IMF | Research | 1
Special Series on COVID-19
The Special Series notes are produced by IMF experts to help members address the economic effects of COVID-19. The views
expressed in these notes are those of the author(s) and do not necessarily represent the views of the IMF, its Executive Board,
or IMF management.
May 6, 2020
Tracking the Economic Impact of
COVID-19 and Mitigation Policies in
Europe and the United States
Sophia Chen, Deniz Igan, Nicola Pierri, and Andrea F. Presbitero
1
This note provides a framework to use high-frequency indicators, such as electricity usage, for
policymakers to assess the economic impact of COVID-19 in close to real time. Further, the note examines
the link between economic activity and mitigation efforts to help policymakers better understand the
possible path of economic activity as lockdown measures are relaxed. We find that:
(1) Electricity usage in Europe declined by 10–15 percent (more in harder-hit countries) during the acute
phase of the pandemic—historically, a 1 percent drop in electricity usage has been associated with 1.3–1.9
percent drop in output.
(2) The decline in electricity usage and job losses in the U.S. are larger in states with a lower share of jobs
that can be done from home. Further, job losses are larger in poorer states and in states that do not have in
place laws for paid sick days.
(3) The heterogeneous impact of COVID-19 is mostly captured by changes in people’s observed mobility,
while, so far,
(4) there is no robust evidence of additional impact from the adoption of de jure non-pharmaceutical
interventions such as school and business closures and shelter-in-place orders.
1
We thank Allen Boddie, Christian Bogmans, Nigel Chalk, Giovanni Dell’Ariccia, Hamid Faruqee, Fah Jirasavetakul, Linda Kaltani, Laurent Kemoe, Marco
Marini, Sole Martinez Peria, Florian Misch, Andrea Pescatori, Daniel Rodriguez, Alberto Sanchez, Andre Santos, Emil Stavrev, Ara Stepanyan, Jim Tebrake,
Petia Topalova, Patrizia Tumbarello, Ruud Vermeulen, Jing Zhou for help with data and helpful discussions. We are grateful to Dalya Elmalt and Mu Yang Shin
for excellent research assistance and to Alberto Sanchez and the BigData@Fund community for signaling useful data sources. For more information, contact
DIgan@imf.org.
IMF | Research | 2
I. HIGH-FREQUENCY INDICATORS: WHY AND HOW?
The COVID-19 pandemic is causing economic disruption at unprecedented speed and scale. For instance, thirty
million Americans filed for unemployment in the past six weeks. It took about a year to reach that number in the
wake of the Lehman Brothers’ bankruptcy. Against this background, the relatively slow frequency of most
macroeconomic indicators represents a challenge for policymakers tasked with mitigating the economic impact
of the crisis and charting a way to the recovery phase.
High-frequency indicators, such as electricity usage, can complement traditional measures of economic activity
in helping policymakers tailor their responses to “flatten the recession curve” (Gourinchas 2020). We exploit the
variation in the timing and intensity of COVID-19 outbreak across different locations—and, hence, the policy-
induced and/or voluntary changes in behavior. We use two proxies for the outbreaks: the number of COVID-19
cases or deaths and the Google Community Mobility index (Google, 2020). The latter measures the number of
times individuals visit public places (transit stations, workplaces, retail stores and recreation places, and
groceries and pharmacies) and thus captures mitigation efforts through reduced mobility. We investigate how
economic conditions deteriorate across European countries and U.S. states as the outbreaks spread and/or
people stay at home. Although we do not fully disentangle the direct effect of the pandemic from that of
mitigation efforts and heightened uncertainty, we provide early evidence on the role of Non-Pharmaceutical
Interventions (NPIs), such as school and business closures and shelter-in-place orders.
We measure economic conditions with electricity usage and unemployment insurance (UI) claims. Electricity is
an input in most of economic activity and is difficult to substitute in the short run. Electricity usage is a very
useful high-frequency indicator of economic fluctuations (Chen et al. 2019, Cicala 2020). UI claims are available
at weekly frequency for all U.S. states and closely track labor market developments, so that an increase in UI
claims is one of the earliest signs of rising unemployment and a weakening economy (Wolfers 2020). We
complement our analysis with data on hours worked from more than 100,000 local businesses from the time-
tracking tool Homebase. These indicators are available with a short time lag, so they can be used to track
economic activity as close as possible to “real time.” We focus on these indicators rather than other high-
frequency measures, such as hotel reservations or flight cancellations, as we aim to capture the overall pace of
economic activity rather than focus on the hardest-hit sectors.
