Inefficient Sovereign Defaults Abstract Many argue that financial crisis resolution should become more efficient and opponents mainly emphasize the detrimental effects on incentives ex-ante of greater efficiency ex-post. However, renegotiation theory suggests that debtors should be able to “buy their way back to efficiency” ex-post and for this privilege creditors can potentially charge them so much that ex-ante incentives are not distorted. Thus, making crisis resolution efficient could be “reform without losers” (Lau et al., 2000). We argue, however, that creditors may not be able to help themselves ex-post without also helping the debtor: even if the debtor defaults it is still a high return country so efficiency warrants that it receives a large capital inflow. Under a weak assumption this raises its welfare. Thus, contracts which cannot be renegotiated and cause a sure efficiency loss in case of default can still be efficient ex-ante. 1. Introduction Financial crises in recent decades were associated with substantial output losses (Sturzenegger and Zettelmeyer, 2006; Paoli et al., 2006). However, the efficient response ex-post is to separate efficiency and distribution as much as possible. In particular, rather than disturb the debtor economy by cutting trade or financial relations, or engage in costly bargaining, efficient rescheduling would maximize the continuation value of the defaulting economy and let creditors collect their claims over time with minimal disturbance. There are, roughly, two sides in this debate. One side wants precisely to minimize the efficiency costs of crises once they occur and argues that the recovery of the deadweight losses can be used to make both creditors and debtors better off. For example, rapid rescheduling, new loans or debt reduction can ensure continued gains to financial trade, improve the country’s incentives to invest or reform, and reduce negotiation costs and uncertainty (Krugman 1988a,b; Krueger, 2002; Roubini and Setser, 2003). In contrast, the other side sees output losses as a necessary evil when agency problems allow countries to affect their likelihood of default and sovereign immunity makes contracts imperfectly enforceable. If crises did not lead to output losses lenders would not trust borrowers in the first place (Dooley and Verma, 2001; Atkeson, 1991; Shleifer, 2003; Pitchford and Wright, 2007, Bulow, 2002). 2 However, it is unclear if this is really a zero-sum debate. In particular, inefficient crisis resolution means that one side suffers losses not gained by the other side. Therefore, by a Coasean argument renegotiation should allow debtors to “buy their way back to efficiency” by giving creditors a share of the social gain. Indeed, if creditors can collect the entire efficiency gain the debtor’s ex-ante incentives and its credit ceiling will be unchanged (Kletzer and Wright, 2000). Thus, in the terminology used by Lau et al. (2000) to describe China’s dual-track economic transition, there should be scope for “reform with losers”. The “penalties” paid by the debtor might take the form of lump sum payments, a rise in future interest rates, lower terms of trade or returns to domestic factors employed abroad, or higher returns (e.g. less taxation) to foreign factors employed at home. Alternatively, the debtor could supply political or military assistance to creditor countries whose governments can then transfer the gain to private creditors (Cole and Kehoe, 1996). But if national bankruptcies can be resolved more efficiently while preserving ex-ante incentives why are they not? This paper suggests an answer based on two general propositions. First, any inefficiencies ex-post will provide incentives for the parties to renegotiate. Second, in practice even colluding creditors cannot appropriate the whole efficiency gain because restoring efficiency means restoring efficiency in the debtor’s economy and under plausible conditions this will raise the debtor’s welfare. Consequently, if creditors can help themselves ex-post they will also be helping the debtor. In turn the debtor the will have a lower credit ceiling ex-ante and prefer a renegotiation-proof contract. Thus, the paper supports the “necessary evil” side of the 3 rescheduling debate but strengthens the argument by explaining why a continuation agreement cannot be designed so harshly from the debtor’s perspective that renegotiation leaves it only indifferent. Necessary ex-post inefficiencies can be modeled in several ways, for example using information asymmetries and wars of attrition (Bulow and Klemperer, 1997; Alesina and Drazen, 1991), costly and protracted bargaining (Pitchford and Wright, 2007)1, or a setting where the gain to addressing the inefficiencies is rising over time for one or more agents with veto-power (Velasco, 1998; Labán and Sturzenegger, 1994, 1998). In these approaches, however, creditors are fighting over a fixed pie of repayments so the models are silent on the changes in capital flows which are a defining feature of crises. Therefore instead we model inefficiencies as the result of a common pool problem among collecting creditors. The common pool regime arising after default leads to overextraction of the debtor’s output and therefore a drop or reversal of the previous capital inflow. Implicitly we thus assume that creditors in the common pool regime cannot coordinate on efficient rescheduling.2 Beyond accounting for capital flow reversals the common pool approach is also consistent with the creditor “panics” which are a stylized fact of debt crises in recent decades (Morris and Shin, 2004; Radelet and Sachs, 1998; Fischer, 1999; Calvo and 1 Asymmetric information has also been used to explain the inefficiencies of real conflicts (wars), see Fearon (1995). 2 Apart from giving commitment not to renegotiate lending with many dispersed creditors can lower risk premia in a stochastic world since each lender can diversify. On the other hand poor incentives to acquire information may raise premia. The only difference between common and private access we need for the paper is in terms of renegotiation costs. 4 Mendoza, 1996; Diamond and Dybvig, 1983) and we can rationalize these panics from an ex-ante perspective. Furthermore, based on their data set Tomz and Wright (2005) argue that defaults are best viewed as a temporary suspension of sovereign capital flows followed by resumption of lending on terms similar to those of pre-existing loans. This is also consistent with the common pool approach. 2. A common pool model of crisis resolution We first model the effects of default and then, in the next subsection, consider how default outcomes affect the optimal contract given the inability of creditors in the common pool regime to renegotiate. 2.1. Debt collections after default To model the post-default regime we slightly adapt the dynamic common pool model used in Tornell and Velasco (1992) to explain capital flight from developing countries.3 However, unlike in their paper and most others which analyze the effects of common access to a resource here the first-best policy may not be private access. Instead, common pool access ex-post default is part of the optimal debt contract because it commits lenders not to renegotiate. 3 Dutta and Rowat (2006) present a more general commons model with a capital market. 5 There is a fixed number of international creditors n ≥ 2 receiving the competitive interest rate r However, if the country fails to meet its financial commitments creditors collect the debtor’s output under common pool conditions. We assume that the country’s entire capital stock is borrowed. I later discuss this assumption in detail. The lifetime utility of a creditor is ∞ Uj =∫ σ σ −1 0 c j (t ) (σ −1) / σ e −δt dt , j = 1...n (1) with c(t ), δ ∈ [0,1] and σ > 0 denoting consumption in period t , the common discount rate, and the common intertemporal elasticity of substitution. For σ = 1 the flow utility is ln c(t ) . The capital stock of the debtor country follows kɺ(t ) = y (k (t )) − d j (t ) − ∑ d i (t ), (2) k (0) > 0 (3) i≠ j where y (k ) is a standard neoclassical production function with diminishing returns to capital due to a fixed non-tradable input such as labor. It also satisfies the Inada conditions lim y ' (k ) = ∞ and lim y ' (k ) = 0 kɺ(t ), k (t ), d j (t ) and k →0 k →∞ ∑ d (t ) i are the change i≠ j 6 in the capital stock, the level of the capital stock, the real debt payment claimed by creditor j , and the total real payments claimed by the other creditors at time t . Finally, creditors can deposit funds abroad at the international interest rate r . The foreign savings asset f has entirely private access and follows fɺ j (t ) = rf j (t ) + d j (t ) − c j (t ) (4) where we assume a zero starting stock for simplicity. However, the creditors can potentially go short in the foreign asset. We show now, like Tornell and Velasco, that despite the costless extraction, the common pool regime, and the positive return to the international asset there remains an interior equilibrium where creditors do leave some capital in the country. The creditors play