Sovereign Debt: Is to Forgive to Forget?

That reputation for repayment cannot sustain sovereign lending at all, and that where lending is sustained by direct sanctions, reputation adds nothing. The abstract states both halves: “International lending to a less-developed country cannot be based on the debtor’s reputation for making repayments. That is, loans to LDCs will not be made or repaid unless foreign creditors have legal or other direct sanctions they can exercise against a sovereign debtor who defaults. Even if some lending is feasible because of direct sanctions, having a reputation for repayment in no way enhances a small LDC’s ability to borrow.” The introduction makes the same point in terms of what must be available to creditors: loans are possible only if creditors have legal rights such as the ability to impede a country’s trade or to seize its financial assets abroad – which the authors identify as “the real reason why a defaulter suffers reduced access to capital markets” – or else can threaten the debtor’s interests outside the borrowing relationship, for instance by persuading their own governments to intervene militarily.

Because the proof is an arbitrage argument, it needs almost no assumptions about preferences, and the authors say so. The setting is “a small country that faces competitive, risk-neutral foreign investors,” small in the sense that it cannot affect the world interest rate r, inhabited by “a single, infinitely-lived representative agent.” On preferences: “Since the proof of our theorem is based on an arbitrage argument, it is not necessary to place any restrictions on the agent’s utility function other than that she prefers having more to having less.” Output depends on investment and on exogenous, serially independent shocks; net exports are output minus consumption minus investment, and can be used either to pay creditors or to accumulate assets abroad. Everything relevant is public: shocks and actions “can be observed by everyone; there is no private information about aggregate variables” in the baseline. The authors also decline to model the gains from capital-market access formally, noting only that “the main benefits all have to do with consumption smoothing” – avoiding having to match the timing of import expenditures and export receipts, maintaining consumption while exploiting high-yielding domestic investment, and insuring against shocks such as terms-of-trade uncertainty. One substantive assumption is that the market value of a claim on the country’s entire future gross income is finite, whose “force… is to rule out any ‘Ponzi’-type reputational contracts, under which a borrower can always expect, in present value terms, to be a net importer of capital over some finite horizon.”

One where the only sanction is exclusion from future reputation contracts, and nothing else. “In a pure reputation-for-repayment (‘reputation’) contract a country’s foreign creditors have no effective legal recourse in the event of default. They cannot interfere with the country’s trade; they cannot even seize any financial assets it may hold abroad. The worst fate that can befall a country which defaults on a reputation contract is that it will never again be allowed to write reputation contracts.” The key structural point follows immediately: “However, the defaulting country cannot be cut off from international capital markets entirely. Though it may no longer be able to borrow for domestic investment, it can still buy consumption-insurance contracts by paying cash in advance.” Formally, the authors need only that the implicit contract specify a state-contingent payment for every shock history and be an equilibrium – “it must be in the country’s interest to honor the contract in every possible state of nature. In particular, the country must never have an incentive to default on its reputation contract and switch completely over to cash-in-advance contracts” – and they explicitly refuse to pin down the supporting beliefs: “For our purposes, it is not necessary to ask what set of off-the-equilibrium-path beliefs might support the contract, nor is it important to ask whether the contract is optimal in any sense.” Candidate equilibria could be trigger strategies as in Eaton-Gersovitz or could involve lenders with imperfect information about the country’s utility function; the theorem covers both.

Investors can commit even when countries cannot, because investors are subject to their own courts. “Implicitly, we are assuming that there are foreign investors who can make commitments. These commitments are enforced by the legal system in investors’ countries. Thus a small country can hold foreign assets such as bank accounts, treasury bills, stocks and other state-contingent assets. Of course, it can also stockpile reserves of precious metals and foreign currency.” The contract’s two conditions are that the risk-neutral investor earn the market rate of return, and that “there can be no state of nature in which the country is called upon to make positive payments” next period. The authors give the second an intuitive reading: “If one thinks of the initial payment A as being collateral, then condition (7) can be interpreted as saying that the country’s collateral must be sufficient to cover its losses on the contract even in the worst possible state of nature.” Two further details matter for the argument’s reach: a cash-in-advance contract “can always be indexed to all the same variables as the implicit reputation contract,” so it replicates the insurance the reputation contract provided; and multi-period versions are unnecessary – “It can be shown, however, that multi-period cash-in-advance contracts are superfluous.”