2
II. COVID-19 AND ECONOMIC ACTIVITY IN EUROPE
We compare current electricity usage to the same day of the same week in 2019 for 31 European countries.
3
Since early March, electricity usage has been declining in most countries, despite lower energy prices: electricity
consumption was about 5 percent lower than in 2019 during weekdays in the median country in the sample
(Figure 1). The decline has accelerated in April reaching 10–15 percent, and roughly twice this median value in
Italy—the first European country to experience an extensive outbreak and one of the hardest-hit so far.
Cross-country analysis shows that countries with a more severe outbreak, as measured by deaths per capita,
and a sharper decline in people’s mobility have reduced their electricity consumption more (Figure 2). This result
is robust to controlling for the share of manufacturing in national production and weather conditions. The
estimated coefficient suggests that during the acute stage of the pandemic, a doubling of the COVID-19
2
Proprietary consumer data or asset prices can also provide useful information (Baker et al. 2020a, 2020b; Alfaro et al. 2020).
3
Since electricity usage exhibits substantial day-of-the-week fluctuations, we compare each day to the same day of the week in 2019. So, we compare
electricity usage on Tuesday March 31, 2020, with that on Tuesday April 2, 2019 rather than Sunday March 31, 2019. We use daily data from ENTSO-E.
IMF | Research | 3
outbreak is associated with a decrease in electricity usage of approximately 2.4 percent. This is a non-trivial
amount, given that the number of cases doubled every 2 to 3 days during the early phase of the pandemic.
FIGURE 1. COVID-19 and Electricity Usage in Europe: Time Series
A. Weekly, weekdays only
B. Daily
Source: ENTSO-E, ECDC.
Note: The figure plots the relationship at weekly (panel A) or daily (panel B) frequency between the year-on-year change in electricity usage and the number of
COVID-19 deaths across countries up to April 18. Panel A: change in electricity is defined as the year-on-year difference with respect to same week of the year
in 2019. We exclude consumption during weekends and national holidays. To keep the number of working days within a week balanced between 2019 and
2020, when we exclude a holiday, we also exclude the same day of the same week of the previous/following year. Panel B: change in electricity is defined as
the year-on-year difference with respect to the same day of the week and some week of the year of 2019. In most European countries Easter was celebrated on
April 12 in 2020 and April 21 in 2019, which is the “equivalent” of April 19, 2020 on the same-day of the same-week scale. We partially correct for this by
extending the weekend to the first day after Easter where it is a national holiday. We observe a drop of electricity usage in the days heading to Easter, perhaps
because of vacations and business closures. Therefore, part of the drop observed in the days around April 12 (and the increase around April 19) is due to this
Easter effect.
FIGURE 2. COVID-19 and Electricity Usage in Europe: Cross-Section
A. Electricity usage and COVID-19 deaths
B. Electricity usage and mobility
Source: ENTSO-E, ECDC, Google Community Mobility Reports.
Note: Panel A plots the percent change in daily electricity usage relative to the same day of the same week in 2019 and the number of COVID-19 deaths per
capita in 32 continental European countries. Panel B plots the percent change in weekly electricity usage relative to the same day of the same week in 2019 and
the percent change in visits public places (retail and recreation, grocery and pharmacy, transit stations, and workplaces) within a geographic area relative to the
pre-COVID-19 period. Bubble size corresponds to the number of COVID-19 deaths per capita. In both charts, the solid line plots a linear fit and the gray area
shows the 95 percent confidence interval bands. The sample is the week ending on April 11, 2020.
IMF | Research | 4
Policymakers could monitor electricity usage to gauge the economic impact of COVID-19.
4
For instance,
approximately 30 percent of electricity is used by households in Europe. Therefore, assuming that neither the
mix of input used in productive processes, nor the amount of electricity consumed domestically have changed
during the pandemic,
5
a 1 percent drop in electricity usage would correspond to a 1.43 (=1/0.70) percent drop in
production. Alternatively, we estimate the elasticity of electricity with respect to GDP using annual data and
exploiting banking crises as shocks to economic activity (Table 1). We obtain coefficient values ranging from
0.53 to 0.78, implying that, historically, a 1 percent drop in electricity usage is associated with 1.3 to 1.9 percent
drop in output. These estimates are within the range of estimates from the literature (Stern 2018).
