Economic Modelling 109 (2022) 105794

Contents lists available at ScienceDirect

Economic Modelling
journal homepage: www.journals.elsevier.com/economic-modelling

The COVID-19 pandemic, consumption and sovereign credit risk:

Cross-country evidence☆
Xiangchao Hao a, b, Qinru Sun a, Fang Xie c, *
a
Institute of Finance, School of Economics, Nankai University, Tianjin, PR China
b
The Collaborative Innovation Center for China Economy, Nankai University, Tianjin, PR China
c
School of Finance, Tianjin University of Finance and Economics, Tianjin, PR China

A R T I C L E I N F O A B S T R A C T

JEL classification: Many recent studies investigate the economic effect of the COVID-19 pandemic in multiple aspects, while whether
G01 and how the sovereign credit risk reacts to the shock is still underexplored. Using a sample of forty developed and
G15 developing countries and employing staggered difference-in-differences models, we find that the sovereign credit
H31
risk measured by sovereign credit default swap spreads significantly increases after the COVID-19 pandemic
Keywords: outbreak, and the adverse effect is more pronounced for short-term credit risk. The reason is that the pandemic
COVID-19 pandemic
causes severe concerns about aggregate consumption contraction in addition to the fiscal capacity and the
Sovereign credit risk
Credit default swap
volatility of exports. We also find that fiscal stimuli stabilizing consumer spending alleviate the adverse effect of
Consumption contraction the COVID-19 pandemic, while debt relief does not matter. Overall, practitioners and policy makers should attach
more importance to consumption and its recovery during the pandemic when making decisions.

1. Introduction resilience to economic shock (Augustin et al., 2021), we focus on the role
of consumption in explaining the effect of the COVID-19 pandemic on
According to the World Health Organization (WHO) report, the new sovereign credit risk for two reasons. First, consumption is one of the
coronavirus pneumonia (COVID-19) broke out in early 2020 and rapidly important determinants of sovereign credit risk (Augustin and Tedongap,
developed into a global epidemic that infected more than 80 million 2016; Chernov et al., 2020) and declines sharply during the global
people in over 200 countries and regions by the end of 2020, causing epidemic. As consumption is a driver of economic growth, its contraction
severe economic contraction.1 The COVID-19 pandemic has inspired a may further undermine a country's creditworthiness and hence trigger
large body of literature to investigate its comprehensive effects on eco- debt default in extreme cases. Baker et al. (2020b) argue that the
nomic uncertainty (Altig et al., 2020), unemployment, consumption (Cox COVID-19 pandemic-induced economic downturn is greatly different
et al., 2020; Chetty et al., 2020), and the stock market (Baker et al., from the 2008 financial crisis because the pandemic shock hits house-
2020a; Ding et al., 2021; Liu et al., 2021b), as well as the effect of gov- holds' income and spending much more quickly. Theoretically, con-
ernments’ policy responses. The deterioration of economic development sumption contraction impairs a country's economic health and amplifies
also arouses concerns about sovereign credit risk, yet only a few studies the reaction of sovereign credit risk to the COVID-19 pandemic. Second,
have explored the reaction of sovereign credit risk to the pandemic most countries have implemented stimulus packages to weather the
(Augustin et al., 2021). pandemic shock on household spending, but their effects are still unclear
Although the extant literature stresses the role of fiscal capacity or in the literature on sovereign credit risk. Specifically, we ask three

☆
We gratefully thank Professor Sushanta Mallick (the editor) and two anonymous reviewers for their valuable comments and suggestions. Hao acknowledges the
financial support from the National Natural Science Foundation of China (#72172064; #71772090). All remaining errors are ours.
* Corresponding author.
E-mail addresses: [email protected] (X. Hao), [email protected] (Q. Sun), [email protected] (F. Xie).
1
The report of International Labor Organization shows that 114 million jobs were reduced in 2020 compared to 2019. According to World Economic Outlook 2020
published by International Monetary Funds, global gross domestic product (GDP) of 2020 decreased by 3.5%. Several advanced economies such as the United States
and Japan experienced the deepest economic recession since 2008 global financial crisis, with
3.5% and
4.8% GDP growth rate in 2020, respectively. In addition,
the pandemic pushed up government debt dramatically. The statistic of Institute of International Finance shows that globally, the ratio of government debt to GDP
increased from 90% in 2019 to around 105% in 2020.

https://doi.org/10.1016/j.econmod.2022.105794
Received 11 August 2021; Received in revised form 4 December 2021; Accepted 6 February 2022
Available online 8 February 2022
0264-9993/© 2022 Elsevier B.V. All rights reserved.
X. Hao et al. Economic Modelling 109 (2022) 105794

questions in this study: does the outbreak of the COVID-19 pandemic that consumption is important to understand the relationship between
significantly increase sovereign credit risk? Does the pandemic shock the COVID-19 pandemic and sovereign credit risk, while its role during
undermine sovereign creditworthiness through the consumption the pandemic is still underexplored. Thus, we complement Augustin et al.
contraction? Can fiscal stimuli aimed at stabilizing consumer spending (2021) by adding a new explanation of how the COVID-19 pandemic
mitigate the adverse effect of the global pandemic on sovereign credit affects sovereign credit risk, which is useful for practitioners and
risk? policy-makers to optimize their decisions.
This paper studies the above questions using weekly data from forty Finally, considering that the pandemic does not break out at the same
countries over the period from January 2019 through November 2020. time across countries, we construct a staggered difference-in-differences
Following previous studies, we use sovereign credit default swap (CDS) (DID) model to address endogeneity issues and use econometric methods
spreads to measure sovereign credit risk.2 Given that the consumption developed recently by Borusyak et al. (2021), Callaway and Sant’Anna
recovery of a country usually takes time after a negative shock (Augustin, (2021) and Goodman-Bacon (2021) to address the timing heterogeneity
2018), we also explore the changes in the spreads of 6-month and 1-year issues of the treatment effect. The staggered DID model is widely used in
sovereign CDS respectively in addition to the spreads of 5-year sovereign economic studies, while recent advances in econometric theory question
CDS. This allows us to investigate the differential effects of the COVID-19 its robustness and develop different modified econometric methods. To
pandemic on a country's short-term and long-term credit risks. our knowledge, our study is one of the early to introduce them in eco-
Our major findings are threefold. First, both the short-term and long- nomic studies.
term sovereign credit risks of a country significantly rise after the The rest of the paper proceeds as follows. Section 2 reviews the
outbreak of the COVID-19 pandemic, especially the former. Specifically, related literature. Section 3 presents our methodology, including the
5-year CDS spreads increase by 4.2% on average, while 1-year and 6- econometric models, the sample construction, the data sources and
month CDS spreads increase by 8.7% and 13.5%, about 2.1 and 3.2 summary statistics. Section 4 presents the regression results and analyses,
times of 5-year CDS spreads' change, respectively. The results suggest and Section 5 concludes.
that the adverse impact is more pronounced for the short-term sovereign
credit risk than for the long-term credit risk. In other words, the market is 2. Literature review
more concerned about the deterioration of sovereign credit risk in the
short run. Second, using retail sales growth rate and pre-pandemic saving In the wake of the COVID-19 pandemic outbreak, a great number of
rate to measure the consumption shock, we find that a country's sover- studies have examined its economic effects at both the macrolevel and
eign credit risk rises tremendously if it suffers a severe consumption microlevel in the past one and a half years. This paper only focuses on
contraction. The amplification effect of consumption still exists after studies that explore the effects on consumption and financial market risk.
controlling for the potential effects of fiscal capacity and trade volatility. Regarding the effect on consumption, the literature finds that the
Finally, the effect of income support policy depends on its coverage on COVID-19 pandemic causes severe consumption contraction worldwide.
income losses of the unemployed. Specifically, the adverse effect of the Although using different datasets,3 Baker et al. (2020b), Chetty et al.
pandemic on sovereign credit risk is less pronounced in countries (2020), and Cox et al. (2020) all find that the overall consumption in the
implementing a more generous income support policy. United States during the early stage of the pandemic decreases signifi-
Our study mainly contributes to the literature in three aspects. First, cantly,4 and they report several different findings. Chetty et al. (2020)
we extend the literature by exploring the differential effects of the find that the consumption contraction in the early month mainly relates
COVID-19 pandemic on long-term and short-term sovereign credit risks. to high-income households and then spreads to low-income households
The existing literature widely explores the effect of the COVID-19 due to the consequent large job losses, while Baker et al. (2020b) and Cox
pandemic on macroeconomic downturns (Baker et al., 2020b; Bartik et al. (2020) find that both the consumption contraction and its recovery
et al., 2020; Chetty et al., 2020; Cox et al., 2020; Forsythe et al., 2020; are more significant for low-income households than for high-income
Karger and Rajan, 2021) and financial markets (Baker et al., 2020a; households. Cox et al. (2020) also find that the spending decline is
Fahlenbrach et al., 2021; Ramelli and Wagner, 2020; Delis et al., 2021; especially significant for nonessential items, namely, restaurants, hotel
Ding et al., 2021; Ftiti et al., 2021), while only a few studies explore the accommodations, and clothing and department stores. Using transaction
effect on sovereign credit risk (Daehler et al., 2021). Unfortunately, most data from UnionPay cards and QR scanners, Chen et al. (2021) find that
previous studies focus on the long-term sovereign credit risk measured by daily offline spending on both goods and services in China declines by
the 5-year CDS spreads. However, Augustin (2018) stresses that a more than 40% in the early stage. Likewise, using bank accounts,
negative country-specific shock increases sovereign credit risk more for Andersen et al. (2020) and Kubota et al. (2021) report that household
short maturities and less for long maturities. Thus, we extend the liter- spending falls significantly during the pandemic in Danish and Japan,
ature and provide effective evidence to Augustin (2018) by examining respectively.
the adverse effect of the COVID-19 pandemic. However, the existing literature reports mixed evidence of the
Second, we document that the outbreak of the COVID-19 pandemic effectiveness of fiscal stimulus payments. Regarding the 2020 CARES Act
affects sovereign credit risk through consumption contractions in addi- and its follow-on policies in the United States, most studies find that
tion to limited fiscal capacity and export volatility. The limited literature stimulus payments to low-income households are useful to reverse con-
seldom investigates how the COVID-19 pandemic affects sovereign credit sumption contractions sharply (Baker et al., 2020b; Chetty et al., 2020;
risk, except for Augustin et al. (2021), who investigate the role of fiscal Cox et al., 2020; Karger and Rajan, 2021). Kubota et al. (2021) find that
capacity. They argue that limited fiscal capacity impairs resilience to spending increases exclusively in the first month after the Japanese
economic shocks and increases a country's default risk during the government launches a universal cash entitlement program. Liu et al.
pandemic. However, in addition to fiscal capacity, country-specific fun- (2021a) find that the large-scale digital coupon program issued by the
damentals such as consumption are also important determinants of sov-
ereign credit risk (Augustin, 2018). Furthermore, regarding the
pandemic shock, consumption recovery is important for economic 3
Chetty et al. (2020) obtain consumer spending data from Affinity Solutions
growth (Baker et al., 2020b) and hence is one of the focuses of fiscal relief
and CoinOut, small business revenue data from Womply, employment data from
policies (Chetty et al., 2020; Cox et al., 2020). All these results suggest
Intuit, Earnin, and Kronos, job posting data from Burning Glass, respectively. In
contrast, Baker et al. (2020b) and Cox et al. (2020) rely on transaction data from
SaverLife and Chase, respectively.
2 4
See Longstaff et al. (2011) and other studies for the advantage of CDS Cox et al. (2020) find that consumption expenditure in the United States
spreads as the proxy for sovereign credit risk in detail. decreases by more than 35% in the second half of March.

