Tax revolts and sovereign defaults

Arce, Fernando; Morgan, Jan; Werquin, Nicolas Working Paper Tax revolts and sovereign defaults IDB Working Paper Series, No. IDB-WP-1719 Provided in Cooperation with: Inter-American Development Bank (IDB), Washington, DC Suggested Citation: Arce, Fernando; Morgan, Jan; Werquin, Nicolas (2025) : Tax revolts and sovereign defaults, IDB Working Paper Series, No. IDB-WP-1719, Inter-American Development Bank (IDB), Washington, DC, https://doi.org/10.18235/0013572 This Version is available at: https://hdl.handle.net/10419/324816 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by/3.0/igo/ Tax Revolts and Sovereign Defaults Fernando Arce Jan Morgan Nicolas Werquin WORKING PAPER No IDB-WP-1719 InterA merican Development Bank Department of Research and Chief Economist June 2025 * InterA merican Development Bank ** Massachusetts Institute of Technology *** Federal Reserve Bank of Chicago Tax Revolts and Sovereign Defaults Fernando Arce* Jan Morgan** Nicolas Werquin*** InterA merican Development Bank Department of Research and Chief Economist June 2025 Cataloging-in-Publication data provided by the Inter-American Development Bank Felipe Herrera Library Arce, Fernando. Tax revolts and sovereign defaults / Fernando Arce, Jan Morgan, Nicolas Werquin. p. cm. — (IDB Working Paper Series ; 1719) Includes bibliographical references. 1. Financial crisis-Mathematical models-Argentina. 2. Tax and expenditure limitations-Mathematical models-Argentina. 3. Fiscal policy-Mathematical models-Argentina. I. Morgan, Jan. II. Werquin, Nicolas. III. Inter-American Development Bank. Department of Research and Chief Economist. IV. Title. V. Series. IDB-WP-1719 http://www.iadb.org Copyright © 2025 Inter-American Development Bank ("IDB"). This work is subject to a Creative Commons license CC BY 3.0 IGO (https://creativecommons.org/licenses/by/3.0/igo/legalcode). The terms and conditions indicated in the URL link must be met and the respective recognition must be granted to the IDB. Further to section 8 of the above license, any mediation relating to disputes arising under such license shall be conducted in accordance with the WIPO Mediation Rules. Any dispute related to the use of the works of the IDB that cannot be settled amicably shall be submitted to arbitration pursuant to the United Nations Commission on International Trade Law (UNCITRAL) rules. The use of the IDB's name for any purpose other than for attribution, and the use of IDB's logo shall be subject to a separate written license agreement between the IDB and the user and is not authorized as part of this license. Note that the URL link includes terms and conditions that are an integral part of this license. The opinions expressed in this work are those of the authors and do not necessarily reflect the views of the Inter-American Development Bank, its Board of Directors, or the countries they represent. Abstract* Political crises often coincide with fiscal crises, with complex causal dynamics at play. We examine the interaction between tax revolts and sovereign risk using a quantitative structural model calibrated to Argentina. In the model, the government can be controlled by political parties with different preferences for redistribution. Households may opt to revolt in response to the fiscal decisions of the ruler. While revolts entail economic costs, they also increase the likelihood of political turnover. Our model mirrors the data by generating political crises concurrent with fiscal turmoil. Specifically, we find that our model aligns closely with the conditions observed during the Macri administration (2015-2019). We find that left-leaning parties are more prone to default upon entering office, while right-leaning parties issue more debt. Our framework explains the high deficits observed during the Macri administration as well as the sovereign default that occurred immediately after the left regained power. JEL classifications: E32, E44, F41, G01, G28 Keywords: Civil unrest, Financial crises, Sovereign default, Redistribution ∗ We thank Manuel Amador, Cristina Arellano, Gadi Barlevy, Javier Bianchi, Emilio Espino, Stelios Fourakis, Mike Golosov, Grey Gordon, Tim Kehoe, Ilenin Kondo, Andy Neumeyer, Juan Pablo Nicolini, Leonardo Martinez, Matias Moretti, Francisco Roldan, Cesar Sosa-Padilla, Monica Tran-Xuan, and Jing Zhang, as well as, seminar participants at the NBER Emerging Markets Spring Meeting 2024, the SED 2024, SCIEA 2024, and the IMF Fiscal Affairs Workshop, for their useful comments. We are also thankful to Sophia Lansell and Caydn Hillier for their excellent research assistance. The views expressed herein are those of the authors and not necessarily those of the Federal Reserve Bank of Chicago, the Federal Reserve System, or the Inter-American Development Bank, its Executive Board, or its Management. All errors are our own. 1 Introduction Fiscal decisions are made by public officials aligned with specific factions of the nation. In practice, tax reforms and fiscal adjustment decisions often provoke civil unrest. While the literature on sovereign default has explored various facets of political risk since the seminal work of Hatchondo et al. (2009) and Hatchondo and Martinez (2010), the theoretical insights from Acemoglu and Robinson (2001) on how the risk of civil conflict constrains governmental decisions have yet to be integrated into a quantitative model of sovereign risk. This paper aims to fill that void. Protests differ from other forms of political backlash along crucial dimensions relevant to sovereign risk. Unlike elections, protests allow opposition members to express their dissatisfaction immediately, often leading to subsequent declines in reelection rates for the incumbent (Madestam et al. (2013)). Moreover, protests disrupt economic activity for both demonstrators and the wider economy, introducing significant strategic considerations for all involved parties. Protesters must be willing to bear the costs, while governments may adjust fiscal policy to mitigate unrest. Furthermore, since protests can alter the reelection prospects of incumbents, they can also shape lenders’ expectations regarding future government preferences, directly impacting the pricing of public debt. We highlight the significance of these dynamics by examining Argentina, particularly during the presidency of Mauricio Macri from December 2015 to December 2019, while also showing that the relationship between political and sovereign risk holds for a cross-section of countries. We show that episodes of heightened civil unrest, which we refer to as revolts in this paper, are associated with increases in sovereign spreads, i.e., fiscal crises. We develop a quantitative sovereign debt model in the tradition of Eaton and Gersovitz (1981) and Arellano (2008), supplementing it with a non-linear tax framework inspired by Heathcote et al. (2017) and incorporating civil conflict dynamics from Acemoglu and Robinson (2001). Heterogeneous households participate in production and are ruled by political parties with different preferences for redistribution that alternate in office. Furthermore, households can strategically respond to the government’s fiscal choices by staging revolts. Revolts lower productivity but decrease the probability that the incumbent party will stay in power. As in the data, we note that political and fiscal crises often coincide. Our model uncovers two channels linking revolts and sovereign risk. First, political conflict can elevate default risk by increasing the probability of transitioning from a ruling party with a low default rate to one with a higher rate. This mechanism generates a testable prediction that coincides with the Argentinean experience in 2020. Defaults are more likely to follow 1 a transition from right-wing to left-wing governance, and protests make these transitions more likely. Second, revolts can mitigate default risk by shortening the expected tenure of incumbents who choose to default. Indeed, we find that in equilibrium, revolts are more common during defaults than under repayment, allowing them to function as an endogenous default cost. The lenders anticipate this additional cost, and offer a more favorable bond price schedule to an incumbent in our baseline model relative to an environment with political turnover but without revolts. Our quantitative analysis shows that, while the two channels offset each other in terms of default frequency, revolts lead to increased overall indebtedness. In equilibrium, governments utilize the fiscal space created by revolts to take on more debt. Although both parties borrow more, right-wing incumbents—who face a higher frequency of protests—are notably more prone to increase deficits. As a result, it becomes more likely that a left-wing government will inherit a debt load misaligned with its distributional priorities, prompting it to default. Related literature This paper connects to several strands of the literature on sovereign default, political economy, and public finance. Our primary contribution lies within the political economy of sovereign default, focusing on the risks associated with political turnover. 1 Our paper builds on the literature, originating with Cuadra and Sapriza (2008) and Hatchondo et al. (2009), that studies the interaction between political economy and sovereign default—to which we add a redistributive motive for fiscal policy. The exogenous component of political turnover in our model resembles Hatchondo et al. (2009) and Cotoc, Johri and Sosa-Padilla (2021). In these papers, however, the asymmetry in default behavior across parties stems from exogenous differences in their discount factors or probability of reelection. 2 By contrast, in our model, both parties share identical discount factors, default costs, and exogenous probability of staying in or regaining power—parties differ only in their preferences regarding the redistributive effects of fiscal policy. Thus, absent endogenous revolt choices, both parties in our model would default at the same rates. In other words, our paper endogenizes differences in default rates across parties without hard-wiring this result via differences in exogenous party characteristics. The recent works of Andreasen et al. (2019), Azzimonti and Mitra (2023b), and Scholl 1 Our model will, however, also draw elements from the strand of the literature that focuses on sovereign reputation (Amador and Phelan (2021), Fourakis (2023), Morelli and Moretti (2023). 2 Asymmetric defaults are also present in the model of endogenous political turnover of Chatterjee and Eyigungor (2019), where it arises from the ability to misappropriate public funds for private gain and from informational frictions concerning the effect of government policies. These issues are outside the scope of our paper. 2 (2024) are the most closely aligned with our paper. 