Sovereign debt responses to the COVID-19 pandemic

We utilize the global natural experiment created by the COVID-19 outbreak to identify sovereign borrowing capacity in time of need and its determinants. First, we demonstrate that the pandemic creates exogeneous shocks to sovereign borrowing needs—governments borrowed more when hit by more severe pandemic shocks. Second, we show that credible fiscal rules enhance sovereign borrowing capacity, while unsustainable debts in terms of high debt-to-GDP ratio, rollover risk, and sovereign default risk weaken it. Third, we find that, in response to the same pandemic shock, sovereign spreads increase more in emerging economies than advanced economies though the former borrow less during the pandemic. Finally, further analysis reveals that pegged exchange rate regimes, open capital accounts, and monetary dependence improve emerging economies' borrowing capacity.


Introduction
Governments' capacity to borrow in time of need is critical for pursuing optimal fiscal policies and avoiding sovereign debt crisis. To what extent governments can borrow what they need varies across countries, and uneven access to credit implies different fiscal policies and sovereign risks. To maintain fiscal and debt sustainability, it is important to understand what determines sovereign borrowing capacity in time of need (sovereign borrowing capacity hereafter). However, this question is hard to answer because potential determinants, such as fiscal rules that discipline public spending and borrowing in different stages of business cycles, and debt burdens that capture how much governments have borrowed, are mostly endogenous-they are derived from economic conditions that also affect sovereign borrowing capacity. It is also difficult to identify sovereign borrowing capacity, as the fiscal shock that triggers the funding need is often endogenous to sovereign borrowing capacity. Nevertheless, the COVID-19 outbreak worldwide has created exogenous shocks to public finance that facilitate the identification of sovereign debt burden in most economies, the increasing reliance on short-term debt financing, and the growing likelihood of sovereign default risk in the pandemic (Arellano et al., 2020). We find that governments with higher debt-to-GDP ratios, greater reliance on short-term debt financing, and default records after World War II (WWII) borrow less in response to pandemic shocks. This suggests that debt overhang, rollover risk and sovereign risk that undermine debt sustainability also hamper sovereign borrowing capacity in the pandemic.
We further show that the sovereign debt response to pandemic shocks is stronger in advanced economies than in emerging economies, suggesting that in difficult times, the former have higher sovereign borrowing capacity than the latter. How can emerging economies improve their borrowing capacity in crisis? Motivated by the growing evidence that global financial cycles are driving international lending (Rey, 2015), we delve into the policy trilemma to understand its impacts on sovereign debt capacity in the pandemic. Focusing on the subsample of bonds issued by emerging economies, we document evidence that pegged exchange rate regimes, open capital accounts, and surrender of monetary autonomy all enhance the sovereign debt response to pandemic shocks, indicating that these policy tools strengthen sovereign borrowing capacity.
One may be concerned that some economies exhibit greater sovereign borrowing capacity in the pandemic because they are willing to pay higher funding costs. To mitigate this concern, we explore the response of sovereign spreads to pandemic shocks. We obtain country-level sovereign spreads from Du and Schreger (2016), who compiled and constructed local-currency yield curves based on actively traded sovereign bonds. Our DID analysis shows that the sovereign spreads widen significantly in response to pandemic shocks, which is consistent with Gilchrist et al. (2022) that global financial risk affects international co-movements of sovereign risks. Further analysis reveals that this result is driven by emerging economies, which have difficulty borrowing more during the pandemic. This provides international evidence to support the theory of Arellano et al. (2020) that pandemic shocks increase sovereign default risks, which limit sovereign borrowing capacity. In advanced economies where governments are able to issue more bonds to finance unexpected fiscal shocks, we find no evidence that pandemic shocks increase sovereign spreads. This mitigates the concern that sovereign borrowing capacity in the pandemic is driven by rising sovereign spreads.
We conduct further analysis to understand whether the pandemic shock increases sovereign borrowing by triggering unprecedented fiscal expenditure and/or by cutting fiscal revenues. Governments that undertake policy actions to combat COVID-19 incur more public expenditure than governments that opt for laissez-faire responses. Our empirical analysis reveals that on average, governments adopting COVID-19-related policies on economic support, stay-at-home requirements, and public health information campaigns borrow 27-35% more than those that eschew such policy actions. This provides evidence that the pandemic shock increases sovereign borrowing by increasing public expenditure. However, we find that vaccination policy reduces the sovereign debt response to pandemic shocks, possibly because it increases the probability of returning to normality and lowers the expectation of future public spending.
Economies disrupted more by the pandemic suffer greater decline in fiscal revenues. To measure such a pandemic-driven shock, we take the differences in fiscal revenues, GDP growth, employment, and current account surplus for 2020 forecast by the IMF's World Economic Outlook (WEO) before and after the outbreak of COVID-19. The presumption is that without pandemic shocks, 2020's fiscal revenues forecast in October of 2019 and 2020 would have been broadly similar. We explore whether the pandemic shock increases sovereign borrowing by undermining fiscal revenue, but find no statistically significant evidence that economies contracted more severely in the pandemic borrow more. We further check whether a large fiscal surplus, current account balance, international reserve, and sovereign wealth fund may provide stronger buffers to pandemic-driven fiscal shocks and therefore reduce the need to borrow new funds during the pandemic. We find no statistical evidence to support such a conjecture.
Finally, we employ alternative measures of COVID-19 shocks to check the robustness of our key results, conduct spatial regression to account for spatial contagion of the coronavirus, illustrate the variations in sovereign borrowing capacity across regions and over time, deal with potential endogeneity issues related to COVID-19-induced policies with an instrumental variable (IV) approach, control for additional variables, and perform various falsification tests to support our findings.

Literature
This study is closely related to the literature on sovereign borrowing capacity, especially during difficult times. Governments' capacity to borrow from the debt market varies substantially with sovereign credit risk (Reinhart et al., 2003), global financial risk and monetary policy (Gilchrist et al., 2022;Uribe and Yue, 2006), business cycles (Ottonello and Perez, 2019), the interaction between political conflicts and institutional transparency (Pancrazi and Prosperi, 2020), levels of financial development (Rebelo et al., 2022), and the capitalization of domestic banks (Crosignani, 2021), among many other factors. Sovereign borrowing capacity is particularly relevant in time of need, when governments require urgent fiscal stimulus, or run into unexpected fiscal deficits, among many other scenarios. We utilize the exogeneous fiscal shocks created by the COVID-19 outbreak to identify sovereign borrowing capacity and its determinants. We add to this strand of the literature by uncovering new evidence that credible fiscal rules strengthen sovereign borrowing capacity, while threats to debt sustainability weaken it.
In time of need, sovereign borrowing capacity also interacts closely with sovereign credit risk. Public debt increases sovereign credit risk (Bianchi et al., 2019), which may lead to a self-fulfilling debt crisis that further tightens credit constraint. This is especially challenging after the COVID-19 outbreak, when governments need to trade off pandemic mitigation against fiscal distress. Arellano et al. (2020) model sovereign debt and default in the presence of epidemiological dynamics, focusing on emerging economies. They derive theoretically that policy actions to mitigate the pandemic increase sovereign default risks, which in turn limit sovereign borrowing capacity. Emilio et al. (2022) arrive at the same conclusion, based on a model with sovereign default and distortionary fiscal and monetary policies. Many other theoretical works also explore the interaction between sovereign debt and sovereign risk in different contexts (Aguiar and Amador, 2014;Aguiar and Gopinath, 2006;Arellano et al., 2019;Chatterjee and Eyigungor, 2012;Yue, 2010). We provide empirical support for these theories by documenting causal evidence in emerging economies that sovereign spreads widen in difficult times when sovereign borrowing is badly needed.
This paper also accords with the literature on policy actions during the COVID-19 pandemic, which increase the need for sovereign borrowing. Gourinchas et al. (2021) show that fiscal policy in the pandemic preserves employment in demand-constrained sectors and slightly offsets the global economic downturn. Guerrieri et al. (2021) derive theoretically that fiscal policy is less effective during the pandemic because some sectors are shut down, a finding endorsed by the empirical evidence of Auerbach et al. (2021). Alvarez et al. (2020) and Eichenbaum et al. (2021) consider the trade-off between lives and livelihoods to derive optimal policies. Regardless of their effectiveness, COVID-19-related policy actions are financially costly. How governments finance their deficits driven by these policies has important implications for future tax policy (Barro, 1974), inflation (Lustig et al., 2008), and fiscal austerity , which, in turn, shape household and business behaviors, thus affecting the speed and strength of economic recovery. A growing number of theoretical studies propose innovative ideas to finance the massive fiscal shocks from a sustainability perspective (Bianchi et al., 2020;Brunnermeier et al., 2022;Galí, 2020;Jiang et al., 2019;Reis, 2021). Our study complements this strand of the literature by first revealing that governments issue bonds in response to the pandemicdriven fiscal shocks, which enables us to identify sovereign borrowing capacity, and by showing how various policies in place affect the sovereign borrowing capacity. We also highlight that containment and health policies are quite costly, which increase sovereign borrowing needs significantly.
Finally, this study fits into the burgeoning literature that explores the economic impacts of COVID-19. Studies have shown that the pandemic crisis amplifies economic uncertainty (Altig et al., 2020;Baker et al., 2020a), reduces total factor productivity and economic growth Bonadio et al., 2021;Çakmaklı et al., 2020), changes household consumption and saving behaviors (Baker et al., 2020b;Cavallo, 2020;Chen et al., 2021;Coibion et al., 2020b;Cox et al., 2020;Goolsbee and Syverson, 2021), disrupts the labor market (Coibion et al., 2020a), and undermines corporate profits and stock returns (Carletti et al., 2020;Ding et al., 2021), among many other effects. Of particular relevance to our paper are the small number of studies that investigate the impact of the pandemic crisis on corporate access to the credit market. Acharya and Steffen (2020) show that firms draw down bank credit lines to raise cash in response to capital shortfall driven by the pandemic (Carletti et al., 2020), and only firms with the highest credit rating can borrow from debt or equity markets as the pandemic continues. Fahlenbrach et al. (2020) point out that access to credit improves firms' performance during the pandemic. However, it is difficult for businesses to raise funds, as the pandemic triggers capital outflows from the corporate bond market and reshuffles capital allocation, which is largely restored by the Federal Reserve's bond purchase program (Falato et al., 2021;Haddad et al., 2021;Nozawa and Qiu, 2021). Feyen et al. (2020), Kose et al. (2021), and World Bank (2021) warn of potential default and macroeconomics risks, while Bolton et al. (2020), Liu et al. (2020), and Truman (2020) call for policy actions to mitigate the risks. Governments face a similar need to access credit, and a similar difficulty in doing so, during the pandemic, but little is known about how they manage this challenge. Fuje et al. (2021) show that Debt Service Suspension Initiative reduces sovereign financing costs during the pandemic. We add to this strand of the literature by evaluating how sovereign debt financing responds to pandemic shocks, and further explore what determines governments' capacity to finance their policy actions to combat COVID-19.

