B31 /^EVWEV 01 M415 working paper department of economics EMERGING MARKET CRISES: AN ASSET MARKETS PERSPECTIVE Ricardo J. Caballero Arvind Krishnamurthy No. 99-23 October, 1999 massachusetts institute of techinology 50 memorial drive Cambridge, mass. 02139 WORKING PAPER DEPARTMENT OF ECONOMICS EMERGING MARKET CRISES: AN ASSET MARKETS PERSPECTIVE Ricardo J, Caballero Arvind Krishnamurthy No. 99-23 October, 1999 MASSACHUSEHS INSTITUTE OF TECHNOLOGY MEMORIAL DRIVE CAMBRIDGE, MASS. 02142 50 Emerging Market An Ricardo Crises: Asset Markets Perspective J. Arvind Krishnamurthy* Caballero This draft: October 08, 1999 Abstract The whole sharp between a mild downturn and a deep difference fire sales of domestic assets in the balance sheets of We Why and how do we argue that a poor contractual build a stylized model that explains the dramatic short run shift in the interest parity condition international collateral. weak domestic sector. a parsimonious account of these and related phenomena present in offers emerging markets. and the ensuing collapse doesn't the private sector take appropriate precautions to avoid the dear consequences of crises? In this paper environment rates, both the financial and non-financial And why such crises materialize? and possibly exchange the occurrence of crisis is The collateral, or —or — fire sale in terms of the weakening of a country's lack of precautions, on the other hand, is a consequence of underdeveloped domestic financial markets, which drive a wedge between the private and social valuation of assets appealing to foreign investors. Moreover, domestic banks are the mechanism by which firms with purely domestic collateral are able to access international markets. Thus, the banks' endogenous collapse as domestic assets depreciate, sharply reduces access to international financial markets and may in so doing Keywords: trigger a feedback process with multiple equilibria features. Capital flows, fire sales, financial constraints, contractual and corporate governance problems, balance sheets, wasted collateral, domestic and foreign spreads, excessive leverage, collateral underprovision, real depreciation, banks. 'Respectively: MIT and NBER; Northwestern University. Bhagwan Chowdhry, Bcngt Holmstrom, Andres tional the for financial support, was accomplished, First draft: EFG-NBER meetings, WFA meetings, the CEPR LACEA/UTDT Summer Camp in Interna- Financial Markets (Gerzensee), and the Economics and Finance (Buenos Aires) NSF thank Daron Acemoglu, V.V. Chari, Velasco, Luigi Zingales and seminar participants at the Fed (Board of Governors), Harvard, IMF, MIT, Northwestern, Summer Symposium on We for their August 1998. for their comments and suggestions. Ricardo Caballero thanks and the AEI and Research Department at the IMF, where part of this work hospitaUty and support. E-mails: caball@mit.edu, a-krishnamurthy@nwu.edu. Digitized by the Internet Archive in 2011 with funding from Boston Library Consortium IVIember Libraries http://www.archive.org/details/emergingmarketcrOOcaba Introduction 1 Following the recent experiences in emerging markets, the merits of an open capital account are once again being called into question. inflows Weighed against the obvious the specter of capital market instability. is was dramatically worsened by the reluctance internal and external demand fell, in East Asian countries of international investors to provide the re- smooth the impact of a combination of sources needed to As The downturn benefits of sustained and external internal —and capital flows external credit in general many leading and currency depreciations took unexpected proportions, asset deflation observers to assert that [the declines were] "well beyond by any reasonable reassessment of economic fundamentals, even (IMF World Economic Outlook, May The — van- As symptoms ished, bringing the real sector, especially nontradeables, to a virtual halt.^ and consequences, deficiencies. what was justifled in the light of the crisis." 1998, p.4.) emphasis from problems of the current account to those of the capital ac- shift in count has led to a rethinking of the traditional "goods markets" view of external A misaligned real exchange rate ternal imbalance imbalances. and is no longer the central element in understanding an ex- in guiding policy responses to avert Although certainly an important factor economy, this relative price ingredient asset prices and is and we in the long - a run readjustment of an and our link which, in our view, movements is in at the core ability to control them.^ present an "asset markets" view of external crises, in which asset values their feedback into real activity take center stage. financial links, harmful consequences of these often subordinate to the short run their feedback into real activity of understanding the magnitude of crises In this paper crises. We show that weak international when combined with underdeveloped domestic financial markets, provide a parsimonious account of the chief features of systemic crises and the scenarios where they develop. The paper proceeds • First, 'For the we provide a first round of in three steps: framework that captures the main features of stylized crisis-countries was probably higher since "errors and — Indonesia, Korea, Malaysia, the Phihppines and Thailand — net capital flows declined from 73 billion dollars in 1996 to inflows real minus 11 billions in 1997. and omissions" went from The -8.5 billions in actual decline in net 1996 to -19.5 billions in 1997. ^The literatures on debt crises, currency speculation, and bank-runs, offer important insights on different aspects of assets markets during crises. more likely by the issues we emphasize Most of the effects in these literatures are either in this paper. For papers on currency speculation (1994, 1996); on debt crises see, e.g., Calvo (1988), Giavazzi amphfied or made see, e.g., Obstfeld and Pagano (1990), Cole and Kehoe (1996); on bank runs see the canonical Diamond and Dybvig (1983) and the apphcations to open economies by Goldfajn and Valdes (1998) and Chang and Velasco (1998a). interesting discussion Also see Calvo (1998) for an overview and on implications and sources of capital market crises in small open economies. financial variables during a crisis. A weakening of international financial links - represented in our model as a small negative shock to international collateral- can