
2019.

FinDep External financial dependence of U.S. firms by 3-digit SIC codes. This is Tong and Wei
an industry-level median of the ratio of capital expenditures minus cash (2008).
flow over capital expenditures. Cash flow is defined as the sum of funds
from operations, decreases in inventories, decreases in receivables, and
increases in payables. Capital expenditures include net acquisitions of
fixed assets. Source: Rajan and Zingales (1998).

DemSen Demand sensitivity is a sector-level index on the sensitivity to demand "

shocks, based on stocks’ response to the 9/11/2001 shock.

OverA The ratio of a company's overheads cost (other operating expenses) to Bureau van Dijk,
its total assets in year 2019. ORBIS, and own
calculation.
CashFlowA The ratio of a company's cash flow to its total assets in year 2019. "
Tobin's Q Total market value of common equity divided by total book value of "
assets in year 2019.

55
Table A2: Number of firms by country
Number of Number of
ID Country ID Country
firms firms

1 Argentina 124 41 Malta 26

2 Australia 625 42 Mauritius 40
3 Austria 45 43 Mexico 109
4 Bangladesh 184 44 Mongolia 19
5 Belgium 99 45 Morocco 45
6 Bolivia 24 46 Netherlands 94
7 Bosnia and Herzegovina 84 47 New Zealand 95
8 Brazil 343 48 Nigeria 82
9 Bulgaria 82 49 North Macedonia 55
10 Canada 660 50 Norway 157
11 Chile 173 51 Oman 58
12 China 9804 52 Pakistan 303
13 Colombia 45 53 Panama 47
14 Côte d'Ivoire 21 54 Paraguay 24
15 Croatia 75 55 Peru 95
16 Cyprus 44 56 Philippines 160
17 Denmark 113 57 Poland 469
18 Ecuador 127 58 Portugal 37
19 Egypt 110 59 Qatar 20
20 Finland 129 60 Republic of Korea 1350
21 France 468 61 Romania 196
22 Germany 436 62 Russia 481
23 Greece 128 63 Saudi Arabia 112
24 Hong Kong 139 64 Serbia 147
25 Hungary 21 65 Singapore 426
26 India 1057 66 Slovakia 26
27 Indonesia 533 67 South Africa 146
28 Ireland 52 68 Spain 204
29 Iran 196 69 Sri Lanka 103
30 Israel 264 70 Sweden 549
31 Italy 271 71 Switzerland 148
32 Jamaica 40 72 Thailand 572
33 Japan 2724 73 Tunisia 35
34 Jordan 63 74 Turkey 171
35 Kazakhstan 48 75 Ukraine 129
36 Kenya 27 76 UAE 50
37 Kuwait 65 77 United Kingdom 694
38 Lithuania 20 78 Uzbekistan 82
39 Luxembourg 46 79 Vietnam 1134
40 Malaysia 694 80 Zimbabwe 22

All 28915

56
Table A3. Social distancing and firm performance during COVID-19
This table reports the results estimating ∆𝑦𝑦𝑖𝑖𝑖𝑖,𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 = 𝜗𝜗𝑗𝑗 + ∅. 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑐𝑐 + 𝜏𝜏. 𝑋𝑋𝑖𝑖𝑖𝑖,𝑃𝑃𝑃𝑃𝑃𝑃 + ∇. 𝑍𝑍𝑐𝑐,𝑃𝑃𝑃𝑃𝑃𝑃 + 𝜀𝜀𝑖𝑖𝑖𝑖,𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 and ∆𝑦𝑦𝑖𝑖𝑖𝑖,𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 = 𝜗𝜗𝑗𝑗 + ∅. 𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝑗𝑗 × 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑐𝑐 + 𝜏𝜏. 𝑋𝑋𝑖𝑖𝑖𝑖,𝑃𝑃𝑃𝑃𝑃𝑃 + ∇. 𝑍𝑍𝑐𝑐,𝑃𝑃𝑃𝑃𝑃𝑃 + 𝜀𝜀𝑖𝑖𝑖𝑖,𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶
where 𝑖𝑖 stands for firm, 𝑗𝑗 for sector, and 𝑐𝑐 for country. ∆𝑦𝑦𝑖𝑖𝑖𝑖,𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 is the change in performance ratios for firm 𝑖𝑖 in country 𝑐𝑐 between 2020 and 2019. We use, alternatively, change in asset turnover
ratio [∆(𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆)], change in profit margin [∆(𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃)], change in interest coverage ratio [∆(𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼)], and change in probability of default [∆(𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃)]. 𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝑖𝑖𝑛𝑛𝑛𝑛𝑗𝑗 is industry j’s degree of
sensitivity to a pandemic from Kóren and Petö (2020). 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑐𝑐 is a vector of variables represent government stimulus packages in country c. 𝑋𝑋𝑖𝑖𝑖𝑖,𝑃𝑃𝑃𝑃𝑃𝑃 is a vector of firm-level explanatory variables,
computed as of 2019. 𝑍𝑍𝑖𝑖𝑖𝑖,𝑃𝑃𝑃𝑃𝑃𝑃 is a vector of country-level explanatory variables, computed as of 2019. We include sector fixed effects (𝜗𝜗𝑗𝑗 ) at the three-digit NAICS in all regressions. See Appendix,
Table A1 for detailed definition of variables. Regressions are estimated using OLS. The statistical inferences are based on clustered standard errors at the country level (associated t-values reported
in parentheses). ***, **, and * denote statistical significance at the 1%, 5%, and 10% levels, respectively.