Data
We describe the sources and definitions of key variables in this section and summarized them in Appendix Table 1.

Sovereign bonds
We collect data on sovereign bonds issued from January 1, 2019 to June 30, 2021 from Refinitiv. To enable a pre-and postpandemic comparison, we require an economy to be an active player in the sovereign bond market, issuing bonds in at least three of the 12 months in 2019. To link bond issuance with COVID-19 shocks, we also require an economy to have at least a six-month track record of COVID-19 cases in our sample period. This leaves us with a sample of 87 distinct economies issuing 21,391 bonds 1 over 30 months. Appendix Table 2 provides a list of these economies, divided into advanced and emerging economies.

Sovereign spreads
The sovereign spreads, which are the yields for sovereign bonds relative to the US treasury bonds of the same tenor, capture funding costs. We obtain bond-level yield to maturity upon issuance from Refinitiv and daily estimates of the US treasury yield curve from Gürkaynak et al. (2007), and take the difference between the two to obtain bond-level sovereign spreads. We only have such data for 210 bonds in our sample as data on yield upon issuance are very limited.
To overcome the small sample issue, we obtain economy-level sovereign spreads data from Du and Schreger (2016), which adjusts for the exchange rate risk and term structure of outstanding sovereign bonds denominated in their local currencies. Du and Schreger (2016) consolidated the economy-level zero-coupon local-currency yield curves from central banks, Bloomberg, and self-estimation based on actively traded sovereign bonds, and then calculated sovereign spreads as the difference between zero-coupon local-currency yields of an economy and the US, for tenors varying from three months to 10 years. We focus on local-currency sovereign spreads because the majority (89%) of bonds in our sample is denominated in local currency. From governments' perspective, their funding costs are best captured by yield curves and sovereign spreads associated with issuing bonds' denominated currency. For robustness checks, we also use sovereign spreads based on USD yield curves, which are essentially deviations from covered interest rate parity (CIP).
Based on the data construction in Du and Schreger (2016), local-currency sovereign spreads are the same for members of the European Union (EU) that use the same currency (Euro). Other than the EU members, there are 25 economies in our main sample that have data on economy-level sovereign spreads, which are highlighted in Appendix Table 2. Our sample for economy-level sovereign spreads is from January 1, 2019 to March 9, 2021, the latest update from Du and Schreger (2016) as the time of writing.

COVID-19 infections
Daily data on confirmed COVID-19 cases and tests in each economy are obtained from the Center for Systems Science and Engineering (CSSE) at Johns Hopkins University. Following Ding et al. (2021), we measure the pandemic shock, Covid i, t , by the growth of the COVID-19 infection rate, which is calculated as the logarithmic difference in the ratio of total confirmed cases to total tests in economy i between day t and t − 30. We also employ alternative measures of pandemic shocks, calculated respectively as the logarithm of one plus new COVID-19-related new cases, new deaths, hospitalized patients, and ICU patients, per million population.

COVID-19 policies
The data on whether governments adopted policies to combat the COVID-19 pandemic and to what extent are from the Oxford COVID-19 Government Response Tracker (OxCGRT). The variations in the timing of policy adoptions and terminations enable us to evaluate the role of various policies in driving sovereign debt responses to pandemic shocks through increased public spending. We are also able to diagnose whether differences in policy stringency across economies and over time matter, in addition to their relative presence. Our analysis focuses on income support to vulnerable households (economic policy), stay-at-home requirements (containment policy), public health information campaigns (healthy policy), and vaccination policy.
To tackle potential endogeneity in the COVID-19 policy, we construct two IVs based on the global distribution of COVID-19 policies. The first is Policy i,t−30 ROW , the average stringency of a particular policy (economic, containment, healthy, or vaccination) in the rest of the world excluding economy i in the past month. The second is GovernmentResponse i,t−30 ROW , the intensity of overall government policy actions in the rest of the world orthogonalized with respect to Policy i,t−30 ROW so as to capture the role of other policies. We argue in more detail later and test formally that the proposed IVs satisfy the relevance criteria as a result of global policy coordination and externality during the pandemic. They are also likely to satisfy the exogeneous criteria, as governments are unlikely to consider other countries' COVID-19 policies when making their borrowing decisions.
2.1.5. Economic shocks due to COVID-19 COVID-19 has triggered economic shocks such as business shutdowns, worker layoffs, and firm exits, which narrow the tax base. To uncover the pandemic-driven economic shocks, we compare macroeconomic indicators forecast by the IMF's WEO before and after the outbreak of COVID-19. If the forecasting method is consistent and the forecasting bias is persistent over time, the difference in the forecasts conducted in different periods of time regarding the same economic indicator in the same period should reflect the new information derived during the interval. For example, in the absence of new information, the WEO's forecasts in 2020 and 2019 about 2020's government revenue should be largely the same. Thus, the forecasting differences reflect new information arrived at between 2019 and 2020, which could serve as a proxy for the economic impact of the pandemic in our context. In particular, we calculate the government revenue loss in 2020 due to COVID-19 as the difference between government revenue for 2020 projected in October 2020, after COVID-19 hit the economy, and government revenue for 2020 projected in October 2019, when there was no COVID-19, normalized by the latter. As alternative measures of a contracted tax base, we also calculate the pandemic-driven shocks to GDP, unemployment, investment, and current account balance in a similar manner.

Pre-pandemic policies and economic conditions
To explore the determinants of sovereign borrowing capacity and its cross-sectional heterogeneity, we collect data on fiscal regimes, economic policies, and economic conditions that already existed before the pandemic.
2.1.6.1. Fiscal rules. We obtain the national fiscal rules from Weber et al. (2012), which record whether an economy has implemented any long-lasting constraints on fiscal policy through numerical limits on budgetary aggregates. We use the latest available data in 2015. The dataset covers debt rules, budget balance rules, expenditure rules, and revenue rules, which are applied to the central or general government or the public sector. The original dataset covers 96 economies, 59 of which overlaps with the 87 economies covered in our main sample.
2.1.6.2. Sovereign default. It takes a long time for an economy to graduate from serial defaults, which is the norm for emerging economies (Reinhart et al., 2003). Governments that defaulted before are more likely to default again, especially in bad times (Aguiar and Amador, 2014). We therefore use the sovereign default record following WWII and immediately prior to the outbreak of COVID-19, from Reinhart and Rogoff (2009), to extrapolate creditors' expectations of sovereign default risk in the pandemic.
2.1.6.3. Foreign exchange regime. Classifications of the foreign exchange rate regime are obtained from Ilzetzki et al. (2019). An economy is considered to have a fixed exchange rate system if its fine classification code in 2019 is below 9, which covers currency union, currency board arrangement, and de facto peg, with a crawling band not wider than ±2%.
2.1.6.4. Capital account openness. Our measure of capital account openness is from Chinn and Ito (2006). The normalized Chinn-Ito index ranges from 0 (closed) to 1 (open). We are interested in how capital mobility affects access to the credit market in difficult times, and we define KA Open as a dummy that equals 1 if the normalized Chinn-Ito index in 2019 is 1, which indicates a fully open capital account that allows for perfect capital mobility.
2.1.6.5. Debt structure. To understand whether high indebtedness before the pandemic limits governments' access to the credit market, we utilize data on central government debt from the World Bank's Quarterly Public Sector Debt (QPSD). This database also records short-term and long-term public debt, which can be further divided by creditors and currency denomination. This enables us to analyze how debt level, debt maturity, and the resulting rollover risk shape sovereign borrowing capacity during the pandemic.
2.1.6.6. Economic buffers. We obtain fiscal surplus, current account (CA) balance, and international reserve, all normalized by GDP, from Refinitiv. Sovereign wealth fund data is from Factset. We focus on data in 2019, before the outbreak of COVID-19, to evaluate how the pre-existing macroeconomic conditions may provide a buffer against pandemic shocks.
2.1.6.7. Spatial matrix. We apply two spatial matrices to capture the global transmissions of the coronavirus. The first is crossborder air travel passengers in 2010, which is estimated by Mao et al. (2015) based on detailed flight information. We normalize the bilateral passenger flows across borders by destination's total incoming passengers from the rest of the world. The bilateral trade flows in 2019, obtained from IMF, is handled in the same manner.

Summary statistics
To illustrate the fiscal shocks that occurred during the pandemic, Table 1 summarizes the mean and standard deviation of fiscal revenue, fiscal expenditure, and fiscal deficit, all normalized by GDP, across economies from 2019 to 2021. The 2021 data is projected by the WEO. The average fiscal revenue as a ratio of GDP in our sample dropped only slightly after the outbreak of COVID-19. The average fiscal expenditure as a ratio of GDP increased from 31% in 2019 to 36% in 2020, a 16% growth, and was projected to remain high in 2021 (34%). Stable fiscal revenue and rising fiscal expenditure contribute to widening fiscal deficit -the average fiscal deficit in our sample grew from around 3% before the COVID-19 pandemic to about 8% during the pandemic.
The adverse fiscal shocks triggered by the pandemic call for more debt finance. Fig. 1 shows that almost all economies in our sample accumulated more debt in 2020 after the outbreak of COVID-19, as their fiscal deficits widened. The fitted line exhibits an upward trend, providing preliminary evidence that governments borrow more to cope with pandemic-driven fiscal shocks. Fig. 2 plots the total amount of newly issued sovereign bonds and average COVID-19 infection rates for all economies in our sample from 2019 M1 to 2021 M6. We observe a spike in bond issuances immediately after the outbreak of COVID-19. Moreover, the increase in bond issuances closely tracks the surge in the COVID-19 shock, with a very short lag, which provides suggestive evidence that pandemic shocks increase sovereign debt finance.

