A Constant Recontracting Model of Sovereign Debt

The paper opens from the observation that “despite the significant sanctions at the disposal of lenders, many borrowers have been able to consistently negotiate for reduced repayments,” and that rescheduling “appears to be a process of constant rebargaining rather than a one-time partial repudiation of debt” (Abstract, Sec. I). The authors argue the “most important limitation” of the existing sovereign-lending literature (citing Eaton, Gersovitz, and Stiglitz 1986 for an overview) is that it fails to provide a satisfactory theory of partial default – models that restrict debtors to a binary choice between full repayment and complete repudiation cannot explain the empirical regularity of repeated, partial reschedulings (footnote 1 cites 1983-85 rescheduling statistics: 18-19 countries rescheduling $51-136 billion per year).

The paper’s core insight is that banks cannot restrict a debtor country to a binary choice by making take-it-or-leave-it offers, because the country can always make counteroffers – so the banks “cannot always credibly refuse to bargain.” If bargaining were limited to banks’ take-it-or-leave-it offers, banks could extract the country’s full cost of seizure (Sec. III), but once bargaining is modeled as bilateral using Rubinstein’s (1982) alternating-offers framework, repayments are strictly lower. The paper models the country as producing good D each period, which can be consumed, stored (depreciating at rate gamma), or traded for good F at price P > 1; while in default, banks can seize up to a fraction beta of goods in transit (net benefit to banks is a fraction alpha, strictly less than beta, of the seized value, reflecting seizure costs), but have no means of seizing assets within the country itself (Sec. II).

Banks and the country take turns making offers over the division of each period’s output, with offers made every h periods (banks move first) and an agreement reached immediately (bargaining is always efficient – no delay occurs in equilibrium). The perfect-equilibrium division (eq. 9, in the continuous-bargaining limit h -> 0) depends on the country’s and banks’ relative “rates of impatience” – gamma + delta for the country (delta its discount rate, gamma the storage depreciation rate) and gamma + r for the banks (r the world interest rate) – and falls into one of three regions depending on which of three right-hand-side expressions is smallest: the bargaining region (both sides’ shares are in inverse proportion to their impatience, and banks’ receipts are linear in the gross value of tradeables, not merely the gains from trade P-1, because the country’s threat to consume or ship without an agreement is not credible once continued bargaining dominates it); the autarky-constrained region (gains from trade are small and seizure costs are high enough that the country’s threat to simply consume domestically becomes credible, so banks settle for a payment just below the full gains from trade); and the punishment-constrained region (the banks have very little seizure power, so the bargaining outcome coincides exactly with the take-it-or-leave-it outcome of Eaton and Gersovitz 1981, Sachs 1984, and others – the bank gets beta*P and the country cannot negotiate anything lower).

The paper shows that outside the punishment-constrained region, increasing beta has no effect on repayments in the bargaining region, and raising the net seizure parameter alpha similarly does nothing as long as alpha stays positive (Sec. III) – “it is completely credible for the banks to threaten maximal seizure activities if the country tries to ship without a rescheduling agreement” regardless of the exact magnitude of alpha, so long as it is positive, which is sufficient to pin the bargaining-region outcome; the banks’ difficulty is that they cannot initiate seizure until the country initiates trade, which is why the size of alpha (beyond its sign) does not matter to the bargained outcome.

In the (static) model of Sections II-III, a higher, perfectly anticipated world interest rate r always makes the country worse off, both because it raises the banks’ discount rate on future repayments and because, within the bargaining region, it strengthens the banks’ bargaining position (a more impatient bank negotiates harder) (Sec. III, end). The maximum amount the country can borrow is given by eq. (10), the present value of the stream of repayments in eq. (9), and this maximum loan is unambiguously decreasing in r through both channels.

