Sovereign Debt

Enforcement, not solvency. The opening sentence of Section 1 is that “the defining feature of sovereign debt is the limited mechanisms for enforcement,” and that this “distinguishes sovereign debt from private debt, whether domestic or international.” The asymmetry is legal: “A private agent or corporation, at least technically, is always subject to a legal authority. Sovereign nations are not.” International bonds and bank loans are issued in New York or London and so fall under the jurisdiction of the place of issue, but if a sovereign misses a contracted payment, creditors are left “relying only on overseas legal instruments and reputational considerations.” Everything in the chapter follows from that single friction: the mechanisms that induce repayment, and the consequences of their weakness for risk sharing, growth, and investment, are named as the chapter’s major theme.

Non-contingency, subjective default, and the absence of seniority. The standard sovereign contract “is typically non-contingent” – a pre-determined, non-state-contingent sequence of payments in a defined currency – with a few exceptions the chapter names, including some Brady restructurings and recent Argentine and Greek issues. But it argues the notional non-contingency “obscures a richer contracting space that comes about through maturity structure, renegotiation, rescheduling, and ‘haircuts.’” Two reasons are offered for why contracts are not written contingent in the first place: moral hazard, if the government can manipulate the actual or reported behaviour of macro aggregates, and non-verifiability of the true state to creditors. Because payments are often renegotiated under the threat of default, “with creditors accepting less in place of none,” the chapter says the fact of renegotiation “makes the definition of default rather subjective.” Finally, “as a rule, there is no strict seniority in sovereign debt issues (with a few ‘de facto’ exceptions, like credit extended by the IMF),” which opens the door to dilution of existing creditors by new issuance – and, run backwards, to the Bulow-Rogoff (1991) result that voluntary buybacks are a transfer to bondholders.

Regularity of default, its cyclical timing, heterogeneous haircuts, long renegotiations, the shape of the spread curve, and the fact that growth coincides with falling sovereign indebtedness. Taking them in the chapter’s own order (Section 2): (1) Default “happens with regularity throughout history,” most countries able to borrow internationally have defaulted at least once, serial default “is the norm rather than the exception,” and defaults come in waves – the Great Depression, the 1980s Latin American crisis, the then-ongoing European crisis. (2) Default “often occurs in bad times, but with exceptions”: in Tomz and Wright’s 175-country sample, output is on average 1.6 percentage points below trend at the start of a default, but “more than one-third of their 169 default episodes began when income was at or above trend,” and countries frequently fall below trend without defaulting, so “a recession is neither necessary nor sufficient for default.” (3) Haircuts are heterogeneous: Sturzenegger and Zettelmeyer’s six-country sample gives losses relative to face value “ranging from roughly 30 percent in Uruguay to over 60 percent for some bond series in Argentina and Russia,” while Benjamin-Wright and Cruces-Trebesch find “investor losses of roughly 30 to 40 percent on-average, and again with considerable heterogeneity.” (4) Renegotiation is slow: across ninety default episodes in seventy-three countries, 1989-2005, restructurings take “eight years on average,” longer negotiations produce larger losses, and “the median country exits restructuring carrying 5 percent higher debt-to-GDP loads then at the time of default.” (5) On spreads (1990-2009 emerging-market sample): spreads over US bonds are higher at longer maturities, all spreads rise in crises but short-term spreads rise relatively more so the curve “inverts,” new issuance shortens, and global factors – US equity returns, US corporate bonds, the VIX – explain a large fraction of common variation, suggesting bondholders are compensated for aggregate as well as idiosyncratic risk. (6) On growth: contrary to the standard open-economy prediction, high-growth emerging markets have been net capital exporters (Gourinchas and Jeanne’s “allocation puzzle”), and Aguiar and Amador (2011) show the pattern is driven by government net foreign assets, not private flows, which behave conventionally.

