key: cord-0784916-brr2x5b9
authors: Paczos, Wojtek; Shakhnov, Kirill
title: Defaulting on Covid debt
date: 2022-01-20
journal: Journal of International Financial Markets, Institutions and Money
DOI: 10.1016/j.intfin.2022.101516
sha: 9fb679a8fc3e36296a4dccadd92c4d67198b5639
doc_id: 784916
cord_uid: brr2x5b9

The COVID-19 pandemic causes sharp reductions in economic output and sharp increases in government expenditures. These increase the riskiness of sovereign debts, especially in emerging economies. This paper proposes a framework to study debt sustainability. The economy is subject to productivity and expenditure shock. The government sets distortionary labour taxes and decides whether to repay its past domestic and foreign obligations. Foreign default is more likely after a negative productivity shock, while domestic default is more likely after a negative expenditure shock. This mechanism finds support in the data. Recent proposals that would ease the burden of foreign debt after COVID-19 would not prevent a wave of domestic defaults.

In response to the COVID-19 pandemic, governments around the globe imposed containment measures. These lockdowns reduce economic activities and increase budget deficits. Due to the commitment problem, emerging economies find it hard to issue debt to smooth the impact of the shock. This paper studies the effects of output and government expenditure shocks on domestic and foreign debt sustainability.

The International Monetary Fund projects that the world output will contract by 3.0% in 2020. To mitigate the economic costs of lockdown measures, governments provide financial rescue packages at an unprecedented scale. Figure 1 plots the expected scale of the real GDP growth against the expected primary deficit changes between 2019 (green stars) and 2020 (red circles). In 2020 emerging markets on aggregate are expected to lose 5.7 percentage points of GDP growth and to have a primary deficit larger by 4.3% of GDP relative to 2019.

The new fiscal measures, which include additional spending and forgone revenue and loans, equity, and guarantees, account for 5.1% of GDP in emerging economies (IMF, 2020) . The arrows show the average changes for country groups, and all point towards the southeast: lower growth and larger deficits.

How to finance such unprecedented deficits and stimulate economies? Governments stockpiled debt to finance rescue and recovery packages. Elevated debt levels coupled with falling output and rising government expenditure bring back the question of debt sustainability. Economists and policymakers call for urgent measures to be taken. The G-20 countries have suspended interest rate payments on bilateral debts. Bolton et al. (2020) argue for a broad "debt standstill" that will include private creditors and will be available to a large set of countries. The total stock of debt under consideration is $3 trillion.

The discussion focuses on external debt. Yet, the external debt accounts only for 40% of the total public debt in emerging economies, while the remaining debt is owed to domestic investors. Table 1 summarizes the stylized facts on the size and composition of public debt in 24 emerging economies in 2019. It also shows that in emerging markets, debt owed to domestic investors is usually denominated in local currency, and debt owed to foreign investors is usually denominated in foreign currency. 1

Defaults on domestic debt are not uncommon and are associated with economic disrup-1 The literature distinguishes foreign and domestic debt on either legal or residency or currency basis.

Here we apply the residency definition. For discussion, see Panizza, Sturzenegger and Zettelmeyer (2009 in macroeconomics (Reinhart and Rogoff, 2011a) . Domestic and foreign debts are hardly similar. Foreign debt involves transferring resources into and out of an economy, which can help to achieve consumption smoothing over the business cycle. Domestic debt cannot accomplish this, as its issuance and repayment occur within an economy: domestic borrowing does not bring in additional resources.

We contribute to the literature by proposing a model to analyze foreign and domestic debt sustainability simultaneously. The economy is subject to two simultaneous shocks, and the government has limited commitment. After observing the shock realizations, the J o u r n a l P r e -p r o o f Journal Pre-proof government decides on distortionary labour tax and whether to default on its foreign or domestic obligations (or both). If it defaults, the economy is subject to an output penalty in the form of reduced productivity. Otherwise, it has to increase taxes to finance repayment.

Thus, repayment imposes endogenous distortion on the economy. However, in the case of domestic debt, this distortion is partially mitigated because resources flow back to domestic households. As a result, foreign default is more likely after a negative output shock, and domestic default is more likely after a negative expenditure shock.

We collect a dataset of default episodes, debt levels, tax rates and macroeconomic aggregates and test the model predictions in the past data. We find strong support for three model predictions: i) that fluctuations in the GDP growth are associated with the probability of foreign default, ii) that fluctuations in the government expenditure are strongly associated with the probability of domestic default, and iii) that higher foreign debt increases the risk of foreign default, but not the risk of domestic default. Additionally, our model predicts that after any type of default, there is a fall in the tax rate, for which the data also offers moderate support.