3 Similar to these papers, we present a model that integrates political constraints, redistributive policies, and the dynamics of sovereign default. Our model complements this literature in two main ways. First, protests endogenize the likelihood of political turnover and simultaneously constrain government policies. They serve as an economically costly means for households to penalize incumbents. Thus, the mere threat of protests influences the strategic decisions of the government, even though our framework is flexible enough to generate revolts along the equilibrium path at rates consistent with empirical data. By contrast, Andreasen et al. (2019) and Azzimonti and Mitra (2023b) impose by construction that government fiscal policies satisfy political feasibility constraints (e.g., parliamentary approval, or corruption in the form of financial incentives channeled to certain group leaders). Therefore, coerced political transitions never actually happen in equilibrium in these papers. 4 Our political economy setting is closer in this sense to that of Scholl (2024), in which the probability of elections is exogenous but the election outcome depends on the redistributive choices made by the incumbent party. Instead, we endogenize turnover via economically costly and strategic revolts, rather than peaceful transfers of power through scheduled elections. Second, Andreasen et al. (2019), Azzimonti and Mitra (2023b), and Scholl (2024) all assume that government debt has a short (one period) duration. By contrast, we allow governments to issue long-term debt. Beyond mere realism, incorporating long-term maturity to analyze defaults in an environment with political frictions yields three important implications. First, it allows us to quantitatively approximate the levels and volatility of debt and spreads observed in empirical data. 5 Second, long maturities imply that the borrowing policies of the opposition party affect the pricing of the incumbent’s debt. In a one-period debt model, only the default policies (and not the borrowing choices) of the opposition influence the price of the incumbent’s debt. Political frictions thus interact with both default and debt-dilution incentives. Third, long debt maturities reduce exposure to rollover risk. A model where the entire stock of debt is due each period can overestimate the impact of political turnovers on default risk. In Section 7.1, we demonstrate that right-to-left transitions are the primary channel through which political risk can escalate default risk, emphasizing the importance of accurately estimating this channel. Finally, our paper builds on an extensive theoretical literature regarding the economic 3 See also their related work, Azzimonti and Mitra (2023a), Scholl (2017), and Scholl and Hermann (2024). 4 In addition, redistribution in our model is subject to an efficiency-equity trade-off that affects aggregate output and, consequently, the government’s repayment capacity. This variation in repayment capacity leads to the “political defaults” proposed by Hatchondo et al. (2009), which are defaults that occur only because of political turnover and are, in our setting, facilitated by protests. 5This result is established in Chatterjee and Eyigungor (2012) and Hatchondo and Martinez (2009). 3 effects of political risk and conflict, particularly in relation to taxation and redistribution; e.g., Alesina and Tabellini (1989), Acemoglu and Robinson (2001), Battaglini and Coate (2008), Yared (2010), Acemoglu et al. (2011), Scheuer and Wolitzky (2016), Dovis et al. (2016), Barbera and Jackson (2020), among others. In particular, recent contributions to this field have focused on models of regime change or global coordination games to study protests. While these issues are certainly important, they fall outside the scope of our present analysis. Furthermore, the analytical insights presented in these studies are derived within deliberately simplified models. By contrast, our paper is quantitative in nature and aims to simultaneously match the data on interest rate spreads, defaults, political transitions, and protests. 2 Empirical Motivation 2.1 Cross-country Evidence Following the literature, we document the positive relation between political risk and spreads across different countries. Hatchondo and Martinez (2009) were first to highlight the importance of political risk measured by the International Country Risk Guide (ICRG) indicator, which they interpret as capturing the effect of governmental turnover on sovereign spreads. They study the 2001 Argentine default episode, pointing to outcomes of high government turnover driven by popular dissatisfaction. Similarly, Trebesch (2019) uses ICRG and length of renegotiation to argue that intense political turmoil makes restructuring more difficult. We confirm the positive correlation between political risk and spreads in the cross-country panel regression of countries with a credit rating below A-, presented in Table 1.6 The positive association between political risk and sovereign spreads persists even after accounting for macroeconomic fundamentals (such as Current Account Balance, Reserves, Real Gross Domestic Product (GDP) growth, and Primary Balance), as well as time and country fixed effects, as shown in Table 1. In Appendix G, we show that this relationship is also present regardless of the party in power, albeit with a stronger effect when the incumbent is a right-wing party. Figure 1shows the fitted values of the regression with all the aforementioned macroeconomic controls and the fixed effects. Additionally, in Figure 2, we further investigate this relationship in the data by focusing 6 The source for credit default swaps (CDS) spreads is CMAN obtained from Bloomberg Per Security Data and the Bloomberg terminal, used with the permission of Bloomberg. This regression only includes countries whose foreign-exchange debt was rated below Aby Standard and Poor’s in July 2024. We conduct the same exercise while keeping in countries of all ratings in Appendix Gand find similar results. 4 rates are monotonically increasing (decreasing) with income. 9 Notice that if the tax schedule is progressive, low-income households receive transfers, i.e., τ ( y ) < 0, whenever y < τ1/τ1 0 . Thus, for a given rate of progressivity τ1> 0, a higher parameter τ0 maps into both a higher base and a higher level of transfers to low-income households. Households are hand-to-mouth. When making their decisions, they take wages and all other aggregate states of the economy—including the fiscal package proposed by the party in power—as given. A household of type iwith wage wisolves Ui= max C,N ui(C, N) (1) s.t. C=wiN−τ(wiN). We obtain that the optimal choice of labor supply is given by Ni= [ψi(1 −τ1)]ψ/(1+ψ). Thus, labor supply is decreasing in the rate of progressivity τ1 . It does not explicitly depend on the wage rate wi because the income and substitution effects on labor supply offset each other with a log utility of consumption. We let yi = wiNi be the labor income and Ui the indirect utility of households of type i. 3.2 Representative Firm A representative firm hires both types of households to produce the final good. The production function is CES with an elasticity of substitution between the two types of labor η < 1. The total factor productivity α ( A, D,R ) depends on an exogenous and persistent shock A , the government’s current standing with its creditors D ∈ { 0 , 1 } , and the households’ revolt decision R∈ {0,1}. The firm produces Y=α(A, D,R) [(θLNL)η+ (θRNR)η]1/η We assume that the function α ( A, D,R ) is strictly decreasing in both D and R . That is, both defaults—which last until the country recovers its good standing with creditors—and revolts lead to a reduction in the economy’s output. We assume the firm maximizes its flow profits. The labor market is competitive. In 9 The parameter −τ1 is then equal to the elasticity of the household’s “keep rate”, i.e., 1 −τ′ ( y ), with respect to labor income y. 11 equilibrium, wages of households of type iare given by wi=1 + θjNj θiNiη(1−η)/η α(A, D,R)θi.(2) Expression (2) shows that both defaults and revolts lead to income losses for the households. Note also that by influencing the decision of households to revolt, the tax schedule affects not only their disposable incomes (“redistribution”) but also their pre-tax wages (“predistribution”). More generally, wages, labor supplies, and indirect utilities depend on the exogenous shock A, the fiscal package D,τ, and the revolt decision R. 3.3 Parties Party Ideology. There are two political parties, left and right, indexed by i∈ {L, R} . As for households, whenever we refer to party i , we denote by j the opposition, and vice versa. Both parties are benevolent and strategic, but they differ in their redistributive tastes. Specifically, they both evaluate social welfare according to a weighted utilitarian criterion, but they assign different welfare weights to the two types of households. We denote by Ω i|i and Ω j|i the welfare weights respectively assigned to households of type i and j by party i , with Ω i|i> 1 / 2 and Ω i|i + Ω j|i = 1. Thus, the right-wing (left-wing) party places a higher Pareto weight on the high-skilled (low-skilled) households—implying in turn that households R ( L ) optimally support the policies of party R ( L ). Besides these heterogeneous redistributive preferences, both parties are identical; in particular, they have the same discount factor β and face the same exogenous default costs. Political Turnover. Each political party is either the incumbent (in-office) or the opposition (out-of-office). The probability that the incumbent party i stays in power next period depends on the households’ decision to revolt, πi|i ( R ). The probability that the incumbent party i is ousted and replaced by party j is then equal to πj|i ( R ) = 1 −πi|i ( R ). We assume that πi|i ( R ) is strictly decreasing in R . Thus, the probability of reelection, without revolts, is the same for both parties; it is unaffected by the default status, but it is strictly lower if a revolt takes place in the current period.10 10 This reduced form assumption allows the model to match the empirical evidence showing that incumbents facing social unrest are reelected at lower rates (Madestam et al. (2013)). A model where the median voter is dissatisfied with the social unrest and at least partially blames the incumbent for it could rationalize this assumption. 12 Fiscal Policy. We denote by B the current level of debt that the party in power, i , inherits from the previous period. If the country is currently in good standing with its creditors, party i gets to choose whether to default, D ∈ { 0 , 1 } , an end-of-period level of debt B′ , and a tax-and-transfer policy τ consisting of a level of transfers τ0 and a rate of progressivity τ1 . 