6
TABLE 1. Electricity and Output
Source: EAI, ENTSO-E, WEO, Laeven and Valencia (2020).
Note: The table presents the results of running the linear regression:
,
=
,
+
+
+
,
. where c and t indicate a country and a year in our sample,
are country fixed effects, and
capture country-
specific time trends. Estimating the parameter β allows us to infer the unobserved drop in GDP caused by the COVID-19 shock as:
,
=
,
. Estimates of β with OLS are reported in columns (1), (2), and (3) which refer to three different sample periods, all ending in 2019
and starting, respectively, in 2001, 1981, and 1961. As an alternative empirical strategy, we instrument the changes in GDP with the banking crises reported by
Laeven and Valencia (2020). Banking crises are a useful instrument as they are unlikely to affect energy production directly but only through their effect on
economic activity, as they are often followed by sharp recessions. We therefore estimate a two-stage least squares model where we instrument delta logs of
GDP with a dummy equal to one if that country experienced a banking crisis in that year or in the previous two (different timing choices lead to less power in the
first stage). The first stage of the model is presented in columns (4), (5), and (6) for the three different time periods. The second stage results are reported in
columns (7), (8), and (9). All variables are in delta log per capita except for the banking crisis dummies. Robust standard errors are in parentheses. ***, **, and *
denote statistical significance at the 1, 5, and 10 percent level, respectively.
III. COVID-19 AND ECONOMIC ACTIVITY IN THE UNITED STATES
In the United States, average daily electricity usage in early April was 5 percent lower than the same period in
2019. More strikingly, 30 million new unemployment insurance (UI) claims have been filed since the outbreak of
the pandemic. Job losses are concentrated in states that have been hit harder by COVID-19 (Figure 3, panel A),
in line with recent evidence shown for U.S. cities during the 1918 flu pandemic (Correia, Luck, and Verner
2020). Similar evidence is also discussed by Doerr and Gambacorta (2020), Coibion, Gorodnichenko, and
Weber (2020) and Béland et al. (2020).
7
4
Two caveats are in order: extrapolation from cross-sectional results to the aggregate (Nakamura and Steinsson 2018) and potential nonlinearities.
5
These assumptions could be violated because electricity used at home may increase during the lockdown, or because other inputs are not as easy to adjust
as energy. Also important to note that, post-COVID-19, we may be in a new normal where fewer people commute or go out and more work from home, whic h
means the current elasticity could differ from the past.
6
These numbers are estimated at the weekly level. Additional assumptions on the duration and intensity of the outbreak and mitigation measures are needed to
estimate quarterly and annual GDP loss, which is outside the scope of this note. See Section V for additional caveats.
7
In our analysis for Europe, we are not able to look at the labor market given that high-frequency data like weekly UI claim filings are not available.
VARIABLES (in delta log per capita)
(1) (2) (3) (4) (5) (6) (7) (8) (9)
GDP 0.0801 0.2703*** 0.2861*** 0.7756** 0.5287* 0.5602**
(0.101) (0.068) (0.063) (0.333) (0.279) (0.271)
Banking crisis (t to t-2) -0.0322*** -0.0374*** -0.0368***
(0.006) (0.011) (0.010)
Time Frame 2001 to 2019 1981 to 2019 1961 to 2019 2001 to 2019 1981 to 2019 1961 to 2019 2001 to 2019 1981 to 2019
1961 to 2019
Estimator
Observations 694 1,329 1,554 700 1,475 2,065 694 1,329 1,554
F-stat 28 12 12
R-squared 0.131 0.084 0.102 0.322 0.086 0.073
R2-within 0.0008 0.0322 0.0374 0.0473 0.0189 0.0166
Electricity
OLS
GDP
Electricity
OLS (First Stage)
IV
IMF | Research | 5
FIGURE 3. COVID-19, NPIs, Mobility, and Unemployment Insurance Claims in the United States
A. UI claims and COVID-19 deaths
B. UI claims and mobility
Source: US Department of Labor, US Census Bureau, https://covidtracking.com, https://github.com/Keystone-Strategy/covid19-intervention-data, Google
Community Mobility Reports.