2
X. Hao et al. Economic Modelling 109 (2022) 105794

Chinese government effectively stimulates household consumption. spreads in addition to fiscal capacity (Augustin, 2018), but both are
However, Chetty et al. (2020) stress that although stimulus payments to underexplored. Recently, Baker et al. (2020b), Chetty et al. (2020), Cox
low-income households are useful to reverse consumption contraction et al. (2020) and other studies also find that consumption stabilization is
sharply, only a tiny proportion of the increased spending flows to busi- the key to economic recovery from the epidemic. Thus, it is valuable to
nesses most affected by the pandemic shock, suggesting that the recovery study the role of consumption in explaining the response of sovereign
of overall consumption benefits economic sectors asymmetrically. Karger credit risk to the COVID-19 pandemic shock.
and Rajan (2021) find that consumer spending falls to normal two weeks
after receiving fiscal stimulus payments. 3. Methodology
The literature on the effect of the COVID-19 pandemic on financial
market risk extensively investigates the stock market's response. First, the 3.1. Econometric models
literature studies financial factors that amplify or mitigate the effect of
the pandemic on stock price volatility at the firm level. Using a sample of To examine the effect of the COVID-19 pandemic on sovereign credit
61 economies, Ding et al. (2021) find that the supply chain's exposure to risk, we rely on a general DID model rather than examining the rela-
the COVID-19 pandemic and financial conditions (such as cash holdings tionship between the number of COVID-19 cases and sovereign credit
and corporate debt) are key determinants of stock price reactions. risk for two reasons. First, the number of COVID-19 confirmed cases
Ramelli and Wagner (2020), focusing on firms in the United States, find highly depends on testing capability and accuracy, especially in the early
that the major concern of the stock market is firms' supply chain exposure stage. Second, slowing down the growth of COVID-19 cases may not
in the early stage of the pandemic and then turns to financial conditions alleviate the concern about the effect of the COVID-19 pandemic on
when the disease spreads more widely. Furthermore, Fahlenbrach et al. sovereign credit risk because the effect may still exist but not disappear
(2021) also find a smaller stock price volatility for firms with less immediately.5 Following Beck et al. (2010) and other studies, our general
short-term debt in the United States. Xue et al. (2021) find that firms' DID model is as follows:
stock prices in China fall significantly after the COVID-19 pandemic
except those in the information technology and health care sectors. LnðSpreadi;t Þ ¼ α þ βCOVID19i;t þ γXi;t þ Ci þ Qt þ εi;t (1)
Second, the existing literature examines the market-level and sector- Ln(Spreadi,t) is the logarithm of sovereign credit risk, measured by
level reactions of the stock market. Focusing on the U.S. case, Baker et al. sovereign CDS spreads of country i in week t. Prior studies suggest two
(2020a) find that the COVID-19 pandemic leads to unprecedented stock major advantages of using sovereign CDS spreads as proxies for sovereign
market uncertainty compared to other epidemics. Using S&P 500 Index credit risk. First, sovereign CDS spreads reflect sovereign credit risk more
returns, Delis et al. (2021) also document that the pandemic generates a adequately since the sovereign CDS market is usually more liquid than
significant negative impact on market return volatility, which is even the sovereign bond market (Longstaff et al., 2011). Second, CDS spreads
greater than the global financial crisis. Focusing on the Chinese stock on different sovereign entities are comparable due to their consistent
market, Ftiti et al. (2021) find that news about the number of deaths and maturity, restructuring clauses, and currency denomination. We use the
cases related to COVID-19 leads to an increase in stock market volatility spreads of 6-month (Spread6m), 1-year (Spread1y) and 5-year sovereign
and a decrease in liquidity. In addition, Haddad et al. (2021), Kargar et al. CDS (Spread5y) to capture the effect of the COVID-19 pandemic on both
(2021), and O'Hara and Zhou (2021) find that the price and liquidity of short-term and long-term sovereign credit risks.
corporate bonds decline significantly during the pandemic, and the COVID19i,t is a dummy variable that captures whether the COVID-19
transaction cost increases as well. pandemic breaks out in country i. Specifically, if country i reports the first
A growing number of studies investigate the response of the financial confirmed case in week t, COVID19i,t equals 1 in week t and consequent
market to the COVID-19 pandemic by focusing on sovereign and firm weeks and otherwise 0. The coefficient of COVID19i,t, β, the DID esti-
credit risks, which are usually measured by CDS spreads. Using the mator, is our focus. If β is positive and significant, the COVID-19
spreads of 5-year sovereign CDS, the existing literature finds that sov- pandemic increases sovereign credit risk and vice versa. As suggested
ereign credit risk is positively correlated with the number of confirmed by prior studies, X is a vector of control variables, including local and
cases or deaths (Daehler et al., 2021) and the adverse effect is more global factors that affect sovereign CDS spreads (Aizenman et al., 2013;
pronounced in countries with small fiscal capacity and strong relief Jeanneret, 2018; Longstaff et al., 2011). Local economic variables
policies (Augustin et al., 2021) and is stronger before the European include the ratio of quarterly foreign debt to GDP (External debt), quar-
Central Bank's intervention (Ortmans and Tripier, 2021). Regarding terly international reserve-to-GDP ratio (Reserve), percentage changes in
corporate CDS spreads, the literature also reports an adverse effect of the monthly real exchange index (Δ(Exchange)), monthly CPI change rate
pandemic, and it is more significant for nonfinancial firms, especially for (Δ(CPI)), and the logarithm of quarterly GDP (Ln(GDP)). Regarding
firms with severe financial constraints, high stock volatility, low credit global factors, we consider the logarithm of weekly volatility of the S&P
rating and profitability (Hasan et al., 2021; Liu et al., 2021b). The extant 500 index (Ln(VIX)) and weekly 5-year U.S. treasury rate (Ln(Trea-
literature also explores the effect of public containment measures on the sure5y)). Ci and Qt control for country and quarter fixed effects that
reaction of sovereign credit risk to the pandemic. For instance, Andries absorb the unobservable country characteristics and time variation in
et al. (2021) find that stringent nonpharmaceutical measures amplify the sovereign CDS spreads, respectively.
erosion of sovereign creditworthiness under the pandemic shock. Next, to examine the role of consumption, following Ding et al.
Taken together, the existing literature has examined the effect of the (2021), we extend the general DID model by including the interaction
COVID-19 pandemic in multiple aspects and provides rich evidence and term of the consumption variable and COVID-19 as follows:
explanations. Although a few studies examine the response of sovereign
credit risk, none study how the pandemic shock undermines sovereign LnðSpreadi;t Þ ¼ α þ βCOVID19i;t þ θCONSUMi;t  COVID19i;t
creditworthiness, except for Augustin et al. (2021). They argue that þ ωCONSUMi;t þ γXi;t þ Ci þ Qt þ εi;t (2)
limited fiscal capacity impairs resilience to economic shocks and in-
creases a country's default risk during the pandemic. Using data from where CONSUMi,t is the consumption change of country i in week t.
thirty developed countries, they find that sovereign CDS spreads increase Following previous studies, we measure consumption change in two
with the growth rate of COVID-19 infections, and the effect is more
significant for fiscally constrained countries, suggesting that the
pandemic affects sovereign credit risk through the fiscal channel. How-
ever, the literature on sovereign credit risk premiums stresses that eco- 5
Nonetheless, in the robustness checks we regress sovereign credit risk on the
nomic growth and consumption are also important determinants of CDS growth of COVID-19 confirmed cases, and report consistent evidence.