A small open economy with a representative agent and a government that cannot commit, facing risk-neutral international lenders who can commit fully; the government maximises the agent’s discounted utility subject to a resource constraint and a participation constraint at every history. The chapter credits the framework to the closed-economy limited-commitment models of Thomas and Worrall (1988) and Kehoe and Levine (1993), with “key conceptual elements” found in “the seminal sovereign debt paper by Eaton and Gersovitz (1982).” International markets trade a full set of state-contingent claims at risk-neutral prices (justified by the small-open-economy assumption plus diversification), and βR ≤ 1 keeps assets bounded. The canonical punishment for deviation is total exclusion from international asset markets, i.e. the autarky value – “the canonical punishment for deviation in the sovereign debt literature” – which is independent of the debt outstanding at the moment of deviation and weakly increasing in the current endowment and in installed capital.

It treats the critique as decisive against naive autarky and then surveys the four responses that restore sustainable debt. Taking the reputational story literally, the chapter asks why losing reputation should stop a government from continuing to save abroad. Bulow and Rogoff (1989b) build on exactly that: they let the defaulter buy a sequence of cash-in-advance insurance contracts, paying upfront with no future obligation, and show that this is feasible and delivers higher utility “as long as the upper bound on debt is strictly positive” – so the equilibrium debt ceiling collapses to zero. The responses the chapter lists are: appeal to legal enforcement, since the same foreign courts that let creditors commit to cash-in-advance contracts can be used to seize a defaulter’s overseas assets or exports (the route Bulow and Rogoff themselves favour), though the chapter notes “mixed empirical evidence regarding the effect of trade sanctions”; drop creditor commitment altogether, as in Kletzer and Wright’s “anarchy,” where nothing guarantees the foreign side of a cash-in-advance contract either; keep one-sided commitment but let a limited number of foreign banks choose not to offer contracts that disrupt lending relationships (Wright, 2002); or appeal to non-legal spillovers, where lost reputation in debt markets depresses trade, output, or investment (Cole and Kehoe, 1998), or to Amador’s (2012) political tragedy of the commons. In the other direction, Broner-Ventura and Broner et al. argue autarky may be worse than the chapter’s benchmark, because failure to enforce international obligations may come with failure to enforce domestic ones, and because domestic residents – especially net-worth-constrained banks – may hold the defaulted debt.

Limited commitment breaks full risk sharing, and the optimal way to live with it is to tilt consumption towards the future. When the participation constraints never bind, the first-order condition reduces to the full-risk-sharing allocation with consumption independent of the state. When a constraint binds, consumption becomes “strictly increasing in the current endowment.” The mechanism behind back loading is that the participation constraint is forward-looking – its left-hand side is the discounted sum of all future utility – so “at the margin, therefore, consumption in the future is preferable as it relaxes all preceding participation constraints,” and the first-order condition accordingly carries the cumulative sum of every earlier multiplier. The chapter draws two sharply different long-run cases. With βR = 1, consumption never declines, current consumption is a floor for all future consumption, multipliers go to zero, and “in the limit, therefore, the economy achieves perfect risk sharing” – attained for good once the highest endowment state is realised. With βR < 1 the multipliers must be strictly positive infinitely often, so “the combination of front loading due to impatience and back loading due to limited commitment implies consumption will converge to an ergodic distribution, which in general will be non-degenerate.”

No, and the chapter explicitly flags this as a common misreading. Rearranging the binding participation constraint shows that current output must be at least current consumption – “the country does not receive net inflows when its constraint is binding,” or, put another way, “the country is never tempted to renege when it is due payments from the rest of the world.” Combined with the risk-sharing requirement that high states are the states to repay in, this implies that if the constraint binds at a low endowment realisation it also binds at a higher one. The chapter then says directly: “While intuitive, this implication has often led to the confusing and incorrect statement that the model predicts that ‘default happens in high-endowment states.’” What actually happens is that the temptation in a high state is met in equilibrium by reducing the net exports demanded in that state, which by the budget constraint must be balanced by lower net imports elsewhere – so the real content of the constraint is that it “acts as a constraint on borrowing in bad times.”