This paper argues that after a Covid shock; a simultaneous output drop, and a government expenditure hike, even in case of a broad restructuring of foreign debt, governments might still choose to default on their domestic debt obligations selectively.

The model in this paper is in the tradition of the strategic sovereign default framework of Eaton and Gersovitz (1981) . We contribute to the recent studies of the selective nature of sovereign defaults (Erce and Mallucci, 2018 , Niepelt, 2016 , Paczos and Shakhnov, 2016 , Sunder-Plassmann, 2020 . In our paper, similarly to D'Erasmo and Mendoza (2016 Mendoza ( , 2020 the riskiness of domestic debt takes centre stage. In contrast, the risk of domestic default is not driven by distributional incentives but by tax distortions arising in repayment.

We also contribute to the new Covid-macro literature; e.g. Arellano, Bai and Mihalache (2020) embed the epidemiological SIR model into the standard sovereign default model to analyze foreign default risk. The interaction of debt, default, and Covid is also addressed quantitatively in Espino et al. (2020) . These contributions focus on the riskiness of foreign debt. While the model in our paper is paper is more stylized, it provides clear insights into the riskiness of domestic debt. This paper is also related to the literature on distortionary taxation with default (Karantounias, 2017, Pouzo and Presno, 2014) , where a government defaults to mitigate endogenous tax distortions, albeit in a closed economy setting. Finally, J o u r n a l P r e -p r o o f Journal Pre-proof since we solve the static optimal labour taxation problem, our model can be viewed as an extension of the Basic Model in Piketty and Saez (2013) , that allows for defaultable debt.

The Households choose labour supply and consumption to maximize utility subject to budget constraint:

where w is wage and π is a lump sum profit transfered from the representative firm owned by the household. Household's optimality condition reads:

The optimality condition, together with the budget constraint, form the solution to the household's problem.

J o u r n a l P r e -p r o o f

The government later takes the solution to the household's problem as a constraint in its maximization problem; hence we subsequently refer to it as the Implementability Constraint (IC).

Firms produce using Cobb-Douglas production function with labour as the sole input and chose labour demand to maximize static profits:

where A is the total factor productivity (TFP) shock and γ is the output cost in the case of a default:

where γ r = 1 and γ d , γ f < 1. The solution to the firms' problem reads:

Optimal policy: The government acts as a Ramsey planner. To maximize household utility, the planner decides whether to default on foreign and domestic debts (d f , d d ). The planner must respect implementability constraint (IC) (4) and the resource constraint (RC).

The economy is subject to two shocks: the aggregate TFP shock A and the government spending shock g:

J o u r n a l P r e -p r o o f Journal Pre-proof In the primal approach, the households and firm first-order conditions are used to eliminate prices and tax rates. Determining optimal policy is reduced to a simple programming problem in which the choice variables are the allocations.

We assume the non-separable preferences, which is the standard form in the optimal taxation literature (Conesa, Kitao and Krueger, 2009):

Thus, the IC constraints reads:

For each d f = {0, 1} and d d = {0, 1} equations (11) and (14) of emerging economies annual income. We set the target for b f at 5% of y * , where y * is the reference output produced in repayment with A = 1 and g = 0. This gives b f = 0.0205.

For the clarity of exposition of the mechanism in the model, we assume that domestic debt is equal to foreign debt. With this set of parameters, the reference output is equal to y * = 0.418. We plot the solution in the state space (A, g) setting the range for A between 0.85 and 1.15 and for g between 0 and 10% of y * . The two remaining parameters are the productivity losses upon domestic and foreign default (γ d and γ f ). These parameters govern how much debt is sustainable in the repayment equilibrium. Or, equivalently: how big shocks are necessary to trigger defaults. That output contracts after foreign as well as after domestic default has been established empirically, for example in Paczos and Shakhnov (2016) , Reinhart and Rogoff (2011a) . This empirical association can be a manifestation of two causal mechanisms. On the one hand, defaults happen during a series of bad shocks.

On the other hand, they cause disruptions on domestic markets, causing the output to fall further. (Engler and Große Steffen, 2016 , Gennaioli, Martin and Rossi, 2014 , Mallucci, 2015 , Sandleris, 2014 , Sosa-Padilla, 2018 , Thaler, 2020 . In this paper, for the sake of simplicity, we capture the GDP loss mechanism with an exogenous parameter. 3 We parametrize 3 Our paper differs from the cited studies in one crucial respect: a government can issue separate bonds to domestic and foreign investors. Hence, the GDP loss is driven by the former mechanism. the output costs such that domestic and foreign defaults occur within the chosen state-space implying γ d = 0.99 and γ f = 0.94. Later, we show that the numerical parametrization of these two parameters does not qualitatively affect the mechanism described in the model.