lead to falls in asset prices below fundamentals and contractions The in real activity. response of the economy to these shocks can be very non-linear depending on the condition of the aggregate balance sheet of the country. • We move on to develop the dynamic implications of our framework and arrive at our main conceptual oped domestic collateral result. financial We show that domestic agents markets —represented in in economies with underdevel- our model as a scarcity of domestic — systematically underestimate the social value of their access to interna- tional capital markets. As a and experience more costly framework • Finally, the is result, these economies arrive to downturns unprepared, crises. enriched to include a banking system, as banks have played We central roles in recent emerging markets crises. show that the fall in asset prices can compromise the banks' balance sheets, resulting in a domestic "credit crunch" which amplifies the shock to international collateral and brings about the possibility of multiple equilibria. An emerging economy, almost by resources to carry on with as the inadequacy its normal definition, activities. is one which requires substantial foreign Weak international financial links, reflected on (real or perceived) of a country's international collateral, place limits the country's ability to acquire international financing.^ During crises, as triggered by a decline in terms of trade or a tightening in international financial markets, the country's international collateral proves inadequate to finance normal operations. Domestic firms, needing foreign funds, willingly exchange domestic assets for international collateral at sale prices."* Domestic interest rates surge, not to increased default probability, as first form concept myriad for higher for expected inflation or an and second generation models would have ration the scarce international collateral.^ on, as well as the compensate Our quote from the IMF's WEO 'Witness current account expect By if it, but to (1998) earlier stories of sharp declines in "risk-appetite" (a practitioners' reduced expected return) in the aftermath of systemic crises in emerging markets, seem to provide widespread support for this central implication.^ may fire deficits of at largest 15% of GDP as opposed to the 150% The meltdown of GDP that one capital markets were fully integrated. See Kletzer (1988) for a similar argument. a fire sale, we mean a similar to that in Shleifer decline in the price of an asset and Vishny beyond (1993), although in our its fundamentals. This definition model the fire sale price is determined is in equilibrium rather than by the exogenous bidding price of an inferior user. ''The modeling choices highlight this rationing aspect of the rise in interest rates. believe that devaluation This is Though, we certainly and default expectations are important factors as well. first and second generation models are not important, not to argue that the insights of so called and aggregate of balance-sheets firms — those with limited international collateral i.e. nontradeables sectors and small activity, especially in the While shortages of international collateral — is just another facet of this process. can characterize crises, the incidence of these shortages arises from underdeveloped domestic financial markets, which reflect the inad- equacy domestic (real or perceived) of domestic collateral creates Insufficient collateral. a spread between the internal rate of return on a good project in the economy and the rate of return that can be credibly promised to outside domestic financiers. tic spread is just the dual of an This domes- domestic capital allocation process - or poor inefficient In a dynamic setting, the domestic spread leads collateral aggregation in our language. The reasoning is a shortage of international collateral, abating it to an undervaluation of international collateral that exacerbates crises. straightforward. Since in a crisis there is requires an ex-ante decision to carry an extra unit of international collateral into the crisis. However, doing so provides the holder collateral -in fact, the shortfall is less than the full social return on this international proportional to the domestic spread. Anticipation of this reduces the private sector's incentive to accumulate and provision international collateral. The fragility resulting We costly.^ show that from such under-accumulation, or over-leverage, makes this inefficiency is due to a pecuniary externality. Thus, efficiency in domestic financial markets, rather links alone, that may justify capital controls crises it is more the in- than the weakness in international financial and equivalent measures.^ In emerging markets, liquid domestic debt markets are the exception rather than the rule. Invariably, the financial system is bank based. Banks issue claims to foreign investors against balance sheets that are partially backed by international collateral, and lend domestically to those that do not have direct access to international market. This brings a third but to highlight a significant difference between "solvency" type models where interest rate "arbitrage" holds and models where it does not, as Tirole (1998b). Note that class, since return is is the case in ours or in Shleifer and Vishny (1997), or Holmstrom and Diamond-Dybvig type models also belong to the solvency-arbitrage "illiquidity" the real costs of the run validate the run, by which fair in the bad equilibrium as well. Later on we it is often meant that the equilibrium required discuss the natural complementarities between these two classes of models. ^Alternatively, shocks may be if domestic financial markets are extremely primitive, volatility costs due to external replaced by depressed growth and over-exposure to domestic shocks as the economy to aggregate internal resources and use its international collateral. Similarly, limited to start with, international leverage disappears and with shocks. We do not emphasize these cases as our purpose than stagnation and traditional domestic business financial fragihty arises cycles. is it if is international collateral unable is very so does volatility with respect to external to emphasize crises in Also, see Aghion open economies rather et al (1998), for a model where from the mechanics of debt accumulation rather than from the undervaluation of collateral provision. Interestingly, they also find that growth volatility is maximized for intermediate levels is also central to the of financial development. In Caballero and Krishnamurthy (1999) debate on the virtues and we argue that pitfalls of alternative this simple observation exchange rate systems. balance sheet tion. in —that of the banking system— into As the country faces an international crisis tiie story, and and with further amphfica- it asset prices collapse, the deterioration banks' balance sheets limits their ability to intermediate foreign funds into the domestic system, further deepening the crisis, and so on. If this feedback is strong enough, the size depends on the extent of pessimism, as the model exhibits multiple of the crisis The framework we equilibria. use gives central roles to credit market constraints arising from mi- croeconomic contractual problems in the process of creation and aggregation of international collateral.