Empirical analysis
In this section, we analyze what determines sovereign borrowing capacity. As a preliminary step, we estimate sovereign borrowing capacity by evaluating the response of sovereign debt to pandemic shocks. After verifying the COVID-19 shock as a good candidate for an exogeneous shock to the sovereign borrowing need, we proceed to explore how different fiscal regimes, debt sustainability indicators, international economic policies, and pandemic related policies enhance or limit sovereign borrowing capacity. We further investigate the possibility of leveraging bond characteristics to better access debt markets during the pandemic. Finally, we rule out the possibility that some economies exhibit higher sovereign borrowing capacity in the pandemic because they have higher sovereign spreads.

Sovereign debt response to pandemic shocks
The outbreak of the COVID-19 pandemic has created drastic exogeneous fiscal shocks. It significantly disrupts economic activities, which undermines the tax base, leading to a decline in public revenue. Despite falling revenue, governments have to boost public spending to protect lives and jobs, support businesses and investments, and stabilize the macroeconomy, among other factors. By reducing public revenue and increasing public spending, the pandemic has tightened government budget constraints. 2 Given the difficulty in raising public revenue by increasing the tax rate in the short term (Slemrod, 2018), governments have to finance the widened budget deficits with debt. Thus, we expect pandemic shocks that tighten the fiscal budget abruptly and unexpectedly to increase sovereign debt finance. In this section, we first elaborate the methodology of estimating sovereign debt response to pandemic shocks and then discuss the estimation results.

Estimating sovereign debt response
To evaluate the sovereign borrowing capacity in time of need-additional funds that governments can borrow in difficult times relative to normal times, we need a counterfactual of how much governments could have borrowed in the absence of COVID-19. The sudden outbreak of COVID-19 provides a global natural experiment that enables us to identify sovereign borrowing capacity using a DID approach. Given the staggered outbreaks of COVID-19 globally and the heterogeneous exposures to COVID-19 across economies, we employ a staggered DID with continuous treatments. Using continuous treatments also allows us to compare sovereign borrowing capacity across economies, as we are capturing the sovereign debt response to every unit of pandemic shock instead of the total pandemic shock.
3.1.1.1. Economy-level analysis. We first evaluate how aggregate sovereign bonds respond to pandemic shocks using the following DID specification 3 : where y i,m is the logarithm of the total amount of sovereign bonds newly issued by economy i in month m, and f i and f m are economy and time fixed effects, respectively. The dependent variable y i,m could also be the logarithm of either the total number of bonds or the average size of bonds, as well as sovereign spreads. The key explanatory variable Covid i,m−1 is a state-dependent variable, which takes the value of 0 before the COVID-19 outbreak and captures the pandemic shock in economy i in month m after the COVID-19 outbreak.
We essentially compare the difference in sovereign borrowing between economies facing relatively large and relatively small pandemic shocks after the outbreak of COVID-19, with their pre-pandemic levels. The coefficient of interest is β, which captures the response of sovereign debt to a unit of pandemic shock. Note that Eq. (1) is essentially a reduced-form regression-the potentially endogenous fiscal shock is a function of the pandemic shock that enter the regression directly. 4 Given the simultaneous increase in funding need and the historically high levels of sovereign debt, not all governments can borrow what they need during the pandemic crisis. Thus, the change in sovereign borrowing in response to pandemic shocks should reflect borrowing capacity rather than borrowing need. In this case, a more positive β indicates greater sovereign borrowing capacity.
3.1.1.2. Bond-level analysis. Governments may borrow more through issuing more bonds or expanding the size of bonds issued. In the latter case, we can delve into bond-level data to explore the sovereign debt response to pandemic shocks, as larger bonds correspond to more sovereign borrowing. The specification based on bond-level data is: where the dependent variable is the logarithm of Bond i,t k , the dollar amount of sovereign bond k newly issued by economy i on day t, and Covid i,t−1 is the pandemic shock measured by the monthly growth in infection rate. The economy-time fixed effects f i,t m absorb both static and monthly-varying economy-specific factors that may affect sovereign bond issuances, such as institution quality, economic growth, and monetary policy, among other potential confounding factors. The error term is captured by ε i,t . Each bond k is a unique observation that reflects the decision of the governments in economy i on day t. Viewing Eq.
(2) from an economy's perspective, we are essentially comparing the difference in the average bond size between economies with relatively high and low exposures to the COVID-19 shock after the pandemic, relative to pre-pandemic levels. A positive estimated γ means larger bond size in response to pandemic shocks. This indicates greater sovereign borrowing if the number of bonds issued after the pandemic is the same as before the pandemic. We show later that rising sovereign borrowing during the pandemic is driven by expanding the size rather than increasing the number of sovereign bonds; this motivates us to focus on Eq.
(2), which allows us to control for a comprehensive list of confounding factors. 2 We provide formal evidence in Section 3.3 and 4.1 that the pandemic shock increases fiscal shocks by enlarging public spending more than by contracting economic activities. 3 See Abadie (2005) for a similar specification. Such a generalized DID specification with continuous treatments may complicate the interpretation of two-way fixed effects (TWFE) regression estimators because treatment effect (the outcome difference between treated and untreated units) and causal response (the change in outcome in response to an incremental change in treatment intensity) could be different (Callaway et al., 2021). We mitigate this concern by showing that the marginal sovereign debt response to pandemic shocks does not change significantly with the size of pandemic shocks (see Appendix Table 4). We further argue that the strong parallel trends assumption, under which there is no "selection bias" and TWFE estimators can be interpreted as causal response, is likely to be satisfied as pandemic shocks are largely unexpected and so are the fiscal shocks driven by them and the associated sovereign borrowing needs. 4 We can apply reduced-form regression because COVID-19 shock is a good instrumental variable for fiscal shock: it clearly satisfies the relevance criteria as the pandemic has tightened the fiscal budget. It is also likely to satisfy the exogeneous criteria as the pandemic has been largely out of control-otherwise it would not last so long and so pervasively.
H. Zheng Journal of International Economics 143 (2023) 103766 3.1.2. Baseline result on sovereign debt response Table 2 presents estimation result for how sovereign bond issuances respond to every unit of pandemic shock, based on Eqs.
Columns 1-3 show that pandemic shocks increase the aggregate sovereign borrowing (column 1), which is not via raising the number of bond issuances (extensive margin, column 2), 5 but by expanding the average size of bond per issuance (intensive margin, column 3). The result implies that we may extrapolate the trend of aggregate sovereign borrowing from that of average bond size-larger bonds, on average, correspond to more aggregate borrowing. Note that our dependent variable is in logarithm so that we are comparing bond size and aggregate borrowing in relative but not absolute terms, which increases comparability across countries.
Focusing on issuance-level bond data, we confirm in column 4 that governments issue larger bonds in response to pandemic shocks. Our estimation result indicates that one standard-deviation increase in the pandemic shock (41%) enlarges the average sovereign bond size by 6% (= 41% (change in pandemic shock) × 0.145 (coefficient of pandemic shock)) relative to the pre-pandemic periods. This is equivalent to raising an additional $180 million from each bond issuance, given an average bond size of $3 trillion in our sample. Such an increment in bond size is translated into growth in aggregate sovereign borrowing, given that the number of sovereign bond issuances does not change in response to pandemic shocks. Given an average borrowing of $29 trillion per economy per month in our sample, our estimation means that an average economy borrows $1.7 trillion more in response to one standard-deviation increase in the pandemic shock.
Although we have controlled for economy and time fixed effects throughout columns 1-4, one may still be concerned about potential time-varying confounding factors that are correlated with both pandemic shocks and sovereign bond size, leading to a spurious relation between the two variables. To address this concern, we control for economy-time fixed effects to absorb economy-specific and time-varying factors, either observable or unobservable. We show in column 5 that our baseline result remains robust. The coefficient of the pandemic shock in column 5 is quantitatively similar to that in column 4, thus mitigating the concern that economy-level time-varying factors may be driving our baseline result. To deal with issues surrounding zero observations and heteroskedasticity, we follow Silva and Tenreyro (2006) in applying the Poisson pseudo-maximum likelihood (PPML) estimation technique. The result in column 6 of Table 2 delivers the same message that pandemic shocks enlarge the sovereign bond. We perform a comprehensive list of checks in Section 4.5 to confirm the robustness of our result.
To summarize, pandemic shocks enlarge sovereign debt by increasing the size but not the number of bond issuances. From here onwards, we focus on transaction-level issuance data to explore what drives the sovereign debt response to pandemic shocks, and we interpret larger bonds as more sovereign borrowing.

What determines sovereign borrowing capacity?
The pandemic triggers simultaneous increases in sovereign debt finance worldwide, yet not all governments can borrow what they need. Some governments may be excluded from the bond market due to rising default risk (Arellano et al., 2020;  Heterogeneity robust standard error clustered by economy is reported in parentheses. ***, **, and * denote the significance level at 1%, 5%, and 10%, respectively. 5 We follow the estimation procedure proposed by Cohn et al. (2022) to tackle with count data, and document similar evidence that the number of bonds does not respond significantly to pandemic shocks (not reported).