Chapter 3 also provides corroborating empirical evidence that domestic and foreign debt defaults are followed by substantial GDP losses. Although the exact positions of the lines are dependent on parameters, the economic mechanism is not. The domestic debt sustainability line is always steeper than the foreign debt sustainability line. This is a striking result. It says that default on foreign debt is driven primarily by changes in productivity A, while default on domestic debt is driven primarily by changes in government expenditures g.

The intuition is the following. Households treat foreign and domestic default differently.

When a government defaults on foreign debt, the economy suffers from decreased productivity, but households benefit through lower taxes. This trade-off is independent of the level of government expenditure g. Hence, foreign default is primarily driven by productivity shock.

When a government defaults on domestic debt, households pay lower taxes, but they also lose their savings. As a result, they work more and become less responsive to tax changes.

The latter effect is independent of the level of productivity A. Hence, domestic default is primarily driven by the expenditure shock.

We can visualize this intuition with the help of Notes: The order on the x axis is reverse. Low government expenditure corresponds to 0% of GDP and high government expenditure corresponds to 7% of reference GDP. The rest of the parameters are as in Utility decreases in g (right panel). As is the case for labour supply and consumption, utility decreases at the fastest pace in repayment and at the slowest pace in domestic default. When the economy is in a "good" state (A = 0.95 and g = 0), utility is the highest in repayment.

The difference between utility in repayment and utility in domestic default narrows down quickly. As a result, for high levels of g domestic default is the preferred option. The effect is similar at different levels of A. Hence, the domestic debt sustainability line is almost vertical in Figure 1 .

On the other hand, the difference between utility in repayment and utility in foreign default narrows down slowly in g. Thus, a relative position of the two lines in the origin is the prime driver of the repayment vs foreign default trade-off. For high levels of A, repayment is the preferred option, and for low levels of A, foreign default is selected. The effect is similar across expenditure levels. As a result, the foreign debt sustainability line is almost horizontal in Figure 1 . We build a database of default episodes, outstanding debt, government expenditures, taxes (and some other auxiliary variables) by carefully merging and unifying multiple, concurrently available data sources. The data on start and finish dates of default episodes come from the updated database accompanying Reinhart and Rogoff (2011b) , which covers up to 150 countries for the years 1800-2014. We focus on the postwar period. We calculate GDP growth rate as a log difference of the GDP in local currency in fixed prices from the Penn World Tables 8.1 (Feenstra, Inklaar and Timmer, 2015) . This variable is widely available across countries and time. We collect debt levels from both Panizza (2008) Since the model is static, it cannot be estimated using the state-of-art estimation and empirical validation techniques such as indirect inference (Le et al., 2016) . Such techniques only require that the theoretical model can be simulated and so are well suited for dynamic models with rich structure. Our theoretical mechanism can be modelled and interpreted using a simpler, static setup, albeit rich in interactions. Thus, the model cannot be simulated and instead, we test whether the model predictions can find support in the data.

To test Hypotheses 1-3, we estimate the following regression equation:

separately for each type of default, where P r is a probability of default, Def j i,t is the indicator for the beginning of a default episode of type j (foreign or domestic), in the country i in period t, and ∆GDP i,t is the rate of growth of real GDP per capita. ∆G i,t is the change in the ratio of government expenditure to GDP and B f i,t is the level of foreign debt to GDP. We further control for the level of distortions in the economy by the level of corporate income tax τ i,t . 6 Our independent variables are lagged one period to mitigate the possible endogeneity problem. We estimate five regressions for each default: a pooled logit, a random-effects logit and a probit, and a linear probability model and a logit model with country fixed effects. Our estimation sample with the full data coverage is 89 countries for the years 1981-2011, yielding a total of 2391 country-year observations with 17 domestic defaults and 41 external defaults.

Our model implies that foreign default is mainly driven by fluctuations in output; therefore, we expect α f 1 to be significant and negative and α f 2 to be zero. Domestic default is mainly driven by public expenditures; therefore, we expect α d 2 to be significant and positive, and we remain agnostic about α d 1 . Trivially, defaults are more probable with higher debt levels; therefore, we expect both α j 3 to be positive for foreign default, but not domestic default.