3
X. Hao et al. Economic Modelling 109 (2022) 105794

ways. First, we consider the real change in consumption using the year- Table 1
on-year changes in monthly total retail sales of consumer goods (Retail Variable definitions.
sales). Second, we use the saving rate in 2019 (Savings) as another proxy Variables Definition
for consumption change since savings can function as a cushion for the
Spread6m (%) Weekly spreads of 6-month sovereign credit default swaps
abrupt income losses associated with the pandemic and hence alleviate Spread1y (%) Weekly spreads of 1-year sovereign credit default swaps
the decline in household spending (Baker et al., 2020b; Cox et al., 2020; Spread5y (%) Weekly spreads of 5-year sovereign credit default swaps
Kubota et al., 2021). That is, unlike Retail sales, Savings is an indirect COVID19 One for observations of a country after its first case is
measure or expectation of future consumption change. We incorporate confirmed, and zero otherwise
Retail sales (%) Year-on-year changes of monthly retail sales
the interaction term between CONSUM and COVID-19 into Equation (1) Savings (%) The saving rate of a given country in 2019
and focus on the coefficient of the interaction term θ. If θ is negative and Income The mean value of the weekly income support index. The index
significant, we can conclude that the consumption contraction amplifies takes values from 0 to 2, with higher values associated with
the adverse effect of the COVID-19 pandemic. more generous income support.
External debt (%) Quarterly external debt as a percentage of last year GDP
Regarding the effectiveness of fiscal stimuli in restraining the adverse
Reserve (%) Quarterly international reserve as a percentage of last year GDP
effect of the pandemic, we construct a new variable, Income, which is Δ(Exchange)(%) Month-on-month percentage change of monthly real exchange
calculated by the weekly average of the income support index released by rate
the Oxford Coronavirus Government Response Tracker (OxCGRT). The Δ(CPI)(%) Year-on-year percentage change of monthly constant consumer
value of Income ranges from 0 to 2, with higher values related to more price index
GDP (million Quarterly gross domestic product
generous income support for unemployed individuals. We incorporate dollars)
the interaction term of Income and COVID19, Income  COVID19, into VIX Weekly volatility of the S&P 500 index
Equation (1) and construct the following extended model: Treasure5y (%) Weekly 5-year U.S. treasury rate
Num_Cases The number of weekly new COVID-19 cases for each country
LnðSpreadi;t Þ ¼ α þ βCOVID19i;t þ φIncomei;t  COVID19i;t þ δIncomei;t Fisconstraint (%) The difference between revenues and expenditures of a country
as a percentage of GDP of a given country in 2019
þ γX1i;t þ Ci þ Qt þ εi;t Export (%) Year-on-year percentage change of monthly exports
(3) Relief The weekly average of Debt Relief Index obtained from
OxCGRT. The index takes values from 0 to 2, with higher values
In addition to the control variables in X, we also control for the effect related to broader debt relief.
of other fiscal policies on debt relief by adding the interaction term of the
weekly average of Debt Relief Index obtained from OxCGRT (Relief) and
COVID19, namely, Relief  COVID19 in X1. We focus on the coefficient of Table 2
the interaction term φ and expect that income support for unemployed Sample distribution by country.
individuals is effective for weathering the pandemic shock if φ is
Country Observations Country Observation
significantly negative.
Austria 99 Malaysia 99
Belgium 99 Mexico 98
3.2. Sample, data and summary statistics Brazil 99 Netherlands 99
Bulgaria 99 Norway 99
Canada 99 Peru 70
Our sample consists of all countries and regions with data available
Chile 99 Portugal 99
from 1 January 2019 to 30 November 2020. We obtain data from mul- China 99 Romania 99
tiple sources. First, we obtain the spreads of sovereign CDS from Markit. Colombia 99 Russian 99
Markit provides the pricing data of sovereign CDS contracts for which the Cyprus 99 Slovakia 99
Czech 99 Slovenia 99
reference obligation is designated foreign debt of the sovereignty. We use
Denmark 99 South Africa 99
sovereign CDS data denominated in the United States dollar in the basic Finland 99 South Korea 99
regression results and sovereign CDS data denominated in the euro in the France 99 Spain 99
robustness checks to ensure that our results are consistent. Second, we Germany 99 Sweden 99
obtain the data of both new and total confirmed COVID-19 cases for each Greece 99 Switzerland 99
Hungary 99 Thailand 95
country with sovereign CDS data available on the WHO website and the
Ireland 99 Turkey 99
data of macroeconomic factors from the CEIC database and Bank for Israel 99 United Kingdom 99
International Settlements (BIS) database. Finally, we obtain financial Italy 99 United States 99
market data from the Chicago Board Options Exchange and Federal Japan 99
Lithuania 99 Total 3926
Reserve Bank and winsorize them at the 1% level on each tail to control
for the effect of outliers. Then, after merging the above data and keeping
all observations with data available, we construct a sample that covers 40 sovereign CDS spreads by country in Panel B of Table 3. The statistics
developing and developed countries. Table 2 reports the sample show that the spreads with a tenor of six months, one year and five years
distribution. for Demark are lower than those for any other country, while the spreads
We present summary statistics in Panel A of Table 3. The mean for Turkey are the highest regardless of the maturity of CDS contracts.
(median) values of 6-month, 1-year, and 5-year sovereign CDS spreads
are 0.22% (0.10%), 0.27% (0.11%) and 0.67% (0.43%), with standard 4. Empirical results
deviations of 0.40%, 0.48% and 0.79%, respectively. These statistics
suggest an upwards sloping term structure of sovereign CDS spreads. The 4.1. The effect of the COVID-19 pandemic on sovereign credit risk
sample average and median value of COVID-19 are 0.31 and 0, respec-
tively. The mean values of Retail sales, Savings and Income are 0.36%, 4.1.1. Basic results
24.52% and 0.45, respectively. Table 4 reports the regression results of Equation (1). The coefficients
Regarding the control variables, the mean values of External debt, of COVID19 are all positive and statistically significant, at least at the
Reserve, Δ(Exchange) and Δ(CPI)) are 157.72%, 16.34%,
0.07% and 10% level, suggesting that sovereign credit risk increases significantly
1.93%, respectively. On average, quarterly GDP is 434,992.2 million in after the pandemic outbreak. Regarding economic magnitudes, the
U.S. dollars, and the weekly VIX index and the 5-year U.S. treasury rate spreads of 6-month, 1-year, and 5-year sovereign CDS contracts increase
are 21.08 and 1.46%, respectively. We also report the mean values of

4
X. Hao et al. Economic Modelling 109 (2022) 105794

Table 3
Summary statistics.
Panel A: Summary statistics of variables for the whole sample

Variable Observations Min Mean median Max Std. Dev.

Spread6m (%) 3926 0.01 0.22 0.10 2.90 0.40
Spread1y (%) 3926 0.01 0.27 0.11 3.29 0.48
Spread5y (%) 3926 0.07 0.67 0.43 4.36 0.79
COVID19 3926 0.00 0.31 0.00 1.00 0.46
Retail sales (%) 3802
47.96 0.36 1.82 23.46 8.30
Savings (%) 3926 9.89 24.52 25.30 44.18 7.19
Income 3926 0.00 0.45 0.00 2.00 0.76
External debt (%) 3926 0.09 157.72 99.43 885.01 175.14
Reserve (%) 3926 0.53 16.34 10.57 134.63 20.54
Δ(Exchange)(%) 3926
15.11
0.07 0.02 7.06 1.76
Δ(CPI)(%) 3926
3.44 1.93 1.54 20.35 2.52
GDP (million dollars) 3926 5346.47 434,992.20 105,786.40 5,517,583.00 969,381.80
VIX 3926 11.97 21.08 16.25 67.55 11.26
Treasure5y (%) 3926 0.23 1.46 1.61 2.58 0.77

Panel B: The mean values of sovereign CDS spreads by country

Country Spread6m (%) Spread1y (%) Spread5y (%) Country Spread6m (%) Spread1y (%) Spread5y (%)

Austria 0.02 0.03 0.10 Lithuania 0.14 0.17 0.63

Belgium 0.04 0.05 0.16 Malaysia 0.11 0.13 0.65
Brazil 0.58 0.77 1.76 Mexico 0.37 0.49 1.30
Bulgaria 0.17 0.20 0.65 Netherlands 0.03 0.03 0.11
Canada 0.17 0.19 0.32 Norway 0.02 0.02 0.10
Chile 0.13 0.17 0.58 Peru 0.09 0.13 0.58
China 0.08 0.09 0.47 Portugal 0.09 0.12 0.45
Colombia 0.33 0.43 1.21 Romania 0.36 0.41 1.05
Cyprus 0.33 0.43 0.93 Russia 0.39 0.48 1.11
Czech 0.09 0.11 0.39 Slovakia 0.11 0.13 0.45
Denmark 0.02 0.02 0.09 Slovenia 0.24 0.28 0.73
Finland 0.02 0.02 0.11 South Africa 0.90 1.11 2.27
France 0.03 0.05 0.15 South Korea 0.11 0.11 0.31
Germany 0.02 0.03 0.10 Spain 0.10 0.13 0.42
Greece 0.57 0.76 2.07 Sweden 0.03 0.03 0.10
Hungary 0.16 0.19 0.76 Switzerland 0.02 0.03 0.09
Ireland 0.05 0.06 0.25 Thailand 0.08 0.09 0.40
Israel 0.17 0.19 0.62 Turkey 1.99 2.40 3.76
Italy 0.34 0.47 1.18 United Kingdom 0.08 0.10 0.25
Japan 0.04 0.04 0.21 United States 0.04 0.04 0.10