Introducing stochastic interest rates (Section IV, modeling a one-time Poisson-timed permanent interest-rate shock), the paper shows that an unanticipated rise in world interest rates can actually reduce the country’s repayments once it is already “loaned up” – committed to a stream of repayments at least as large as the maximum it can be forced to pay. The mechanism: a higher post-shock interest rate makes the banks more impatient to extract their money and redeploy it elsewhere, and the debtor country, already at its borrowing limit, can exploit this impatience in continued bargaining; since the shock was unanticipated, it was not priced into the original loan terms, so the country captures the benefit rather than having it competed away ex ante. The paper offers a concrete numerical illustration: a country with $100 billion face-value debt paying $5 billion/year at a 7 percent interest rate (2 percent inflation, constant real debt) could see payments fall to $4 billion/year if rates rise to 10 percent, even as the nominal face value of the debt rises “meaninglessly” by the unpaid difference. The authors note this result does not hold for a country not yet loaned up, and emphasize it means observed defaults correlating with rising world interest rates need not reflect countries’ inability to handle higher debt-service burdens – it can instead reflect improved bargaining leverage that lowers actual payments.

The paper identifies two reasons within the scope of the model (Sec. III, end) for why sovereign borrowers will, in equilibrium, sign contracts that formally label their obligations as debt: first, if banks instead held equity in the country’s export industries, a later policy interfering with dividend repatriation or taxing exports would reduce the banks’ share without triggering the same legal recourse a loan default does; second, because the variables that would need to determine state-contingent repayments (the country’s true productive capacity, GNP) may be observable to both sides but not verifiable in U.S. courts (most debtor countries do not publish reliable national accounts), making explicit state-contingent contracts unenforceable and creating the ongoing need for informal rescheduling negotiations to substitute for a contract that cannot be written in advance.

The authors motivate U.S. government interest via a stylized model in which the U.S. is the sole importer of the country’s good D (other foreign suppliers having inelastic supply), so that a suspension of trade imposes a loss of U.S. consumer surplus proportional to the inverse elasticity of U.S. demand (eq. 15-16); the U.S. government is modeled as maximizing the present discounted value of U.S. consumer surplus net of any side-payments it makes (eq. 17). Extending the three-party alternating-offers bargaining to include the U.S. (eqs. 18-19), the paper’s “central conclusion” is that “even if the gains from trade are important to the U.S., the U.S. will generally be better off if it can precommit to stay out of the rescheduling process” – because the banks and the country will always reach an efficient bargain on their own (no interruption of trade occurs), so U.S. side-payments are not needed to secure that outcome. Unfortunately, the paper shows the U.S. typically cannot credibly make this precommitment: when its stakes are of the same order of magnitude as the country’s, the banks and the country can exploit the U.S.’s impatience for trade to resume, bargaining the U.S. into making sidepayments in the unique perfect equilibrium. A further, less intuitive result (Sec. V, end): the country can extract a larger share of total surplus by bargaining against the U.S. with the banks present than by bargaining directly against the U.S. alone, since bank loans let the country more effectively exploit its bargaining power with the U.S. – “in sharp contrast to the popular image of the U.S. and the banks teaming up against the country.”

The paper shows that the specific split of bargaining power between the country and the banks depends on the fraction of offers each side is allowed to make (not necessarily a 50-50 alternation), and illustrates (eq. 24) that if the country can make only one-third of offers instead of one-half, the banks’ equilibrium share rises from roughly one-half to roughly two-thirds of GNP (in the bargaining region). With two separate bank creditors, the paper shows that whether the country bargains with them jointly (each getting half the single-bank outcome) or sequentially alternates offers with each bank separately (reproducing the single, unified-bank outcome exactly) depends on the structure of the bargaining protocol – the model does not endogenize every feature of real-world bargaining structure but isolates the fraction of offers each party is allowed to make as the exogenous primitive driving the resulting split.

The paper contrasts its own trade-sanctions-based collateral mechanism with Eaton and Gersovitz’s (1981) emphasis on the threat of prolonged exclusion from future borrowing as the primary repayment incentive, arguing the latter is hard to square with the data for major 1980s debtors: debt-to-GNP ratios for major problem debtors were “in most cases greater than fifty percent, and in some cases over one hundred percent,” making it “totally implausible” that the discount rates required to rationalize decades of repayment purely to preserve future borrowing access are realistic (Sec. VII, Conclusions). The paper also notes that major debtors ran trade surpluses in 1984, and that U.S. bank loans to Argentina, Mexico, and Venezuela actually shrank (by -3.2 percent, -3.0 percent, and -6.3 percent respectively) over the twelve months ending June 1985 even as 12 percent interest accrued – i.e., banks were receiving net payments precisely when new lending had stopped, which the paper argues is consistent with its own gains-from-trade-collateral mechanism but harder to reconcile with a pure future-access-to-credit motive.