A larger capital stock makes walking away more attractive, so the sovereign under-invests to keep its participation constraint slack – and the distortion is worst when debt is high. With capital as a state variable in the deviation value and V strictly increasing in it, the first-order condition for capital says that if the participation constraint binds in any state tomorrow, then the expected marginal product of capital exceeds its opportunity cost today: “investment is distorted down today if the participation constraint binds in any state tomorrow.” Two corollaries follow. First, capital is never above the first-best level – “over investment in this environment never enhances commitment.” Second, the wedge can be decentralised as a tax on capital income, where the chapter notes “taxes” may in practice mean “statutory taxes as well as bribes, permits and red tape, or anticipated ex post expropriations of capital income by politicians,” citing Argentina’s 2001 crisis and the sequence of expropriations that followed as a case study. Because low output tightens the desire to borrow, low output also leads to low investment, so “limited commitment can prolong the impact of a transitory shock.” With βR = 1 the economy eventually saves enough that investment reaches first best; with βR < 1 “investment cycles indefinitely at a level always strictly below the first best.”

The chapter’s preferred answer is political turnover, which makes the decision-maker effectively quasi-geometric rather than merely impatient. It first notes that simply assuming βR < 1 “is not completely satisfactory as an explanation,” partly because many countries do eventually pay down debt but over a very long period. In the Aguiar-Amador (2011) alternative, political actors rotate in and out of power by a Markov process and an incumbent places a premium θ̃ > 1 on consumption while in office; with iid turnover the incumbent’s normalised objective discounts between today and tomorrow at β/θ < β but between any two future dates at β, so “the incumbent has hyperbolic or quasi-geometric preferences in the spirit of Laibson.” The chapter explains why: the current period is special because the incumbent holds power in it with probability one, while any future date must be down-weighted by the chance of being out of office. With βR = 1 and risk-neutral preferences the multiplier obeys the first-order law λ_{t+1} = (1 − 1/θ)λ_t, so it “converges to zero at a rate 1/θ” – convergence, and hence the speed at which investment reaches first best, is governed entirely by the size of the political distortion, and “a more politically distorted economy grows at a slower rate.” The model also reproduces “debt intolerance”: the more distorted economy must reach a lower steady-state debt level to support the same capital stock, matching Reinhart et al.’s finding that less-developed economies run into trouble at lower external debt than rich ones. The chapter’s calibrated conclusion is hedged – political distortions “do not preclude an economy from eventually achieving high-income status, but does suggest that the process will be a long one.”

By adding a shock to the outside option that lenders cannot observe. In the benchmark, contracts are rich enough that a risk-sharing relationship is never severed in equilibrium. Section 4.1 introduces an iid continuous shock z, independent of output, that enters only the deviation value V(z,s) and is known to the government but not to creditors – interpreted as the economy’s vulnerability to sanctions or financial autarky (for instance the exposure of the domestic banking system), or the political consequences the incumbent faces from the wealth transfers default entails. Because z affects only the outside option, it “is not something that can be credibly revealed absent separation,” which keeps the problem tractable. Writing the creditor’s value as a function of promised debtor utility and differentiating yields a distortion to consumption smoothing equal to the hazard-rate elasticity of default times discounted net payments, so “there is an incentive to distort consumption to reduce default at the margin,” concentrated in states where the default probability is most elastic and repayments largest. Where the country is a net saver it never defaults for any z and consumption is flat. The empirical payoff is that “the independent shock z weakens the correlation between output and default, consistent with the empirical facts that default is more likely in recessions, but occurs with some probability in booms as well.”