H. Zheng
Journal of International Economics 143 (2023) 103766 Bianchi et al., 2019) or falling fiscal and debt sustainability (Aguiar et al., 2009;Reinhart et al., 2012). This is especially the case during the pandemic, when governments all over the world are competing for funds while risk-averse investors are seeking safety. Thus, the sovereign debt response to a unit of pandemic shock, which captures the additional borrowing during the pandemic relative to normal times, well reflects sovereign borrowing capacity. The more governments can borrow from the debt market in response to the same pandemic shock, the greater their borrowing capacity in time of need. We are interested in what affects governments' capacity to borrow from the bond market when facing significant fiscal shocks. This is important for understanding the feasibility of coping with an emergency situation using fiscal policy, which is optimal when conducted countercyclically (Frankel et al., 2013;Ramey, 2019). Given that most governments were already running substantial budget deficits and had historically high debt burdens before the pandemic, without the capacity to borrow new funds during the pandemic, they are constrained from pursuing expansive fiscal policy to save the economy and support the vulnerable, and some may even default.
Despite its importance, what determines sovereign borrowing capacity is difficult to pinpoint, as budgetary and economic policies are mainly endogenous to economic conditions. The interaction between sovereign borrowing capacity and its determinants further complicates the causal identification: governments-that are not rule-based and/or can easily dilute their debt burden that hurt bond holders-are less likely to secure funding in time of distress, which leads to policy adjustments so as to better access credit next time.
The COVID-19 outbreak facilitates resolution of these endogeneity issues. It triggers a drastic and immediate fiscal shock that calls for urgent sovereign debt finance while leaving little time for governments to improve sovereign borrowing capacity overnight. After all, it takes a long time to build up sovereign borrowing capacity that can be drawn upon in time of need-governments need to earn credibility through repeatedly fulfilling their commitments, even in difficult times. Most policy twists and commitments need some time to translate into sovereign borrowing capacity. It would be too late to change these determinants to increase sovereign borrowing capacity in pandemic after the COVID-19 outbreak has created fiscal shocks. Therefore, key determinants of sovereign borrowing capacity in the pandemic should have existed before the pandemic, which addresses the reverse causality concern that sovereign borrowing capacity in the pandemic affects its determinants. The effects of economic conditions underlying sovereign borrowing capacity and its determinants are either net out in our DID regressions, mitigating the concern over omitted variables bias. All else being equal, in response to the same pandemic shock, economies with a positive determinant of sovereign borrowing capacity should be able to borrow more than their peers without such a determinant. That is, determinants of sovereign borrowing capacity should distinguish sovereign debt response to pandemic shocks in economies with and without such determinants.

Estimating determinants of sovereign borrowing capacity
As previously noted, it takes a long time to build up sovereign borrowing capacity that can be drawn upon in time of need. The sudden outbreak of COVID-19 generates immediate fiscal shocks that call for urgent funding, yet it leaves no time for governments to enhance their borrowing capacity, which cannot be done overnight. This provides a unique scenario to explore how various fiscal regimes and economic policies in place affect sovereign borrowing capacity. All else being equal, economies with positive (negative) determinants of sovereign borrowing capacity should respond more (less) positively to pandemic shocks. Utilizing the cross-country variations in various fiscal regimes and economic policies before the COVID-19 outbreak, we can explore how these factors affect sovereign borrowing capacity by estimating the following model: where Determinant i is a potential determinant of sovereign borrowing capacity, which could be an indicator of fiscal rules, debt sustainability, or international economic policy in economy i that existed before the pandemic. The parameter of interest is δ, which captures the impact of Determinant i on sovereign borrowing capacity. If the proposed determinant strengthens (weakens) sovereign borrowing capacity, the estimated δ should be positive (negative) and statistically significant.
In what follows, we focus on fiscal rules, debt sustainability indicators, and international economic policies to explore whether and how they affect sovereign borrowing capacity.

Fiscal rules
A growing number of economies have adopted fiscal rules to contain the pressures of overspending and ensure fiscal responsibility and debt sustainability. These rules set numerical targets on budgetary aggregates for a large share of public finances, covering the central government level at a minimum, and are either fixed in legislation or binding for at least three years (Weber et al., 2012). Feld et al. (2017) and the IMF (2021) document evidence that strong and credible fiscal rules lower the cost of public finance, while Heinemann et al. (2014) highlight the potential of fiscal rules to restore market confidence in historically unstable economies. These studies point to the role of fiscal rules in signaling to investors the governments' commitment to fiscal responsibility and debt sustainability, which may mitigate concerns over sovereign default risk and improve public access to credit. However, it is also possible that rigid fiscal rules prevent governments from borrowing too much, especially when the pre-pandemic fiscal space was already limited. We are interested in how different types of fiscal rules affect sovereign borrowing capacity in times of unexpected fiscal shocks.
3.2.2.1. Debt rules. The debt rule sets an explicit upper limit for the debt-to-GDP ratio. For example, Poland established in its Constitution and Public Finance Act that general government debt cannot exceed 60% of GDP. Out of the 59 economies that have fiscal-rule data in our sample, 42 (71%) had adopted debt rules before the pandemic. 6 Note that our measure of debt rule is de jure; it is important to have an accountable enforcement mechanism to ensure compliance to such a rule. We therefore define the dummy variable for credible debt rules, DR, as 1 if the economy of interest had adopted a debt rule that had a nationwide formal enforcement procedure, and 0 otherwise. We observe that DR = 1 for only six economies (10%) in our sample (see Appendix Table 3 for the list of these economies).
Column 1 of Table 3 shows that governments with credible debt rules (DR = 1) borrow more in the pandemic than those without credible debt rules (DR = 0). In response to one standard-deviation increase in pandemic shock (41%), governments with credible debt rules enlarge their borrowing by 50%, while governments without such rules borrow only 7% more than their pre-pandemic levels. The difference in borrowing capacity between governments with and without credible debt rules (43%) is statistically significant at the 5% level. This provides evidence that adherence to debt rules pays off in difficult times, by granting governments higher borrowing capacity amid global competition for credit.
Note that the average COVID-19 shock for economies with credible debt rules is 0.11%, which is higher than for those without creditable debt rules (0.03%). One may be concerned that the nonlinear response of sovereign debt to pandemic shocks is driving our result. To mitigate this concern, we test for potential nonlinearity by comparing sovereign debt responses to each unit of pandemic shock conditional on the scale of pandemic shocks, classifying into different groups based on different percentiles. We find no statistical evidence in Appendix Table 4 that the sovereign debt response to each unit of pandemic shock varies with the scale of pandemic shocks. Thus, the effect of credible debt rules on strengthening sovereign borrowing capacity in the pandemic is unlikely to be driven by the scale of pandemic shocks.
3.2.2.2. Budget balance rules. The budget balance rule constrains governments from spending excessively beyond their fiscal affordability, considering different stages of business cycles. It disciplines governments in terms of fiscal responsibility by ensuring that during economic expansions, high fiscal revenues are saved rather than spent so as to provide a fiscal buffer during economic recession. For example, in Singapore, the Constitution requires that the government has a balanced budget over its term of office. In our sample, 50 out of 59 economies (85%) adopted budget balance rules before the pandemic. 7 Studies have shown that the presence of a budget balance rule has a positive effect on the economy, leading to a gradual reduction of public debt (Azzimonti et al., 2016). However, implementing such a rule entails complications, leaving room for arbitrage and undermining its effect. As budget balancing is often structural and cyclically adjusted by policy makers (see Chile, for example), this rule may be twisted to justify government overspending. To enhance the credibility of the budget balance rule, some economies, such as Chile and Sweden, have introduced an independent body to set the budget assumptions. We define The fiscal rule dummy variables DR, BBR, RR, and ER equal 1, respectively, if the economy of interest adopted a credible debt rule (DR) that specifies the limit of debt-to-GDP ratio with a formal enforcement procedure, a credible budget balance rule (BBR) that limits the fiscal deficit cyclically monitored by an independent body, a revenue rule (RR) that sets ceilings or floors of government revenues, and an expenditure rule (ER) that limits the size or growth of public spending in 2015. All regressions control for economy-time fixed effects (FE). Heterogeneity robust standard error clustered by economy is reported in parentheses. ***, **, and * denote the significance level at 1%, 5%, and 10%, respectively.
BBR, a dummy variable for a creditable budget balance rule, as 1 if the economy of interest had enforced a budget balance rule through an independent body. Only five economies (8% in our sample)-Belgium, the Netherlands, the United Kingdom, Chile, and Singapore-satisfy these criteria.
Column 2 of Table 3 shows that credible budget balance rules strengthen governments' capacity to raise new funds in the pandemic-the coefficient of the interaction between Covid and BBR is positive and statistically significant. In response to one standard-deviation increase in the pandemic shock (41%), governments with credible budget balance rules (BBR = 1) increase their sovereign borrowing by 29% more than the pre-pandemic level, which is about four times that of governments without credible budget balance rules (BBR = 0). This result provides evidence that credible budget balance rules, which strengthen the credibility of public finance, improve public access to credit in emergency situations.
3.2.2.3. Revenue and expenditure rules. Other than setting the limits of public debt and fiscal deficit, countries can also specify the targets of their revenue and expenditure separately. Revenue rules either impose upper limits on revenue to avoid an excessive tax burden for firms and households or set lower limits on revenue to boost tax collection. Expenditure rules set upper limits on the growth or size of public spending in different categories. In our sample, out of the 59 economies with fiscal rule data, 9 (15%) had adopted revenue rules and 33 (56%) had adopted expenditure rules. 8 Let RR (ER) be a dummy variable that equals 1 if the economy of interest had adopted a revenue rule (an expenditure rule), and 0 otherwise (see Appendix Table 3 for the list of economies with RR = 1 and ER = 1). We show in column 3 of Table 3 that revenue rules enhance sovereign borrowing capacity in the pandemic. However, we find in column 4 of Table 3 that expenditure rules reduce sovereign borrowing capacity in the pandemic. Nonetheless, this is not statistically significant. This identifies the expenditure rule as a double-edge sword-on the one hand, it increases fiscal credibility, which increases sovereign borrowing capacity, on the other hand, it constrains governments from spending beyond pre-specified limits, which limits sovereign borrowing.
3.2.2.4. Summary of fiscal rules. Overall, our analysis reveals that the credible debt rule, the budget balance rule, and the revenue rule all strengthen sovereign borrowing capacity, while the expenditure rule appears to limit governments' borrowing capacity in emergency situations. Note that some economies simultaneously adopted multiple fiscal rules, so a concern may arise over whether the impact of a particular fiscal rule is driven by its concurrence with other rules. Therefore, we pool all rules in one regression and show in column 5 of Table 3 that the results remain robust, and that most fiscal rules enhance sovereign debt capacity in time of need.