^ There are, in principle, many factors behind the incomplete development of both domestic and international financial links in emerging economies, ranging from macroeco- nomic and political instability to Our model emphasizes the weak corporate governance and contractual environment. both because they are latter factors less well understood within the context of sovereign risk and because their importance has been heightened by the recent Asian crisis. Relation to the literature. opment of The Latin American debt crisis of the 1980s led to the devel- of formal models of sovereign debt renegotiation, which posed the central question what can international lenders threaten sovereign countries with fault (see Eaton and Gersovitz (1981), Bulow and Rogoff in the event of a de- (1989)). This is closely related to our identification of international collateral. Since in this literature default and its costs are mostly modeled as a country-wide phenomenon, the question immediately arises on whether the domestic private sector internalizes the effect of their actions on the likelihood of these events. Bardhan (1967), Harberger (1985) and Aizenman (1989) advocated taxing capital flows on the grounds that individual agents do not pay for the increase in a country's aver- age cost of capital brought about by their international indebtedness. While our results on collateral underprovision are related to theirs, our externality stems from imperfections in the domestic financial system rather than from the country's monopsony capital flows power and in international financial martkets or any sovereign Although we touch on many of the insights of the eling and conceptual ^According to the differences with IMF it are large. On constraint.-^" debt-crisis literature, both our mod- these aspects, our approach (1998, pp. 63-64), a distinguishing characteristic of the Asian crisis macroeconomic considerations played a lesser role than in previous crises. Also, see Johnson, ct is contractual environment) in an economy and the incidence of the current crisis. more that purely al. evidence on the correlation between weak corporate governance and institutional environment is (1998) for (i.e. a weak As the corporate finance hterature has demonstrated, weak corporate governance and, more generally, weak contractual enforcement go hand in hand with credit market constraints. '"Similarly, our policy recommendation to increase collateral availability ment in the is — while not its origin — in favor of subsidizing the tradable sector akin to Borensztein and Ghosh's (1989) advocacy for subsidizing invest- export sector. See Fernandez-Arias and Lombardo (1998) for a recent model of sovereign risk and overborrowing. closely related to the new literature and Moore (1997) place the on the role of collateral in collateral role of a fixed input macroeconomics. Kiyotaki —land— and its distribution across agents at the center of their amplification mechanism.^ ^ Krishnamurthy (1998) shows how a richer set of contracts than that allowed by Kiyotaki and Moore shifts the relevant constraint from that which applies to individual agents to the total availability of collateral in the Holmstrom and economy. Tirole (1998a) argue that the stock market alone provide insufficient collective collateral if They not aggregated properly. even when wasted corporate collateral (liquidity) is also may show that not an issue, public supply of it may be required in the presence of large shocks. In terms of our modeling choices, our paper canonical model of liquidity, and particularly to model As in the context of firms. is related to Diamond and Dybvig's Holmstrom and in that paper, firms in Tirole's departure (1983) from this our economy receive a production shock that they must cope with by mortgaging collateral in order to borrow additional The fimds. Analogously, firms in the model of Holmstrom- Tirole can borrow funds either investors. from other firms the first can be obtained from other domestic firms or directly from international latter in the corporate sector or directly and efficiency of the second transaction. transactions encounter frictions in our model. on many differences in terms of substance Indeed, and more In contrast, both generally, while there and modeling aspects with each of the articles and macroeconomics mentioned above, our central theoretical departure with collateral them Intermediation renders transaction frictionless in Holmstrom-Tirole and their macroeconomic analysis focuses on the implementation are from savers. as a group is our focus on the presence of two collateral constraints an international one — and Our underprovision their rich static as well as dynamic —a domestic and interactions. of collateral results are analogous to the free rider problems and un- derprovision of liquidity studied in the banking literature (see Jacklin (1987), Bhattacharya and Gale model is However, the source of the externality behind (1987)). this inefficiency in our the endogenous result of imperfect contract laws and institutions, while theirs the exogenous divergent ex-post valuation of liquidity by consumers. from these models is A is second departure our focus on markets rather than on banking arrangements and their key sequential servicing constraint. While we do believe that bank failures and runs can aggravate of crises crises, as and we discuss in section 4, our their prelude arise even ply liquidity. Moreover, the