11 If the party defaults or has defaulted in the past and has not yet recovered its good standing with creditors, then D = 0 and B′ = 0, i.e., the country cannot borrow on international markets. In this case, the tax schedule τ must balance the current-period budget. We use the shorthand notation Fi≡ {D, B′, τ}to denote the fiscal package chosen by party i. Note the sovereign decides how much debt to issue before knowing whether households will riot—but this outcome will in turn affect the price of borrowing. As we explain below, the level of transfers τ0 needs to adjust ex post in order to ensure that the government budget constraint holds. For the sake of simplicity we omit the argument Rfrom τ0in the sequel. We assume that debt is a long-term contract promising a stream of exponentially declining coupon payments. Specifically, a unit of the bond issued at time t promises to pay (1 − δ ) t+s−1 ( δ + z ) units of the consumption good in period t + s . The price of the newly issued bonds faced by party i , denoted by Qi ( A, B′,R ), is then a function of the exogenous total factor productivity A , the level of debt announced for next period B′ (conditional on which the initial stock of debt B and the tax policy τ are irrelevant), and the households’ revolt decision R∈ { 0 , 1 } . Note that the ideology of the party in office i∈ {L, R} matters explicitly for the bond price, since this variable is persistent and therefore informative about future borrowing and default choices. The government budget constraint then reads 0 = X k∈{L,R} τ(yk(A, D,R, τ1))+Qi(A, B′,R)[B′−(1 −δ)B]−(δ+z)B+κ(B, B′)(1−D). (3) The first term on the right-hand side is the tax revenue, consisting of the sum of taxes levied on the labor incomes yL, yR of the two types of households. Recall that labor income depends on the variables A, D,R via the total factor productivity α ( · ), and on the rate of progressivity τ1 that affects labor supply decisions. The second term on the right-hand side is the government’s debt balance. It is non-zero only if the government is able to borrow, i.e., if it is in good standing with its creditors so that D = 0. The last term in the curly brackets ( κ ( B, B′ ) is a convex portfolio adjustment cost that penalizes the government for large changes in the stock of debt. We add this term to avoid the well-known issue of extreme dilution immediately before a default; at the calibrated values, less than 0.06% of output is 11 The opposition party j does not make a fiscal decision but still receives a utility flow according to its own preferences. 13 spent on this adjustment cost. Note that a revolt has the ex-post effect of altering the price of bonds and tax revenue. However, given a specific revolt choice, the budget constraint (3) renders one of the fiscal variables redundant. The choices of end-of-period debt B′ (possibly constrained to zero if D = 1) and tax progressivity τ1 uniquely determine the level of transfers τ0 . We assume that the government commits to the levels of borrowing and tax progressivity, but once the revolt decision is realized, it adjusts the transfers ( τ0 ) to satisfy the budget constraint. Following a repayment of its debt, the government keeps its good standing and can borrow in the next period. By contrast, after a default, the government is unable to borrow until it recovers its good standing and can re-enter credit markets. We assume that this happens with an exogenous positive probability γ∈ (0 , 1) in each period. Letting G ( D ) denote the probability that the government is in good standing in the current period, we thus have G(D) = 1 if D= 0, and G(D) = γas long as D= 1. Taste Shocks. The final ingredient of the model are two privately observed taste shock vectors ( εg, εh ) that affect the political parties and the households in each period. The government’s shock εg ( B′,τ ) is a vector containing all the potential fiscal packages that are available to the government. 12 The households’ shock εh ( R ) ∈R2 is a two-dimensional vector associated with the costs of revolting and of accepting the fiscal package. We draw the taste shocks from a generalized type-one extreme value distribution with scale parameters µg, µh and correlation coefficients ρg, ρh for the government and the households, respectively, with ρh = 0. The shocks are independently and identically distributed (i.i.d.) over time and uncorrelated to each other. 3.4 Value Functions We write the model in recursive form and “primed” variables (e.g., B′ ) always represent the next-period values. We denote by W the value functions of the political parties and by V the value functions of the households. Whenever a variable has a superscript of the form i|i or j|i , the second variable ( i in these two examples) denotes the party that is currently in office. 12 For computational simplicity, we assume that the potential choices of end-of-period debt B′ and tax progressivity τ1 can only take a finite set of values: B′∈ B ≡ {B1, B2, ..., BNB} and τ1∈ T ≡ {τ1, τ2, ..., τNτ} . There are thus NB×Nτ possible fiscal packages in repayment and Nτ possible packages in default. Hence, εg∈R(NB+1)×Nτ. 14 Parties. Suppose that the country is in good standing and that party i is in power; hence, party j is in the opposition. We let Wi|i and Wj|i denote the value functions of parties i and j respectively, given that i is in office. Moreover, we let Wi|i D and Wj|i D be the values obtained when party irepays the debt (D= 0) or defaults (D= 1). The party in office, i , chooses whether to repay or default, so that its value function satisfies Wi|i(A, B, εg) = max nWi|i 0(A, B, εg) ; Wi|i 1(A, εg)o,(4) where the state variables are the exogenous level of productivity A observed at the beginning of the period (and inherited from the previous period), the initial stock of debt B , and the vector of taste shocks εg . The initial stock of debt becomes irrelevant once the government has decided to default ( D = 1). We denote by Di ( A, B, εg ) ∈ { 0 , 1 } the resulting policy function that determines the decision to default or repay by the party in power i . For the opposition party j , there is no optimization problem: It takes the policy functions of the incumbent iand of the households as given. Its value function is then given by Wj|i(A, B, εg) = (1 − Di(A, B, εg))Wj|i 0(A, B, εg) + Di(A, B, εg)Wj|i 1(A, εg).(5) We now characterize the values of repayment and default. For simplicity’s sake, we only derive those of the party in power, Wi|i 0, Wi|i 1 , and omit those of the opposition, Wj|i 0, Wj|i 1 . Party i makes its fiscal decisions taking into account the households’ reaction function, i.e., whether to revolt or not. However, it cannot perfectly predict whether a revolt will happen, since it does not observe the taste shocks εh . We denote by P ( R|A, Fi ) the probability that households of type j make the decision R∈ { 0 , 1 } given the productivity shock A , which affects their wages, as well as the incumbent party i ’s choice of fiscal package Fi = {D, B′,τ} . In case of repayment, we have Wi|i 0(A, B, εg) = (6) max B′,τX R∈{0,1} P(R|A, Fi)  X k∈{i,j} Ωk|iUk(A, Fi,R) + βEhWi|i′(A′, B′, εg′)A, Ri   +εg(Fi) subject to the budget constraint (3) with D= 0. In case of default, we have Wi|i 1(A, εg) = (7) max τX R∈{0,1} P(R|A, Fi)  X k∈{i,j} Ωk|iUk(A, Fi,R) + βEhWi|i′(A′,0, εg′)A, Ri   +εg(Fi) 15 subject to the budget constraint (3) with D = 1. In these expressions, εg ( Fi ) is the value of the taste shock εg , given the chosen fiscal package Fi = {D, B′,τ} , and the expectation inside the curly brackets is taken over future productivity A′ and the ideology of the next government i′ , given the current productivity shock A and the household decision R . To characterize this continuation value, recall that party i is ousted from office with probability πj|i ( R ) ∈ (0 , 1); thus, when choosing the fiscal package, the incumbent party internalizes that its probability of staying in power is strictly smaller if citizens decide to revolt. In addition, recall that the government will be in good standing with creditors in the next period with probability G ( D ), which is equal to 1 if it repays its debt ( D = 0), but is strictly lower than 1 if it defaults (D= 1). We can thus write EhWi|i′(A′, B′, εg′)i(8) =EA′|A X i′∈{i,j} πi′|i(R)nG(D)Wi|i′(A′, B′, εg′) + (1 − G(D))Wi|i′ 1(A′, εg′)o,  where B′= 0 if D= 1. Households. Given the state of the economy, the fiscal package Fi = {D, B′,τ} chosen by the incumbent government of type i and their taste shock εh , households of the opposite type j=imake the revolt decision.13 They solve: Vj|i(A, Fi, εh) = max R∈{0,1}Vj|i R(A, Fi, εh),(9) where Vj|i 0, V j|i 1 are the values of revolting and accepting the fiscal package, respectively. We denote by Rj ( A, Fi, εh ) the corresponding policy function. By contrast, the households that support the party in office do not make a revolt decision and take all of the policy functions as given. Their value function is then given by Vi|i(A, Fi, εh) = (1 −Rj(A, Fi, εh))Vi|i 0(A, Fi) + Rj(A, Fi, εh)Vi|i 1(A, Fi).(10) 13 This assumption is without loss of generality in a Markov equilibrium. Revolting against a party of the same type as the households is not a credible threat, since it would decrease current wages and increase the odds of a future incumbent who places a lower welfare weight on the household. 16 In turn, the values of revolting and accepting the fiscal package for household jsatisfy Vj|i R(A, Fi, εh) = Uj(A, Fi,R) + βEA′|A X i′∈{i,j} πi′|i(R)Vj|i′(A′,F′i′, εh′) +εh(R),(11) where εh ( R ) is the value of the taste shock associated with the revolt or acceptance decision R . Revolts therefore lead to a decline in wages and hence in utility, in the current period but increase the odds that a household’s preferred party will be in power next period. The value functions Vi|i 0, V i|i 1 for households of type i , who support the incumbent, can be derived analogously. 3.5 Foreign Lenders A continuum of deep-pocketed, risk-neutral, and competitive international lenders can buy the government’s bonds. Lenders have access to a one-period risk-free rate bond that pays interest rate r . As is standard in the literature, lenders are forward-looking and price the risk of default and debt dilution. Moreover, in our environment, lenders also internalize that the government’s redistributive preferences vary by party—and hence change over time—and that revolts decrease the odds of an incumbent staying in power. When party i is in office, the bond price that satisfies the lenders’ zero-profit condition is given by Qi(A, B′,R) = 1 1 + rEA′|A"1−X i′∈{i,j} πi′|i(R)Di′(A′, B′, εg′) (12) ×(δ+z+ (1 −δ)X R′∈{0,1} P(R′|A′,Fi′)Qi′A′, B′′ i′(A′, B′, εg′),R′)#, where B′′ i′ is the policy function that determines next-period’s borrowing of party i′ . The price of debt therefore depends on the probability of reelection of the incumbent, not only because of the default decision next period (as in, e.g., Scholl (2024)), but also because the ideology of the incumbent changes the level of future debt issuance and therefore the probability of future defaults. Long-term debt also implies that future political instability (represented by the probability of future revolts) also has an effect on the price of current bonds. In a model with one-period debt, both of these channels would be absent. 