Debt sustainability
Most economies had accumulated a significant debt burden before the outbreak of COVID-19, which caused various concerns over debt sustainability, depending on debt structure and credit risk. In the following, we seek to understand how debt sustainability, in terms of debt overhang, rollover risk, and sovereign default, affects sovereign borrowing capacity in the pandemic.
3.2.3.1. Debt overhang. Debt overhang arises when existing debt is sufficiently large that creditors expect to be repaid only partially (Krugman, 1988). Sovereign debt overhang constrains governments from raising additional funds to finance new investments and other expansionary fiscal programs, especially in recessions, even if these would be beneficial to economic growth (Aguiar et al., 2009;Reinhart et al., 2012). Indeed, countries with high debt-to-GDP ratios face relatively large and long-lasting output losses after a crisis Romer, 2010, 2019). The debt overhang theory implies that economies already heavily in debt before COVID-19 would have difficulty accessing new credit during the pandemic, which may weaken the sovereign debt response to pandemic shocks. Reinhart et al. (2012) define major public debt overhang in advanced markets as gross public debt exceeding 90% of nominal GDP for at least five years. Aguiar et al. (2009) and Lamont (1995) show that debt overhang is more likely to undermine macroeconomic performance during economic downturns. Guided by these studies, we create a dummy variable Overhang, which equals 1 for economies experiencing debt overhang, with gross central government debt-to-GDP above 90% and GDP per capita declined in 2019, 9 and 0 otherwise. Appendix Table 3 provides the list of economies with Overhang = 1.
To test whether debt overhang reduces sovereign borrowing capacity, we compare economies with and without debt overhang in their debt responses to pandemic shocks, by estimating Eq. (3). Column 1 of Table 4 shows that economies experiencing debt overhang borrow 66% less than their peers in response to the same pandemic shock. This supports our hypothesis that debt overhang undermines sovereign borrowing capacity during the pandemic, which also echoes the theoretical prediction in Janeba and Todtenhaupt (2018).
3.2.3.2. Rollover risk. Short-term debt is less costly but exposes borrowers to rollover risk, especially during crises, when creditors are less willing to renew matured debt because of rising risk aversion, falling collateral values, self-fulfilling prophecy, etc. (Acharya et al., 2011;Cole and Kehoe, 2000;Krishnamurthy, 2010). He and Xiong (2012) prove theoretically that short-term debt exacerbates rollover risk, which is endorsed empirically by Kalemli-Ozcan et al. (2018). We therefore conjecture that governments have lower borrowing capacity in the pandemic if they borrowed more in short-term debt and therefore facing higher rollover risk before the pandemic.
To test whether rollover risk undermines sovereign borrowing capacity in the pandemic, we define Rollover as a dummy that equals 1 if the short-term central government debt as a ratio of GDP is in the top 25% percentile of our sample, and then estimate Eq. (3). Appendix Table 3 provides the list of economies with Rollover = 1. Comparing economies with high and low rollover risk, we find in column 2 of Table 4 that economies exposed to higher rollover risk borrow 66% less than their peers in response to the same pandemic shock. This provides evidence that sovereign borrowing capacity is weaker in economies facing higher rollover risk.

Sovereign default.
Investors prioritize safety over return during crises, as they become more risk averse (see, for example, Zheng, 2020). As a result, economies with a higher probability of sovereign default are less competitive in raising funds during the pandemic crisis. Highly risky borrowers may raise bond yields to a sufficiently high level or seek other more expensive sources of financing to access new funds. However, such additional borrowing may increase sovereign default risk, which undermines the effectiveness of fiscal policy in boosting economic outputs (Bianchi et al., 2019). This, in turn, may threaten debt sustainability, which further limits sovereign debt capacity. Arellano et al. (2020) reached a similar conclusion about sovereign debt and sovereign borrowing in the presence of epidemiological dynamics.
Economies that have defaulted historically are more likely to default again (Reinhart and Rogoff, 2009). According to Reinhart et al. (2003), based on a country's history of debt default and real repayment on internal obligations, one may extrapolate the sovereign borrowing capacity. We therefore expect economies with sovereign default records in their recent history to have lower sovereign borrowing capacity during the pandemic.
Based on the historical crisis data from Reinhart and Rogoff (2009), we define Default as a dummy that equals 1 if the economy under consideration has any sovereign default records after WWII and 0 otherwise (see Appendix Table 3 for the list of economies with Default = 1). Comparing economies with and without post-WWII defaults in their sovereign debt responses to pandemic shocks, we find in column 3 of Table 4 that economies with default records since WWII borrow 30% less than their peers. This provides evidence that historical sovereign defaults weaken sovereign borrowing capacity in the pandemic. Furthermore, there is no evidence that economies that have defaulted after WWII are able to borrow more in the pandemic-the coefficient sum of Covid and Covid ×Default is −0.05, which is not statistically significant (F statistics = 0.17, p-value = 0.68). This suggests that economies that have not undergone any sovereign default since WWII are driving the positive sovereign debt response to pandemic shocks.

Summary of debt sustainability.
To summarize, we find that debt overhang, rollover risk, and sovereign default all weaken sovereign borrowing capacity in the pandemic. The roles of debt sustainability indicators generally remain robust when we pool them together (see columns 4 and 5 of Table 4). Our results highlight the importance of enhancing sovereign debt sustainability to improve access to credit markets in time of need.

Advanced versus emerging economies
Emerging economies typically have higher sovereign credit risk, which limits their access to the debt market, especially during crises. Indeed, Arellano et al. (2020) show that pandemic shocks increase the sovereign default risk in emerging economies, which further limits their sovereign borrowing capacity. To investigate whether advanced and emerging economies exhibit different Overhang and Rollover are dummy variables that equal 1, respectively, if the debt-to-GDP ratio exceeds 90% and the shortterm debt as a ratio of GDP is in the top 25th percentile, conditional on a declined GDP per capita in 2019. Default is a dummy that equals 1 for economies that have defaulted since WWII. All regressions control for economy-time fixed effects (FE). Heterogeneity robust standard error clustered by economy is reported in parentheses. ***, **, and * denote the significance level at 1%, 5%, and 10%, respectively. sovereign borrowing capacity in the pandemic, we estimate Eq. (2) for the two groups separately. The results in columns 1 and 2 of Table 5 show that sovereign debt finance increases in response to pandemic shocks in advanced economies but not in emerging economies. The coefficient of pandemic shocks based on the subsample of emerging economies, though positive, is economically small and statistically insignificant. Given that most emerging economies have defaulted after WWII in one way or another, this finding is consistent with our previous finding in Section 3.2.3 that economies with historical sovereign default have difficulty borrowing during the pandemic.

International macroeconomic policy
If sovereign default is the key that prevents emerging economies from borrowing in time of need, then international macroeconomic policies that are closely related to sovereign risk potentially could help. Sovereign defaults and currency devaluations often come hand in hand (Na et al., 2018;Reinhart, 2002). Pegged exchange rate policy signals governments' commitment to stabilize currency valuation and avoid default. Free international capital mobility allows emerging economies to access a broader credit market while exposing them to serial defaults (Reinhart and Rogoff, 2009). Monetary autonomy enables governments to use monetary policy tools to stimulate the domestic economy, support local currency, and even manipulate the debt burden to reduce sovereign default risk (Bianchi and Mondragon, 2018). We explore whether these international macroeconomic policies affect emerging economies' borrowing capacity during the pandemic.
Column 3 of Table 5 shows that a de facto pegged exchange rate regime significantly improves sovereign borrowing capacity in the pandemic for emerging economies-the coefficient of the interaction between Covid and Peg, a dummy that equals 1 for economies adopting a de facto pegged exchange rate system, is positive and statistically significant. The coefficient of Covid is not statistically significant, which suggests that emerging economies with flexible exchange rate regimes do not increase borrowing significantly during the pandemic. We show in column 4 of Table 5 that the coefficient of the interaction between Covid and KA open, a dummy that equals 1 for economies with open capital accounts, is positive but not statistically significant. Thus, there is no evidence that an open capital account affects sovereign borrowing capacity during the pandemic. According to the trilemma, an economy loses its monetary autonomy if it has a pegged exchange regime and an open capital account. The results in column 5 of Table 5 suggest that the monetary dependence on a leading economy improves sovereign borrowing capacity in the pandemic. This could be due to the presence of credible monetary policy in these emerging economies, which is essentially determined by major central banks such as the Fed and the European Central Bank. Note that major central banks have cut interest rates aggressively and swiftly during the pandemic, and the accommodative monetary policy carried over to emerging economies may also improve their access to the credit market. 10

How do costly COVID policies affect sovereign debt finance?
Governments had adopted various policies to combat the pandemic and mitigate its negative effects. These policies require significant public resources, which are economically costly regardless of their efficacy. All else being equal, governments that adopt COVID-19-related policies face greater budget constraints than governments that eschew these policies. However, such effects may be offset or even reversed if the policies are deemed effective in improving economic and fiscal outlooks. . Heterogeneity robust standard error clustered by economy is reported in parentheses. ***, **, and * denote the significance level at 1%, 5%, and 10%, respectively. 10 We find no evidence that these international macroeconomic policies matter for advance economies' debt response to pandemic shocks (not reported).
To explore how COVID-19-related policies affect the impact of pandemic shocks on sovereign debt finance, we compare economies with and without a specific COVID-19 policy based on Eq. (3). In particular, we replace Determinant i in Eq. (3) with timevarying COVID-19-related policy indicator, Policy i,t , a dummy variable that equals 1 respectively if governments have adopted economic policy, containment policy, health policy, and vaccination policy to combat the pandemic in economy i at time t, and 0 otherwise. We report the estimation results in Table 6 and discuss their roles in the following.