By constructing, at any node with positive reputational debt, a replacement plan funded from the withheld payments that dominates the reputation contract. The statement is that in any sequential equilibrium the market value of reputational debt is non-positive. The proof supposes debt exceeds the relevant bound, has the country “cease payment on its reputation contract and initiate the following sequence of cash-in-advance contracts,” and verifies that the constructed sequence satisfies both investor conditions; it then observes that the country’s outlay under the new plan is weakly smaller period by period, with equality only in the degenerate case, forcing the bound down to zero. The authors give the economic reading plainly: “if one traces out the game tree governed by any reputation contract, there must exist some node at which the country can switch to a sequence of cash-in-advance contracts which dominates the reputation contract. The collateral for the cash-in-advance contracts is drawn from funds the country would otherwise have used to pay back its reputation contract. Despite the fact that the collateral may at first be quite small, it is still sufficient to provide at least as much insurance as the country could have obtained under the reputation contract. A reputation contract can only be equilibrium under the unrealistic assumption that the country is not allowed to hold assets abroad.”

Lending becomes feasible, but its size is set entirely by the punishment, not by reputation. Theorem 2 lets creditors impose a random penalty on a country standing in default, which reduces that period’s output, and shows that sustainable debt is bounded by the expected present value of those penalties. The authors state the separation that matters: “if there are some direct costs which lenders can impose on a country in the event of default, then loans can [be] sustained, but only on the basis of these costs. A good reputation for repaying loans will not in any way enhance a country’s ability to borrow.” In the appendix the object they bound is explicitly the residual: the gap between total debt and the sanction-supported amount “can be thought of as the amount of debt not supportable by direct sanctions, i.e., reputation debt.” And they immediately weaken their own bound: “Actually, the bound given by Theorem 2 may be too high, since countries can typically bargain with their creditors.”

No: it is just another direct sanction, and Theorem 2 caps borrowing at its value. The authors consider a country “playing a tariff supergame, in which either raising tariffs or defaulting on foreign debt triggers a costly trade war,” and concede that “such a mechanism could conceivably support a positive level of lending.” But it does not restore any independent role for repayment reputation: “However, Theorem 2 directly applies to this case. The maximum amount the country is allowed to borrow must be governed strictly by the costs of a trade war. If the costs of a trade war are very small, then the amount the country can borrow is very small.” Their summary is careful about what is and is not being denied: “We do not claim that reputation plays no role in international relations, only that a good reputation for repaying foreign loans does not enhance a small country’s ability to borrow abroad.”

Then the reputational debt is bounded by the cost of switching lenders, which the authors judge small. “Theorems 1 and 2 are based on the standard assumption that the country faces competitive foreign investors. In some sense, the essence of our result is that if there are no gains for the country in dealing with any specific lender, then reputation contracts are impossible. As long as the country faces competitive foreign investors, then any service provided by the current lender (e.g., insurance) can equally well be provided by a new investor.” They allow the practical qualification and then bound it: “It is possible, of course, that in practice there may be some efficiency gain in having the country continue to deal with its current lenders. However, the upper bound on any ‘reputation’ debt is still only the real cost to the country of switching its business to a new set of financial institutions. It seems that this cost cannot be very large relative to the size of most LDC’s foreign debts.”

The authors treat this as the most natural escape route and give three reasons to doubt it. The alternative assumption – Grossman and Van Huyck’s – is that the borrower, lender and all potential lenders observe the shock but it cannot be written into contracts, so the country can never hold foreign assets indexed to it. “It is doubtful that this story can be used to explain reputation contracts of any significant size.” First, the informational asymmetry is implausible in a competitive market: “it is hard to see what kind of shock would be observable to a huge pool of potential (competitive) lenders, but yet cannot be put into contracts. (The concept of observable but not verifiable shocks works better in the context of a bilateral monopoly relationship.)” Second, the shocks are hedgeable elsewhere: “much of the uncertainty that a small country faces is likely to be highly correlated with events elsewhere in the world. Commodity price uncertainty, for example, can clearly be hedged in world asset markets. So, too, can shocks to world demand for the country’s other goods; certainly technology shocks are highly correlated with events elsewhere in the world. Even weather conditions can be highly correlated across countries. As long as the country is able to lend in world capital markets after a default, it ought to be able to construct a portfolio which is highly correlated with [the shock].” Third, some lending needs no contingency at all: “to the extent that foreign loans are used simply to smooth predictable seasonal fluctuations in income, verification is not an issue. (Our theorem encompasses this case.)”