a differential game against each other since the claimant decisions of each affects the decision problems of the rest. We restrict attention to symmetric closedloop Markov equilibria (which imply Markov perfection). Thus, players condition behavior only on current asset stocks and the current behavior of other players. Furthermore, we make the make the plausible assumption that d ' (k ) > 0 . Apart from tractability these assumptions seem appropriate for the financial crises of recent decades where both apparently self-fulfilling crises (e.g. Asia, 1997-8; Mexico, 1994) and fundamentals driven crises (Argentina, 2001) were associated with large 7 creditor “panics” once begun. These panics seemed to reflect high levels of dispersion and anonymity among creditors eroding their capability or incentives to coordinate. On the supply side, on the other hand, it was difficult to limit the payments creditors could collect. Therefore collections occurred in unregulated conditions resembling a common pool situation. Recent proposals to reform the international financial system by using IMF sanctioned moratoriums, majority clauses, or even a sovereign bankruptcy court, reflect precisely such concerns. However, we argue below that these measures may fail to improve efficiency ex-ante. The control problem of creditor j can be described by the current value Hamiltonian4 H= σ σ −1 c j (t ) (σ −1) / σ + λ (t ){y (k (t )) − d j (t ) − (n − 1)d (k (t ))}+ ψ (t ){rf (t ) + d j (t ) − c j (t )} (5) where d (k (t )) is the identical debt collections of the other n − 1 players. The first order conditions for {c j (t ), d j (t )} and the state variables {k (t ), f j (t )} , and the possible equilibria, are described in the appendix. However, a key property of any interior solution is that a creditor collects from the debtor until 4 In an interior equilibrium the capital stock of the debtor does not fall to zero so we do not have to consider a non-negativity constraint on capital. The capital stock only falls to zero if there is a “grab race” equilibrium. We discuss this case later. 8 y ' (k ) − (n − 1)d ' (k ) = r (6) where we have assumed that the left hand side is concave in the capital stock. (6) shows two key results. First, there is excessive debt collection because the creditor income maximizing capital stock k m satisfies y ' (k m ) = r (7) and therefore ke < km (8) where the superscript ‘e’ denotes an equilibrium outcome. Second, debt collections are strategic complements: if other creditors respond more strongly to every capital stock selected by a creditor, so that d ' (k ) rises, then the creditor prefers to lower k . However, the Inada condition on the production function implies that in an interior equilibrium the capital stock of the debtor remains larger than zero. In fact the interior equilibrium can be relatively “friendly” because the creditors can potentially coordinate on non-aggressive reaction functions d ' (k ) . In this case the perceived private return is close to the income maximizing level and a net capital inflow may prevail even after default. The common pool situation is therefore no prisoner’s dilemma but a coordination game.5 However, 5 Dutta and Sundaram (1993) derive conditions for when a tragedy of the commons does or does not have to occur under common pool access. Copeland and Taylor (2004) present a model where regulatory institutions form endogenously under international trade as a function of the costs and benefits of policing 9 comparing (6) and (7) common pool access always reduces the capital stock below the income maximizing level. Also, creditors can potentially coordinate at the other extreme, on the non-interior grab race equilibrium where each creditor tries to cease the entire capital stock immediately. Since this equilibrium will be worse for the debtor, however (see below), we can focus without loss on the interior equilibrium. 