fire sale model points out that the basic features when no agent or institution is committed to sup- mechanism we emphasize boosts and feeds from the likelihood of banking crises themselves. Outline. Section 2 follows this introduction with a description of the basic model and See Bernanke et nomics. al (1999) for a thorough survey of the literature on credit constraints and macroeco- weak connections with international discusses the aggregate implications of kets. Section 3 highlights the implicit transactions in collateral aggregation, frictions in financial mar- and introduces domestic financial markets. Banks as collateral aggregators and as a source of amphfication and multiple equilibria are discussed in section 4. Section 5 concludes, and is followed by an appendix. A 2 Time-to-build model with limited international collat- eral In our view, there are two central ingredients behind the recurrent crises affecting emerging economies: weak international financial links and underdeveloped domestic financial mar- We kets. start with a discussion of the ingredient in isolation, while the next section first brings into the analysis frictions in domestic financial markets. The Economy 2.1 Our goal Time. in this section is to highlight the consequences of suffering insufficient (perceived) international collateral during crises, as well as to understand motives for un- dertaking precautionary measures in anticipation of more stringent international financial conditions. lem in Our timing assumptions are designed to capture these two aspects of the prob- a minimalist framework. The world flexible period and portfolio lasts three periods. when economic allocations. shift resources indexed as t = 0,1,2. Date is the fully agents design the productive structure, ownership structure, Date when Heterogeneity cial frictions. is 1 is the crisis period, when economic agents struggle to from the future to cope with shocks in the present. Date 2 represents the unconstrained future, Agents. Time the economy among agents is is (relatively) rich in resources. required to understand the consequences of finan- We have two dimensions of heterogeneity: consumers and foreign financiers, The domestic economy is one between domestic entrepreneurs- and another one among domestic entrepreneurs. populated by a continuum of unit measure of agents with Cobb-Douglas preferences over date 2 consumption, u'' where, ct is (ctc„)l/^ consumption of tradeable (T) goods at date 2 and c„ tradeables (N) at date N-goods as = e. By 2. We is consumption of non- denote the date 2 real exchange rate of T-goods in units of specifying preferences over date 2 consumption only, we remove from our analysis standard intertemporal substitution and smoothing arguments which are not central to our approach. The world rest of the is represented by foreign investors consumption of tradeable goods at The = Ct,0 + Ct^l + Ct,2may be contrast with domestic agents arises because foreign investors supply tradeable goods in the date They have period. called period as well as during the date 1 crisis endowments of T-goods large preferences over dates, all u^ who have at all dates. pre-crisis we assume Additionally, down that they do not discount the future so that the preference structure pins on to the gross international interest rate at one. At date Endowments and production. 0, ex-ante homogenous domestic agents are endowed with resources of wq T-goods which can be invested or stored. resources by taking on foreign debt of doj has a time-to-build aspect. Investments are at date Let 2. fc„ and the firms in the N-sector N-goods at date of be offset realize 2. is The fc„ + kt make saving T-sector is < l in both N » Rail in - manner - of production units shocks to the N-sectors. is 1, realized and T at kt requires a fraction 1 —p of. A„ = /?„ — r„ > 0, which can we assume result in substantial additional Production per unit of capital in the critical link 1 below. between financing and production any economy have ongoing capital needs (working is our down production crisis. The dual many capital, etc.). units in a potentially of this shutting that the marginal value of capital during a crisis implication of our modeling underlies is down very high. This of the key results in the paper. '^Nonetheless, there are a few interesting peripheral issues that arise once the T-sector idiosyncratic shocks. For example, a Kiyotaki that, p) illustrated in figure to anticipate results, this is At date N and fc„ is -tMiKn- and T-sectors does not Starving firms of capital has the effect of shutting wasteful creating capital of = Vn + ^n"n)kn ^ simply Rt-^^ The time-line Firms Then distressed production units sufficiently profitable, restrict these crisis. and output of kn, in units of the tradeable good, in order to This time-to-build structure underlines a during a 1, of: Because having shocks we 0. rest experience a productivity fall, A„ insights, Production this shortly. spared of further investment and go on to produce Rnkn units Rniynjkn In order to and at dates units of tradeable goods. by reinvesting a fraction On output at date 2 made inherited from date 1, investment totaling a date expand on shall increase these denote the total amount of capital devoted to sectors kt the beginning of date we - They can and Moore (1997) type multiplier may arise. is subject to The domestic economy have no tradable goods at date will 1 to contribute and capital needs can only be accommodated by an increase in imports. In order to finance this increase, domestics must access international capital markets. DateO Date —I Date 2 1 + Borrowing + Repayment Consumption Crisis Investment Budget Constraint: Date T-Production: do f + Wo = + k„ k, k, > R,k, ^ Rnkn N- Production: Time-to-build (INTACT) k„ Prob = p (DISTRESSEDT' Need rnkn+ A„0„ additional investment of e„ k„ Our assumptions on with capital of this kt 1 : K Time Line Foreign investors extend finance only with the security of col- Collateral Constraints. lateral.^"* = RnOn) T-goods Figure k„ collateral are two- fold. First, a firm in the tradeable sector We expects output at date 2 of Rtkt- assume that only a fraction Aj of output can be valued and seized by foreigners in default.