This is the case the authors concede the theorem does not cover, and they mark their own position as a conjecture. “Theorem 1 does not apply directly to the case where the country has private information, though an extension might be possible.” If investors see output but not investment or the shock, a reputational equilibrium can still be written as an implicit contract with payments conditioned on the output history, and the country can still switch to cash-in-advance contracts indexed to output – but the terms it gets will now depend on investors’ beliefs about its capital stock, and “if a default at time s adversely affects investors’ beliefs about the country’s capital stock, this will hurt its ability to get good terms on its cash-in-advance contracts.” The obstruction is technical and honestly stated: “sequential equilibrium places no restrictions on investors’ off-the-equilibrium path beliefs. The resulting multiplicity of sequential equilibria is endemic in models with private information. It seems unlikely that one can obtain a definite result in the present context without applying a refinement of sequential equilibrium, and without fully specifying the country’s utility function and production function. We conjecture that the intuition underlying the present analysis should carry over to private information case, but the question can only be resolved after further research.” They then argue the case is empirically minor rather than logically closed: the shock “is an aggregate shock, and it is hard to argue that aggregate information can be private. We suspect that in the typical LDC, the country’s leaders do not know any more about [the shock] than do the country’s major lenders,” and in any event the shock is largely correlated with external variables, so “the component which can potentially be private information is minor.”

Unobservable preferences are harmless; the reserves assumption the authors reject as economically unjustified. On preferences: “As long as the country’s actions are observable, and as long as investors believe that the country prefers having more to having less, then it does not matter whether investors know the country’s preferences exactly. Theorem 1 would still apply.” On reserves, some authors had assumed a country cannot use its foreign-currency reserves to buy imports needed for investment, so that only new loans can finance them – which creates a gain from repaying in order to receive new credit. The authors reject this on two grounds. First, it smuggles in a direct sanction: “We do not think such loans should be classified as ‘reputational’, since the lender is assumed able to directly interfere with the country’s trade.” Second, it is implausible on its face: “it is hard to justify the assumption that a country cannot buy the same goods with its own foreign-currency earnings that it can buy with lenders’ foreign currency. We believe that the conventional assumption, that what a country can buy depends only on how much money it has available to spend, is the correct one.”

A coordination problem among lenders, and three empirical predictions that do not hold. On coordination: “We have not even mentioned the coordination problem inherent in a reputation model with a huge number of potential lenders. How long will a country which defaults be shut out of credit markets? The greater the length of the ‘punishment period’, the more a country can be lent. But each creditor must know how long other creditors will wait before resuming lending.” The natural institutional fix undermines the framework: “the coordination problem is mitigated if the legal system gives existing creditors equal seniority with any new lenders, as in real-world debt contracts. But then LDC debt contracts should be analyzed as a bargaining problem, not as a reputation-for-repayment problem.” On the evidence: Eichengreen and Lindert-Morton “have shown that, historically, the ability of LDC’s to participate in credit markets does not seem to depend on their past repayment records. It does depend on their volume of trade and their GNP”; Wynne and Winkler document that “it is typically necessary for a country to settle (reschedule) its past defaults before it is allowed access to new loans”; and on jurisdiction, “a legal/bargaining approach predicts that the countries of jurisdiction for international loans will be major creditor countries. Reputation-for-repayment models make the falsifiable prediction that lenders should be equally willing to have contracts adjudicated in debtor-country courts.”

Strong on the theorem, hedged on the empirical magnitudes, and explicit that the enforcement side needs work. They allow that the spanning assumption may fail slightly and argue it does not matter much: “True, the set of assets available to the country in the world’s vast and varied capital markets may not quite span the set of shocks to which an implicit reputation contract can be indexed. But as long as it comes close, the maximum size of reputation contracts is quite limited.” Their positive claim is a belief rather than a proof: “We believe that Western loans to small developing countries in fact depend on the legal and political rights of lenders within their own countries. The reason that an LDC cannot simply default on its loans and switch to cash-in-advance capital market transactions is that existing creditors can seize its assets abroad.” And they are candid that this leaves the quantitative question open: “Admittedly, there are many uncertainties surrounding the actual damage which a lender can inflict on an LDC; it is a gray area of Western law. But if one wants to understand LDC loan contracts, then these costs must be studied further. Reputation for repayment considerations are at most a secondary factor.” A footnote adds where they think the real cost lies: “the main cost of repudiation may well be in lost gains from trade.”

Because a country can bargain over legal claims but not over beliefs, which changes both the empirics and the policy. “If the cutoff is based on lenders’ legal rights, then LDCs can bargain with their lenders, as in Bulow and Rogoff [1987]. A borrower cannot bargain over trigger-strategy beliefs. Because reputation-for-repayment models neglect a country’s ability to bargain, they greatly overstate the ability of lenders to threaten to cut off an LDC from world capital markets. Thus they tend to overstate the empirical importance of capital market cutoffs relative to say, interference with the country’s current account transactions.” On policy, the authors flag two implications: “A Western government policy to force LDC loans through equity markets may be insensitive to the legal reasons why LDC loans have historically been channelled through bank markets and bond markets”; and against the then-current argument “that debt forgiveness schemes may actually harm LDCs by causing them to forfeit their reputation for repayment,” they conclude: “We would argue that if, through bargaining, an LDC can induce its lenders to forgive a portion of its debts, it will gain. Debts which are forgiven will be forgotten.”