Without interaction Interacted with distancing

∆(SaleA) ∆(ProfM) ∆(IntrC) ∆(ProbD) ∆(SaleA) ∆(ProfM) ∆(IntrC) ∆(ProbD)

(1) (2) (3) (4) (5) (6) (7) (8)

FisStimc -0.001* 0.0001 -0.003 -0.0004*** -0.001*** -0.0002 -0.008** -0.0002*

(-1.744) (0.426) (-1.026) (-2.832) (-2.799) (-0.769) (-2.419) (-1.862)
MP_BP c 0.003 -0.004 -0.031 0.000 -0.003 -0.009** -0.074 0.001
(0.954) (-1.326) (-0.941) (0.343) (-0.594) (-2.128) (-1.544) (0.850)
FXIc 0.005 0.001 0.044 -0.001 0.011* -0.000 -0.026 0.001
(0.543) (0.106) (0.719) (-0.628) (1.677) (-0.057) (-0.327) (0.258)

Distancingj x FisStimc 0.004** 0.002*** 0.031*** -0.001***

(2.405) (2.905) (3.193) (-2.875)
Distancingj x MP_BP c 0.036** 0.030** 0.277* -0.004
(2.361) (2.043) (1.808) (-0.845)
Distancingj x FXIc -0.041 0.009 0.476 -0.013**
(-1.064) (0.315) (1.426) (-2.114)

Controls c,pre
(FDI, Trade, NPL, Covid_Severity, √ √ √ √ √ √ √ √
HDI, LOG_GDP)
Controls ic,pre
(Size, Age, CashA, RD_A, Private,
√ √ √ √ √ √ √ √
SaleA, ROA, IntrC, EqitA)

Constant -0.182** 0.065 -1.274 -0.113*** -0.167* 0.072 -1.190 -0.118***

(-2.147) (0.802) (-1.487) (-3.833) (-1.979) (0.882) (-1.388) (-4.017)

Sector FEs Y Y Y Y Y Y Y Y
Country FEs N N N N N N N N

# Countries 80 80 80 80 80 80 80 80
# Sectors 79 79 79 79 79 79 79 79
N 28,860 26,941 27,791 10,023 28,860 26,941 27,791 10,023
Adj. R 2 0.168 0.131 0.089 0.028 0.169 0.132 0.089 0.029

57
Table A4. Social distancing and real firm performance during COVID-19
This table reports the results estimating ∆𝑦𝑦𝑖𝑖𝑖𝑖,𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 = 𝜗𝜗𝑗𝑗 + 𝜗𝜗𝑐𝑐 + ∅. 𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝑗𝑗 × 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑐𝑐 + 𝜏𝜏. 𝑋𝑋𝑖𝑖𝑖𝑖,𝑃𝑃𝑃𝑃𝑃𝑃 + 𝜀𝜀𝑖𝑖𝑖𝑖,𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 where 𝑖𝑖
stands for firm, 𝑗𝑗 for sector, and 𝑐𝑐 for country. ∆𝑦𝑦𝑖𝑖𝑖𝑖,𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 is the change in real performance ratios for firm 𝑖𝑖 in country 𝑐𝑐
between 2020 and 2019. We use, alternatively, change in number of employees [∆(𝑁𝑁𝑁𝑁𝑁𝑁𝑁𝑁𝑁𝑁)] and change in valued added
[∆(𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉)]. Value added is computed as the sum of earnings before taxes, depreciation and labor expense. 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑐𝑐 is a vector
of variables represent government stimulus packages in country c. 𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝑗𝑗 is industry j’s degree of sensitivity to a
pandemic from Kóren and Petö (2020). 𝑋𝑋𝑖𝑖𝑖𝑖,𝑃𝑃𝑃𝑃𝑃𝑃 is a vector of firm-level explanatory variables, computed as of 2019. We include
sector fixed effects (𝜗𝜗𝑗𝑗 ) at the three-digit NAICS level as well as country fixed effects (𝜗𝜗𝑐𝑐 ) in all regressions. See Appendix,
Table A1 for detailed definition of variables. Regressions are estimated using OLS. The statistical inferences are based on
clustered standard errors at the country level (associated t-values reported in parentheses). ***, **, and * denote statistical
significance at the 1%, 5%, and 10% levels, respectively.