^' Second, we assume that foreigners can only value collateral to foreigners. and seize output of the T-sector, while the N-sector provides no Thus, the quantity XtRt^t constitutes the international collateral of the economy. Almost by however the definition, an emerging economy is ability to access international capital in which project returns are high, low. These features are captvired by one is assuming that, Rt This collateral '"We appeal to justification of the may be > but 1 explicit or implicit, as agency conflicts in making XtRt < 1. the case in a surplus-sharing bargaining equilibrium. is this "high-level" assumption. assumption based on costly state verification. It is See the appendix for a deeper costly for a foreign investor to verify and liquidate the output of Rtkt- This implies that only a fraction of the output can be used as Also, see La Porta et al. managers from collateral. (1998) for worldwide evidence on the high positive correlation between corporate governance problems and ownership concentration to prevent (1) (in their interpretation, stealing). 8 higher concentration is needed While the capital market has weak international that the domestic capital market N of output in both is for domestic with A„ collateral, The asymmetric the rest of the world. valuation of goods full value XnRn^n of this output constitutes 1. collateral constraint identifies the domestic Much of the economy as distinct from international economics literature assumes an' asymmetric —foreigners only value T-goods, while domestics value both T and N- In contrast, the collateral restriction goods. until the next section example, a firm in the N-sector with capital of kn (at date 1) expects output at date 2 of Rnkn- Then, = we assume Residents accept as collateral the frictionless. and T-sectors. Thus, links, is that domestic residents extend finance to either tradeable or nontradeable firms, while foreigners accept as collateral only claims the tradeable sector.^^ It is on important to point out that this assumption does not correspond to the empirical restriction that we should only observe foreign credit extended to firms in the T-sector. In our model, foreigners are willing to hold claims on the N-sector but only as long as their market is "liquid, " in the sense that there are to allow an instantaneous trade. 1, The in a crisis state, they will always restriction implied swap these for claims enough claims by the assumption in the T-sector is that at date on the T-sector with a party that does value N-claims ("residents"). The international economics literature has long recognized the importance of interna- tional collateral and its Formal models of relation with a country's tradeable sector.^^ sovereign debt renegotiation are built around the question of what international lenders can threaten sovereign countries with in the event of a default (see Eaton and Gersovitz (1981), Bulow and Rogoff In this literature, international collateral (1989)). is typically taken to be some fraction of exports. While similar in spirit to this identification, our than aggregate. Liquidation and seizure of output foreigners than For instance, they make crisis, that it when it it is it is by the fact that cash revenues back into the country. received. revenues were In many '^See, e.g., N is harder for disadvantage disappears. from exports can be seized before part of the collateral backing the liquidity package instances the bias first is directly justified principles sector to be infinite for foreigners, while Simonsen (1985). microeconomic rather This feature was used by Mexico during the 94-95 made '^This assumption can be justified from output in the is in the non-tradeable sector for domestics, while in the tradeable sector this justified its oil assumption by mandates on foreign by taking the cost of verifying and liquidating it is finite for domestics. on institutional investors (e.g. limits Alternatively, this assumption cations of an an is expanded model where ment and Krishnamurthy is only against a fraction of affect the sector. times." In the churn of any economy, there are Thus, there is (see an economy at date 1 and take two an idiosyncratic shock that demands reinvest- is N p of the firms in the resources than others. to capture the central impli- (1998)).^^ Shocks we described above, there in a fraction mechanism foreigners' collateral-bias Aggregate and idiosyncratic shocks. forms. First, as efficient ^^ and current production of perishable nontradeable goods firms -in either sectorearlier draft, Caballero real estate investments).^^' The idiosyncratic shock reflects "normal some production units which require more a constant need to attract and reallocate resources from some production units to others. Second, we introduce an external shock at date 1 that affects, in the aggregate, the economy's ability to attract resources from foreign investors. XtRth represents international collateral LJ with respect to foreign investors. There are two states of the world at date e {B,G}, differentiated by a low amount of international That of international collateral. collateral 1, and a high amount is, X^ R^ < kt Xt Rt kt Shocks to international collateral can, in principle, arise either as a shock to Rt or a shock to Af. Terms capital of trade shocks are declines in Rt, while a shortage in the supply of international a decline in Aj.^" is To simplify the exposition, '^Of course there are important exceptions (e.g. "too big to we focus only on shocks to At. In fail" utilities), but see, e.g., Kang and Stulz (1997) for systematic Japanese evidence showing that, for small firms, those that are export oriented are favored by foreign investors. See, e.g. Blommestein (1997), for a discussion of how real estate assets considered highly illiquid or exposed to exchange risk are generally avoided (sometimes by foreign institutional investors. rating, are for is crisis sector. by the World Bank, describes firms that borrow in foreign currency as "predominantly large exporting firms with ties to foreign companies, of those firms that crisis..." borrow (box 4.3, page 62). The 1997 The same in foreign currency, 88 percent export, statistics for firms that do not borrow and they have Industrial Survey in Thailand reflected that have an average of 818 employees, a debt- equity ratio of 3.12, a relatively high capacity utilization and optimism during the respectively. by bounded from above by that country's companies which belong to the export "'The September 1998 report on the Asian better adjusted to the by mandate) Interestingly, the very few exceptions to the sovereignty principle, which the rating on debt issued by a country's corporate sector government debt and other in foreign crisis, 70 and 37 percent currency (75 percent of firms) are: 46 percent, 139 employees, 2.36, 61 percent and 19 percent. See Dollar and Hallward-Dreimeier (1998). '^There 1991), is an extensive literature documenting "home bias" in asset holdings matched by asymmetric information explanations reasons behind home bias, as Zhou (see, e.g., 1998). French and Poterba There are also institutional most countries hmit foreign ownership of domestic companies. These range from constraints on a few "strategic" sectors (e.g. across-the-board constraints, as in pre-crisis Korea. ^°A (see e.g. "collateral squeeze" versus a "credit crunch." 