17 3.6 Recursive equilibrium definition Definition 1. Markov Perfect Equilibrium (MPE). An MPE is defined by value functions {Wi|i, Wj|i, Wi|i D, Wj|i D, V i|i, V j|i, V i|i R, V j|i R} , policy functions {Di, B′i,Fi, Ni,Ri} , and prices {wi, Qi}, for all (i, j)∈ {L, R}2and (D,R)∈ {0,1}2, such that: 1. Households’ policy functions solve (1) and (9)-(11). 2. Parties solve the dynamic programming problems (4)-(8). 3. Wages are given by (2). 4. Bond prices are given by (12). The logistic shocks from Dvorkin et al. (2021) allow us to find closed-form solutions for all policy functions, and value functions in expectation of the taste shocks. We solve the model numerically using value function iteration. 4 Quantitative Analysis The model is calibrated at the quarterly frequency using Argentine macroeconomic data. A first set of parameters to values that are either standard in the literature or based on historical Argentine data. We internally calibrate the remaining parameters to match relevant moments for Argentina’s sovereign spreads, political turnover, frequency of revolts, and other business-cycle statistics. Table 2summarizes the parameters set outside the model. We take the first set of parameters from sovereign default models calibrated to Argentina. The quarterly risk-free real interest rate r is set to 0 . 01, a standard value for this period. The inverse Frisch elasticity is ψ = . 5, in line with the values used by Arellano et al. (2017) and Arellano and Bai (2017) on sovereign debt models with labor. The maturity rate δ = 0 . 05 and its coupon value z = 0 . 03 are set to the values used by Chatterjee and Eyigungor (2012) who also study Argentina and match the average maturity of the debt of 5 years and the debt service. Similarly, we assume that the productivity shock follows an AR(1) process given by ln ( At ) = ρAln ( At−1 ) + ϵA t with ϵA t∼N (0 , σA ). Once again, we use Chatterjee and Eyigungor (2012) estimates of parameters of an AR(1) endowment income process on detrended GDP data. We keep the persistence at their values ρA = 0 . 95, and we adjust the volatility of innovation such that the simulated volatility of output matches that of the data: σA = 0 . 03. The reentry parameter is set to γ = . 0385, which corresponds to an average exclusion period 18 from credit markets after default of 6 years and 6 months. 14 We use Morelli and Moretti (2023) estimates of political change in Argentina as our measure of the average probability of reelection without revolts (i.e., πi|i ( R = 0) = . 969). Without revolts, this corresponds to an average tenure in office of 8 years for each political party. 15 The reelection probability under sustained revolts ( πi|i ( R = 1)) is a key determinant of the benefits of revolting and as such will be internally calibrated. We take a second set of parameters from the literature on skill premia and inequality in Latin America. Gallego (2006) analyzes 40 years of skill premium data in Chile following the same method as Krusell et al. (2000). They measure a labor elasticity between skilled and unskilled labor of 1 . 5 that is consistent with η = 0 . 66, in line with estimates for the United States. We use data on hourly wages by education group in Argentina from the SocioEconomic Database for Latin America and the Caribbean (SEDLAC) dataset (SEDLAC (2023)). The data are available biannually from 2003 to 2021. The dataset splits the Argentinean labor force into three groups with different years of formal education (fewer than 8 years, between 8 years and 13, and more than 13 years). For each group, we have their size, hourly wages, labor hours, and finally total net labor income. We use this to divide the labor force into two half-tiles of equal size. We follow Heathcote et al. (2017) and use equation (2) to compute the model prediction for the ratio of hours: NR NL=ψR ψLψ 1+ψ . We use average hours for each type of half-tile in the data to estimate ψR, ψL , finding that NR NL = 1 . 15. Normalizing by the mass of households and using our estimate of ψ , we obtain ψR = 0 . 60 and ψL = 0 . 40. Similarly, we know from equation (2) that the ratio of pre-tax wage in the model is: wR wL=(θR)η (θL)ηNR NL1−η . Using the previous result for the ratio of hours, along with our estimate of η , we estimate θR θL = 2 . 3, and after being normalized to the sum to one, this yields θR = 0 . 70 and θL = 0 . 30. Table 3shows the parameters of the model that we calibrate internally. The stochastic discount factor ( β ) is the same for both parties and the households. We follow Dvorkin et al. (2021) and assume a Generalized Type One Extreme Value distribution with scale parameter 14 This number corresponds to the average length of debt renegotiation period across multiple Argentine defaults and is computed in Chatterjee and Eyigungor (2012) using data from Benjamin and Wright (2009). 15 As in the United States, Argentina’s presidential elections are held every four years, and only one reelection is permitted. 19 Table 2: Parameters Estimated outside of the Model Parameter Value Source/Transition Risk-free rate r=.01 Standard value Inverse Frisch elasticity ψ=.5 Standard value Productivity shock ρA=.95 Chatterjee and Eyigungor (2012) log(At) = ρAlog(At−1) + ϵA tσA=.03 Argentina’s GDP Debt maturity δ=.05 Avg. maturity of debt Debt coupon z= 0.03 Debt service Reentry probability γ= 1/26 Average renegotiation length Reelection odds under stability πi|i(R= 0) = 1 −1/32 Morelli and Moretti (2023) Elasticity of substitution η= 2/3Gallego (2006) Labor productivity θR=.70, θL=.30 Hourly wage premia Taste for effort ψR=.60 Hours top education half-tile Taste for effort ψL=.40 Hours bottom education half-tile σϵG and correlation ρϵG for the fiscal taste shock and scale parameter σϵHH for the revolt decision of the households. 16 We also take the functional form of the portfolio adjustment cost of debt from Dvorkin et al. (2021): ι(B′, B) = ι1exp(ι2|B′−B|)−ι1). At the calibrated parameters, less than 6 e− 4 of output is spent on these costs. To assess the effect of revolts and defaults on productivity, we borrow the functional form of the default costs from Chatterjee and Eyigungor (2012), and assume a similar transformation for the revolt costs:17 α(A, D= 0,R= 0) = A, α(A, D= 0,R= 1) = A−max(ϕR 0A+ϕR 1A2,0), α(A, D= 1,R= 0) = A−max(ϕD 0A+ϕD 1A2,0) ≡αD R0, α(A, D= 1,R= 1) = αD R0−max(ϕR 0αD R0+ϕR 1αD R0 2,0). (13) These transformations add four parameters to calibrate internally. The first two, ϕD 0, ϕD 1 , correspond to the exogenous default costs common in the sovereign default literature. The other two, ϕR 0, ϕR 1 , represent the analogous penalty that the economy suffers during a revolt. 16 The households’ taste shock is a scalar and not a vector; thus, it has no correlation parameter. We allow for the government’s taste shocks to be correlated across fiscal packages (ρϵG= 1) 17 In Appendix Dwe verify that the main results of the paper are robust to alternative specifications of the revolt costs. In particular, we show that our results would hold qualitatively under utility costs of revolts. 20 0.7 0.75 0.8 0.85 0.9 0.95 1 1.05 1.1 1.15 1.2 0.7 0.75 0.8 0.85 0.9 0.95 1 1.05 1.1 1.15 1.2 (a) Borrowing Policy Functions 0.7 0.75 0.8 0.85 0.9 0.95 1 1.05 1.1 1.15 1.2 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5 (b) Tax progressivity. Figure 8: Expected fiscal packages by party Note: Panel (a) depicts the expected end-of-period debt ( B′ ) by party affiliation of the ruler as a function of the initial level of debt while keeping TFP fixed at mean values. Since credit markets are only available in repayment, we only plot the borrowing choices when the ruler’s repayment probability is above 5%. Panel(b) depicts the expected tax progressivity ( τ′ 1 ) by party affiliation of the ruler as a function of the initial level of debt while keeping TFP fixed at mean values. Since the progressivity decision depends on the default decisions, tax decisions are depicted with a solid line when the ruler’s repayment probability is above 5% and with a dotted line when the probability of default is above 5%. Revolt probabilities: Panel (a) of Figure 9depicts the probabilities of revolt against each party as a function of the initial level of debt. In line with previous figures, we differentiate between policy functions when the probability of repayment exceeds 5% (dashed lines) and when the probability of default exceeds 5% (dotted lines). This figure provides intuition for the positive association between political and fiscal crises, highlighting why this effect is particularly pronounced under a right-wing ruler. First, revolt probabilities show a discontinuous increase for both parties when transitioning from repayment to default, as the opportunity costs of revolt are significantly lower in default conditions. Second, even conditional on repayment, the probability of revolt against a right-wing ruler—when there is a positive probability of default—rises with the initial level of debt. This increased risk of revolt occurs because the debt levels at which a right-wing ruler is at risk of default fall entirely within the default sets of the opposition. As a result, the fiscal policies enacted by the right-wing ruler to sustain repayment at these debt levels are particularly intolerable to the opposition. In contrast, we do not observe a similar rise in revolt probabilities against a left-wing ruler near their default threshold, as their repayment decision is consistent with what the right-wing opposition would do in the same scenario. Right-wing governments face favorable schedules, but revolts worsen them: Panel (b) in Figure 9shows the price schedules of government debt, at the average TFP, by party 27 0.7 0.75 0.8 0.85 0.9 0.95 1 1.05 1.1 1.15 1.2 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 (a) Probability of Revolt 0 0.2 0.4 0.6 0.8 1 1.2 0 500 1000 1500 Spread (BPS) (b) Price of debt. Figure 9: Revolt probabilities and spreads Note: Panel (a) depicts the probability of default and revolt by party affiliation of the ruler as a function of the initial level of debt while keeping TFP fixed at mean values. Revolt probabilities depend on the default decisions; as such, we denote with dashed-line revolt probabilities when the ruler’s repayment odds are above 5% and with dotted lines the revolt probabilities when the default odds are above 5%. Panel (b) plots the policy-implied spreads as a function of future debt for both parties in the baseline model under stability (dotted lines) and under revolt (solid lines). We fix TFP at its mean value, and we fix initial debt to the mean level in the ergodic distribution. affiliation of the ruler and by revolt status of the opposition. Consistent with the default sets previously discussed, right-wing governments benefit from a more favorable price schedule compared with their left-wing counterparts. That is, for any given end-of-period debt choice, a right-wing ruler pays a lower interest rate spread relative to the risk-free rate than a left-wing ruler.24 Furthermore, when end-of-period debt is high, right-wing rulers face higher spreads when the households revolt ( R = 1) rather than opt for political stability ( R = 0). This occurs because forward-looking lenders anticipate that revolts heighten the likelihood of political turnover, and the observed debt levels fall within the default sets of the opposition. As a result, lenders demand higher spreads to compensate for this increased political default risk. 25 24 Cotoc et al. (2021) find that a similar result holds at the cross-country level on average when a country leans towards electing left-wing leaders. 