by an average of 13.5%, 8.7% and 4.2%, respectively, after the outbreak

Table 4
of the new coronavirus disease. Furthermore, these results also show that
Basic regression results of the effect of the COVID-19 pandemic outbreak on
the adverse effect of the pandemic shock is more pronounced for short-
sovereign credit risk.
term credit risk than for long-term credit risk, suggesting that the mar-
Variables Ln (Spread6m) Ln (Spread1y) Ln (Spread5y) ket is more concerned about sovereign credit risk deterioration in the
COVID19 0.135*** 0.087** 0.042* short run. This is consistent with Augustin (2018), who predicts that the
(3.688) (2.526) (1.772) negative country-specific shock affects credit risk more for the short term
External debt
0.003***
0.003***
0.002***
and less for the long term.
(-3.576) (-3.367) (-4.137)
Reserve
0.027***
0.026***
0.011*** In addition, most regression results of our control variable are
(-6.859) (-6.991) (-5.632) consistent with prior studies (Aizenman et al., 2013; Jeanneret, 2018).
Δ(Exchange)
0.015***
0.012***
0.007*** For instance, sovereign credit risk is lower for a country with more in-
(-3.950) (-3.337) (-3.073) ternational reserves, higher GDP, and lower currency depreciation and
Δ(CPI) 0.046*** 0.031*** 0.006
(5.721) (4.389) (1.311)
inflation.
Ln (GDP)
0.663***
0.605***
0.533***
(-5.596) (-5.616) (-8.722) 4.1.2. The effect of timing heterogeneity
Ln (VIX) 0.526*** 0.505*** 0.304*** It is worth noting that the general DID model we construct is stag-
(13.039) (13.420) (12.759)
gered since the date of the first confirmed case varies across countries.
Ln (Treasure5y)
0.037
0.021
0.041
(-0.635) (-0.385) (-1.158) Recent econometric theories point out that timing heterogeneity may
Constant 5.020*** 4.584*** 5.173*** bias the two-way fixed effect estimate of the DID model (Borusyak et al.,
(3.653) (3.661) (7.287) 2021; Callaway and Sant’Anna, 2021; Goodman-Bacon, 2021). Thus, it is
Country/Quarter FE Yes Yes Yes necessary to address whether and how our estimates of the general DID
Observations 3926 3926 3926
R2 0.388 0.392 0.417
model are biased. We introduce three new methods developed by recent
literature, while their common requirement on balanced panel data re-
Notes: Ln (Spread6m), Ln (Spread1y) and Ln (Spread5y) are the logarithms of sults in excluding three countries from the sample. We also focus on a DID
weekly spreads of 6-month, 1-year, and 5-year sovereign CDS, respectively.
model without control variables because the conclusion remains un-
COVID19 equals one for observations after the COVID-19 pandemic spreads to a
changed even when incorporating control variables into the model
country. See definitions of other variables in Table 1 t-statistics inferred by robust
standard errors are reported in parentheses. ***, ** and * stand for the signifi-
(Callaway and Sant’Anna, 2021).
cance at the 1%, 5%, and 10% levels, respectively. First, we use Goodman-Bacon (2021) diagnostic to decompose the
aggregate estimate into two components, i.e., the portion related to the

5
X. Hao et al. Economic Modelling 109 (2022) 105794

Table 5 X

2 X
10

Analyses on the effect of timing heterogeneity. LnðSpreadi;w Þ ¼ α þ βw Di;w þ βw Di;w þ Ci þ Qw þ εi;w (4)
w¼
10 w¼0
Panel A: Goodman-Bacon (2021) decomposition

Components Weight Average estimate for two The aggregate where w is the relative week, and Dws are dummy variables equal to one
group countries estimate for observations in the wth week. To avoid multicollinearity, D
1 is
Panel A1: The dependent variable is Ln(Spread5y) excluded. The Callaway and Sant’Anna (2021) estimators are plotted in
Earlier vs Later 0.69 0.03 0.02 Panels A, B, and C of Fig. 1 for long-term and short-term sovereign CDS
Treated spreads, respectively. The coefficients of dummy variables before the first
Later vs Earlier 0.31
0.01 COVID-19 case is reported are all close to zero and statistically insig-
Treated
Panel A2: The dependent variable is Ln(Spread1y)
nificant, suggesting that the parallel trend assumption holds. In contrast,
Earlier vs Later 0.69 0.06 0.07 the coefficients of dummy variables after the COVID-19 pandemic
Treated outbreak grow fast and become statistically significant eight weeks later,
Later vs Earlier 0.31 0.10 suggesting a prominent posttreatment change in sovereign CDS spreads.
Treated
On average, sovereign CDS spreads increase by at least 29% for
Panel A3: The dependent variable is Ln(Spread6m)
Earlier vs Later 0.69 0.13 0.14 short-term credit risk and by 54% for long-term credit risk. Even for those
Treated insignificant coefficients, their standard errors notably increase, pre-
Later vs Earlier 0.31 0.15 senting a different pattern after the COVID-19 pandemic outbreak.
Treated Borusyak et al. (2021) argue that Callaway and Sant’Anna (2021) and
Panel B: The overall group-time treatment effect of Callaway and Sant’Anna (2021) other recently developed methods do not separate testing of the as-
Ln(Spread5y) 0.12 sumptions about pre-trends from the estimation of dynamic treatment
Ln(Spread1y) 0.19 effects under those assumptions and hence cause lower estimation effi-
Ln(Spread6m) 0.17 ciency and higher pretesting bias. Following their method, we further
Notes: Ln (Spread6m), Ln (Spread1y) and Ln (Spread5y) are the logarithms of calculate Borusyak et al.’s (2021) estimators and plot the coefficients in
weekly spreads of 6-month, 1-year, and 5-year sovereign CDS, respectively. Panels D, E, and F in Fig. 1 for long-term and short-term sovereign CDS
spreads. The pre-trend coefficients seem to be greater than the Callaway
comparison between countries early reporting COVID-19 cases and and Sant’Anna (2021) estimators, while they are all statistically insig-
control countries later reporting COVID-19 cases and the portion related nificant, suggesting that the parallel trend assumption holds. The post-
to the comparison between countries later reporting COVID-19 cases and treatment coefficients are greater than the Callaway and Sant’Anna
control countries early reporting COVID-19 cases. Panel A of Table 5 (2021) estimators and statistically significant, with the highest increase
reports the decomposition for three spreads. In Panel A1, the comparison of over 65%. All these results provide consistent evidence of the treat-
between countries early reporting COVID-19 cases and control countries ment effect of the COVID-19 pandemic outbreak on sovereign credit risk.
later reporting COVID-19 cases suggests that sovereign CDS spreads in- Furthermore, the increasing posttreatment coefficients also show that the
crease significantly along with the COVID-19 pandemic outbreak, treatment effects grow over time. In sum, the event study DID analyses
namely, the sovereign credit risk increases, and this treatment effect demonstrate that the parallel trend assumption is effective, and the
accounts for 69%. In contrast, the comparison between countries later treatment effects are significant and grow over time.
reporting COVID-19 cases and control countries early reporting
COVID-19 cases suggests that sovereign CDS spreads decrease signifi- 4.1.3. Further robustness checks
cantly. Since the later observations account for 31%, they finally un- We conduct six additional robustness checks. First, we examine
derestimate the aggregate estimate. The biases for short-term spreads in whether the effect is different between developing and developed
Panel A2 and Panel A3 of Table 5 are the opposite. Nonetheless, countries. Panel A and Panel B of Table 6 report the regression results for
regardless of the direction and the degree of potential biases from timing developed and developing countries, respectively. In Panel A, the results
heterogeneity, the aggregate estimate presents a consistent relationship for developed countries are very similar to those for the whole sample in
between the COVID-19 pandemic and sovereign CDS spreads, i.e., sov- Table 4. In Panel B, most coefficients of COVID19 are significant, and
ereign CDS spreads increase after the COVID-19 pandemic outbreak. their values are also close to those in Table 4. It seems that the COVID19
Second, we follow Callaway and Sant’Anna (2021) to calculate an coefficients for developed countries are greater than those for developing
overall group-time average treatment effect to correct for potential bia- countries, while the χ2 statistics are 0.32, 0.94 and 0.21, respectively,
ses.6 Initially, we divide the sample countries into different groups based suggesting that there is no systemic difference between the adverse ef-
on when a country reports the first COVID-19 case. Then, we calculate fects for these countries.
the average treatment effect for each group. Finally, we average the Second, we use two alternative proxies for sovereign credit risk. One
treatment effects of all groups with the weight of group size to obtain the is the spreads of sovereign CDS contracts denominated in the euro, and
overall group-time average treatment effect. In Panel B of Table 5, the the other is the change in spreads of sovereign CDS contracts denomi-
Callaway and Sant’Anna estimates of the overall group-time average nated in the United States dollar. Panel A and Panel B of Table 7 report
treatment effect are 0.12, 0.19 and 0.17 for long-term and short-term the regression results of the two alternative proxies. The results show that
CDS spreads, respectively. These results further suggest that the almost all coefficients of COVID19 are significantly positive, suggesting
two-way fixed effect estimates of the DID model are systemically that our main results are robust.
underestimated, but our conclusions from Table 4 remain unchanged. Third, we employ three groups of alternative proxies for the COVID-
Third, we conduct an event study DID with a window of 21 weeks, 19 pandemic shock.7 Following Augustin et al. (2021) and other studies,
i.e., starting from 10 weeks prior to the week when the first COVID-19 we first replace COVID19, the dummy variable, with the logarithm of the
case is reported and ending 10 weeks after that. Following Callaway number of new COVID-19 confirmed cases (Num_Cases) plus one. Next,
and Sant’Anna (2021), the econometric model is as follows: we use the weekly average of the Stringency Index (Stringency) released
by OxCGRT as the second alternative proxy. Stringency ranges from 0 to
100, with higher values related to stricter travel restriction measures in

6 7
More details see Callaway and Sant’Anna (2021). We thank an anonymous reviewer for the valuable suggestions.