10 oil and telecommunications), limits as in the U.S., to fact, any shock to international Firms can Financial structure. We foreign investors. either by domestic Since free, we can likely to is be a combination of these two shocks. and date raise finance at date assume that all 1 from either domestic or finance must be default free and fully collateralized - collateral in the case of domestics, or international collateral in the case of foreign investors. contracts. collateral all Without loss of generality, this allows us to focus firms are homogeneous will involve free debt and since debt contracts are default at date borrowing from foreign investors. also restrict attention to date borrowing between domestics on default no net transfers at either date Any date or at a later date and can therefore be ignored.^^ Balance Sheets and Financial Contracts 2.2 Balance Each entrepreneur has a firm sheets. in each sector. While he can issue claims separately on each firm, the entreprenem: consolidates their financial resources in case of We distress. firms at date Date start our characterization of equilibrium 1, and by highlighting their dependence on asset prices and exchange decisions result in firms arriving at date 1 with installed capital of kn foreign debt of doj. At date that escapes the shock is 1, intact a firm that receives a shock The balance (I). kn and foreign debt of doj. However, not value. by studying the balance sheets of On all is and rates. and kt distressed (S), while a firm sheet of an intact firm has assets oikt and of these assets have international collateral one hand, incentive constraints determine that only a fraction At of the shares On of kt are internationally pledgeable. the other, shares of /c„ possibly may be converted domestically into international collateral at a discount. LetL'^ be the market gross interest rate applied to a -collateralized- claim world Date u!. 2 output of ^^^ on one unit of N-goods can be exchanged at date at date 2, in state of the 1 for ^^ units of date T- 1 goods, or equivalently, claims on date 2 T-goods. Then, the international collateral value of the assets of this firm is, XtRtkt The debt of this firm with respect to foreigners of the intact firm as the difference between <it The + = >^rRtkt its + ^. is dgj. Hence assets and we can define the equity value liabilities, in terms of T-goods, as ^-doj. (2) condition of balance sheets play a more central role for distressed firms, since determines their debt capacity (and from This assumption among domestic is relaxed in section 4 now we use these terms interchangeably). when we introduce domestic banks and agents. 11 it A therefore lose homogeneity distressed firm suffers a production shock that requires goods in order to attain units of N-goods at date Rnkn The debt the firm realizes a lower output of Rn{dn)kn- Qs -\ + i^tt-t it 2. to reinvest resources of Investing 6nf^n < fcn capacity of such a firm compromise balance As we show below, reinvestment for the distressed firm at date Debt contracts. Lenders at date do not know and therefore if a firm will be distressed or intact. 1 is non- verifiable. Thus lenders on being distressed or intact. Purely for simplifying asstmiption that debt contracts are also not written contingent on the aggregate shock when appendix, are preserved even collateral inter- 1. are unable to write debt contracts contingent we make a N High this translates directly into constrained Moreover, the realization of the idiosyncratic shock at date expositional purposes, is: [i] exchange rates decrease the value of sheets. T- means that do,/- ^^^^ This expression hints at the sensitivity of balance sheets to asset prices. est rates or depreciated A:„ — in fact, most of our results, as we show in the contracts are fully contingent. Lenders impose the no default constraint that, This ensures that firms always have sufficient resources to repay debts at date no reinvestment takes place. Debt levels above this 1 even if Hmit introduce considerations of debt overhang and bankruptcy costs which, while relevant and interesting, are not central to our main concerns. ^^ The than foreign debt constraint requires that (worst-outcome) marketable assets be greater liabilities. N-assets. This It is is important to note that these marketable assets includes both T- and the sense in which the collateral assumption that imply that foreign credit 2.3 Date its will only we have made does not be extended to T-firms. Microeconomic Decisions and Equilibrium 1 Consider the problem of a distressed firm in raising funds to alleviate problem. production shock. A choice of O^kn will result in output at date 2 oi Rn{6n)kn N-goods. Sincc T-goods e'^ is the real exchange rate at date 2, Rth T-goods and the value of this output in is, This constraint is tighter than it needs to be since the firm in principle could borrow against the pledgeable component of the reinvestment option. since even the tighter constraint is non-binding in Absent renegotiation, equilibrium (see below). 