25 A similar mechanism is present in reverse for the left-wing party in Appendix B. That is, revolts lower borrowing costs for left-wing governments. In the Appendix, we also show that, as in the data, most revolts have a very small effect on the spreads outside of fiscal crises. 28 7 Equilibrium Ergodic Distribution After presenting the policy functions for each potential incumbent, we now turn to the description of the ergodic equilibrium. Table 5provides a detailed breakdown of revolts and defaults in the ergodic distribution by party. We calculate the proportion of time spent in repayment ( D = 0) and default ( D = 1), both in total and when disaggregated by incumbent type. Additionally, we report the frequency of revolts ( R = 1) unconditionally and conditionally depending on the default status and the ideology of the ruler. Two asymmetries between the parties become apparent once again. Left-wing governments default more: At the ergodic distribution, 27 . 7% of periods are spent in default; specifically, 15 . 9% of periods are spent in default under a left-wing incumbent and only 11 . 8% of them are spent in default under a right-wing incumbent. Thus 72 . 3% of periods are spent in repayment, with this share almost evenly split between the two types of rulers. The fact that left-wing governments are more associated with defaults arises from the interaction between asymmetries in defaults sets (Figure 7) and political transitions. P[D= 0] P[D= 1] P[R= 1] P[R= 1|D = 0] P[R= 1|D = 1] Right-wing ruler 36.1 11.8 20.5 17.2 29.3 Left-wing ruler 36.2 15.9 16.1 11.0 29.3 Total 72.3 27.7 36.6 28.2 58.7 Table 5: Frequency of Each State in Percent at the Ergodic Distribution Note: The first and second columns report the frequency of repayment and default states, at the ergodic distribution. The third column shows the unconditional frequency of revolts, while the fourth and fifth columns report the conditional frequency of revolts within repayment and default states, respectively. The bottom line row the total frequencies at the ergodic, while the first and second rows show the decomposition by party ideology of the ruler. Revolts are more common during defaults: When credit markets are accessible, revolts occur 28% of the time. In contrast, during default periods, the frequency of revolts rises to 59%. Additionally, right-wing incumbents face a higher likelihood of revolts in all states, with this disparity being particularly pronounced during repayment periods. This asymmetry in revolt frequency ultimately leads to differences in the time each party spends in office. Although the exogenous productivity cost of revolts is independent of the incumbent’s ideology, the opportunity cost of revolting is higher for more productive (right-leaning) households, especially during repayment periods when productivity is relatively high. During default periods, average productivity is lower, which narrows the difference in the opportunity cost of revolting across all parties. 29 7.1 Political Defaults: Right-to-left-to-default Transitions Political transitions from right to left are associated with a heightened risk of default, as right-wing rulers can sustain higher levels of debt compared with their left-wing counterparts. To illustrate this, we examine default patterns following political turnovers within the ergodic distribution. Figure 10 traces the duration of each party’s tenure from the moment it assumes power until either a default occurs or power is transferred to the opposition. By this definition, most tenures last no longer than 60 periods. 10 20 30 40 50 60+ Periods since party change 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Cumulative fraction Default Political Default Left Turnover (a) Left Takeover 10 20 30 40 50 60+ Periods since party change 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Cumulative fraction Default Political Default Right Turnover (b) Right takeover Figure 10: Dynamics of default following a switch in the ideology of the ruling party Note: Plotted are the cumulative fractions of left-wing(right-wing) governments that have defaulted across 60 periods since coming into power. Each tenure is started by the left (right) coming into power in repayment. The cumulative fraction is computed by isolating events from a 100,000-period simulation, yielding 1,783 (1,586) takeovers by the left (right). Areas in purple correspond to states in which both types of incumbents choose to default. Areas in blue (red) are political turnovers where only left-wing (right-wing) incumbents default. We plot the cumulative number of tenures ending in turnover (gray shaded areas) and defaults over time for left-wing (panel (a)) and right-wing (panel (b)) governments. A distinction is made between political defaults (blue and red) and regular defaults (purple). Following Hatchondo et al. (2009), we define a political default as a default that occurs in the first period after a political turnover and in a state where the previous incumbent would have opted to repay. This exercise reveals a stark asymmetry: over 15% of transitions from right to left result in immediate defaults, almost all of which are political. These defaults account for nearly a third of all defaults observed under left-wing governments. 26 In contrast, no political defaults are observed when a right-wing party takes power. The similarity in the size of the purple 26 When a left-wing government inherits a substantial debt burden, its alternative to default is deleveraging through austerity measures. The dynamics of such policies under left-wing rule are explored in Appendix C. 30 areas in both panels suggests that political defaults are the primary driver of the asymmetry in default frequencies between left-wing and right-wing governments outlined in Table 5. Political defaults allow our model to contribute to the policy debate (IMF (2013) and Guzman et al. (2016)) that argues that countries default “too late.” 27 Political defaults occur when servicing the debt requires regressive policies that are acceptable to the right but not to the left. In the lead-up to such a default, a planner with the left’s preferences would have defaulted before the political transition takes place. In the Argentinean context this is reminiscent of the 2020 default, but also of the the rapturous applause in Congress after the motion to default passed in 2001(La-Nacion (2001)). 7.2 Revolts as an Endogenous Default Cost In our model, revolts are more than twice as likely to occur during default periods than during repayment periods. This feature effectively transforms revolts into an endogenous cost of default. Incumbents who default should expect a shorter tenure. To assess the strength of this deterrent, we compare the default dynamics of our baseline model with those of a model with only exogenous political turnover. To simplify the comparison, we set the exogenous probability of turnover to the values we use in our baseline under stability. Figure 11, panel (a) compares the default sets of a left-wing government in the baseline model (dark blue) and in the counterfactual model with exogenous turnover (light blue). The results indicate that allowing for revolts reduces the size of the default set. In other words, for certain combinations of productivity and initial debt, a left-wing incumbent would choose to default in the model with exogenous transitions but would opt to repay in our baseline model. In panel (b), we examine this mechanism in even more detail by focusing on one such initial state: we fix the productivity at its mean level and the initial debt to 96% of mean output at the ergodic distribution. The first pair of bars shows the difference between the value of repayment and the value of default for the baseline model (dark blue) and the model with only exogenous turnover (light blue). As expected, this difference is positive in the baseline (indicating that repayment is preferred) and negative in the exogenous turnover model (indicating that default is preferred). The second and third pairs of bars decompose this difference into the contributions of current-period utility and expected continuation values in the future, respectively. The results show that revolts have a limited effect on current-period values but significantly 27 A related branch of the academic literature has focused on default renegotiations that are inefficiently long; see, for instance, Bai and Zhang (2012), Benjamin and Wright (2009), and Asonuma and Trebesch (2016). 31 (a) Default Sets Left-wing Rule Total value Utility Continuation -0.06 -0.04 -0.02 0 0.02 0.04 0.06 (b) Repayment minus default Figure 11: Default decisions: Baseline and model with exogenous turnover Note: In panel (a) we plot the default sets of the baseline model (dark blue) alongside the default sets (light blue) of the model with exogenous turnover. In panel (b), we plot the difference between the value of repayment and default while fixing the taste shocks at zero, the TFP shock at its mean value, and the initial level of debt at 96% of mean output in the baseline. In this state the value of repayment is higher than the value of default in the baseline (dark blue)) economy, while the opposite is true for the economy with exogenous political turnover (light blue). We further decompose the total difference in values between the differences in current utilities and the differences in continuation values. impact continuation values. In the model with only exogenous transitions, default yields a higher value in both the current period and the future. By contrast, in the model with revolts, defaulting still provides higher utility in the current period, but these gains are outweighed by the losses in continuation values. This is because defaulting raises the likelihood of facing a revolt, which in turn reduces the probability of remaining in office.28 7.3 Implications of Revolts and Turnover for Debt and Default Risk To evaluate the quantitative impact of revolts on sovereign risk, we examine various specifications of the baseline model. Table 6provides a summary of the aggregate moments across six different model configurations. The first two specifications are models without political turnover, representing either a permanent rule by the right-wing party or the left-wing party. These models reveal that 28A potential concern is that in our model, reelection probabilities depend not only on default status but also on party affiliation. Moreover, revolts lead to a lower average tenure in office at the ergodic distribution compared with the model with purely exogenous turnover considered here. We address these issues in Appendix E, where we explore alternative specifications of the exogenous turnover model. Our analysis confirms that the results are primarily driven by differences in reelection rates based on default status. 