2.2. Ex-ante versus ex-post efficiency This section considers lending with two types of creditors: a single creditor (or equivalently a group of creditors which can coordinate to maximize their joint incomes after default) and a group of non-coordinating creditors which will collect the debtor’s capital stock under common pool conditions, as described above, if there is default. Suppose first the case of a single creditor. With a single creditor it is possible to renegotiate the contract ex-post and avoid the common pool inefficiency. The renegotiation leads to k m . Consequently, creditors will leave the debtor with a larger capital stock than under the common pool regime. However, except under extreme and we believe unrealistic conditions this rise in the capital stock will leave the debtor better off as well. A simple way to model this is to assume that the true production function is (1 + a ) y (k ) , a > 0 , but the debtor can always consume ay (k ) or invest it for returns creditors cannot access (for example, in non-tradables production, the informal sector, or use of a resource. Thus common access, private access, or regulated common access can emerge. Future research could construct a similar model for international lending. For example bank lending in the 1970s and 1980s worked different institutionally from bond lending in the 1990s. 10 third country assets), so the function y (k ) is now the output accessible to the creditor(s). a could reflect moral hazard or adverse selection problems giving rise to information rents or sovereign immunity. The debtor is clearly better off when left with the stream { y (k m )} than with the stream { y (k e )} after default so common pool access ex-post will raise the credit ceiling ex-ante. This raises ex-ante efficiency for a credit constrained country. More formally suppose that the debtor, risk neutral for simplicity, can appropriate (1 + r )k (0)∆t (9) by failing to repay the creditor(s) for ∆t units of time before the creditor(s) begin(s) the punishment. Postponing the punishment by ∆t is necessary because the model is in continuous time and creditors can freely collect capital after default. The debtor discounts the future at rate δ . Now a loan of k (0) can be enforced with a single creditor if, assuming only financial means to punish are available, (1 + r )k (0)∆t ≤ e −δ∆t ∞ ∞ ∫ [(1 + a ) y (k (0)) − (1 + r )k (0)]e −δs dt − ∫ ay (k m (t ))e −δs dt t =0 t =0 (10) and with common access in the punishment phase due to lack of coordination and appropriate contract design if 11 ∞ ∞ (1 + r )k (0)∆t ≤ e −δ∆t ∫ [(1 + a ) y (k (0)) − (1 + r )k (0)]e −δt dt − ∫ ay (k e (t ))e −δt dt t =0 t =0 (11) and clearly the right hand side is larger in the second equation for any k (0) since ke < km . Due to the competitive lending market the debtor receives all gains to trade no matter what the post default regime. However, post-default a single or coordinated creditor can extract more from the country on the margin than can each creditor in an uncoordinated group in a common pool regime. However, the higher and more efficient capital stock inevitably helps the debtor. Any new capital stock less than each creditor’s income maximizing level is not a Nash equilibrium. 2.2.3. Lump-sum fines removing the debtor’s surplus Kletzer and Wright (2000) present a renegotiation-proof international lending model in which they show that a contract exists such that, following default, lending remains efficient subject to rationality and no-default constraints but all gains to trade now accrue to creditors, that is, the continuation value of the debtor is zero and at the corner of the Pareto frontier of feasible continuation values. This is particularly significant because the default punishment replicates the punishment from future exclusion from the credit market without being inefficient off the equilibrium path (hence the equilibrium is renegotiation-proof). The punishment is implemented by the debtor making a lump sum 12 payment after default equal to its continuation value from the relationship, after which the relationship resumes. This raises the question if a similar contract will work here: suppose that after default, with either one or many creditors, the debtor must pay its continuation value ay (k i ) / δ , i = e, m . Then its post-default payoff would be zero in present-value terms and the feasible loan k (0) would rise. More importantly, the debtor would be indifferent between borrowing from a single creditor or a group of uncoordinated creditors since the punishment and in turn the credit ceiling is the same. Additionally, since lenders prefer a single lender to punish off the equilibrium path presumably lending should not involve common pool conditions after default but an orderly ‘fine’ payment followed by resumed lending. The threat to not resume lending before the fine is paid would be credible at every instant if the creditor expects the fine to be paid at the next instant and the debtor finds it rational to pay the fine since otherwise lending will not resume (as an analogy, in a two-player war of attrition (continue, quit) at every instant is a perfect equilibrium). Could such a ‘fine’ clause in the (implicit) contract make the common pool regime unnecessary? First, in the model currently the country has no production without foreign capital so its income would be zero and paying ay (k i ) / δ would then be infeasible. This liquidity problem does not arise in Kletzer and Wright (2000) due to their endowment output assumption. Realistically, here as well, the debtor should earn income without foreign capital. However, suppose (1) that with the maximum fine the country can finance, it still cannot finance the present value of continued lending with a single creditor ay (k m ) / δ . After all, this value of access to foreign capital now and in all future could be substantial and the real counterpart of the fine is net exports, the value of which may be limited. Additionally, (2) even if the debtor can pay off ay (k m ) / δ with its future 13 net export stream, it may not be able to do so right away or even with a continuous payment stream: if net exports nx accrue every instant, then the debtor can pay off its agency rent continuously if nx ≥ ay (k m ) and otherwise it cannot. If it can not, the most ~ efficient capital level k post-default which leaves the debtor no surplus solves ~ ~ nx = ay (k ) and such k < k m may be implemented by an appropriate choice of n , the number of creditors who would engage in common pool debt collection after default. Thus, the common pool regime can remain relevant by raising the debt ceiling ex-ante. Additionally, while there is continuing debate, the evidence that creditors confiscate net export earnings for debt repayment is limited (Mitchener and Weidenmier 2005, Sandleris 2006, Martinez and Sandleris 2004). 2.3. Is the common pool always better? In the perfect information model considered so far dispersed creditors and complete commitment not to renegotiate the contract after default is always better. In practice, however, involuntary or strategic defaults do happen - due perhaps to asymmetric information (Atkeson, 1991), non-representative governments, or maturity mismatches in a stochastic model. Because defaults do happen “on reality’s equilibrium path” ex-ante efficiency and ex-post efficiency must be traded off. An easy way to model this is to put a cap on the collection function d (k ) creditors can employ after default. The optimal cap balances increased lending ex-ante when the cap is higher with a rising common pool inefficiency cost ex-post, and the cap can accommodate both extremes. To implement such a cap the market can use the contract specification or policy makers can use the design of the international financial system (Pitchford and Wright, 2007). 14 3. Conclusion Providing incentives for countries to repay national debts requires a punishment for default. However, this alone does not mean that defaults should cause inefficiency because efficient resolution coupled with a redistribution of the gains to continued trade away from the debtor can, in principle, replicate the penalty from the inefficient resolution. However, ex-post efficiency warrants leaving the debtor with a high capital stock and this is likely to raise its welfare. Therefore, ex-post creditors cannot benefit themselves without helping the debtor as well. In turn, to raise its credit ceiling ex-ante a debtor may prefer creditor commitment not to renegotiate. One commitment device is to use a dispersed anonymous creditor base with common pool access to debtor funds after default. This can explain the frequency of creditor “panics” and output losses following financial crises and predicts a trade-off between ex-post and ex-ante efficiency in practice. 15 References Alesina, A., Drazen, A., 1991. Why are Stabilizations Delayed? American Economic Review no. 81 5 1170-88 Atkeson, A., 1991. International Lending with Moral Hazard and Risk of Repudiation. Econometrica, 59, 4, 1069-89 Bulow, J., Klemperer, P., 1997. The Generalized War of Attrition. American Economic Review 89, 1, 175-89 Bulow, J., First World, 2002. Governments and Third World Debt. Brookings Papers on Economic Activity I, 229-255 Calvo, G. A., Mendoza, E. G., Mexico’s Balance of Payments Crisis: A Chronicle of a Death Foretold. Journal of International Economics 41, 3, 235-64 Copeland, B. R., Taylor, M. S., 2004. Trade, Tragedy and the Commons. NBER Working Paper 10836 Diamond, D., Dybvig, P., 1983. Bank Runs, Deposit Insurance and Liquidity. Journal of Political Economy 91, 401-19 16 Dooley, M. P, Verma, S., 2001. Rescue Packages Output Losses Following Crises. NBER Working Paper 8315 Dutta, P. K., Sundaram, R. K., 1993. The Tragedy of the Commons? Economic Theory 3, 413 - 426 Dutta, J., Rowat, C., 2006. The Road to Extinction: Commons with Capital Markets. Working Paper, University of Birmingham Fearon, J. D., 1995. Rationalist Explanations for War. International Organization. 