6
X. Hao et al. Economic Modelling 109 (2022) 105794

Fig. 1. Event study DID of COVID-19 pandemic on long-term and short-term sovereign credit risk. Plots in Panel A, B and C are based on Callaway and Sant’Anna
(2021) (CS) estimators with a simultaneous confidence interval at the 95% level, while plots in Panel D, E and F are based on Borusyak et al. (2021) (BJS) estimators
with a point confidence interval at the 95% level. The vertical axis presents the CS and BJS coefficients, and the horizontal axis presents the relative weeks.

7
X. Hao et al. Economic Modelling 109 (2022) 105794

Table 6
The robustness check on the effect for developed and developing countries.
Variables Panel A: Developed countries Panel B: Developing countries

Ln (Spread6m) Ln (Spread1y) Ln (Spread5y) Ln (Spread6m) Ln (Spread1y) Ln (Spread5y)

COVID19 0.144*** 0.101*** 0.053* 0.121* 0.064 0.066*

(3.498) (2.640) (1.867) (1.849) (1.035) (1.805)
External debt
0.003***
0.003***
0.001*** 0.010*** 0.009***
0.003***
(-3.984) (-3.252) (-2.711) (4.838) (4.766) (-2.584)
Reserve
0.010
0.003 0.018***
0.034***
0.024**
0.031***
(-0.808) (-0.272) (3.483) (-3.065) (-2.383) (-4.569)
Δ(CPI) 0.041*** 0.055*** 0.024*** 0.126*** 0.096*** 0.056***
(2.729) (3.823) (2.718) (5.294) (4.264) (3.995)
Δ(Exchange) 0.030*** 0.029*** 0.013***
0.013***
0.010**
0.005*
(3.662) (3.632) (3.227) (-2.783) (-2.268) (-1.786)
Ln(GDP) 0.288 0.502** 0.484***
0.102
0.086
0.461***
(1.316) (2.282) (3.955) (-0.624) (-0.573) (-5.631)
Ln(VIX) 0.312*** 0.318*** 0.160*** 0.471*** 0.410*** 0.200***
(6.040) (6.450) (5.392) (6.041) (5.696) (4.713)
Ln(Treasure5y)
0.203***
0.196***
0.174***
0.212*
0.241**
0.187***
(-2.753) (-2.766) (-3.951) (-1.944) (-2.347) (-3.019)
Constant
5.175**
7.803***
6.626***
0.191 0.011 6.609***
(-1.974) (-2.974) (-4.503) (-0.095) (0.006) (6.598)
Country/Quarter FE Yes Yes Yes Yes Yes Yes
Observations 2195 2195 2195 1288 1288 1288
R2 0.352 0.367 0.431 0.457 0.455 0.471

Notes: Ln (Spread6m), Ln (Spread1y) and Ln (Spread5y) are the logarithms of weekly spreads of 6-month, 1-year, and 5-year sovereign CDS, respectively. COVID19 equals
one for observations after the COVID-19 pandemic spreads to a country. See definitions of other variables in Table 1 t-statistics inferred by robust standard errors are
reported in parentheses. ***, ** and * stand for the significance at the 1%, 5%, and 10% levels, respectively.

Table 7
The robustness check on alternative proxies for sovereign CDS spreads.
Variables Panel A: Using spreads of sovereign CDS contracts denominated in the euro as the Panel B: Using changes in sovereign CDS spreads as the dependent
dependent variable variable

Ln (Spread6m) Ln (Spread1y) Ln (Spread5y) Δ(Spread6m) Δ(Spread6m) Δ(Spread1y)

COVID19 0.127*** 0.069* 0.014 3.564* 4.082** 2.613***

(3.183) (1.811) (0.559) (1.828) (2.405) (3.132)
External debt
0.002***
0.003***
0.003*** 0.021 0.010 0.015
(-3.041) (-3.530) (-4.368) (0.584) (0.279) (1.101)
Reserve
0.035***
0.028***
0.013*** 0.476 0.215 0.134
(-7.926) (-7.019) (-6.054) (1.314) (0.646) (0.949)
Δ(CPI)
0.012***
0.012***
0.007***
0.078
0.024
0.018
(-2.615) (-2.935) (-2.779) (-0.116) (-0.040) (-0.053)
Δ(Exchange) 0.053*** 0.034*** 0.009*
0.766**
0.785***
0.399**
(5.742) (4.188) (1.819) (-2.516) (-2.650) (-2.253)
Ln (GDP)
0.392***
0.503***
0.469*** 3.520 4.380 1.392
(-2.979) (-4.281) (-6.915) (0.672) (0.888) (0.561)
Ln (VIX) 0.548*** 0.539*** 0.339*** 25.482*** 23.671*** 13.398***
(11.898) (13.030) (12.490) (9.182) (9.119) (9.837)
Ln (Treasure5y)
0.031
0.034
0.056 17.341*** 17.287*** 8.328***
(-0.467) (-0.572) (-1.404) (4.265) (4.560) (3.969)
Constant 1.692 3.305** 4.307***
219.056***
218.534***
104.215***
(1.104) (2.420) (5.464) (-3.324) (-3.508) (-3.293)
Country/Quarter FE Yes Yes Yes Yes Yes Yes
Observations 3364 3364 3364 3414 3414 3413
R2 0.921 0.941 0.969 0.161 0.164 0.224

Notes: Ln (Spread6m), Ln (Spread1y) and Ln (Spread5y) are the logarithms of weekly spreads on 6-month, 1-year, and 5-year sovereign CDS, respectively. Δ(Spread6m),
Δ(Spread1y) and Δ(Spread5y)are the percentage changes in weekly spreads on 6-month, 1-year, and 5-year sovereign CDS, respectively. COVID19 equals one for ob-
servations after the COVID-19 pandemic spreads to a country. See definitions of other variables in Table 1 t-statistics inferred by robust standard errors are reported in
parentheses. ***, ** and * stand for the significance at the 1%, 5%, and 10% levels, respectively.

response to the COVID-19 pandemic. Finally, using the weekly averages and Retail are negative, all more statistically significant and with absolute
of Google mobility trend data, we construct six additional variables: values decreasing over maturities.
Residential, Workplace, Transit, Outdoor, Grocery, and Retail. These Fourth, we investigate the potential impact of the sovereign CDS
represent the changes in the number of visitors to residential areas, contract liquidity.8 Since detailed sovereign CDS volume data are not
workplaces, transit stations, parks and outdoor spaces, grocery and
pharmacy stores, and places of retail and recreation, compared to the
number of visitors before the pandemic, respectively. Residential is
8
positively associated with the intensity of the pandemic shock, while the We thank an anonymous reviewer for the valuable suggestions. Besides re-
other five variables are negatively associated. The regression results in sults in Table 9, we also collect the data of all sovereign CDS contracts during
the sample period from Markit, and summarize the distribution of their matu-
Table 8 show that the coefficients of Num_Cases, Stringency and Residential
rities. The statistics show that the proportion varies little across maturities,
are positive, and the coefficients of Workplace, Transit, Outdoor, Grocery
providing supplemental evidence. To save space, we do not report the statistics.

8
X. Hao et al. Economic Modelling 109 (2022) 105794

Table 8 Table 9
The robustness check replacing COVID19 with alternative proxies. The regression results of 3-year and 10-year sovereign CDS spreads.
Variables Ln (Spread6m) Ln (Spread1y) Ln (Spread5y) Variables Ln (Spread3y) Ln (Spread10y)