12 this omission is not important In order to retool a fraction T-goods. ff^kji can do It collateral at date 1 that it 6"^ of the distressed unit, the firm must raise finance and reinvest this in two ways. may have some First, the firm can use to borrow directly from foreigners. That international is the firm can always raise directly, </ < The latter quantity can raise at date 1. The rest must come from distressed firm can use intact firms. doj. always positive (see below), and represents the is investors since their capacity to A - >^tRth intact firms, borrow abroad at date its minimum that a firm which also have access to foreign also 1 is X^Rtkt — doj. N-collateral to access the international collateral of the > In equilibrium, the latter discount the N-collateral at a rate of L'^ providing international collateral. Through this "credit chain," the distressed firm to aggregate the international collateral of the raise resources for date 1 reinvestment. Let d-^ j: debt (in imits of y- = s.t. {i) max,^,<i^^^,<i^_^ (z) is the foreign and domestic 1. We can then write the + ^^^^S^-doj-dlf-d-,, = X^Rth - doj + "^j^lil"" > dlf + §f (ts 6^^K = dl^ + %i {ii) reflects resources received by the firm at date is this to foreigners to 1. a balance sheet constraint (net marketable assets greater than while constraint able R^kt K< iiii) liabilities), d'^^ represent in as, (ii) Constraint and T-goods) contracted by a distressed firm at date problem of a distressed firm (PI) economy and pledge is 1 that 1 in new investment must be fully new paid with the taking on debts of d^y and d^^. Constraint {in) purely technological. An intact firm at date distressed firm. 1 has only one decision: Suppose that the firm accepts claims at date T-goods) in return for making a date 1 contribution of ^if =max,^_^ Vr (P2) R^kt problem. At date either distressed or intact, Thus the decision at date (P3) 0, + X^Rtkt s.t. Date how much , 1 oixid (face value of date 2 ^+x^-^ - doj - 5r > 1 can expect to find is, s.t. E.E{G,B} ^'^ipVs^ doj kn extend to the itself as good or the bad aggregate state of the world. in either the maxfc„,fc,,rf„^ it then, a firm looking forward to date and finance will < + min^{X'^ Rth + kt = Wo + doj. 13 (1 + - pWH ^} where vr'^ represents the probabiUty that state Market clearing Equilibrium. uj occurs. domestic debt market at date in the 1 (capital letters denote aggregate quantities) requires that the aggregate amount of domestic debt taken on by distressed firms is fully funded by intact firms: ^M = P<d Therefore, market clearing, Dla = Xr,„ determines the gross discount rate The real exchange rate at date (4) L'^. 2, e'^, is determined by goods-market clearing at date 2. Given the Cobb-Douglas preference structure, the exchange rate only depends on aggregate consumption. The first order condition for a representative consumer yields, e^ = 9^, c: (5) with G;: An = {Rn + p^n{On " economy equilibrium of this Cr = RtKt - and 1))^-^- and date consists of date 1 pe';:Kn. (contingent) decisions, {kn,kt,doj) and {9';^,d'^j,d'^^a,d'^^a)^^^^'^\ respectively, and prices (e",L'^)'^s^'^'^>. Decisions are solutions to the firms' problems (PI), (P2), market clearing conditions, Crises 2.4 With the and Fire and (4) and (P3) given prices, (5), hold. we can now basic setup behind us, 1, characterize crises. as stated in (PI). Consider the problem Halting the productivity decline of a production unit in the N-sector generates a date 2 return of 1 At these Sales of a distressed firm at date date prices. A„ N-goods, but requires a investment of one T-good. Since foreign lenders charge a gross interest rate of one on collateralized loans, the firm foreign lenders as long as ^ > will 1 choose to pledge and 6':^ < l- all " much > 1 from as possible => e^ d^f For financial requirements above this lateral. Since foreigners do not accept = level, its international collateral to Our assumption that A„ allows us to focus on the case where saving distressed units so that firms borrow as of is >> always profitable and — p) ^> 1, foreigners: min[kn, XI^Rth - d^j]. the firm has to mortgage (6) its this collateral directly, distressed firms 14 Rn{l domestic col- must borrow from intact firms in order to access foreign resources. In equilibrium, domestic collateral discounted at the rate of 1/L^ so that the firm sets as long as An/e'^ choice > L^ and < 0^!^ The 1. firm domestic debt as high as possible its the marginal product of investing for the firm if and debt indifferent in its investment scale is is is equal to the domestic discount rate: < <_d However > (4? if the marginal product of investing L'^), this - L"^ max[fc„ The expression holds with equality. 0]. above the domestic discount rate strictly is cally held international collateral, d^r, needs to this, d^f, distressed firms' demand domesti- for be paired to intact firms' supply of consider the problem of an intact firm at date 1, The as portrayed in (P2). it. For firm lends to the distressed firm at a gross interest rate of L'^. It accepts debt claims at date 2 with x^^ and pays face value of solution to this problem for this equilibriimi value L'^, l ^ determined by market clearing condition, is international collateral as long as L'^ is fiat L = at Demand, on the other hand, when reinvestment mild —and is 1 is > and -^. The 1. = (4). Distressed while intact firms supply all their Figure 2 portrays a simple supply and demand vertical pdi^d-, when all international collateral is used up. mildly declining and equal to the marginal product incomplete, and less interesting rate induced gets from foreigners) of xl^<XtRtkt-doj. firms issue domestic debt claims totaling Di^d diagram. Supply it straightforward: is L- = The with T-goods (which — decline falls is sharply as full reinvestment due to the depreciation by increased reinvestment. The sharp decline, fact that protecting the capital of financially distressed is financed. in the long-run on the other hand, companies is A„/e The exchange reflects the a very high return project.^^ The figure illustrates In panel so there is Since did two scenarios. (a), excess supply of international collateral in the domestic market = max[A;„ — {X^Rtkt — doj), 0], we can Having a menu of reinvestment opportunities is would certainly smooth the sharpness of the that there are distinct times when - => L'^ for instance decline. But the = 1. 1. (7) by having more heterogeneity among firms essential point captured slack rapidly runs out for a large 15 = rewrite this condition as, pKn < X^RtKt - Do J - and hence L'^ number by our assumption of companies. Notice that the right hand side of this inequality economy once date collateral in the is the aggregate amount of international We foreign debts have been paid. refer to this constraint as the international collateral constraint. L= L= L= e L= 1 1 I I ' 1 e„=i Panel (a) Figure Panel (b) illustrates a 2: : firms, the latter : Supply Shortage Equilibrium and the International Collateral Constraint crisis of international collateral, ceases to hold. In this case, there interest rate rises Panel (b) Excess Supply is excess demand for which occurs when condition funds at date 1 (7) and the domestic above one. Moreover, as intact firms lend as much as they can to distressed compete away pKr, all borrower's surplus: > X^RtKt - Do J Let us define the international spread, s'i, 1 (8) as s'^f=L'^Figure 3 illustrates that as the date ^L^ = ^. 