32 in the absence of political turnover, the incumbent’s ideology does not significantly alter the average levels of debt, spreads, and defaults. This is because both parties face identical exogenous default costs and discount factors. Table 6: Moments Comparison between Models Model specification Debt Spread Freq. default Revolts Share in power(Right) Permanent left-wing 85.0 6.5 4.0 - 0.0 Permanent right-wing 86.7 6.3 3.9 - 100.0 Exogenous turnover 81.4 7.1 4.3 - 50.0 Revolts only in repayment 55.2 10.1 5.1 19.7 47.5 Revolts only in default 118.4 6.1 3.6 13.9 50.9 Baseline 89.6 7.3 4.4 36.6 47.9 Data 88.8 8.4 4.9 39.7 46.4 Note: We compare key moments of the data and the baseline model with the five alternative model specifications. The moments are computed using 100,000 simulations for each model specification. Revolts are not possible in the first three specifications and are thus not reported. We then return to the model with only exogenous political turnover discussed in the previous subsection. The reelection probability is kept at the value under stability in our baseline, and therefore each party averages an eight-year term in office. Relative to the models with permanent rule by one party, this specification shows lower levels of debt, higher spreads, and more frequent defaults. As in Hatchondo et al. (2009) we find that political turnover on its own can increase the frequency of sovereign defaults. 29 These additional “political defaults” happen in their framework when an impatient government takes over from a more patient one. We add to their insights by showing that political defaults can arise because of ideological differences in the efficiency-equity trade-off. Moreover, since in our calibration revolts against a right-wing ruler are more common, revolts can exacerbate default risk by increasing the likelihood of a political default.30 The final two specifications involve models where revolts are restricted to either repayment or default states. When revolts are allowed only during repayment periods, spreads increase, defaults become more frequent, and the economy maintains significantly lower levels of debt. In contrast, when revolts are allowed solely during default periods, higher levels of debt are sustained with lower default frequencies. In the first scenario, defaulting becomes more attractive as it can extend the incumbent’s tenure. In the second scenario, revolts 29 This result is also found in the sovereign default literature with reputation. In this literature, one type of government is assumed to be more myopic than the other. See, for instance, Morelli and Moretti (2023), Amador and Phelan (2021), Alfaro and Kanczuk (2005), and Cole et al. (1995). 30 Besides the differences in discount factors of Hatchondo et al. (2009), asymmetric default rates by party can emerge because of variations in exogenous reelection probabilities (as in Cotoc et al. (2021)) or in debt restructuring (as in Cruces and Trebesch (2013)). 33 serve only as an additional cost of default, as they present a new threat to the incumbent’s tenure. This threat enhances commitment, allowing the economy to sustain more debt. In our model, revolts are always possible but endogenously observed more frequently during defaults, effectively acting as an endogenous default cost. Consequently, relative to the model with only exogenous turnover, our baseline specification sustains significantly higher levels of debt at the cost of a modest increase in spreads and default frequency. These quantitative results indicate that revolts in our baseline model create additional fiscal space relative to the model with only exogenous turnover. However, this additional fiscal space is primarily utilized to increase the debt by around 5% of output. The next section explores the role of each party in contributing to this additional borrowing. 7.4 Revolts and Deficits: A Theory of “Gradualism” (a) Default Sets for the Right-wing Party 0 0.5 1 1.5 0 500 1000 1500 Spread Right-wing government, baseline Right-wing government, model with exogenous turnover Right-wing government, model with revolts only in repayment Right-wing government, model with permanent parties (b) Right-wing price schedule. Figure 12: Default sets and comparison of policy-implied spreads across model variants. Note: In panel (a) we plot the default sets implied by the policy functions corresponding to our baseline and to the model with exogenous turnover for the right-wing party exclusively. This is done by adopting the previous definition under which the shaded area represents an ex-ante probability of default that is greater than 0.5, conditional on being in good standing initially. In panel (b). we plot the schedule for the baseline model(solid line), the model with exogenous turnover(dashed line), the model without revolts in default(dashed line with asterisks), and the model with permanent types. For this purpose, we fix productivity and the initial debt at their mean value. Figure 12, panel (a), illustrates a comparison of the default sets for a right-wing ruler in our baseline model versus the model with only exogenous turnover. As before, revolts are associated with a smaller default set. This implies that, for a given debt level and productivity, a model accounting for revolts predicts lower default risk, which should influence lender 34 pricing. Panel (b) validates this by showing the price schedules faced by a right-wing ruler under the different model specifications outlined in section 7.3. Including revolts in the model leads to more favorable borrowing terms. However, as Table 6shows, the average level of spreads and the default frequency at the ergodic distribution in the baseline model are slightly higher than those in the model with only exogenous turnovers. This difference can be attributed to the fact that the average debt level in the baseline model is significantly higher than in the economy with exogenous turnovers. Indeed, both parties—but particularly right-wing ones—take advantage of the improved bond price schedules enabled by the revolts to issue more debt relative to an economy with only exogenous turnovers. 0 10 20 30 40 50 60 Periods since reentry 0 0.2 0.4 0.6 0.8 1 1.2 Left Baseline Left Exogenous Turnover (a) Debt under a Left-wing Party 0 10 20 30 40 50 60 Periods since reentry 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 Right Baseline Right Exogenous Turnover (b) Debt under a Right-wing party. Figure 13: Debt levels at the ergodic distribution after reentry. Note: Plotted in panel (a) (panel (b)) are the average levels of debt for left-wing (right-wing) governments that reenter with zero debt at period zero and remain in power and repayment up to 60 periods later at the ergodic distribution of our baseline model. The averages are computed by isolating events from a 100,000-period simulation. In both panels, we also compute the hypothetical borrowing decisions that the same party would have made in the model with only exogenous turnover given the same productivity shock. Figure 13 illustrates the contributions of each party to the additional borrowing observed in our baseline model. It shows the evolution of debt at the ergodic distribution following reentry to credit markets (starting with zero debt) for both left-wing (panel (a)) and rightwing (panel (b)) governments. As a benchmark, we compute counterfactual debt levels using the borrowing rules from the model with only exogenous turnover, applied under identical shocks and party conditions. The figure reveals that both parties issue more debt, though the increase is more pronounced for right-wing governments. Two mechanisms drive this behavior. First, as discussed in Section 7, right-wing governments in the baseline model face a 35 higher frequency of revolts during repayment periods. As a result, Figure 13 suggests that these governments borrow more to align fiscal policies with opposition preferences and mitigate unrest. This dynamic mirrors Argentina’ss “Gradualism” approach under the Macri administration, where concerns about social unrest led to fiscal concessions that expanded the budget deficit, when borrowing costs were still low. Second, the figure aligns with the political economy theory of “Starving the Beast.” 31 This theory posits that right-wing governments deliberately create larger budget deficits to compel future left-wing administrations to implement spending cuts after a turnover. Although the original theory does not address defaults, our model suggests that a left-wing party inheriting a large debt burden must choose between political default or adopt austerity measures to restore sustainability. In Appendix C, we confirm this by examining left-wing governments that inherit high debt levels without resorting to default. Finally, in Appendix Fempirically tests two key predictions derived from our model. First, we confirm that political transitions from right-wing to left-wing governments are associated with higher default risk, as evidenced by movements in interest rate spreads during these transitions. Second, we demonstrate that right-wing governments tend to issue more debt compared to their left-wing counterparts. 8 Conclusion We develop a quantitative model of sovereign debt that incorporates parties with distinct preferences for redistribution, political transitions, and social unrest. In our framework, revolts impose economic costs but increase the probability of political turnover. Governments with different ideologies strategically choose fiscal policies to satisfy their constituents and extend their time in office, while households strategically decide when to revolt. We calibrate the model to the economic and political context of Argentina. Specifically, we find that our model aligns closely with the conditions observed during the Macri administration (2015-2019), a period characterized by social unrest under a right-wing government, a subsequent political transition from right to left, and a sovereign default. Our model is quantitatively consistent with several empirical patterns. First, it captures the positive relationship between political crises and sovereign risk, with political instability leading to higher interest rate spreads, especially under right-wing leadership. Second, we find that right-wing rulers face more frequent protests and that protests are more common 31 For a political economy perspective, see Bartlett (2007), and for an economic application, see Martin (2019). 