Note: Panel A plots the total number of unemployment insurance claims per capita (in logs) and the total number of COVID-19 deaths per capita, at the state
level, from March 8 to April 25, 2020. The solid line plots a linear fit. The slope is 0.10 (s.e.=0.04). Panel B plots the total number of unemployment insurance
claims per capita (in logs) and the percent change in visits to various places (grouped under four categories: retail & recreation, grocery & pharmacy, transit
stations, and workplaces) within a geographic area relative to the pre-COVID-19 period, at the state level, from March 8 to April 25, 2020. The solid line plots a
linear fit. The slope is -0.017 (s.e.=0.004). States are divided between early (red labels) and late (blue labels) NPI adopters. The NPIs considered are social
distancing, closure of nonessential services, closure of public venues, school closures, and shelter-in-place orders. A state is considered an early NPIs adopter
if all these five policies have been implemented within a week from the day in which the first death in the state has been recorded.
Exploiting both the time and cross-sectional dimensions of the data, we see that as the number of COVID-19
cases increases, electricity usage decreases while UI claims increase. The results are economically significant.
For electricity usage, the average elasticity is 0.009 for all U.S. continental states and 0.022 for the top 5 states
with most COVID-19 cases, indicating that during the acute stage of pandemic, a doubling of the number of
cases leads to a decrease in electricity usage of 0.9 percent among continental U.S. states and 2.2 percent
among the top 5 states with most COVID-19 cases.
8
For UI claims, the average elasticity is 0.11, indicating that
a doubling of the COVID-19 positive cases is associated with 11 percent more claims. However, this elasticity
has weakened over time—during March 7–21, the elasticity was close to 0.3—suggesting that the labor market
has reacted very fast to the outbreak and the related social distancing measures.
The economic reaction is heterogenous across U.S. states, reflecting economic and institutional characteristics.
For a given number of COVID-19 cases, UI claims increased more in poorer states, in states with a lower
employment share in hotels and leisure, and lower share of jobs that can be done from home, and in states that
do not have laws for paid sick days (Figure 4, panel A—these differences are statistically significant). The
impact on electricity usage is also stronger among states with lower shares of jobs that can be done at home
(Figure 4, panel B). These results are robust to controlling for the share of manufacturing in state economy and
weather conditions.
8
As with the cross-country regressions for Europe, there are likely nonlinearities involved.
IMF | Research | 6
FIGURE 4. COVID-19, Unemployment Insurance Claims, and Electricity Usage: Panel Regressions
A. COVID-19 and UI claims
B. COVID-19 and electricity usage
Source: US Department of Labor, US Energy Information Administration, Bureau of Economic Analysis, https://covidtracking.com,
https://www.kff.org/other/state-indicator/paid-family-and-sick-leave/, Dingel and Neiman (2020).
Note: Results of estimating equation:
,
=
+
+
,
+
,
+
,
. S is a U.S. state, t is a week between March 7 and April 25, 2020
(Panel A) or a day between March 3 and April 4, 2020 (Panel B).
,
is the number of unemployment insurance claims in a that week (in logs) (Panel A) or
electricity usage (Panel B, in logs of MWHs, relative to the same day of the week of the same week in 2019),
,
is the number of COVID-19 cases in the
previous week (Panel A) or the day (Panel B) (in logs),
is a vector of state-level characteristics,
and
are, respectively, state and week fixed effects. The
sample is a balanced panel with t=7, n=51 (Panel A), or t=49, n=50 (Panel B). The top bar plots the coefficient of the baseline regression (), while the other
bars plot the coefficients ( + ) separately for states with and without paid sick days laws; low and high GDP per capita; low and high share of jobs that can be
done from home; and low and high share of employment in hotels and leisure. Low is defined as below the first quartile of the state distribution. Each bar also
shows the associated 90 percent confidence intervals. Standard errors are clustered by state.
IV. COVID-19, ECONOMIC CONTRACTION, AND MITIGATION EFFORTS
The positive association between COVID-19 cases or deaths and economic contraction can be explained by the
extent and effectiveness of mitigation efforts—which we capture by changes in mobility relative to January 2020.