Panel A: replacing COVID19 with the number of new COVID-19 confirmed cases COVID19 0.052* 0.045**
(Num_Cases) (1.833) (2.455)
Ln(1 þ Num_Cases) 0.051*** 0.047*** 0.031*** External debt
0.002***
0.002***
(8.257) (8.297) (8.531) (-3.949) (-3.800)
Panel B: replacing COVID19 with Stringency Index (Stringency) Reserve
0.009***
0.013***
Stringency 0.006*** 0.005*** 0.004*** (-3.462) (-9.235)
(9.525) (9.491) (10.896) Δ(Exchange)
0.011***
0.006***
Panel C: replacing COVID19 with Google mobility data by metric (-3.763) (-3.134)
Residential 0.017*** 0.016*** 0.013*** Δ(CPI) 0.007
0.000
(7.693) (7.715) (10.328) (1.210) (-0.106)
Workplace
0.008***
0.008***
0.005*** Ln (GDP)
0.668***
0.423***
(-8.391) (-8.537) (-10.241) (-8.441) (-9.112)
Transit
0.006***
0.005***
0.004*** Ln (VIX) 0.383*** 0.195***
(-9.653) (-9.873) (-12.104) (13.463) (10.621)
Outdoor
0.002***
0.002***
0.001*** Ln (Treasure5y)
0.052
0.007
(-8.306) (-8.255) (-9.001) (-1.240) (-0.264)
Grocery
0.007***
0.006***
0.005*** Constant 6.060*** 4.656***
(-5.613) (-5.189) (-6.812) (6.609) (8.616)
Retail
0.005***
0.005***
0.004*** Country/Quarter FE Yes Yes
(-9.890) (-10.222) (-12.737) Observations 3926 3900
R2 0.479 0.401
Notes: Ln (Spread6m), Ln (Spread1y) and Ln (Spread5y) are the logarithms of
weekly spreads of 6-month, 1-year, and 5-year sovereign CDS, respectively. Notes: Ln (Spread3y) and Ln (Spread10y) are the logarithms of weekly spreads of
Num_Cases is the number of weekly new COVID-19 cases for each country. 3-year, and 10-year sovereign CDS, respectively. COVID19 equals one for ob-
Stringency is the weekly average of Stringency Index released by OxCGRT. Resi- servations after the COVID-19 pandemic spreads to a country. See definitions of
dential, Workplace, Transit, Outdoor, Grocery, and Retail are constructed based on other variables in Table 1 t-statistics inferred by robust standard errors are re-
Google mobility data, and they are the weekly averages of the changes in the ported in parentheses. ***, ** and * stand for the significance at the 1%, 5%, and
number of visitors to residential areas, workplaces, transit stations, parks and 10% levels, respectively.
outdoor spaces, grocery and pharmacy stores, and places of retail and recreation
compared to the number of visitors before the outbreak of the pandemic, results in Table 10 show that the COVID-19 pandemic has led to a larger
respectively. COVID19 is replaced with one of the alternative proxies each time. increase in sovereign CDS spreads for high-risk-exposure countries than
See definitions of other variables in Table 1 t-statistics inferred by robust stan-
for other countries. That is, liquidity and credit risks, especially of spe-
dard errors are reported in parentheses. ***, ** and * stand for the significance at
cific industries, play an important role in pushing the increase of sover-
the 1%, 5%, and 10% levels, respectively. To save space, results of control var-
iables and statistics are not reported.
eign credit risk after the pandemic outbreak.
Sixth, the standard errors of CDS spreads may be correlated because
many countries are affected, either directly or through the negative
available publicly, we rely on Pan and Singleton (2008) who show that
spillover due to the interconnectedness of the economies. To control for
liquidity distribution varies across CDS contracts for different maturities.
the potential correlation, we adjust for t-statistics by clustering standard
They find that sovereign CDS contracts at maturity points between 1 and
errors by group. Countries are divided into six groups, namely, Africa,
10 years are all traded actively, while the trading of non-sovereign CDS
Asia, Europe, the Middle East, North America, and South America, since
centres on a 5-year contract. The more even distribution of sovereign CDS
countries geographically close are usually in the same trading block or
contracts suggests that its potential impact is less likely to change our
share close economic relationships. The results in Table 11 show that the
main findings. Nonetheless, we further provide additional regression
coefficients of COVID19 are all positive and significant at least at the 10%
results of 3-year and 10-year sovereign CDS contracts. The results in
level, suggesting that our results remain unchanged.
Table 9 show that the coefficients of COVID19 for 3-year and 10-year
sovereign CDS spreads are significantly positive and greater for
short-term contracts, suggesting that our results are robust. 4.2. Analyses on the role of consumption
Fifth, the COVID-19 pandemic may increase firms' liquidity and credit
risks, especially those in industries exposed to the shock, thereby exac- Panel A of Table 12 reports the estimation results of Equation (2). In
erbating sovereign credit risk.9 This indicates that the impact of the Columns (1) to (3), the coefficients of Retail sales  COVID19 are all
COVID-19 pandemic may vary in countries with different liquidity risks statistically negative, suggesting that when retail sales do not contract,
and credit risks. To capture the differential impacts, we compare the namely, Retail sales is zero, the spreads of sovereign CDS on average in-
regression results of subsamples based on different risk measures. First, crease by 11.7%, 7.4% and 3.7% for Spread6m, Spread1y and Spread5y,
we use the ratio of current liabilities to total liabilities, and Altman Z respectively. In contrast, when retail sales contract most, namely, Retail
score in 2019 at the country level to measure liquidity risk and credit risk sales is
47.9%, the spreads of sovereign CDS on average increase by
before the pandemic, respectively. Second, we use the proportion of the 23.95%, 23.95% and 38.32% for Spread6m, Spread1y and Spread5y,
output of hotel, accommodation, and transportation industries that are respectively. These results mean that, on average, the increase in sover-
more likely to be hit by the pandemic as another measure for liquidity eign CDS spreads in the worst case is two to ten times the increase
and credit risks. If a country's output depends more on these industries without consumption contraction. In Columns (4) to (6) of Panel B, given
before the pandemic, the increase in liquidity and credit risks in such that savings rates at the country level are predetermined, Savings is
industries may cause deeper concern about massive default. Then, we dropped in a country-fixed effect model due to multicollinearity. The
define countries in the top quartile of risk indicators as high-risk- results show that sovereign CDS spreads increase more for countries with
exposure countries and others as low-risk-exposure countries. The lower savings rates during the pandemic. To address the risk of spurious
correlation by including the interaction term, we further conduct re-
gressions by group. We define countries in the top quartile of Retail sales
or Savings as countries suffering small consumption shocks and others
9 suffering big consumption shocks. The results in Panel B show that the
We thank an anonymous reviewer for the valuable suggestions on liquidity
and credit risks as well as standard errors correction. adverse effect of the pandemic is more significant in countries with large

9
X. Hao et al. Economic Modelling 109 (2022) 105794

Table 10
The regression results of countries with different risks.
Variables Countries with low-risk-exposure before the pandemic Countries with high-risk-exposure before the pandemic

Ln (Spread6m) Ln (Spread1y) Ln (Spread5y) Ln (Spread6m) Ln (Spread1y) Ln (Spread5y)

Panel A: Liquidity risk measured by the ratio of current liabilities to total liabilities in 2019
COVID19 0.078* 0.051 0.060* 0.530*** 0.372*** 0.133***
(1.798) (1.221) (1.912) (6.665) (5.724) (3.802)

Panel B: Credit risk measured by Altman Z-score in 2019

COVID19 0.117*** 0.072* 0.039 0.214*** 0.165** 0.076*

(2.862) (1.842) (1.383) (2.720) (2.337) (1.782)

Panel C: Risk measured by the proportion of the output of industries exposed to the pandemic

COVID19 0.105** 0.042 0.017 0.208*** 0.190*** 0.144***

(2.463) (1.050) (0.640) (3.068) (2.866) (3.015)

Notes: Ln (Spread6m), Ln (Spread1y) and Ln (Spread5y) are the logarithms of weekly spreads of 6-month, 1-year, and 5-year sovereign CDS, respectively. COVID19 equals
one for observations after the COVID-19 pandemic spreads to a country. See definitions of other variables in Table 1. Countries with low risk exposure and countries with
high risk exposure are defined based on the ratio of current liabilities to total liabilities, Altman Z-score and the proportion of the output of hotel, accommodation, and
transportation industries to the GDP in 2019, respectively. t-statistics inferred by robust standard errors are reported in parentheses. ***, ** and * stand for the sig-
nificance at the 1%, 5%, and 10% levels, respectively. To save space, results of control variables and statistics are not reported.

a given country in 2019, namely, the fiscal balance. Generally, Fiscon-

Table 11
straint is negatively associated with financial constraints. The other is
The regression results of clustering standard errors across countries sharing close
Export, measured by the percentage changes in monthly exports. Then,
trading relationship.
we incorporate them and the interaction term between them and
Variables Ln (Spread6m) Ln (Spread1y) Ln (Spread5y) COVID19 into Equation (2). Likewise, since Fisconstraint is also pre-
COVID19 0.136*** 0.088** 0.042* determined during the sample period, the estimation is also dropped.
(805.826) (16.210) (9.914) Panel C of Table 12 reports the regression results of alternative
External debt
0.003**
0.003**
0.002***
explanation tests. Most coefficients of the interaction term of Fisconstraint
(-19.320) (-24.198) (-262.021)
Reserve
0.027***
0.026**
0.011** and COVID19 are significantly negative, especially in Columns (4) to (6),
(-111.619) (-38.235) (-15.868) suggesting that the adverse effect of the pandemic on sovereign CDS
Δ(Exchange)
0.013
0.010
0.006 spreads is weaker for less financially constrained countries. The co-
(-2.281) (-1.641) (-1.770) efficients of the interaction term of Export and COVID19 are significantly
Δ(CPI) 0.046** 0.031*** 0.005
(39.557) (580.795) (5.013)
negative in Column (3) and Column (6), suggesting that the pandemic
Ln (GDP)
0.660*
0.603**
0.532 only affects long-term sovereign CDS spreads through export changes.
(-7.141) (-34.351) (-6.202) These results show that the pandemic affects sovereign credit risk
Ln (VIX) 0.529** 0.508** 0.306** through the deterioration of fiscal capacity and export accounts of a
(38.616) (37.214) (63.615)
country. However, all the coefficients of the interaction term between
Ln (Treasure5y)
0.038**
0.022**
0.041*
(-13.665) (-11.349) (-21.764) consumption variables and COVID19 are still significantly negative,
Constant 4.982 4.554** 5.154 suggesting that the consumption channel exists.
(4.336) (17.101) (5.197)
Country/Quarter FE Yes Yes Yes
Observations 3926 3926 3926
4.3. The effectiveness of fiscal stimuli related to consumption recovery
R2 0.433 0.442 0.460
In this section, we investigate whether fiscal stimuli aimed at con-
Notes: Ln (Spread6m), Ln (Spread1y) and Ln (Spread5y) are the logarithms of
sumption recovery, namely, income support for unemployed individuals,
weekly spreads of 6-month, 1-year, and 5-year sovereign CDS, respectively.
COVID19 equals one for observations after the COVID-19 pandemic spreads to a
can mitigate the adverse effect of the pandemic on sovereign credit risk.
country. See definitions of other variables in Table 1 t-statistics inferred by Panel A of Table 13 reports the regression results of Equation (3). The
clustering standard errors across countries sharing close trading relationship are coefficients of COVID19 are still statistically positive, and the coefficients
reported in parentheses. ***, ** and * stand for the significance at the 1%, 5%, of the interaction term Income  COVID19 are negative and significant at
and 10% levels, respectively. the 1% level regardless of what sovereign CDS spread is used, suggesting
that fiscal stimuli alleviate the adverse effect of the pandemic shock. In
consumption shocks, suggesting that our results are robust. addition, we also find the coefficients of the interaction term Relief 
However, the findings above have to exclude alternative channels COVID19 are all positive, suggesting that debt relief package exaggerates
through which the pandemic shock affects sovereign credit risk. On the rather than mitigates the adverse effect, in particular for long-term sov-
one hand, Augustin et al. (2021) propose a fiscal channel and argue that ereign CDS spreads. Likewise, we also address the risk of spurious cor-
sovereign CDS spreads are more sensitive to COVID-19 cases for finan- relation by conducting regressions by group. We first calculate the
cially constrained governments. On the other hand, given that trade sample average of Income for each country. Then, we define countries in
volatility is one of the important determinants of sovereign credit risk the top quartile of income support as high coverage countries and others
(Hilscher and Nosbusch, 2010), the COVID-19 pandemic inevitably as low coverage countries. The results in Panel B of Table 13 suggest that
damages international trade and may affect sovereign CDS spreads countries offering more generous income support for unemployed in-
through the trade channel. To exclude these two potential explanations, dividuals suffer less from the erosion of sovereign creditworthiness,
we construct two new variables. One is Fisconstraint, measured by the suggesting that our results remain unchanged.
difference between revenues and expenditures as a percentage of GDP of We further identify whether the relief on income losses eases the
pandemic-induced erosion of sovereign creditworthiness by stabilizing