Incomplete markets create an incentive to wait for a good state, and multiple creditors create an incentive to hold out; collective action clauses can make delay worse, not better. The chapter is explicit that the earlier complete-markets readings of default – Grossman-Van Huyck’s excusable default, Bulow-Rogoff’s renegotiation under threat of sanctions – while “useful as foundations for rich risk-sharing possibilities implemented with non-contingent contracts… do not speak to the delays observed in actual default episodes,” and that the eight-year average default episode “call[s] for a richer model of debt renegotiation.” In Bi and Benjamin-Wright, once state-contingent assets are gone the debtor can pledge only a non-contingent amount, so “there is therefore an incentive to delay… until expectations of future output are high in order to credibly pledge these endowment streams.” With many bondholders, Pitchford and Wright show that “the last to agree to a settlement has disproportionately large bargaining power due to its ability to veto the entire settlement,” and limited commitment matters here too because the borrower cannot commit to treat hold-outs more harshly. CACs are the obvious fix, but Pitchford and Wright show “somewhat paradoxically that this may serve to increase delay,” since negotiation is costly and only a subset of holders is needed, inviting free-riding on negotiation costs. Bolton and Jeanne add the ex ante trade-off: harder restructuring raises ex post costs but may induce repayment that would not otherwise happen – yet individual creditors have an incentive to make their own bonds hard to restructure, and a government tempted to dilute will issue harder-to-restructure new bonds, so “in equilibrium, this exernality generates bond issues that are excessively difficult to renegotiate, potentially lowering ex ante welfare.”

The level of inherited debt does. The chapter builds a two-period example using “the timing introduced by Cole and Kehoe (2000),” in which the price schedule is offered before the within-period default decision – the distinction from the quantitative literature’s convention being “whether the price schedule… is conditional or unconditional on non-default within the period.” Proposition 1 defines two debt thresholds from primitives and states that if inherited debt lies in the interval between them, “there exists at least two equilibria, one of them with q(b′) = 0 for all b′, and another one with an equilibrium schedule such that q(b/(1+β)) = β.” In the first, lenders will buy no new bonds at any quantity, existing holders demand immediate repayment, and default is the government’s best response – which confirms the zero price. In the second, the government issues at the full-commitment price and repays. The important boundary is the lower one: “if initial debt is low enough, the government is not subject to self-fulfilling crises.” The chapter is careful that the multiplicity region need not be non-empty – “Proposition 1 does not guarantee that b < b̄” – but shows there is a threshold default cost τ* above which it is. Embedding this in a dynamic model with a constant positive crisis hazard while in the zone, Cole and Kehoe’s main result is that “the government has an incentive to save its way out of the crisis zone,” with the caveat that if initial debt is very large the transition may be long enough that remaining in the zone indefinitely is preferable.

A reason not to save, and an ambiguous verdict on nominal debt. Conesa and Kehoe add income shocks: if the economy is in recession facing a constant hazard of recovery, saving deepens the consumption cost of the recession, so “the government may opt to remain in the crisis zone hoping that a recovery occurs before a debt crisis” – “gambling for redemption.” On nominal bonds, Aguiar et al. (2012) show “issuing nominal bonds has an ambiguous effect on vulnerability to a self-fulfilling debt crisis.” If the government’s non-crisis commitment to low inflation is strong, nominal debt is usefully state-contingent: real returns are high in good times and partly inflated away in a crisis, and since creditors prefer partial repayment to outright default, this beats real bonds. If that commitment is weak, the temptation to inflate exists even in normal times, gets priced into lower bond prices ex ante, and “this effect may be large enough to dominate, generating a larger crisis zone for nominal bonds” – which the chapter offers as a rationalisation of why weak-inflation-regime emerging markets borrow in foreign currency while the US, UK, and Japan issue large amounts of domestic-currency debt cheaply. Chatterjee and Eyigungor’s calibration finds that “even a small likelihood of a rollover crisis implies that a country would seek to limit its short-term debt to the point where the probability of a rollover crisis is endogenously reduced to a small value,” and that long-maturity bonds reduce how much debt reduction the sovereign has to rely on.

It replaces state-contingent claims with a single defaultable bond, and then needs an impatient sovereign and state-contingent default costs to generate realistic debt and default rates. The paradigm is Eaton-Gersovitz: a small open economy trades a one-period non-contingent bond with risk-neutral lenders, and “the only state contingency spanned by the asset markets is through the option to default.” Default costs output a fraction τ of income and excludes the country from markets, with re-entry at hazard λ and (in the simplest versions) a 100 percent haircut. The chapter notes a foundational quantitative problem: financial exclusion alone sustains very little debt, because aggregate consumption fluctuations at business-cycle frequencies have small welfare consequences (Lucas’s calculation, extended to a small open endowment economy by Aguiar and Gopinath). Two fixes recur. First, calibration “typically involves βR significantly below one,” motivated by an impatient decision-maker facing political turnover, which offsets the commitment-driven incentive to save. Second, Arellano (2008) lets the output cost τ vary with the endowment, which “makes default more responsive to output shocks and therefore relatively less responsive to outstanding debt, lowering the elasticity of the bond-price schedule as well as making non-contingent debt more attractive as an insurance option”; Mendoza and Yue later derive such a nonlinear cost endogenously from trade credit. The chapter also shows algebraically why volatility matters: the bond-price elasticity is a ratio of the marginal to the average default probability, so if the endowment distribution has substantial mass below the default threshold (a fat lower tail), the ratio falls and “volatility generates more frequent default.”