49, 379-414 Fischer, S., 1999. delivered On the Need for an International Lender of Last Resort, paper at the AEA/AFA meeting, January, http://www.imf.org/external/np/speeches/1999/010399.htm Kletzer, K. M., Wright, B. D., 2000. Sovereign Debt as Intertemporal Barter. American Economic Review, 90, 3, 621-39 Krueger, A., 2002. A New Approach to Sovereign Debt Restructuring. Unpublished Manuscript, International Monetary Fund, Washington, DC 17 Krugman, P., 1988a. Financing vs. Forgiving a Debt Overhang: Some Analytical Notes. NBER Working Paper 2486 Krugman, P., 1988b. Market-based Debt-Reduction Schemes. NBER Working Paper 2587 Labán and Sturzenegger, 1998. Fiscal Conservatism as a Response to the Debt Crisis. Journal of Development Economics 45, 2, 305-24 Labán, R., Sturzenegger, F., 1994. Distributional Conflict, Financial Adaptation and Delayed Stabilization, Economics and Politics 6, 3, 257-76 Lau, L. J., Qian, Y, Roland, G., 2000. Reform Without Losers: An Interpretation of China’s Dual-Track Approach to Transition. Journal of Political Economy, 108, 1, 12043 Morris, S., Shin, H. S., 2004. Coordination Risk and the Price of Debt. European Economic Review 48, 1, 133-53 Mitchener K J, Weidenmier M D (2005) Supersanctions and Sovereign Debt Repayment. NBER Working Paper 11472 Paoli, B. De, Hoggarth, G., Saporta, V., 2006. Costs of Sovereign Default. Bank of England Financial Stability Paper No. 1, July 18 Pitchford, R., Wright, M. L. J., 2007 Restructuring the Sovereign Debt Restructuring Mechanism. Working Paper, University of California Los Angeles Radelet, S., Sachs, J. 1998 The Onset of the East Asian Financial Crisis. Working Paper Harvard Institute for International Development Roubini, N., Setser, B., 2003. Improving the Sovereign Debt Restructuring Process: Problems in Restructuring, Proposed Solutions, and a Roadmap for Reform. Paper prepared for the conference on “Improving the Sovereign Debt Restructuring Process”, co-hosted by the Institute for International Economics and Institut Francaises des Relations Internationales, Paris, March 9 Sandleris G (2006) Information, Investment and Credit. Working Paper, Johns Hopkins University Sandleris G, Martinez J V (2004) Is it Punishment? Sovereign Defaults and the Decline in Trade. Unpublished Paper Columbia University Shleifer, A., 2003. Will the Sovereign Debt Market Survive? American Economic Review 93, 2, 85-90 Sturzenegger, F., Zettelmeyer, J., 2006. Debt Defaults and Lessons from a Decade of Crises. Cambridge: MIT Press 19 Tomz, M., Wright, M. L. J., 2005. Sovereign Debt, Defaults, and Bailout. Working Paper, Stanford University Velasco, A., 1998. The Common Property Approach to the Political Economy of Fiscal Policy. In Sturzenegger,F., Tommasi, M., 1998 The Political Economy of Reform. MIT Press, Cambridge 20 Appendix Tornell and Velasco (1992) consider the same model with a linear production technology, y = ak , and solve for three equilibria where creditors collect a constant share of the capital stock d = βk . One equilibrium is interior and the others are at the maximum and minimum collection rates allowed in their paper. We now describe the solution to the model in the main paper. Interior equilibrium The necessary first order conditions for the controls {c j (t ), d j (t )} and state variables {k (t ), f j (t )} in (5) in an interior equilibrium are c −j 1 / σ = ψ (a1) λ =ψ (a2) λ (δ + (n − 1)d ' (k (t )) − y ' (k ) ) = λɺ (a3) 21 ψ (δ − r ) = ψɺ (a4) and the transversality conditions lim λ (t )k (t )e −δt = 0 , limψ (t ) f j (t )e −δt = 0 t →∞ t →∞ (a5) and (2) – (4). Since the Hamiltonian maximized with respect to the controls max H is c j ,d j concave in the state variables {k (t ), f j (t )} , as long as the constraint set is convex the first order conditions are sufficient for a maximum. The maximized Hamiltonian is concave since the first and second principal minors of the associated Hessian are negative and zero, respectively, and the Hessian therefore negative semi-definite. The stock growth constraints define a convex set if y (k (t )) − (n − 1)d (k (t )) is convex in k (t ) , which is true since we assume in the text that y (k (t )) − (n − 1)d (k (t )) is concave. (a1) and (a2) state that the value of consuming a unit of output today must be equal to the value of