Number of Employees Value Added

∆(NumbE) ∆(ValuA)
(1) (2)

Distancingj x FisStimc 0.111*** 0.192***

(5.761) (2.969)
Distancingj x MP_BPc 0.015 0.264
(0.042) (0.436)
Distancingj x FXIc 0.111 -1.486
(0.222) (-1.068)

Controls ic,pre
(Size, Age, CashA, RD_A, Private,
SaleA, ROA, IntrC, EqitA) √ √

Constant 4.170*** 0.618

(4.578) (0.300)

Sector FEs Y Y
Country FEs Y Y

# Countries 80 80
# Sectors 79 79
N 20,793 13,090
Adj. R 2 0.085 0.123

58
Table A5. Other robustness tests

External finance dependence at the NAICS 3-

Baseline results for restricted dataset Baseline results with the US data
digit level

∆(SaleA) ∆(ProfM) ∆(IntrC) ∆(ProbD) ∆(SaleA) ∆(ProfM) ∆(IntrC) ∆(ProbD) ∆(SaleA) ∆(ProfM) ∆(IntrC) ∆(ProbD)

(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)

Distancingj x FisStimc 0.007*** 0.001 0.034*** -0.001** 0.003** 0.002** 0.041*** -0.001* 0.003** 0.001** 0.029*** -0.001**
(6.368) (1.552) (3.039) (-2.660) (2.224) (2.522) (3.914) (-1.937) (2.294) (2.109) (2.648) (-2.535)
Distancingj x MP_BPc 0.033 0.010 0.311 -0.004 0.015 0.027* 0.208 -0.002 0.029* 0.020 0.184 -0.005
(1.026) (0.310) (1.477) (-0.864) (0.923) (1.665) (1.258) (-0.556) (1.814) (1.435) (1.302) (-1.036)
Distancingj x FXIc -0.070 -0.012 0.081 -0.008 -0.005 0.028 0.848** -0.003 -0.007 0.031 0.724** -0.008
(-1.069) (-0.354) (0.095) (-1.205) (-0.131) (1.000) (2.104) (-0.446) (-0.170) (1.159) (2.026) (-1.303)
Other sector characteristics
FinDepj x FisStimc -0.000 -0.000 -0.000 0.000
(-1.342) (-1.144) (-0.615) (0.209)
FinDepj x MP_BPc 0.001** 0.001** 0.014** -0.000
(2.198) (2.138) (2.387) (-0.517)
FinDepj x FXIc 0.001 0.000 -0.003 -0.000
(0.839) (0.391) (-0.192) (-0.198)
Controls ic,pre
(Size, Age, CashA, RD_A, Private,
√ √ √ √ √ √ √ √ √ √ √ √
SaleA, ROA, IntrC, EqitA)

Constant -0.015 -0.021 -0.530 -0.015* -0.037 -0.162*** -1.017*** 0.018 0.003 -0.071 -0.439 -0.021**
(-0.331) (-0.328) (-0.788) (-1.944) (-1.059) (-5.851) (-4.004) (1.064) (0.111) (-1.530) (-0.969) (-2.547)

Sector FEs Y Y Y Y Y Y Y Y Y Y Y Y
Country FEs Y Y Y Y Y Y Y Y Y Y Y Y

# Countries 80 80 80 80 80 80 80 80 81 81 81 81
# Sectors 79 79 79 79 79 79 79 79 79 79 79 79
N 9,200 9,200 9,200 9,200 25,571 23,832 24,642 8,929 29,964 27,890 28,830 10,857
Adj. R 2 0.200 0.177 0.100 0.028 0.168 0.143 0.088 0.031 0.177 0.141 0.091 0.037