Because the burden of a net repayment is heavier when the endowment is low, given incomplete markets. The chapter warns against concluding that default in bad states is solely an artefact of an output cost increasing in the endowment – true in the final-period decision of its two-period example, and an important contributing mechanism, but not the whole story. In an infinite-horizon setting, concavity of utility means “the marginal burden of net repayment of debt is higher the smaller the current endowment, all else equal,” and Arellano (2008) uses this with iid shocks to show that if default is optimal at some endowment it is also optimal at any lower one, holding debt fixed. The step that makes this bite is that default only occurs when the country cannot obtain a net inflow by borrowing – otherwise it could borrow, consume more, and default later – and “the fact that a country may be required to make a net payment even in the lowest endowment states reflects the incompleteness of the asset markets.” Conversely, a high endowment makes default unattractive because it forecloses carrying that endowment forward through asset markets. Under persistent shocks the picture is muddied by the effect of the current realisation on future bond prices, but “in the calibrated models popular in the literature, the consistent prediction is that conditional on b there is a threshold endowment above which the country repays and below which the country defaults.”

They introduce dilution and richer hedging, and they change what the sovereign is punished for. Hatchondo-Martinez use geometrically declining coupons and Chatterjee-Eyigungor a “perpetual youth” constant maturity hazard; both make maturity a stationary parameter so bonds issued at different dates are homogeneous. The bond-price schedule is no longer given by the simple monotone operator, because a bond’s return now also depends on next period’s price absent default – the capital gain or loss over the bond’s life. Because that capital loss is an implicit transfer to the government, and because the models give bondholders no trigger strategies or sanctions against dilution (as distinct from outright default), the sovereign has a standing incentive to dilute; the pari passu clause means new bonds are not junior. Arellano and Ramanarayanan add both short- and long-term debt, which the chapter says does two things: it widens insurance possibilities, and it “allows a richer enforcement mechanism,” because in the one-bond incomplete-markets model the government is punished only for missing a payment, whereas a maturity portfolio can replicate a pattern of near- and long-term payments that better mimics the complete-markets allocation. Their model rationalises maturity shortening ahead of crises, since short debt’s advantage is largest when near-term default is likely – though the chapter notes Broner et al. read the same pattern as changing lender risk premia rather than borrower hedging, and that both accounts deliver the observed yield-curve inversion.

Microfoundations, equilibrium selection, and the mapping between model objects and measured data. Section 7 lists the insights the surveyed models deliver – reputation versus legal enforcement, debt overhang onto investment, growth, and volatility, the slow graduation from serial default and the role of political frictions, unobserved shocks as a limit on risk sharing, vulnerability to self-fulfilling crises, renegotiation delay, and quantitative fit – and then sets out what is missing. More progress is needed “on mapping the theoretical models to the data,” including decentralisations that rely on realistic assets, legal mechanisms, and reputational concerns, “combined with a coherent theory of equilibrium selection as many models in the literature support multiple equilibria.” The quantitative literature has begun matching prices and quantities “but often relying on ad hoc assumptions that restrict equilibrium objects such as financial contracts and the output costs of default.” The chapter also raises a measurement problem it takes seriously: Dias et al. note that observed debt stocks mix face and market values across differing maturities, durations, and coupons, which are “difficult to aggregate into a parsimonious set of state variables that appear in a quantitative model,” so care is needed “that we are comparing conceptually similar objects.”