increasing either of the two asset stocks by one unit and (a3) and (a4) define the optimal growth rate of the identical shadow values of the two capital stocks. For the debt claim the return for each agent is only y ' (k ) − (n − 1)d ' (k ) < y ' (k ) : due to the common 22 resource regime capital on the margin is perceived to yield the social return net of the rise in collection by other creditors it leads to. Combining (a1)-(a4) implies cɺ j cj = −σ (δ + (n − 1)d (k (t )) − y ' (k ) ) = −σ (δ − r ) so that c j (t ) = c j (0)eσ ( r −δ )t (a6) and from (2) t k (t ) = k (0) + ∫ y (k ( s )) − nd (k ( s ))ds (a7) 0 and from (4) t f j (t ) = f j (0)e rt + ∫ (d (k ( s)) − c j ( s ))e r (t − s ) ds s =0 23 t = f j ( 0) e + rt ∫ d (k (s))e s =0 r (t − s) ds − c j (0)(eσ ( r −δ ) t − e rt ) σ (r − δ ) − r (a8) using (a6). Now using the transversality condition for the international asset in (a5) to find c j (0) we get t c j (0)(eσ ( r −δ )t − e rt ) limψ (t ) f j (t )e −δt = limψ (0)e (δ − r )t e −δt f j (0)e rt + ∫ d (k ( s ))e r (t − s ) ds − t →∞ t →∞ σ ( r δ ) r − − s =0 t c j (0)(e (σ ( r −δ ) − r ) t − 1) − rs =0 = limψ (0) f j (0) + ∫ d (k ( s ))e ds − t →∞ − − σ r δ r ( ) s =0 (a9) which says that the present discounted value of assets net of consumption converges to zero. From (1) and (a6) bounded lifetime utility requires σ (r − δ ) − r < 0 (a10) which we will assume. Therefore in (a9) 24 c j ( 0) e ( σ ( r − δ ) − r ) t goes to zero and therefore t c j (0) = (r − σ (r − δ )( f j (0) + lim ∫ d (k ( s ))e − rs ds ) t →∞ s =0 = ( r − σ ( r − δ ) k ( 0) (a11) where the last equation follows by arbitrage: all assets earn a private return r so the value of what is collected equals the value of the initial asset stock k (0) . While the capital stock of the debtor is inefficiently low in equilibrium (see (6)) there does not have to be a net capital outflow: even ensuring y ' (k ) − (n − 1)d ' (k ) = r can require a capital inflow. Agents then take a short position in the international asset. Grab race equilibrium 25 If the equilibrium is not interior it is either a grab race or creditors cannot supply enough capital because they cannot go short in the foreign asset. We first consider the grab race equilibrium. Here each creditor wants to appropriate the entire capital stock immediately because others are expected to do so. This lowers the creditor payoff for a given capital stock since in (a11) each gets only k (0) / n and the capital stock is instantly depleted. Equilibrium when creditors cannot go short in the foreign asset If creditors cannot take a short position in the foreign asset and this constraint is binding – the creditors may be hedge funds who wish to limit their exposure – then y ' (k ) − (n − 1)c' (k ) > r Since agents would like to take a short position in the foreign asset but cannot the economy resembles a closed economy and c j (t ) = d j (t ) so (5) simplifies to H= σ σ −1 c j (t ) (σ −1) / σ + λ (t ){y (k (t )) − c j (t ) − (n − 1)c(k (t ))} (5’) with first order conditions 26 c −j 1 / σ = λ (a1’) λ (δ + (n − 1)c' (k (t )) − y ' (k ) ) = λɺ (a3’) and the transversality condition. lim λ (t )k (t )e −δt = 0 t →∞ Then again cɺ j cj = −σ (δ + (n − 1)c' (k (t )) − y ' (k ) ) and c j (t ) = c j (0)eσ ( rˆ −δ )t , rˆ = y ' (k ) − (n − 1)d ' (k ) (a6’) and capital accumulates according to 27 t k (t ) = k (0) + ∫ y (k ( s )) − nc(k ( s ))ds (2’) 0 And from the transversality condition lim λ (t )k (t )e −δt t →∞ = lim λ (0)e t →∞ (δ − rˆ ) t e −δt t k (0) + ∫ y (k ( s )) − nc(k ( s ))ds 0 t c j (0)(e (σ ( rˆ −δ ) − rˆ ) t − 1) = λ (0) lim k (0) + ∫ y (k ( s )) − (n − 1)c(k ( s ))e − rˆs ds − t →∞ ˆ ˆ − − σ r δ r ( ) 0 (a9’) and similarly to before to ensure bounded utility we assume σ (rˆ − δ ) − rˆ < 0 (a10’) which means that in (a9’) c j (0)e (σ ( rˆ −δ ) − rˆ ) t 28 converges to zero and so t c j (0) = (rˆ − σ (rˆ − δ ) lim k (0) + ∫ y (k ( s )) − (n − 1)c(k ( s ))e − rˆs ds t →∞ 0 t = (rˆ − σ (rˆ − δ ) lim k (0) + ∫ y (k ( s ))ds − (n − 1)c j (0)e (σ ( rˆ −δ ) t − rˆ ) s ds t →∞ 0 t c j ( 0) = (rˆ − σ (rˆ − δ ) lim k (0) + ∫ y (k ( s ))ds − (n − 1) t →∞ ˆ ˆ − − ( r σ ( r δ ) 0 (using symmetry in consumption) so t 1 c j (0) = (rˆ − σ (rˆ − δ ) k (0) + lim ∫ y (k ( s ))ds t →∞ n 0 (a11’) 29 which can be used in (a6’) to get consumption as a function of time. Substitution into (2’) then gives a single first-order differential equation in the capital stock which can be solved in principle after substituting specific functional forms for y (k ) and using c j (t ) = d j (t ) for all creditors. 30