36 Other data sources : Our cross-country regressions use data on External Debt, Gross Domestic Product (GDP), Current Account Balance, Reserves, and Primary fiscal balance. The external debt data is from the Joint External Debt Hub of the World Bank, International Monetary Fund (IMF), and Bank of International Settlements (BIS). GDP data in national currency and U.S. dollars are from the World Bank’s Global Economic Monitor, Haver/National Account sources, and National Account sources in Global Financial Data. Current Account Balance, Reserves, and Primary fiscal balance are from the IMF International Financial Statistics data set. Party affiliation data is from the Inter-American Development Bank’s (IADB) 2020 Database of Political Institutions. Default data is from Asonuma and Trebesch (2016) Monthly Default and Restructuring Database. Finally, the credit ratings used to select countries for the regression and event analysis, are from Standard and Poor’s country ratings for foreign exchange debt extracted in July 2024. We use data on hourly wages from the SEDLAC (2023) dataset, extracted in April 2023. 43 B Revolts and Spreads: Data and model 1995-1 1995-7 1996-1 1996-7 1997-1 1997-7 1998-1 1998-7 1999-1 1999-7 1999-12 0 250 500 750 1000 1250 1500 1750 2000 Spread (bp) 22 24 26 28 30 32 Po itica Risk E)ent Type Protest Strike Both Protesters ( 100 ( 1, 000 ( 10, 000 ( 100, 000 2nd term E ection EMBI≳ Po itica Risk (a) 1995-1999. 2000-1 2000-7 2001-1 2001-7 2002-1 2002-7 2003-1 2003-6 0 1000 2000 3000 4000 5000 6000 7000 8000 Spread (bp) Defaul) 20 25 30 35 40 45 50 Poli)ical Ri(k Even) T-pe Pro)e() S)rike Bo) Pro)e()er( ≳ 100 ≳ 1 , 000 , 10, 000 , 100, 000 De la R ≳ ua Puerta Rodr ≳ iguez Default Cama  no Du alde EMBI+ Political Risk (b) 2000-2003. 2003-6 2004-1 2004-7 2005-1 2005-7 2006-1 2006-7 2007-1 2007-7 2007-12 0 1000 2000 3000 4000 5000 6000 7000 Spread (bp) Default 28 30 32 34 36 38 40 Political Ri)k Event Type P(ote)t St(ike Both P(ote)te() ≳ 100 ≳ 1 , 000 ≳ 10, 000 ≳ 100, 000 Inauguration Restructuring begins Deal with IMF Election EMBI+ Political Risk (c) 2003-2007. 2008 2009 2010 2011 2012 2013 2014 2015 2016 0 250 500 750 1000 1250 1500 1750 2000 Spread (bp) Defa)l( 30 32 34 36 38 40 Poli(ical Ris Even( Type Pro(es( S(ri e Bo(h Pro(es(ers ≳ 100 ≳ 1 , 000 ≳ 10, 000 ≳ 100, 000 1st≳Term Election 2nd≳Term 1st≳missed pa,ment Election EMBI + Political Risk (d) 2008-2015. Figure 14: Spreads, political risk, and daily protests Note: Left vertical axis measures the interest rate spread using the EMBI+ in basis points. The right vertical axis corresponds to one hundred minus country risk from the ICRG database. This is an index of political risk with high values representing higher levels of political instability. Blue and orange circles correspond respectively to protests and general strikes mentioned in the Dow Jones Factiva dataset and are associated with fiscal reforms. The size of the dot corresponds to the highest protest size recorded. We extend the daily analysis conducted for the Macri administration, shown in Figure 4, to the administrations that governed Argentina between 1995 and 2015. The results, plotted in Figure 14, reveal that the 2001 default also experienced similar frequent episodes of social unrest, government turnovers, and rapid increases in spreads. However, it is important to note that outside of these default crisis periods, most protests seem to only have a small impact on spreads. In the model, protests influence the reelection probability of the incumbent. Consequently, they can induce an increase in spread when a right-wing government is in office, as illustrated in Figure 9in the main body of the paper. Conversely, protests can lead to a decline in 44 spreads when a left-wing government is in office, as depicted in panel (a) of Figure 15, where the y-axis is chosen to highlight this gap. Although a significant change in spreads exists for certain portions of the price schedule, the spreads paid under stability and revolt also coincide for substantial portions of the debt support. At the ergodic distribution, we find evidence that revolts affect spreads at a similar rate as the one observed in the data. In panel (b) of Figure 15, we plot the difference in spreads between stability and revolt over 300 periods of the ergodic distribution. The results align with the data, showing that most revolts change the interest rate spread by less than 30 basis points, with a few rare instances having a significant impact. On average, revolts increase spreads by 8.6 basis points, with a standard deviation of 29.8 basis points. 0 0.2 0.4 0.6 0.8 1 1.2 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 Spread (BPS) 104 (a) Left-wing gap in spreads. (b) Revolts and spreads at the ergodic. Figure 15: Effect of revolts on spreads in the baseline model. Note: Panel (a) plots the policy-implied spreads as a function of future debt for both parties in the baseline model under stability(dotted lines), and under revolt(solid lines). For this purpose, we fix TFP to the mean level, and we fix initial debt to the mean level in the ergodic. This figure is identical to Figure 9with the exception of the y-axis. Panel (b) plots a stylized simulation that is meant to be an example of the median effect of revolts on spreads. The simulation starts in power and lasts for 300 periods. The initial debt level is fixed to B = 1 . 23, which corresponds to the median level of debt for the right-wing party in the ergodic distribution. The initial productivity level is fixed to the mean level, which is in the middle of the productivity grid. We also verify that out measure of protests is consistent with the alternative measure obtained by Barrett et al. (2022) with the Dow Jones Factiva Data for Argentina. 16 presents our measure of protests for Argentina at the quarterly level described in 14 alongside the Index of Social Unrest computed by Barrett et al. (2022). The results of the main body of the paper are validated by this alternative measure. 45 1995 1997 2000 2003 2006 2009 2012 2015 2018 2021 0 200 400 600 800 1000 Barrett et al . (2022) Social Unrest Index Default 0 2 4 6 8 10 12 14 16 E2en0s per q1ar0er E2en0 Type Pro0es0 Bo0h S0rike Pro0es0ers ≳ 100 ≳ 1 , 000 ≳ 10, 000 ≳ 100, 000 Lef0 Righ0 Cen0er Menem 2nd Term De la R ≳ ua 3 Turnovers Duhalde N . Kirchner C. Kirchner 1st Term C. Kirchner 2nd Term Macri Fernandez Figure 16: Index of social unrest and frequency of protests in Argentina 1995-2021 Note: We aggregate at the quarterly level the index of Social Unrest of Barrett et al. (2022) for Argentina, using the updated numbers from September 2024. C Austerity policies under a left-wing ruler When a left-wing government takes power and opts against a political default, it must still navigate the substantial debt burden left by its right-wing predecessor. The only way to restore the debt to a level more aligned with its ideology is through austerity measures. In Figure 17, we present the average evolution of various aggregates around a political turnover from a right-wing to a left-wing government that does not involve a default. 35 Panel (a) depicts the evolution of the debt-to-average-output ratio, panel (b) shows the evolution of aggregate consumption ( CH + CL ) relative to average output, and panel (c) illustrates the evolution of the consumption share of typeL households as a percentage of total consumption (CL CL+CH). Panel (a) shows that the left-wing government begins its term by reducing the debt by approximately 4% of average output. The subsequent rise in debt issuance after the first five periods is partially explained by a selection bias, as economies experiencing positive productivity shocks are overrepresented in our sample. Panels (b) and (c) illustrate how austerity is implemented. The simultaneous decline in aggregate consumption and the increase in the consumption share of low-income households are consistent with higher progressivity ( τ1 ) and increased overall tax levels ( τ0 ). In other words, austerity is achieved through disproportionately higher taxes on high-income households. Given the efficiency-equity trade-off in our model, these policies also result in lower total output, leading to a political 35 The event analysis focuses on episodes without defaults for 10 periods before and after the political turnover. Additionally, we exclude episodes where an additional turnover occurred within these 20 periods. 46 -10 0 10 Period 0.92 0.93 0.94 0.95 0.96 0.97 0.98 0.99 1 Average debt level Left-wing takeover (a) Debt to mean output. -10 0 10 Period 0.9 0.92 0.94 0.96 0.98 1 1.02 1.04 1.06 1.08 1.1 Average total consumption Left-wing takeover (b) Aggregate consumption -10 0 10 Period 0.24 0.25 0.26 0.27 0.28 0.29 0.3 0.31 0.32 0.33 Average tax progressivity Left-wing takeover (c) Redistribution Figure 17: Austerity policies under a left-wing government Note: Plotted are episodes of political transition from a right-wing government to a left-wing government in the ergodic distribution of the baseline model. The transition happens at the period labeled as 0. The episodes are defined as chains in which the right-wing party has been in repayment for at least 10 periods, to be then replaced by a left-wing government that remains in repayment for at least 10 periods after the transition. Plotted are the associated series of debt relative to average GDP in the economy, total consumption relative to average GDP, and consumption of the L-type household as a share of total consumption. recession. D Alternative revolt costs We show here that our results are broadly robust to different specifications of the revolt costs (13). In all the specifications we keep the values of ϕR 0 and ϕR 1 at their calibrated values from section 4. We compare the moments under different specifications of the revolt costs to the baseline, and the model with only exogenous transitions in Table 7. In the first alternative specification, we alter only the cost of revolts during defaults. 