This is a de facto measure of mitigation efforts and captures de jure NPIs, such as school closures and shelter-
in-place orders, but also compliance and voluntary social distancing by the public. We find that mobility is
positively associated with electricity usage (Figure 2, panel B) and negatively associates with the number on UI
claims (Figure 3, panel B). In contrast, the relationship between de jure NPIs and economic contraction is
weaker. The cross-sectional correlation between the electricity usage across European countries and the
stringency of mitigation policies (Hale et al. 2020) is statistically significant only in the early weeks of the
pandemic but not in April. In the United States, the timing of de jure NPIs is not significantly associated with the
number of UI claims per capita, whether we control for the size of the local outbreak and other state-level
characteristics or not (Table 2).
In other words, de jure NPIs are only part of the story. Compliance and voluntary social distancing matter. This
is also in line with the Swedish experience, albeit the situation is still unfolding: the observed decline in electricity
usage in Sweden—which has adopted relatively less strict mitigation policies but where many have been
practicing social distancing by choice—is fairly similar to that in neighboring countries although the decline in
mobility is smaller.
9
9
In a similar vein, personal vehicle travel declined both in states that imposed stay-at -home orders early in March and in those that imposed such orders later,
although the decline in the former was slightly more (Cicala et al. 2020).
IMF | Research | 7
TABLE 2. COVID-19, NPIs, and Unemployment Insurance (UI) Claims in the United States
Source: US Department of Labor, US Census Bureau, https://covidtracking.com, https://github.com/Keyst one-Strategy/covid19-intervention-data.
Note: The table reports the estimated coefficient of a regression which the total number of unemployment insurance claims per capita (in logs) is function of
NPIs, the total number of COVID-19 death per capita, per capita GDP (in logs), the employment share in hotels and leisure, and a dummy for the presence of
paid sick days laws, at the state level. The sample is a cross-section of 51 U.S. states, with variables measured from March 8 to April 25, 2020. The NPIs
considered are: (i) social distancing, (ii) closure of nonessential services, (iii) closure of public venues, (iv) school closures, and (v) shelter-in-place orders. For
each NPI, a state is considered an early adopter if the policy has been implemented within a week from the day in which the first death in the state has been
recorded. A state is considered an early NPIs adopter if all these five policies have been implemented within a week from the day in which the first death in the
state has been recorded. Results obtained excluding control variables are qualitatively and quantitatively similar.
Furthermore, using daily data for a large sample of local businesses from Homebase (Bartik et al. 2020), we find
that the sharp decline in hours worked—relative to January—begins well before the introduction of de jure NPIs
at the state level (Figure 5, panel A).
10
Similarly, by the time stay-at-home orders were adopted in Europe, the
decrease in mobility and electricity usage was already sizeable (Figure 5, panel B). Interestingly, relative to the
COVID-19 caseloads, mobility dropped earlier in the United States than in Europe although NPIs were adopted
around the same phase of the epidemic. The United States reached 1,000 COVID-19 cases 11 days after
Europe. The first stay-at-home order in the United States (in California) was issued 10 days after the first stay-
at-home order in Europe (in Italy). But mobility in the United States fell by 20 percent compared to January 2020
just 4 days after Europe (Figure 6). Moreover, the early NPIs—school closures in many cases—triggered the
drop in mobility and economic activity in Europe (Figure 5, panel D) but even they seem to have been
anticipated in the United States (Figure 5, panel C).
A likely explanation of this difference is that Americans “learnt” from the European experience and practiced
voluntary distancing and closures before de jure NPIs were adopted. Increased news coverage on COVID-19
during the second week of March is also consistent with this increased “awareness” explanation: on March 11,
for instance, the WHO declared COVID-19 a pandemic, the NBA suspended its games, and Hollywood star Tom
Hanks revealed that he had tested positive.
These findings suggest that avoiding or delaying NPIs may not fully shield an economy from the COVID-19
shock,
11
and that the depression of economic activity may persist even after mandatory lockdown measures are
lifted if people continue to voluntarily limit their mobility.
10
Sectors that are hit harder and earlier by the pandemic, such as restaurants, may be overrepresented in this data source.
11
This could be because people’s behavior changes even in the absence of mandatory restrictions and/or due to spillovers from other regions (for instance,
through supply chain disruptions or reduced demand for travel).