10
X. Hao et al. Economic Modelling 109 (2022) 105794

Table 12
Analyses on the role of consumption and alternative explanations.
Variables (1) (2) (3) (4) (5) (6)

Ln (Spread6m) Ln (Spread1y) Ln (Spread5y) Ln (Spread6m) Ln (Spread1y) Ln (Spread5y)

Panel A: Tests of the role of consumption

COVID19 0.117*** 0.074** 0.037 0.211*** 0.185*** 0.156***
(3.181) (2.107) (1.570) (3.430) (3.195) (3.919)
Retail sales  COVID19
0.005*
0.005*
0.008***
(-1.676) (-1.894) (-5.085)
Savings  COVID19
0.003*
0.004**
0.005***
(-1.814) (-2.254) (-4.394)
Panel B: Regression results by group

Variables Countries with big consumption shock Countries with small consumption shock

Panel B1: Dividing sample countries based on Retail sales

COVID19 0.167*** 0.116*** 0.053* 0.026 0.002 0.012
(4.006) (2.857) (1.827) (0.354) (0.025) (0.402)
Panel B2: Dividing sample countries based on Savings
COVID19 0.148*** 0.104** 0.068** 0.003
0.066
0.029
(3.380) (2.490) (2.327) (0.046) (-1.095) (-0.901)
Panel C: Tests of alternative explanations: the fiscal capacity and the trade volatility

COVID19 0.106*** 0.064* 0.014 0.175*** 0.142** 0.070*

(2.805) (1.773) (0.622) (2.957) (2.510) (1.737)
Retail sales  COVID19
0.005*
0.005**
0.009***
(-1.754) (-2.014) (-5.793)
Savings  COVID19
0.003*
0.003**
0.003**
(-1.697) (-2.095) (-2.411)
Fisconstraint  COVID19
0.008
0.008
0.017***
0.016***
0.017***
0.021***
(-1.560) (-1.538) (-6.412) (-3.356) (-3.732) (-8.346)
Export  COVID19
0.001
0.001
0.001*
0.001
0.001
0.001***
(-1.355) (-0.985) (-1.950) (-0.895) (-0.772) (-2.805)

Notes: Ln (Spread6m), Ln (Spread1y) and Ln (Spread5y) are the logarithms of weekly spreads of 6-month, 1-year, and 5-year sovereign CDS, respectively. COVID19 equals
one for observations after the COVID-19 pandemic spreads to a country. Retail sales is calculated by the monthly growth rate of social retail sales. Savings is the saving
rate in 2019 for each country. Fisconstraint equals fiscal surplus/GDP in 2019 for each country. Export is the percentage changes in monthly exports. See definitions of
other variables in Table 1. Countries in the top quartile of Retail sales or Savings are countries with small consumption shock and others are countries with big con-
sumption shock. t-statistics inferred by robust standard errors are reported in parentheses. ***, ** and * stand for the significance at the 1%, 5%, and 10% levels,
respectively. To save space, results of control variables and statistics are not reported.

Table 13
Analyses on the role of income support policy in mitigating the adverse effect of the COVID-19 pandemic.
Variables Ln (Spread6m) Ln (Spread1y) Ln (Spread5y)

Panel A: Analyses on the role of income support policy

COVID19 0.246*** 0.196*** 0.110***
(5.729) (4.867) (4.138)
Income  COVID19
0.263***
0.250***
0.214***
(-5.468) (-5.608) (-6.692)
Relief  COVID19 0.030 0.021 0.079***
(0.692) (0.551) (2.847)
Panel B: Regression results of income support policy by group

Variables Low cover High cover Low cover High cover Low cover High cover

COVID19 0.204*** 0.015 0.139***
0.025 0.080***
0.024

(4.295) (0.287) (2.949) (-0.536) (2.769) (-0.653)

Notes: Ln (Spread6m), Ln (Spread1y) and Ln (Spread5y) are the logarithms of weekly spreads of 6-month, 1-year, and 5-year sovereign CDS, respectively. COVID19 equals
one for observations after the COVID-19 pandemic spreads to a country. Income and Relief range from 0 to 2, with higher value representing for more generous income
support and debt relief, respectively. See definitions of other variables in Table 1. Countries in the top quartile of income support are high coverage countries and others
are low coverage countries. t-statistics inferred by robust standard errors are reported in parentheses. ***, ** and * stand for the significance at the 1%, 5%, and 10%
levels, respectively. To save space, results of control variables and statistics are not reported.

consumer spending. Generally, the relief on income losses is greater, and 5. Conclusion
the consumption contraction is less. Thus, if consumption matters, the
marginal effect of consumption contraction on sovereign CDS spreads is This paper studies the effect of the COVID-19 pandemic on sovereign
weaker for countries with more generous coverage on income losses. credit risk measured by sovereign CDS spreads, and examines whether
Table 14 reports the regression results of two subsamples using Retail the COVID-19 pandemic affects sovereign credit risk through the con-
sales and Savings as proxies for the consumption change, respectively. sumption channel. Using data from forty developing and developed
Overall, the results in Table 14 show that income support for unemployed countries and constructing staggered DID models, we find that the
individuals better mitigates the adverse effect in countries with lower COVID-19 pandemic shock leads to a significant rise in sovereign credit
coverage of income support than in other countries. risk. Averagely, after the report of the first confirmed case, 6-month, 1-

11
X. Hao et al. Economic Modelling 109 (2022) 105794

Table 14
The regression results of two subsamples constructed based on Retail sales and Savings, respectively.
Variables Ln (Spread6m) Ln (Spread1y) Ln (Spread5y)

Low cover High cover Low cover High cover Low cover High cover

Panel A: The regression results of two subsamples using Retail sales as the proxy for the consumption change
COVID19 0.125** 0.067 0.070 0.020 0.026 0.018
(2.545) (1.259) (1.445) (0.413) (0.898) (0.495)
Retail sales  COVID19
0.006*
0.005
0.006**
0.003
0.009***
0.006**
(-1.901) (-0.647) (-2.106) (-0.418) (-4.817) (-2.170)
Retail sales 0.004***
0.009*** 0.004***
0.008***
0.000
0.004***
(5.175) (-8.788) (5.413) (-8.926) (-0.015) (-9.534)
Panel B: The regression results of two subsamples using Savings as the proxy for the consumption change

COVID19 0.367***
0.194* 0.298***
0.211** 0.237***
0.035

(4.776) (-1.874) (3.912) (-2.285) (4.281) (-0.636)
Savings  COVID19
0.006*** 0.009**
0.006*** 0.008**
0.006*** 0.001
(-2.933) (2.557) (-2.971) (2.465) (-3.874) (0.320)

Notes: Ln (Spread6m), Ln (Spread1y) and Ln (Spread5y) are the logarithms of weekly spreads of 6-month, 1-year, and 5-year sovereign CDS, respectively. COVID19 equals
one for observations after the COVID-19 pandemic spreads to a country. Retail sales is calculated by the monthly growth rate of social retail sales. Savings is the saving
rate in 2019 for each country. See definitions of other variables in Table 1. Countries in the top quartile of income support are high coverage countries and others are low
coverage countries. t-statistics inferred by robust standard errors are reported in parentheses. ***, ** and * stand for the significance at the 1%, 5%, and 10% levels,
respectively. To save space, results of control variables and statistics are not reported.