Economic policy
To support the economy, some governments offer cash transfer and debt contract reliefs to households. Theoretically, Gourinchas et al. (2021) show that the fiscal stimulus offsets the pandemic-driven economic downturn, while Guerrieri et al. (2021) argue that lockdown policies reduce the effectiveness of fiscal policy. Auerbach et al. (2021) provides empirical support to these predictions showing that fiscal multiplier is generally low, especially when subject to strong stay-at-home orders. Indeed, Coibion et al. (2020b) show that only 40% of the fiscal transfer is (planned to) turned into consumption. We therefore conjecture that the fiscal stimulus' impact on improving economic outlook, if any, is not as pronounced as its impact on enlarging fiscal shocks. In this case, the related economic policy should strengthen the sovereign debt response to pandemic shocks.
The coefficient of the interaction term (Covid × Policy) in column 1 of Table 6 is positive and statistically significant, which provides evidence that financial support to households strengthens sovereign debt response to pandemic shocks. In particular, in response to a standard-deviation increase in the pandemic shock (41%), governments that provide income support borrow 11% (= 41% (change in pandemic shock) × 0.272 (coefficient of Covid × Policy)) more than those not pursuing such a policy action. The coefficient of Covid is economically small and statistically insignificant, suggesting that the pandemic shock has little impact on sovereign debt finance in economies without such income support. This implies that the impact of pandemic shocks on increasing sovereign debt finance is concentrated upon economies that use the fiscal budget to protect households from pandemic-driven income shocks. However, we find no evidence in Panel A of Appendix Table 5 that other economic policies to combat COVID-19 affect the impact of pandemic shocks on sovereign debt finance.

Containment policy
To curb the spread of the disease, many governments had closed borders, suspended businesses, and/or ordered citizens to stay at home. These containment policies require substantial funding and manpower. Even when conditional on their effectiveness in saving lives, optimal containment policy exacerbates recessions (Alvarez et al., 2021;Eichenbaum et al., 2021) and undermines fiscal multiplier (Guerrieri et al., 2021;Auerbach et al., 2021). Thus, we expect containment policies to increase fiscal spending and reduce prospective fiscal revenues, which then enhance sovereign debt response to pandemic shocks.
Column 2 of Table 6 shows that the stay-at-home policy, which recommends or requires people to stay at home, enhances the sovereign debt response to pandemic shocks. In response to a standard-deviation increase in the pandemic shock (41%), governments that implement a stay-at-home policy increase their borrowing from the bond market by 13% more than their peers. We document similar evidence in Panel B of Appendix Table 5, indicating that other containment policies, such as restricting international travel and internal movement within a country, cancelling public events, and closing schools and workplaces, heighten the impact of pandemic shocks on sovereign debt finance.

Health and vaccination policies
To manage the health crisis, governments urge citizens to caution the coronavirus, wear mask, and get vaccinated, among others. These health policies increase public expenditure on the one hand, and improve the economic and fiscal outlooks on the other hand. For example, fast vaccine rollouts have reduced the infection rate and accelerated border reopening in many Policy is a dummy variable that equals 1 in columns 1-4, respectively, if governments provided income support to households, imposed stay-at-home requirements, launched public information campaigns, and adopted a vaccination policy to boost immunity to COVID-19. All regressions control for economy-time fixed effects (FE). Heterogeneity robust standard error clustered by economy is reported in parentheses. ***, **, and * denote the significance level at 1%, 5%, and 10%, respectively. economies (Barro, 2022), which alleviate the threat of pandemic shocks to the economy (Çakmaklı et al., 2021). Conditional on effective policy, governments may anticipate looser budget constraints in the future as the economic recovery lessens public spending and broadens the tax base. Thus, health policies could either strengthen or weaken sovereign debt response to pandemic shocks depending on their economic costs and benefits. We find in column 3 of Table 6 that public health information campaigns, which urge caution in relation to COVID-19, strengthen the sovereign debt response to pandemic shocks. In response to a standard-deviation increase in the pandemic shock (41%), newly issued sovereign bonds grow by 7% in economies with public health information campaigns but drop by 8% in economies without such a policy, creating a 15% difference. 11 We find similar evidence for health policies regulating access to testing for COVID-19 infections, presented in Panel C of Appendix Table 5.
In contrast to economic, containment, and health policies, the presence of a vaccination policy weakens the impact of pandemic shocks on enlarging sovereign debt finance. Column 4 of Table 6 shows that the coefficient of the interaction between Covid and vaccination policy is negative and statistically significant. In particular, sovereign borrowing declines by 30% in response to a standard-deviation increase in the pandemic shock (41%) after the vaccine becomes available, and increases by only 7% in economies that have not yet provided vaccines. 12 Intuitively, vaccine provision alleviates the concern over being affected by COVID-19 or having significant negative health consequences, such as severe viral infection and even death. Despite the ongoing pandemic, vaccinated people are more confident about resuming business as usual, such as working on-site and dining in restaurants, which gradually restores normalcy and cultivates optimism about economic recovery. Expecting COVID-19 to weigh less on the economy after mass public vaccination, governments may anticipate looser budget constraints as the tax base broadens and public spending designed to contain the pandemic shrinks, which then lowers the sovereign debt response to pandemic shocks.

Stringency of pandemic-related policy actions
So far, we have focused on the existence, but not the stringency, of policy actions. Would the stringency of COVID-19-related policies affect the impact of pandemic shocks on sovereign debt finance? To answer this question, for those economies in which policy actions to combat the pandemic are in place, we further compare economies with relatively stringent and relatively loose policies and check whether they respond differently to pandemic shocks. We find no statistical evidence in Appendix Table 6 that stricter COVID-19-related policies change the sovereign debt response to pandemic shocks, conditional on the existence of these policies. In short, it is the presence but not the stringency of COVID-19 policies that affect the sovereign debt response to pandemic shocks. A possible explanation is that the adoption of a policy is costly per se, but the relative stringency of the policy, which mostly reflects policymakers' attitude and stance, does not create much additional costs.

Endogeneity of policy choices
One may argue that COVID-19 policies are endogenous, as governments have to consider the source of financing before rolling out these policies. We argue that the urges to save lives and support the vulnerable underlying these policies are exceptional and beyond economic considerations-otherwise they would not have been implemented so swiftly despite foreseeable costs and unclear benefits. We further address this endogeneity concern by applying an IV approach and considering the potential interactions between pandemic shocks and COVID-19-related policies.
3.3.5.1. IV approach. Our proposed IVs for each policy measure are Policy ROW and GovermentResponse ROW , which respectively capture the average stringency of this policy and the other policies in the rest of the world. These two IVs are likely to be relevant due to the coordination and externality of global COVID-19 policies. A government's decision to save economic growth at the cost of higher infection rates could significantly undermine other governments' efforts to save lives, as the virus transmits across borders (Zivin and Sanders, 2020). Governments are subject to peer pressure when dealing with coronavirus contagion. Further, they need to coordinate internally among public sectors to trade off pandemic mitigation against economic losses (Alvarez et al., 2021;Eichenbaum et al., 2021). Therefore, the COVID-19-related policies in one economy are likely to relate to those in the rest of the world.
We further argue that the proposed IVs are likely to satisfy the exogeneous criteria for two reasons. First, governments are unlikely to directly incorporate other countries' COVID-19 policies into their borrowing decisions. Second, we rule out the possibility that COVID-19-related policies in the rest of the world affect sovereign borrowing in the economy of interest via the spatial contagion of coronavirus in Section 4.5.2. Thus, it seems that the proposed IVs can only affect a particular economy's borrowing by influencing its COVID-19 policies.
We report the firstand second-stage estimation results in Appendix Table 7. The under-and weak-identification tests suggest that the proposed IVs are relevant and not weak for economic and vaccination polices. The second-stage estimation results show that the economic policy increases while the vaccination policy mitigates the sovereign debt response to pandemic shocks, consistent with our previous findings. However, the roles of containment and health policies become statistically insignificant. Note that the proposed IVs are not relevant to these policy choices, so we shall interpret the second-stage regression results in columns 2 and 3 of Appendix Table 7 with caution. 3.3.5.2. Pandemic shocks unrelated to COVID-19 policies. The observed pandemic shocks could be affected by COVID-19 policies as well as location-specific factors (Guaitoli and Pancrazi, 2021). To filter out the relatively exogeneous pandemic shocks, we take the residual from the regression of the observed pandemic shocks on its potential determinants, controlling for economy and time fixed effects: where POLICY i,t is the vector of COVID-19 policies, f i and f t m are economy and time fixed effects respectively. We rename the predicted residualε i,t as Covid i,t

Residual
, which measures the pandemic shocks not related to COVID-19 policies, and re-evaluate the roles of COVIDrelated policies by replacing Covid i,t in Eq. (3) with Covid i,t Residual .
We document similar evidence in Appendix Table 8 that economic and containment policies enhance sovereign debt response to pandemic shocks while vaccination policy weakens it. This mitigates the concern that our findings are contaminated by the relation between pandemic shocks and COVID-19 related policies.

Summary
To summarize, we find that vaccination policy lessens the sovereign debt response to pandemic shocks, while other COVID-19related policies strengthen the response. Moreover, it is the adoption rather than the stringency of these policies that makes a difference to sovereign debt finance. These results remain generally robust when we consider all policy actions together using the consolidated government response index and related measures (see Panel D in Appendix Table 5). After employing an IV approach to address the potential endogenous policy choices and tackling potential interactions between pandemic shocks and policy actions, the roles of economic and vaccination policies remain robust. 13 These findings suggest that pandemic related policies are costly, which increases the sovereign borrowing needs.

Leveraging on bond characteristics
Would certain bond characteristics enable governments to raise more funds in difficult times, i.e., because they hedge inflation, offer a level of safety, or have longer maturity preferred by governments conditional on no default (Aguiar et al., 2019)? Here, we are interested in whether governments may leverage various bond characteristics, such as market of issuance, currency denomination, tenor, and inflation protection option, to facilitate their access to the debt market during the pandemic. To address this question, we replace Determinant i in Eq. (3) with various bond characteristics.
The results in Table 7 show that governments borrow significantly more via USD-denominated bonds in response to pandemic shocks (see column 3). This chimes with the safe-haven nature of USD-denominated assets during crises (Maggiori et al., 2020), which attracts investors and enables governments to raise more funds. We also find that bonds issued in foreign markets, denominated in foreign currency, matured in at least five years, and associated with inflation protection appear to improve governments' access to the debt market. However, these results are not statistically significant at the conventional level (5%).