Dependent variable: Unemployment insurance claims per capita (logs) (1) (2) (3) (4) (5) (6)
Early NPIs 0.0406
(0.075)
Early nonessential service closure 0.0205
(0.071)
Early public venue closure 0.0169
(0.113)
Early social distancing -0.0780
(0.072)
Early school closure 0.0954
(0.177)
Early shelter in place
0.0700
(0.073)
Covid-19 deaths per capita (logs) 0.1184*** 0.1181** 0.1169*** 0.1089** 0.1145*** 0.1209***
(0.043) (0.044) (0.044) (0.045) (0.042) (0.043)
Observations 51
51 51
51 51
51
Controls Yes Yes Yes Yes Yes Yes
R
2
0.260 0.257 0.256
0.273 0.264 0.270
IMF | Research | 8
FIGURE 5. COVID-19, NPI Timing, Mobility, and Economic Activity
A. Shelter-in-place orders in the United States
B. Stay-at-home orders in Europe
C. School closures in the United States
D. School closures in Europe
Source: https://covidtracking.com, https://github.com/Keystone-Strategy/covid19-intervention-data, Google Community Mobility Reports, Homebase, ECDC,
ENTSO-E, Hale et al. (2020).
Note Panels A and C plot the changes in hours worked for a large sample of small businesses and in mobility (both relative to the pre-COVID-19 period) for the
median U.S. state, and the cumulative number of COVID-19 deaths for all U.S. states in the sample. The x-axis is the number of days before/after the
introduction of NPIs (shelter-in-plac e in Panel A and school closures in Panel C). The sample only includes states that have adopted the policy by April 30.
Figures based on other NPIs, such as closure of non-essential business or public venues, are qualitatively and quantitatively similar. Panels B and D plot the
median change in electricity usage—with respect to the previous year—the median change in mobility relative to the pre-COVID-19 period, across European
countries, and the cumulative number of COVID-19 deaths for all European countries. The x-axis reports the number of days before/after the introduction of
NPIs (stay-at-home orders in Panel B and school closures in Panel D). The sample only includes European countries that have adopted the policy by April 10.
NPIs introduction and classification is based on Hale et al. (2020).
V. CAVEATS
The use of cases or death counts as a measure of the COVID-19 shock at the local level comes with a few
caveats. The numbers of cases depend on testing policies and capabilities whereas the number of deaths
depends on healthcare capacity and demographics. Importantly, the risk that cases and deaths are
undercounted due to local protocols could be non-trivial. Moreover, the exact reasons why some areas have
been experiencing earlier or more intense outbreaks are still largely unknown. Therefore, hardest-hit areas
might be different from other areas and, importantly for any empirical analysis, what makes an area susceptible
to large outbreaks could be correlated with what also makes the economic impact sizeable (e.g., the prevalence
of nonessential service jobs, industry composition, etc.). The use of high-frequency indicators is also subject to
several caveats. For instance, there are many elements besides economic conditions that affect electricity
consumption, such as energy efficiency, electricity prices, and weather conditions. Also, measurement error
may be present in mobility data.
IMF | Research | 9
FIGURE 6. COVID-19, NPI Timing, and Mobility in Europe versus the United States
Source: https://covidtracking.com, Google Community Mobility Reports, Homebase, ECDC, ENTSO-E.
Note The chart plots the cumulative number of COVID-19 cases and changes in mobility relative to the pre-COVID-19 period in the United States and Europe.
The vertical lines are March 9 and March 19, 2020—the dates when state-at-home orders were issued in Italy and California, respectively.
V. KEY TAKEAWAYS
Our analysis relies on the heterogeneous timing and intensity of the COVID-19 outbreak across different
European countries and U.S. states to provide some useful indications to guide policymaking.
First, the sharp decline in electricity usage and the unique spike in UI claims highlight that this crisis is novel not
only for its magnitude, but also for the speed at which the economy and specifically the labor market are
affected. These numbers are a call for an unprecedented policy response, which should be more similar in spirit
to the reaction to wars and natural disasters, rather than a standard macroeconomic stimulus to support
demand. A mix of monetary, fiscal, and financial measures should be aimed at minimizing disruptions and
scarring from the lockdown, by providing sizable, targeted support to households and businesses to cope with
the “hibernation” of the economy and to be able to jump-start soon after the health crisis will be over.