59
Figure 1a. Cumulative fiscal stimulus (% of GDP) from January to December 2020 by country

Qatar
Sri Lanka

75th percentile
25th percentile
Bangladesh
Jamaica
Kuwait
Vietnam
Egypt
Cyprus
Mexico
Pakistan
Kazakhstan
Philippines
Kenya
Jordan
Paraguay
UAE
Panama
Tunisia
North Macedonia
Saudi Arabia
Ukraine
Colombia
Indonesia
Netherlands
Chile
I.R.Iran
Russia
Côte d'Ivoire
Argentina
Morocco
Romania
Serbia
China
Uzbekistan
Croatia
Zimbabwe
Greece
India
Slovakia
Ireland
Bulgaria
Malaysia
Austria
Lithuania
Norway
South Africa
Mongolia
Portugal
Sweden
Israel
Finland
Malta
Turkey
Hong Kong
Switzerland
Luxembourg
Poland
Thailand
Bolivia
Republic of Korea
Brazil
Denmark
Singapore
Belgium
UK
Canada
Spain
Australia
Peru
New Zealand
France
Mauritius
Japan
Germany
Italy

0 4 8 12 16 20 24 28 32 36 40 44

Cum FisStim (% GDP)

Sources: IMF and own calculations.

60
Figure 1b. Cumulative of change in monetary policy (basis point) from January to December 2020 by country

Zimbabwe
Ukraine
Pakistan
Egypt
Paraguay
Mexico
Mongolia
South Africa
Colombia
Sri Lanka
Brazil
Peru
Philippines
Russia
Uzbekistan
Vietnam
Hong Kong
Qatar
Oman
Canada
Jordan
Kenya
Mauritius
Norway
Tunisia
Poland
Bangladesh
Chile
Indonesia
Malaysia
Saudi Arabia
UAE
India
Kuwait
Romania
Serbia
Morocco
New Zealand
North Macedonia
Republic of Korea
Thailand
Australia
UK
China
Côte d'Ivoire
Kazakhstan
25th percentile
75th percentile

Israel
Denmark
Cyprus
Nigeria
Turkey

-1,100 -1,000 -900 -800 -700 -600 -500 -400 -300 -200 -100 0 100 200 300 400

Cum MP_BP (Basis point)

Sources: IMF and own calculations.

61
Previous volumes in this series
1013 Banking in the shadow of Bitcoin? Raphael Auer, Marc Farag, Ulf
May 2022 The institutional adoption of cryptocurrencies Lewrick, Lovrenc Orazem and
Markus Zoss
1012 It takes two: Fiscal and monetary policy in Ana Aguilar, Carlos Cantú and
May 2022 Mexico Claudia Ramírez

1011 Big techs, QR code payments and financial Thorsten Beck, Leonardo
May 2022 inclusion Gambacorta, Yiping Huang,
Zhenhua Li and Han Qiu
1010 Financial openness and inequality Tsvetana Spasova and Stefan
March 2022 Avdjiev

1009 Quantitative forward guidance through Boris Hofmann and Fan Dora Xia
March 2022 interest rate projections

1008 Deconstructing ESG scores: Torsten Ehlers, Ulrike Elsenhuber,

March 2022 How to invest with your own criteria Anandakumar Jegarasasingam and
Eric Jondeau
1007 Cross-border regulatory spillovers Pierre-Richard Agénor, Timothy P
March 2022 and macroprudential policy coordination Jackson and Luiz A Pereira da Silva

1006 Estimating conditional treatment effects Alessandro Barbera, Aron Gereben

February 2022 of EIB lending to SMEs in Europe and Marcin Wolski

1005 The NAIRU and informality in the Mexican Ana María Aguilar-Argaez,
February 2022 labor market Carlo Alcaraz, Claudia Ramírez and
Cid Alonso Rodríguez-Pérez
1004 Original sin redux: a model-based evaluation Boris Hofmann, Nikhil Patel and
February 2022 Steve Pak Yeung Wu

1003 Global production linkages and stock market Raphael Auer, Bruce Muneaki
February 2022 co-movement Iwadate, Andreas Schrimpf and
Alexander Wagner
1002 Exorbitant privilege? Quantitative easing and Viral V Acharya, Ryan Banerjee,
February 2022 the bond market subsidy of prospective fallen Matteo Crosignani, Tim Eisert and
angels Renée Spigt
1001 Unequal expenditure switching: Raphael Auer, Ariel Burstein,
February 2022 Evidence from Switzerland Sarah Lein and Jonathan Vogel

1000 Dollar beta and stock returns Valentina Bruno, Ilhyock Shim and
February 2022 Hyun Song Shin

999 Shadow loans and regulatory arbitrage: Amanda Liu, Jing Liu and Ilhyock
February 2022 evidence from China Shim

All volumes are available on our website www.bis.org.