Sovereign spreads and pandemic shocks
Sovereign borrowing capacity should reflect the ability to borrow additional funds without having to tolerate higher sovereign default risk than when maintaining the same debt level. One may argue that some governments have greater borrowing capacity in the pandemic because they pay higher financing costs. To mitigate this concern, we evaluate the response of sovereign spreads, the difference in the local-currency yield curves between an economy and the US, to pandemic shocks. If rising financing costs are driving sovereign borrowing capacity, we should observe that sovereign spreads increase in response to pandemic shocks.
The outbreak of COVID-19 is accompanied by global finance turbulence, which is expected to increase sovereign credit risk and widen sovereign spreads, according to Gilchrist et al. (2022). Indeed, Arellano et al. (2020) and Emilio et al. (2022) show theoretically that pandemic shocks increase sovereign default risks. Bianchi et al. (2019) reach the same conclusion when considering the impact of fiscal stimulus on sovereign default. Therefore, we expect sovereign spreads to widen in the pandemic crisis.
To test this hypothesis, we replace the dependent variable in Eq.
(2) with bond-level sovereign spreads. Despite the small sample size, we show in column 1 of Table 8 that the sovereign spreads increase significantly by 0.14 percentage points (= the pandemic shock (41%) times the coefficient (0.352)) in response to a standard-deviation increase in the pandemic shock (41%). Repeating the analysis using economy-level sovereign spreads, we document similar evidence in column 2 of Table 8 that pandemic shocks increase sovereign spreads. Given the rising global financial risk accompanying the pandemic crisis, our finding is consistent with Gilchrist et al. (2022) that global financial risks increase sovereign spreads and enhance their international co-movements.
We next delve into the subsamples of emerging and advanced economies. Columns 3 and 4 of Table 8 show that sovereign spreads increase in response to pandemic shocks in emerging economies, but not in advanced economies. This suggests that 13 We acknowledge the challenge of finding proper IVs for pandemic-related policies, which limits our interpretation of the estimation results related to containment and health policies.

H. Zheng
Journal of International Economics 143 (2023) 103766 the positive response of sovereign spreads to pandemic shocks documented in column 2 of Table 8 is mainly driven by the subsample of emerging economies. Our finding provides empirical support for Arellano et al. (2020) that sovereign default risk increases in emerging economies after the COVID-19 outbreak. Note that our measures of sovereign spreads based on localcurrency yield curves include not only sovereign default risk but also exchange rate risk. To ensure that our result is not driven by the exchange rate risk, we repeat the analysis using CIP deviations, which reflect mostly sovereign default risk, and document similar results in Appendix Table 9 that funding costs in emerging economies increase in response to pandemic shocks. Connecting the difference in sovereign spreads between emerging and advanced economies documented in Table 8 with the difference in sovereign borrowing capacity documented in Table 5 yields very interesting conclusions: during the pandemic crisis, advanced economies are able to borrow more from the bond market without paying higher funding costs, while emerging economies are not able to raise new funds, despite offering higher yields. These results mitigate the concern that some economies are able to raise more funds because they offer higher sovereign yields, which could challenge our measures of sovereign borrowing capacity and their comparability across economies-because the willingness to pay for accessing additional funds varies. Our results concerning emerging economies are consistent with Arellano et al. (2020) and Bianchi et al. (2019), that rising sovereign risk in time of fiscal shocks undermines the sovereign borrowing capacity.

Further analysis
In this section, we conduct additional analysis to (i) understand additional sources of sovereign borrowing needs driven by fiscal structure or economic fundamentals, (ii) explore the heterogeneity in sovereign borrowing capacity in the pandemic across countries and over time, and (iii) perform various robustness checks. , denominated in USD, matured in >5 years, and associated with inflation protection. All regressions control for economy-time fixed effects (FE). Heterogeneity robust standard error clustered by economy is reported in parentheses. ***, **, and * denote the significance level at 1%, 5%, and 10%, respectively. Column 1 is based on bond-level sovereign spreads upon issuance, while columns 2-4 are based on economy-level sovereign spreads. Columns 2-4 use the full sample, the subsample of emerging economies, and the subsample of advanced economies, respectively. Heterogeneity robust standard error clustered by economy is reported in parentheses. ***, **, and * denote the significance level at 1%, 5%, and 10%, respectively.

Shrinking tax base as additional sources of fiscal shocks
The public spending designed to contain the pandemic and the shrinkage in tax bases due to contraction in economic activities during the pandemic both contribute to fiscal shocks, driving the need for sovereign borrowing. The unprecedented and abrupt policy actions and economic contractions would not have been possible in the absence of the global pandemic crisis. We have shown in Section 3.3 that costly COVID-19 policy actions that magnify fiscal shocks strengthen sovereign debt response to pandemic shocks. In this section, we explore the role of shrinking tax base in shaping sovereign debt response.
While public expenditure has risen substantially, fiscal revenue, in either a dollar amount or as a ratio of GDP, has dropped since the outbreak of COVID-19, though by a relatively small magnitude (see Table 1). The pandemic-driven economic shock is a double-edged sword for sovereign debt finance. On the one hand, it narrows the tax base, as businesses are shut down and production is disrupted, leading to loss of revenue (Alvarez et al., 2020;Eichenbaum et al., 2021). On the other hand, it undermines governments' capacity to repay debt, which constrains them from borrowing additional funds in the pandemic (Arellano et al., 2020;Emilio et al., 2022).
To understand which effect dominates, we first seek to uncover the pandemic-driven economic shock from the WEO's forecasts of the 2020 economic conditions in its October reports in 2019 and 2020. Clearly the WEO's forecast in 2020 incorporated the shock of the COVID-19 pandemic, but its forecast in 2019 did not, as nobody was expecting the world to be hit by COVID-19. Assuming consistent forecasting bias, the difference between the 2020 and 2019 forecasts of economic conditions in 2020 is a good proxy of the economic shock driven by the pandemic crisis. We therefore measure the pandemic-driven revenue loss by the percentage change in forecasted revenue in 2020 relative to that in 2019. If the revenue loss enhances the need for sovereign debt financing more than it depresses the capacity for public financing, economies that suffer from greater revenue loss should borrow more in response to pandemic shocks.
We compare economies with above-and below-median loss in fiscal revenue, and find no evidence in column 1 of Table 9 that revenue loss changes the sovereign debt response to pandemic shocks. Government revenues are positively related to growth in GDP, payroll, investment, and current account balance. As a robustness check, we use a similar methodology to delve further into pandemic-driven shocks in relation to these macroeconomic variables. Again, we find no statistical evidence that they affect the sovereign debt response to pandemic shocks (see columns 2-5 in Table 9). Overall, we find no evidence that pandemic shocks increase sovereign debt by shrinking the tax base.

Economic buffers
One may be concerned that our measures of sovereign borrowing capacity could be contaminated by economic buffers accumulated before the pandemic that mitigate the need for sovereign borrowing. Economies with a large fiscal surplus, current account balance, international reserve, and sovereign wealth fund (SWF) before the outbreak of COVID-19 may draw from their savings to finance surging public expenditure during the pandemic. These sources of revenue, saved in more prosperous times, may, on the one hand, reduce the need to borrow from the debt market and, on the other hand, improve access to capital by reducing the likelihood of sovereign default. To see which effect dominates, we evaluate how these economic buffers affect the sovereign debt response to pandemic shocks.
We show in Appendix Table 10 that fiscal surplus, current account balance, international reserve, and SWF all mitigate the sovereign debt response to pandemic shocks. However, the result is not statistically significant. 14 Seemingly, the diverged effects of fiscal space and macroeconomic buffers offset each other. It is also possible that these buffers are not sufficient to support the sudden increases in public spending, or that tapping into the debt market is easier and faster than drawing funds from these sources in emergency situations. Regardless of the rationales behind our finding, it rules out the possibility that sovereign borrowing needs are driving our result-otherwise, sovereign debt response to pandemic shocks should be statistically less pronounced in economies with stronger economic buffers saved before the pandemic.

Dynamic sovereign debt responses to pandemic shocks
Just as the coronavirus surges in several waves, so do governments' policy actions to combat the pandemic. Some of the costs of COVID-19-related policies are upfront, such as the research and development of vaccines, while others come later, such as vaccine purchase during the vaccination stage. To consider the roles of multiple waves and the different timings of pandemic-related fiscal shocks, we allow the sovereign debt response to pandemic shocks to vary over time. Focusing on the time window from three months before to 16 months after the COVID-19 outbreak in each economy, we follow Agarwal et al. (2007) to estimate the dynamic responses using the following event-study specification: where T i is the month of the initial COVID-19 outbreak in economy i, which varies across economies. The lagged variable Covid i,T i −h , which captures the pandemic shock before the COVID-19 outbreak, should equal 0 in reality but was hypothetically set to the initial pandemic shock in economy i (Covid i,T i −h = Covid i,T i ) to perform falsification tests. If the sovereign debt response is driven by some factors existed before the COVID-19 outbreak, the coefficient of the hypothetical pandemic shock, γ −h , should be statistically significant. Otherwise, if γ −h is not statistically significant, this means economies exposed to different pandemic shocks are not statistically different in sovereign borrowing before the COVID-19 outbreak, that is, they are comparable before the pandemic. This would justify our application of the DID approach and confirm the impact of pandemic shocks on sovereign borrowing. The variations of γ j over time, for j > 0, uncover the dynamic sovereign debt response to pandemic shocks. Fig. 3 illustrates the dynamic sovereign debt responses to pandemic shocks from three months before to 16 months after the COVID-19 outbreak. We also overlap the chart with the time trends of global COVID-19 cases per million population to broadly showcase key stages of the pandemic crisis. We observe no evidence that sovereign debt responds significantly to the hypothetical COVID-19 shocks before the actual COVID-19 outbreak (T + 0, around 2020 M2 15 ). It means that sovereign debt finance is comparable across economies before the COVID-19 outbreak, which justifies our DID application and supports the causal impacts of pandemic shocks on sovereign borrowing.
Moreover, the sovereign debt response to pandemic shocks varies with key stages of the pandemic crisis. In response to a 10percentage-point increase in pandemic shock, sovereign debt increases significantly by 2.2% following the initial COVID-19 outbreak, by another 5.1% when rising cases and deaths trigger worldwide lockdown (T + 3, around 2020 M5), become relatively muted during the second half of 2020 when governments start to roll out vaccination policies, and surge again by 14.7% following the Omicron variant, which is much more transmissible and causes more severe disease (T + 13, around 2021 M3). Note that sovereign debt finance tends to increase in periods of unexpected escalation of the pandemic crisis such as the initial outbreak and the Omicron variant when policy actions were largely absent or took time to affect the pandemic. This mitigates the concern that our result is driven by the potential impact of policy on pandemic mitigation.
To address potential concerns over path-dependent and expected pandemic shocks, we further estimate dynamic effects within narrow windows around the first wave of COVID-19 in 2020 M3 and the outbreak of Omicron variant in 2020 M12, during which the evolution of pandemic shocks was relatively unexpected and exogenous. For this alternative estimation, T i in Eq. (4) is set constantly to 2020 M3 or 2020 M12 for all economies, and the pandemic shock before period T i was hypothetically set to the pandemic shock in economy i in 2020 M3 or 2020 M12 (Covid i,T i −h = Covid i,T i ). We focus on two months before and six months after the major waves of unexpected pandemic shocks. Appendix Fig. 1 corroborates our previous finding that sovereign debt increases significantly after but not before the first wave of COVID-19 or the outbreak of Omicron variant. This again mitigates the concern that the documented sovereign debt response is driven by factors other than pandemic shocks.