Second, although COVID-19 is a truly global shock, regions and countries where the outbreak is more sizeable
experience significantly more severe economic losses. This underlines the importance of preventions, early
responses, and other health measures to contain the outbreak at the local and national level.
Third, the early evidence suggesting that the heterogeneous impact of COVID-19 is mostly due to observed
mobility instead of the adoption of de jure NPIs is a warning against optimistic projections that the economic
recovery will start once NPIs are officially lifted, unless workers and consumers feel that the danger stemming
from the epidemic is effectively under control. As countries start looking into reopening the economy, analyses
such as ours could guide decisions not only on the pace and breadth of lifting mitigation policies but also on
other measures that may be needed to restore confidence and trust for people to get back to pre-COVID-19
behaviors.
IMF | Research | 10
REFERENCES
Alfaro, Laura, Anusha Chari, Andrew N. Greenland, and Peter K. Schott, 2020, “Aggregate and Firm-Level Stock Returns During
Pandemics, in Real Time,” NBER Working Paper No. 26950.
Baker, Scott R., R.A. Farrokhnia, Steffen Meyer, Michaela Pagel, and Constantine Yannelis, 2020a, “How Does Household Spending
Respond to an Epidemic? Consumption During the 2020 COVID-19 Pandemic,” NBER Working Paper No. 26949.
Baker, Scott R., Nicholas Bloom, Steven J. Davis, Kyle J. Kost, Marco C. Sammon, and Tasaneeya Viratyosin, 2020b, “The Unprecedented
Stock Market Impact of COVID-19,” NBER Working Paper No. 26945.
Bartik, Alexander W., Marianne Betrand, Feng Lin, Jesse Rothstein, and Matt Unrath, 2020, “Labor Market Impacts of COVID-19 on Hourly
Workers in Small- and Medium- Size Businesses: Four Facts from HomeBase Data,” Chicago Booth Rustandy Center, Working
Paper.
Béland, Louis-Philippe, Abel Brodeur, and Taylor Wright, 2020, “The Short-Term Economic Consequences of COVID-19: Exposure to
Disease, Remote Work and Government Response,” IZA DP No. 13159.
Chen, Wei, Xilu Chen, Chang tai Hsieh, and Zheng (Michael) Song, 2019, “A Forensic Examination of China’s National Accounts,” NBER
Working Paper No. 25754.
Cicala, Steve, 2020, “Early Economic Impacts of COVID-19 in Europe: A View from the Grid,” unpublished manuscript.
Cicala, Steve, Stephen P. Holland, Erin T. Mansur, Nicholas Z. Muller, and Andrew J. Yates, 2020, “Expected Health Effects of Reduced Air
Pollution from COVID-19 Social Distancing,” unpublished manuscript.
Coibion, Oliver, Yuriy Gorodnichenko, and Michael Weber, 2020, “Labor Markets During the COVID-19 Crisis: A Preliminary View,” NBER
Working Paper No. 27017.
Correia, Sergio, Stephan Luck, and Emil Verner, 2020, “Pandemics Depress the Economy, Public Health Interventions Do Not: Evidence
from the 1918 Flu,” unpublished manuscript.
Dingel, Jonathan I. and Brent Neiman, 2020, “How many jobs can be done at home?” NBER Working Paper No. 26948.
Doerr, Sebastian and Leonardo Gambacorta, 2020, “Identifying regions at risk with Google Trends: The impact of Covid-19 on US labor
markets,” BIS Bulletin No. 8.
Gourinchas, Pierre-Olivier, 2020, “Flattening the pandemic and recession curves,” in Richard Baldwin and Beatrice Weder di Mauro (eds),
Mitigating the COVID Economic Crisis: Act Fast and Do Whatever It Takes, CEPR Press.
Hale, Thomas, Sam Webster, Anna Petherick, Toby Phillips, and Beatriz Kira, 2020, Oxford COVID-19 Government Response Tracker,
Blavatnik School of Government.
Nakamura Emi and Jón Steinsson, 2018, “Identification in Macroeconomics,” Journal of Economic Perspectives, 32(3): 59-86.
Stern David I., 2018, “Energy-GDP Relationship,” in Macmillan Publishers Ltd (eds) The New Palgrave Dictionary of Economics. Palgrave
Macmillan, London.
