Research
Publications
On the Welfare Losses from External Sovereign Borrowing – PDF – Journal link – Slides – Code
Aguiar, Mark, Manuel Amador, and Stelios Fourakis (2020): “On the Welfare Losses from External Sovereign Borrowing,” IMF Economic Review, 68(1): 163-194.
Abstract: This paper studies the losses to the citizenry when the private agents discount the future at different rates from their government. In the presence of such a disagreement, the private sector may prefer an environment in which the government is in financial autarky. Using a sequence of sovereign debt models, the paper quantifies the potential welfare losses that citizens suffer from the government’s access to international bond markets. While the environment is not necessarily comprehensive, the analysis provides a counterweight to proposals that are designed to ease market access for sovereign borrowers.
Working Papers
The Puzzling Behavior of Spreads During Covid (with Loukas Karabarbounis) – PDF
Abstract: Advanced economies borrowed substantially during the Covid recession to fund their fiscal policy. The Covid recession differed from the Great Recession in that sovereign debt markets remained calm and spreads barely responded. We study the experience of Greece, the most extreme manifestation of the puzzling behavior of spreads during Covid. We develop a small open economy model with long-term debt and default, which we augment with official lenders, heterogeneous households and sectors, and Covid constraints on labor supply and consumption demand. The model is quantitatively consistent with the observed boom-bust cycle of Greece before Covid and salient observations on macro aggregates, government debt, and the sovereign spread during Covid. The spread is stable despite a rise in external borrowing during Covid, because lockdowns were perceived as transitory and the bailouts of the 2010s had tilted the composition of debt at the beginning of Covid away from defaultable private debt. The ECB's policy of purchasing debt in secondary markets during Covid did not stabilize spreads so much, but allowed the government to provide transfers that reduced inequality.
Sovereign Default and Government Reputation – PDF
Abstract: In this paper, I build a flexible theoretical model of sovereign borrowing, default, and renegotiation with borrower reputation. There is asymmetric information about the government's "type", and reputation is the market belief that it is "responsible" and therefore less likely to default. I calibrate the model using data on how countries' credit histories affect the prices they face and validate its predictions about the effects of borrowing on interest rate spreads in the data. Using the model, I show that countries that have recently defaulted have poor reputations because they rapidly run up their debts prior to default, not because the default decision itself is revealing. I also show that, for countries facing non-trivial levels of default risk, the reputational benefits of repayment are less than $0.5$ basis points of consumption. Policies that disrupt the signalling motives induced by asymmetric information, such as transparency initiatives or fiscal rules, can have substantial negative implications for welfare (losses of $0.23\%-0.85\%$ of permanent consumption), because they lead to increased overborrowing by the government.
Abstract: In this paper, I document that, during the height of the Eurozone Debt Crisis in Spain, 1.) Spanish government bonds became substantially less liquid and less traded on secondary markets, 2.) the first appearance of this phenomenon lagged far behind the initial jump in interest rate spreads in late 2008, and 3.) it persisted throughout the period of peak interest rate spreads and only subsided after the worst of the crisis had passed. I argue that these facts are related and best explained by a model of sovereign default that features secondary markets in which it is possible that some traders have private information. I then build a model in which some traders have private information about the country's future economic conditions and show that this allows the model to reproduce both the delayed reaction of bid-ask spreads as well as their peak and behavior during the height of the crisis. Using the model, I measure the losses to investors associated with variation in liquidity during debt crises. Finally, I validate the model by showing that the model's predicted relationship between current, realized bid-ask spreads and future values of GDP allows me to forecast GDP significantly better than a standard, benchmark forecast.
Sovereign Debt Auctions with Strategic Interactions – PDF
Abstract: In this paper, we compare how different protocols for sovereign debt auctions affect borrowing, the cost of debt and welfare, when default risk is a concern. To do so, we build a theoretical model of sovereign borrowing and default with auctions and asymmetric information. We calibrate the model to the Portuguese economy under the discriminatory price protocol, prior to its bailout in 2011. Using the calibrated model we perform a counterfactual exercise by changing the auction protocol. We find that the uniform price protocol yields higher welfare than the discriminatory price protocol, and that these gains are highest during crises (up to 0.6% of permanent consumption). This result is consistent with the observed switch to a uniform price protocol (for long term debt). Our accounting for dynamic effects is crucial for this result. Given standard values for risk aversion of the borrowing country, the discriminatory price protocol performs better than the uniform under a single auction setting. Once we allow for repeated auctions, however, the uniform price protocol is preferred under the calibrated model. In fact, when default risk is a concern, the uniform price protocol provides better incentives for borrowing over time, as well as protecting investors from static dilution within an auction. Both lead to much better prices for the government, which more than justifies forgoing the insurance mechanism provided by the discriminatory protocol.
Sovereign Default under Imperfect Information – PDF
Abstract: In this paper, I document that, during the Eurozone Debt Crises, 1.) forecasts of output were persistently biased upwards, 2.) the afflicted countries all saw steep increases in their government debt to GDP ratios and their external government debt to GDP ratios, and 3.) spreads reacted slowly to these increases. I argue that these three facts are related and connect them through a model of sovereign default which features incomplete information with respect to the persistent component of output. I then show that the inclusion of information imperfections allows the model to produce patterns during and before crises which better match the patterns in the data than the benchmark model.
Work in Progress
Optimal Joint Bond Design (with Eduardo Davila and Charles-Henri Weymuller)
The Welfare Effects of Bailouts
Inequality, Redistribution, and Sovereign Risk (with Monica Tran-Xuan)
Computing Sovereign Debt Models: Why So Hard?
Joint with Mark Aguiar and Manuel Amador
Abstract: Sovereign debt models with long-duration bonds are notoriously hard to compute. Using a simplified environment of the standard Eaton & Gersovitz (1981) model with outside option shocks, we show that equilibria in pure strategies may not exist, explaining the lack-of-convergence issues encountered in the quantitative literature. We propose an algorithm for computing mixed-strategy equilibria. For some parameterizations, we uncover millions.
Long Term Debt Models: Solution Methods Matter
Abstract: Over the last decade, long term debt has become a standard feature in quantitative studies of sovereign default. A wide variety of numerical solution methods have been used in solving these models, but little is known about the relative performance of various methods and their sensitivity to parameters and grid specifications. In this paper, I provide a survey of the commonly used methods and then test their performance. For long run simulations (often used to calculate the moments used to estimate the model), I find that results (both within each method, varying grid fineness, and across methods) are relatively stable. However, for short run simulations (often used to analyze the model's prediction about a specific country's experience at a specific time of interest), the results can vary dramatically. In some cases, the results even change qualitatively. Using these results, I provide guidance on when each method is likely to be robust and when it is probably the case that results are not robust to the choice of solution method (and specific parameters of that solution method, such as grid fineness).