47 Table 7: Comparison between different revolt cost specifications. Model specification Debt Spread Freq. default Revolts Share in power(Right) Baseline 89.6 7.3 4.4 36.6 47.9 Exogenous turnover 81.4 7.1 4.3 - 50.0 Revolt-Default cost 89.6 7.2 4.4 36.6 47.9 Utility cost 82.5 5.5 3.6 38.4 47.4 Note: We compare key moments of the baseline model and the model with only exogenous turnover to two alternative revolt cost specifications. The moments are computed using 100,000 simulations for each model specification. The Revolt-Default costs are still productivity costs as in the baseline model, but the additional cost of revolt under default is a function of total productivity ( A ) as opposed to default productivity ( αD R ). The utility cost specification imposes additive utility costs proportional to productivity. These costs are not a function of the default status and affect only the protesters, but not the households of the other type. Instead of the fourth expression in (13) we use : α(A, D= 1,R= 1) = αD R0−max(ϕR 0A+ϕR 1A2,0), This specification increases revolt costs during defaults relative to the baseline nonlinearly with productivity. However, given that defaults will endogenously occur only when productivity is low, the aggregate moments are very close to the ones in the baseline model. In the second alternative specification, we use instead a utility cost of default, as opposed to a productivity cost. However, to keep things comparable, we still make this utility cost proportional to productivity. 36 A household of type i will therefore have the following static utility in both default and repayment states : ui(C, N)−R×max(ϕR 0A+ϕR 1A2,0) Compared to our baseline model, this specification exhibits a similar frequency of revolts and right-wing governments in power. While the debt level at the ergodic distribution is lower, both the frequency of defaults and the spreads are also reduced. When we compare this specification to the model with only exogenous turnover, we still observe a higher level of debt issuance, albeit at lower spreads and with fewer defaults. The decrease in debt relative to our baseline model is expected, as output is higher when revolts do not impact total factor productivity (TFP). Revolts are more effective as an endogenous default cost when their costs are modeled in terms of utility losses for the protestors. 36 Utility costs of this form are now common in the sovereign debt literature. See for instance Aguiar and Amador (2013), Bianchi and Sosa-Padilla (2023), and Roch and Uhlig (2018). 48 E Alternative models with only exogenous turnover We show here that the differences between our baseline model and the model with only exogenous transition introduced in section 7.2 are primarily driven by the asymmetric reelection odds that default and repayment entail for the ruler. Specifically, we solve alternative specifications of the model with only exogenous turnover and compute the moments of Table 6for each of them. We show that specifications that do not allow for reelection odds that depend on the default status sustain substantially less debt than our baseline model. Table 8, summarizes the ergodic distribution moments for the baseline model, the model with only exogenous turnover, and various alternative specifications of the latter. For clarity, we refer to the exogenous turnover model presented in the main paper as the No-Revolts model. In this model, the probability of reelection is set to the value under stability in the baseline model, resulting in a higher average reelection probability than the average reelection odds at the ergodic distribution of the baseline model. In the first alternative specification, we keep reelection probabilities constant across states but match the average probability of reelection at the ergodic distribution of the baseline. The resulting moments are very similar to those in the No-Revolts model. In the second specification, we allow reelection probabilities to differ by party but keep them constant across default or repayment states. Again, the average debt level and spreads remain close to those in the No-Revolts model. In the third version, we set equal exogenous reelection probabilities for each party, but let them depend on the default or repayment status, using conditional means from our baseline. The resulting moments more closely resemble those of the baseline model than the No-Revolts model. Finally, in the fourth specification, we allow reelection probabilities to vary by party and by default status. The moments of this specification are the closest to our baseline but still underestimate the level of debt issuances by 3% of mean output. This suggests that the reelection odds channel is not the only way revolts lead to higher borrowing. For additional discussion about the role of revolts for borrowing see section 7.4. F Predictions of the model in the data We verify several model predictions using Argentinean and cross-country data. In Section 7we explain that right-wing governments sustain higher levels of debt, while left-wing governments default more frequently upon entering office. One implication of the first observation is 49 Table 8: Comparison to alternative specifications of the model with only exogenous turnover Model specification Debt Spread Freq. default Share in power(Right) Baseline 89.6 7.3 4.4 47.9 No-revolts model 81.4 7.1 4.3 50.0 Average reelection odds 81.4 6.7 4.1 51.0 Odds vary by party only 81.3 6.7 4.2 48.4 Odds vary by default status only 85.1 7.1 4.3 50.1 Odds vary by party and default status 86.43 6.9 4.1 47.7 Note: We compare moments of the baseline model and the model with exogenous turnover (No-revolts) to two alternative specifications of the model with only exogenous transitions. The moments are computed using 100,000 simulations for each model specification. The average reelection odds model keeps the probability of turnover constant at the average of the ergodic distribution of the baseline model. The fourth line specifies a model where the probability of turnover varies by party but is exogenously fixed to the conditional means measured at the ergodic. The fifth line specifies a model where the probability of turnover varies by default status but is exogenously fixed to the conditional means measured at the ergodic. The sixth line specifies a model where the probability of turnover varies by default status and party but is exogenously fixed to the conditional means measured at the ergodic. that we should expect more debt issuances under right-wing administrations compared to left-wing ones. Additionally, the second observation suggests that, in the absence of defaults, we should anticipate larger increases in spreads following a Right-to-Left transition than after a Left-to-Right. Here, we provide suggestive evidence supporting the validity of these two predictions in the data. Data Model −2 −1 0 1 2 3 4 5 Average Yearly Change in Level of Debt to GDP So rce Left Right (a) Argentina changes in debt-to-GDP. Data Model −3 −2 −1 0 1 2 3 4 5 6 Average Yearly Change in Level of Debt to GDP Source Left Right (b) Cross-country changes in debt-to-GDP. Figure 18: First difference in annual debt-to-GDP by ideology of the executive. Note: In panel (a) We plot the average annual change in debt as a percentage of GDP in Argentina sorted by ideology of the executive. We use annual central government debt as a percentage of GDP from the IMF Global Debt Database. We combine this data with political information from the DPI. This restricts our study to years between 1975 and 2020. We use the party orientation of the executive to assign each year to a political leaning. In panel (b). We plot the average annual change in debt as a percentage of GDP for our cross-country sample grouped by which party was in power. We use the same data and methods as panel (a). We exclude years in which a country was in default as determined by Asonuma and Trebesch (2016) Monthly Default and Restructuring Database. 50 Figure 18 depicts the evolution of the debt-to-GDP ratio by party affiliation of the incumbent executive from 1975 to 2020, excluding defaults. The analysis is conducted at an annual frequency, using party affiliation data from the DPI and annual central government debt as a percentage of GDP from the historical IMF Global Debt Database. Panel (a) displays the first difference in levels for Argentina and the baseline model, while panel (b) presents the same statistics across different countries. The results are qualitatively consistent with the model predictions. In the model, left-wing governments that do not default must initiate their tenures with austerity measures (see Appendix C), resulting in slightly negative changes in the stock of debt an observation that aligns with the data from Argentina. In the cross-country data, we find increases in the debt-to-GDP ratio for both types of governments, but the increases are less pronounced under left-wing administrations. 3 Months 6 Months 12 Months Length Before and After 0 50 100 150 200 250 Change in Spreads Around Turnover (Basis Points) Source Right to Left Left to Right (a) Data changes in spreads after turnover. 3 Months 6 Months 12 Months Length Before and After 0 50 100 150 200 250 Change in Spreads Around Turnover (Basis Points) Source Right to Left Left to Right (b) Model changes in spreads after turnover. Figure 19: Changes in the average spread around a political turnover. Note: In panel (a) We plot the average change in spreads around political turnover. We use DPI data to find political turnovers. We use the election date as the turnover moment and conduct a monthly analysis. For each turnover, we exclude the three months immediately preceding and immediately following the election month. We then took the average spread for the given period before and after and took a difference. We then took an average of these differences grouped by turnover direction. In panel (b). We plot the average spread changes around political turnover generated by our model, excluding periods of default and one quarter before and after the transition. Figure 19 shows the average change in the CDS spread around a political turnover. We identify Right-to-Left and Left-to-Right turnovers using the DPI dataset, with the election day marked as the day of turnover. To mitigate anticipation effects in the data, we calculate the average spread while excluding the quarter before and after the turnover in both the model and the actual data. Outside of this exclusion period, we determine the difference between the average spread three months (as well as six months and twelve months) after the election and the average spread three months (six months, and twelve months) prior to 51 the election. 37 Consistent with the model, we observe that average spreads tend to increase more following Right-to-Left turnovers than after Left-to-Right transitions. 37 We only perform this analysis on Argentinean data for the 2020 turnover, which we examine in detail in the paper. The other two turnovers for which we have spread data (2003 and 2015) occurred while Argentina was in default and engaged in litigation with its creditors. 52