Tables 3 and 4 report the estimation results for foreign and domestic default respectively. third source as a supplement for missing countries and data points. For overlapping country-years, we take the tax rate claimed by at least two sources. If there is a missing data period and the tax rate before and after is the same, we interpolate this tax rate throughout the period. 6 The rationale for using corporate tax rate is its broad coverage in the available datasets. The minor point is that corporate taxes are, to the best of our knowledge, always flat-rate taxes -unlike personal income taxes. In addition, empirical studies show that labour bears between 50 per cent and 100 per cent of the burden of the corporate income tax: https://taxfoundation.org/labor-bears-corporate-tax/ J o u r n a l P r e -p r o o f Journal Pre-proof (3), the identification is, by construction, only through defaulting countries, which yields a large drop in observations. *p < .10, **p < .05, ***p < .01

Source: Authors' calculations.

A faster GDP growth per capita is associated with a lower probability of foreign default but does not statistically affect the probability of domestic default. These results are robust across five methodological specifications. An increase in the government expenditure-to-GDP is associated with the higher probability of domestic default but does not statistically affect the probability of foreign default. A positive association with domestic default probability is statistically significant at the 5%-level in three specifications and at the 10%-level in the remaining two. The level of foreign debt is associated with a higher probability of foreign default in the three specifications. The results are less clear-cut for the domestic default: in four specifications, the association is not statistically significant.

To test Hypothesis 4 we estimate the following regression equation:

where δ j i,t are indicators for a default of j-type (domestic and foreign) in country i in J o u r n a l P r e -p r o o f Journal Pre-proof period t and the other variables are already familiar from equation (15). Our model predicts that after a default tax rate should go down so that β 1 and β 2 are negative. Table 5 reports the estimation results of equation (16). Interestingly, the corporate tax rate goes down after domestic default but not after foreign default. This result is robust for controlling for GDP growth and the growth of public expenditures (in column 2) and also for the level of debt (in column 3).

The empirical exercise shows that the data offers strong support to Hypotheses 1-3 and some support to Hypothesis 4. Thus, we conclude that the data corroborates the mechanism described in the model.

Our dataset allows us also to test the relevance of the assumption that defaults lead to losses in the TFP. We estimate the following regression equation:

J o u r n a l P r e -p r o o f Journal Pre-proof where the dependent variable is the growth rate of GDP, and independent variables are domestic and foreign default, change in the ratio of government expenditure to GDP and the levels of domestic and foreign debt to GDP. All regressors are lagged one period to avoid reverse causality, and the estimation is executed using country fixed effects. Table 6 presents the results. Indeed, one year after domestic default, GDP growth falls on average by 3-6 percentage points, and after foreign default, it falls by about 1.5 percentage points.

An increase in the government expenditure ratio and in both debt levels is also strongly associated with slower GDP growth one year later.

This paper illustrates the mechanism that drives strategic foreign and domestic default in the presence of two shocks. Foreign default is more likely after a negative productivity J o u r n a l P r e -p r o o f Journal Pre-proof shock, and domestic default is more likely after a negative government expenditure shock.

This mechanism finds strong support in the data. A Covid shock, which reduces output and increases government expenditure, brings the economy closer to a total default. Even in the case of well designed foreign debt restructuring and "standstill" programmes, we can still expect a wave of domestic defaults. 

Deadly Debt Crises: COVID-19 in Emerging Markets

Tracking Global Demand for Emerging Market Conesa

Distributional incentives in an equilibrium model of domestic sovereign default

History remembered: Optimal sovereign default on domestic and external debt

Debt with Potential Repudiation: Theoretical and Empirical Analysis

Sovereign risk, interbank freezes, and aggregate fluctuations

Selective sovereign defaults

Seigniorage and Sovereign Default: The Response of Emerging Markets to COVID-19

Louis Working Papers

The Next Generation of the Penn World Table

Sovereign Default, Domestic Banks, and Financial Institutions

World Economic Outlook Update

Optimal time-consistent taxation with default

Tax Rates Online

Testing macro models by indirect inference: a survey for users

Reserve System (U.S.) International Finance Discussion Papers 1153

A General Equilibrium Model of Sovereign Default and Business Cycles

Society for Economic Dynamics Meeting Papers 635

European University Institute Economics Working Papers

Domestic And External Public Debt In Developing Countries

The Economics and Law of Sovereign Debt and Default

Optimal labor income taxation

Optimal Taxation with Endogenous Default under Incomplete Markets

The Forgotten History of Domestic Debt

From Financial Crash to Debt Crisis

Sovereign Defaults, Credit to the Private Sector, and Domestic Credit Market Institutions

Sovereign defaults and banking crises

Infation, default and sovereign debt: The role of denomination and ownership

Sovereign Default, Domestic Banks and Exclusion from International Capital Markets

World Tax Database