Sovereign debt responses across regions
We further explore how sovereign borrowing capacity varies across regions. Fig. 4 shows that all regions except for Latin America are able to borrow significantly more during the pandemic. In response to a 10-percentage-point increase in the Loss is a dummy variable that equals 1 in columns 1-5, respectively, if the macroeconomic indicator specified in the top row is below the sample median, indicating more significant loss, and 0 otherwise. All regressions control for economy-time fixed effects (FE). Heterogeneity robust standard error clustered by economy is reported in parentheses. ***, **, and * denote the significance level at 1%, 5%, and 10%, respectively. 15 Note that when calculating the COVID-19 shocks, we use the monthly growth in the COVID-19 infection rate. The first observation of COVID-19 shock measures is essentially in 2020 M2, even though the data on COVID-19 cases starts from 2020 M1.
pandemic shock, economies in Oceania (Australia and New Zealand) borrow 7.5% more than in the pre-pandemic periods on average-the highest in our sample, followed by Africa (2.3%), North America (1.9%), Europe (1.6%), and then Asia (1.2%).

Robustness checks
In the following, we employ alternative measures of pandemic shocks, consider spatial contagion of coronavirus, and control for additional variables to perform additional robustness checks.

Alternative measures of pandemic shocks
We check the robustness of our key results using alternative measures of COVID-19 shocks, including the logarithm of one plus COVID-19-related new cases, new deaths, ICU patients, and hospitalized patients per million population, respectively. Consistent with our baseline result, Appendix Table 11 shows that sovereign debt increases in response to different measures of pandemic shocks.
Note that the number of new cases may vary with disease severity-some cases may not be detected due to mild symptoms or lack of testing technology-making this variable less comparable across countries. The number of new deaths may best reflect the severity of the disease but their variations are quite limited and may vary with demographic structures, as the elderly are more vulnerable to coronavirus. Compared to hospitalized patients, the number of ICU patients better reflects the severity of the disease and is more comparable across countries, because patients can only be sent to ICU when they are extremely ill, whereas hospitalization may be voluntary. We therefore focus on the number of ICU patients to further explore the robustness of our key results related to the determinants of sovereign borrowing capacity.
Consistent with the previous results, we show in Appendix Table 12 that fiscal rules improve sovereign borrowing capacity in the pandemic, while debt overhang and default risk undermine it. Although debt rollover risk weakens sovereign borrowing capacity in the pandemic, the result is not statistically significant as before.

Spatial contagion of pandemic shocks
The geographic transmission of coronavirus through global networks of, for example, passengers and goods flows, could lead to spatial correlations of pandemic shocks. This may trigger concern that an economy can be hit by its own pandemic shocks as well as other economies', which spread across borders through international linkages. To address this issue, we perform spatial regressions and conduct falsification tests.
4.5.2.1. Spatial regression. Global linkages in passenger and trade flows directly contribute to the global transmission of coronavirus. We first perform spatial regression using data on bilateral air travel passenger flows before the COVID-19 outbreak, considering the spatial contagion of pandemic shocks and the spatial structure of the error term. The estimation result reported in column 1 of Appendix Table 13 show that the direct impact of domestic pandemic shocks on increasing sovereign debt remains robust after controlling for spatial contagion of coronavirus. The indirect impact of foreign pandemic shocks on domestic sovereign debt channeled through global passenger flows is positive but not statistically significant. These findings mitigate the concern that our baseline results are driven by the geographic spread of coronavirus. When measuring spatial contagion using bilateral import weighted by an economy's total import, we document similar results in column 2 of Appendix Table 13-the direct impact of domestic pandemic shocks rather than the indirect impact of foreign pandemic shocks contributes to the sovereign debt response to pandemic shocks.

Falsification tests.
We further conduct falsification tests to mitigate the concern that our results are driven by spatial contagion. For each economy, we calculate the average COVID-19 shocks in the rest of the region and the rest of the world, denoted as Covid ROR and Covid ROW respectively. Should coronavirus contagion drive our result, the coefficient of Covid ROR and Covid ROW should have been statistically significant. However, we observe no statistical evidence in Appendix Table 14 that either Covid ROR or Covid ROW affect sovereign debt finance.

Additional control variables
Could changing bond characteristics drive the response of sovereign bond issuances to pandemic shocks? We control for a series of bond-level characteristics, such as bond tenor, currency denomination, and coupon class, and show in Appendix Table 15 that our baseline result-sovereign debt increases in response to pandemic shocks-remains intact. We further show in Appendix Table 16 that most sovereign bond characteristics do not change in response to pandemic shocks. This evidence rules out the possibility that bond-level characteristics could drive the positive sovereign debt response to pandemic shocks.

Conclusion
The COVID-19 outbreak has created a global natural experiment to identify sovereign borrowing capacity as well as its determinants. We have measured sovereign borrowing capacity by investigating the sovereign debt response to pandemic shocks, comparing the difference in sovereign bond issuances between economies exposed to relatively severe and relatively mild pandemic shocks after the COVID-19 outbreak, with that before the pandemic. We first demonstrated that pandemic shocks increase sovereign borrowing by creating exogeneous fiscal shocks, and then proceeded to explore how various fiscal regimes and economic policies affect sovereign borrowing capacity. We found that strong and credible fiscal rules-especially debt and budget balance rules that directly discipline debt sustainability and fiscal responsibility-strengthen sovereign borrowing capacity during the pandemic. Further analysis revealed that debt overhang due to historically high debt burden, rollover risk grounded in a large fraction of short-term debt, and sovereign default risk built up before the outbreak of COVID-19 all limit sovereign borrowing capacity, preventing governments from borrowing additional funds to finance pandemic-driven fiscal shocks. Finally, we showed that advanced economies exhibit greater borrowing capacity than emerging economies but have lower sovereign spreads during the pandemic. Interestingly, our analysis suggests that emerging economies can better access the bond market during the pandemic if they have a pegged exchange regime, an open capital account, and no monetary autonomy.
Our findings have important policy implications. First, our study speaks to the current debate on whether governments should relax or suspend existing fiscal rules to accommodate expansive fiscal policies (see, for example, Medas et al., 2022). While policies should be adaptive, certain principles are worth sticking to, and fiscal rules are one of these according to our study. Giving up fiscal rules that discipline public spending and fiscal budgeting would result in lower sovereign borrowing capacity. This implies that governments may not be able to borrow more even if new fiscal regimes allow them to.  Second, we challenge the policy prescription of issuing more debt to finance expansive fiscal policies when debt sustainability is already at risk. According to our study, excessive borrowing increases sovereign default risk and limits governments' capacity to finance fiscal stimulus. Should governments default as a result of unsustainable debt, it will take a very long time to restore sovereign borrowing capacity, which then constrain fiscal policy. It is important for policymakers to consider the risk impacts of sovereign borrowing when making public financing decisions.

Africa
Third, our study warns of substantial sovereign credit risk and fiscal distress building up in emerging economies during the pandemic. Unlike advanced economies, emerging economies have difficulty accessing credit in the pandemic, despite paying higher funding costs. As the pandemic continues, they would become more and more vulnerable as sovereign credit risks spiral up. The vicious circle of rising sovereign risks and falling borrowing capacity may lead to debt crisis in some emerging economies. Our finding also implies that expansive fiscal policies during recessions are not feasible in emerging economies unless they save sufficiently in good times to be used in difficult times.
Fourth, our study uncovers new benefits of surrendering monetary autonomy and pursuing exchange rate stability. For emerging economies that rely on fiscal policies more than monetary policies to stimulate economic growth, our finding implies that surrendering monetary autonomy and pursuing exchange rate stability improve their borrowing capacity in time of need, which enable them to better finance fiscal shocks.
Overall, our findings illustrate the importance of maintaining fiscal responsibility and debt sustainability to access funds when in need, which is important for fiscal policy and sovereign risk. There are several caveats to our study. First, by focusing on sovereign bonds, we ignore other channels through which governments may finance their fiscal deficits, such as investment returns, bank loans, and money-financed fiscal transfers, as proposed by Galí (2020). Second, this study explores how governments finance their policy actions but does not address how they may repay these debts in the future. These caveats point to several interesting avenues of future study. First, would money finance be a better option than debt finance? Unlike debt finance through loans and bonds, money finance, which simply transfers helicopter money from central banks to governments, neither leads to distortionary tax hikes nor incentivizes manipulation of inflation. However, it undermines central bank independence, a principle that is fundamental for monetary credibility and macroeconomic stability. It would be interesting to explore the costs and benefits of money finance relative to debt finance in emergency situations. Second, how will governments repay these debts accumulated during the pandemic, and what are the implications for future borrowing? Governments may raise tax or cut public expenditure to increase fiscal surplus, increase inflation to mitigate the debt burden, or issue new debt to finance old debt. These policy actions, which aim to improve sovereign solvency and debt sustainability in the future, have important implications for the path to economic recovery, which, in turn, shapes sovereign debt capacity when the next crisis occurs.

Disclosure
The author, Zheng Huanhuan, declares that (s)he has no relevant or material financial interests that relate to the research described in this paper.

Declaration of Competing Interest
None.