year and 5-year CDS spreads increase by 13.5%, 8.7% and 4.2%, Augustin, P., 2018. The term structure of CDS spreads and sovereign credit risk.
J. Monetary Econ. 90, 53–76.
respectively, presenting a descending trend over maturities of sovereign
Augustin, P., Sokolovski, V., Subrahmanyam, M.G., Tomio, D., 2021. In sickness and in
CDS contracts. Further analyses show that the sovereign credit risk rises debt: the COVID-19 impact on sovereign credit risk. J. Finan. Econ. Forthcoming.
more if a country suffers greater negative growth of retail sales during the https://doi.org/10.1016/j.jfineco.2021.05.009.
pandemic or has a lower aggregate savings rate before the pandemic Augustin, P., Tedongap, R., 2016. Real economic shocks and sovereign credit risk.
J. Financ. Quant. Anal. 51 (2), 541–587.
outbreak. These results show that the consumption contraction is Baker, S.R., Bloom, N., Davis, S.J., Kost, K., Sammon, M., Viratyosin, T., 2020a. The
important to explain the effect of the COVID-19 pandemic in addition to unprecedented stock market reaction to COVID-19. Rev. Asset. Pricing. Stud. 10,
the fiscal capacity and the volatility of exports. We also find that the 742–758.
Baker, S.R., Farrokhnia, R.A., Meyer, S., Pagel, M., Yannelis, C., 2020b. Income, liquidity,
adverse effect is less significant for countries with high coverage of in- and the consumption response to the 2020 economic stimulus payments. Available at:
come support for the unemployed, and hence add new evidence of the SSRN: https://ssrn.com/abstract¼3592174 https://doi.org/10.2139/ssrn.3587894.
effectiveness of consumption-stabilized fiscal stimuli. Bartik, A.W., Bertrand, M., Cullen, Z.B., Glaeser, E.L., Luca, M., Stanton, C.T., 2020. How
are small businesses adjusting to COVID-19? early evidence from a survey. Available
Our results demonstrate that the pandemic-induced macroeconomic at: SSRN: https://ssrn.com/abstract¼3574741 https://doi.org/10.2139/ssr
shock significantly damages a country's creditworthiness through the n.3574741.
consumption contraction. It implies that the aggregate consumption re- Beck, T., Levine, R., Levkov, A., 2010. Big bad banks? The winners and losers from bank
deregulation in the United States. J. Finance 65 (5), 1637–1667.
covery plays a crucial role in weathering the crisis associated with the Borusyak, K., Jaravel, X., Spiess, J., 2021. Revisiting Event Study Designs: Robust and
COVID-19 pandemic. Furthermore, the increase of sovereign CDS spreads Efficient Estimation arXiv preprint arXiv:2108.12419. https://arxiv.org/pdf/2108.
is more likely to push up the borrowing cost in the international financial 12419.pdf.
Callaway, B., Sant'Anna, P.H.C., 2021. Difference-in-differences with multiple time
markets, and thereby damage the financing capacity of a country. In
periods. J. Econom. 225, 200–230.
addition, the fiscal stimulus policy is a double-edged sword, and its Chen, H., Qian, W., Wen, Q., 2021. The impact of the COVID-19 pandemic on
impact on sovereign credit risk is complicated. Fiscal stimuli may in- consumption: learning from high-frequency transaction data. AEA Pap. Proc. 111,
crease sovereign credit risk by adding more debt burden, but those 307–311.
Chernov, M., Schmid, L., Schneider, A., 2020. A macrofinance view of U.S. sovereign CDS
related to income support for the unemployed may alleviate the adverse premiums. J. Finance 76 (5), 2809–2844.
effect of the COVID-19 pandemic because they accelerate the recovery of Chetty, R., Friedman, J.N., Hendren, N., Stepner, M., 2020. The economic impacts of
consumer spending. Overall, practitioners and policy makers should COVID-19: evidence from a new public database built using private sector data.
Available at: https://doi.org/10.3386/w27431.
realize the complexity of the effect of the COVID-19 pandemic when Cox, N., Ganong, P., Noel, P.J., Vavra, J.S., Wong, A., Farrell, D., Greig, F.E., Deadman, E.,
optimizing their decisions. 2020. Initial impacts of the pandemic on consumer behavior: evidence from linked
income, spending, and savings data. Brookings Pap. Econ. Activ. 35–69.
Daehler, T.B., Aizenman, J., Jinjarak, Y., 2021. Emerging markets sovereign CDS spreads
Declaration of conflicting interest during COVID-19: economics versus epidemiology news. Econ. Modell. 100, 105504.
Delis, M.D., Savva, C.S., Theodossiou, P., 2021. The impact of the coronavirus crisis on
The authors declare that they have no known competing financial the market price of risk. J. Financ. Stabil. 53, 100840.
Ding, W., Levine, R., Lin, C., Xie, W., 2021. Corporate immunity to the COVID-19
interests or personal relationships that could have appeared to influence pandemic. J. Financ. Econ. 141, 802–830.
the work reported in this paper. Fahlenbrach, R., Rageth, K., Stulz, R.M., 2021. How valuable is financial flexibility when
revenue stops? evidence from the COVID-19 crisis. Rev. Financ. Stud. 34 (11),
5474–5521.
References
Forsythe, E., Kahn, L.B., Lange, F., Wiczer, D., 2020. Labor demand in the time of COVID-
19: evidence from vacancy postings and UI claims. J. Publ. Econ. 189, 104238.
Aizenman, J., Hutchison, M., Jinjarak, Y., 2013. What is the risk of European sovereign Ftiti, Z., Ameur, H., Louhichi, W., 2021. Does non-fundamental news related to COVID-19
debt defaults? fiscal space, CDS spreads and market pricing of risk. J. Int. Money matter for stock returns? evidence from Shanghai stock market. Econ. Modell. 99,
Finance 34, 37–59. 105484.
Altig, D., Baker, S., Barrero, J.M., Bloom, N., Bunn, P., Chen, S., Davis, S.J., Leather, J., Goodman-Bacon, A., 2021. Difference-in-differences with variation in treatment timing.
Meyer, B., Mihaylov, E., Mizen, P., Parker, N., Renault, T., Smietanka, P., J. Econom. 225, 254–277.
Thwaites, G., 2020. Economic uncertainty before and during the COVID-19 Haddad, V., Moreira, A., Muir, T., 2021. When selling becomes viral: disruptions in debt
pandemic. J. Publ. Econ. 191, 104274. markets in the COVID-19 crisis and the fed's response. Rev. Financ. Stud. 34,
Andries, A.M., Ongena, S., Sprincean, N., 2021. The COVID-19 pandemic and sovereign 5309–5351.
bond risk. N. Am. J. Econ. Finance 58, 101527. Hasan, I., Marra, M., To, T., Wu, E., Zhang, G., 2021. COVID-19 pandemic and global
Andersen, A., Hansen, E.T., Johannesen, N., Sheridan, A., 2020. Consumer responses to corporate CDS spreads. Available at: SSRN: https://ssrn.com/abstract¼3858059
the COVID-19 crisis: evidence from bank account transaction data. Covid Econ. 7, https://doi.org/10.2139/ssrn.3858059.
92–118.

12
X. Hao et al. Economic Modelling 109 (2022) 105794

Hilscher, J., Nosbusch, Y., 2010. Determinants of sovereign risk: macroeconomic Liu, Y., Qiu, B., Wang, T., 2021b. Debt rollover risk, credit default swap spread and stock
fundamentals and the pricing of sovereign debt. Rev. Finance 14 (2), 235–262. returns: evidence from the COVID-19 crisis. J. Financ. Stabil. 53, 100855.
Jeanneret, A., 2018. Sovereign credit spreads under good/bad governance. J. Bank. Longstaff, F.A., Pan, J., Pedersen, L.H., Singleton, K.J., 2011. How sovereign is sovereign
Finance 93, 230–246. credit risk. Am. Econ. J. Macroecon. 3 (2), 75–103.
Kargar, M., Lester, B., Lindsay, D., Liu, S., Weill, P., Zuniga, D., 2021. Corporate bond O'Hara, M., Zhou, X.A., 2021. Anatomy of a liquidity crisis: corporate bonds in the
liquidity during the COVID-19 crisis. Rev. Financ. Stud. 34, 5352–5401. COVID-19 crisis. J. Financ. Econ. 142, 46–68.
Karger, E., Rajan, A., 2021. Heterogeneity in the marginal propensity to consume: Ortmans, A., Tripier, F., 2021. COVID-induced sovereign risk in the euro area: when did
evidence from COVID-19 stimulus payments. Available at: SSRN: https://ssr the ECB stop the spread? Eur. Econ. Rev. 137, 103809.
n.com/abstract¼3612828 https://doi.org/10.2139/ssrn.3612828. Pan, J., Singleton, K.J., 2008. Default and recovery implicit in the term structure of
Kubota, S., Onishi, K., Toyama, Y., 2021. Consumption responses to COVID-19 payments: sovereign CDS spreads. J. Finance 63 (5), 2345–2384.
evidence from a natural experiment and bank account data. J. Econ. Behav. Organ. Ramelli, S., Wagner, A.F., 2020. Feverish stock price reactions to COVID-19. Rev. Corp.
188, 1–17. Financ. Stud. 9, 622–655.
Liu, Q., Shen, Q., Li, Z., Chen, S., 2021a. Stimulating consumption at low budget: evidence Xue, F., Li, X., Zhang, T., Hu, N., 2021. Stock market reactions to the COVID-19 pandemic:
from a large-scale policy experiment amid the COVID-19 pandemic. Manag. Sci. the moderating role of corporate big data strategies based on Word2Vec. Pac. Basin
Forthcom. https://doi.org/10.1287/mnsc.2021.4119. Finance J. 68, 101608.

13