1. realization of A( 16 falls, eventually the relevant constraint shifts from (7) to (8) For small aggregate shocks and declines in Af, while and the international spread jumps from zero to richer specification.^^ demand jump in spreads curve, the sharp nonUnearity in all the countries involved in the Tequila, Figure dramatically 3: Interest on financing. This when there it illustrates is would survive a prices has occured crises.^^ 1 * Rates and International Collateral good realizations of At the supply of international attractive terms from certainly due to the stylized Asian and post-Russian 1 "''For is Without exception, the non-linear response of asset L= all 1.-' critical level, a contraction in international collateral causes spreads to blow up. While the abrupt nature of our collateral the economy does not suffer at initial leverage, beyond a certain ^— collateral fully protects their returns is plentiful and borrowers get very from re-investment. The situation changes a shortage of international collateral, as credit is difficult to obtain and all re-investment returns are transferred to domestic liquidity holders. ^°For instance, suppose there was heterogeneity in reinvestment opportunities such that the demand curve was smoother. is plentiful, shifts in At willl result in collateral 1 is In the region where At no changes in short supply, shifts will result in are used to either finance domestic are sufficient funds, are in short supply it is all demand in spreads. much is (i.e. heterogeneous A„) high and international collateral However, when At is low and international bigger changes in spreads. Economically, funds at date or to purchase tradeable goods (or assets) abroad. the foreign interest rates which set domestic interest rates. However, surplus shifts to suppliers and interest rates on non-heterogeneous demand, heightens this contrast. ^^Without a surge in nominal interest rates, the fire sale 17 is become very reflected as a sharp responsive. When there when funds Our assumption exchange rate depreciation. As the international spread rises, domestic asset prices collapse. There is domestic assets because the domestic opportunity cost of holding these assets credit high is when Foreigners are unable to capture these high returns because at times of scarce. is afire sale of they only hold and arbitrage claims backed by international collateral. While their crises arbitrage during normal times keeps the international spread at zero, That the international collateral constraint binds. it is immaterial when the interest-parity-condition shifts is, than international arbitrage, holds. until domestic equilibrium, rather Fire sales bring about a sharp decline in the balance sheets of distressed firms. In terms of the program (PI), investment at date 1 is no longer determined by technological constraints but by the balance sheet constraint of (i). This shortage of international resources firms — the rise in of their production shock, thus all L^ is and competitive Proposition Proposition occurs In 1 when = summarizes the main = 1. Conversely, in the - {XtRtKt - Doj) results be in the international collateral constraint, this case international region fire sales is made (9) . upto this point: The economy can (Regions). 1 6'^ — are collateral aggregation, so that 0;^K„ 1. distressed simply the mechanism by which limited international collateral consistent with full bind. When, pKn < X^RtKt — Dqj, resolved. is the mechanism by which the aggregate either directly, or through perfect aggregation of the intact firms' collateral able to fund 1 shift is one of two regions at date pKn < X^RtKt — Dqj, spreads are zero and there is full 1. Region does not project completion, Q'1^ = Region 2 occurs when the international collateral constraint does bind, in which case international spreads jump to ^— 1 > and distressed projects are downsized: 9t^Kn = \{XtRtKt-DoJ)<Kr^. 3 Underdeveloped Domestic In this section setup. We do FinanciEil we introduce underdeveloped domestic this to on international study the date 1 effect collateral provisioning. the supply decisions at date Markets financial markets into the previous of poor domestic collateral on date That is, we isolate the effect of date 1 decisions demand on 0. Distressed firms use their domestic N-coUateral to aggregate the international collateral in the economy. (domestically), acteristic of economy. By assuming that we have rendered of the output in the N-sector can be collateralized this transaction frictionless. an emerging economy Weak all is that it However an important char- does not allocate capital efficiently within the corporate governance or a weak domestic banking sector are as likely to 18 hinder domestic collateral aggregation as they are to limit the economy's access to foreign capital. In this section, ket not only amplify we show that and make more collateral constraints into the domestic capital mar- costly the scenarios described above, but also create pecuniary externalities which make crises more severe and likely than is socially optimal. Basic Setup and Domestic Spreads 3.1 The environment described in the previous sections needs to accommodate weak domestic at date 2 links. A firm in the N-sector with output oi can pledge a fraction XnRnkn, with A„ order to receive financing. We be modified only slightly to retain the same < 1, Rnkn N-goods as collateral to another domestic in bias with regards to foreigners - they only accept claims on the T-sector up to a fraction At of output. Return to the balance sheet of a distressed Balance Sheets. firm, modified domestic debt constraint. Since the firm can only pledge a fraction A^ of balance sheet, or debt capacity, q, Its date 1 - K" =max,^ s.t. {{') (u) (iii) The maximum amount (i') and (ii) is + At HtLt N-output, its to: (lUj do,/- jj^ + ^%">*^ - do J - d^^ - d^^ (^ = XfRtkt - doj + ^"^jfp''" > d^f + e^K = d^j + f^ Rtkt 0t < ^ 1 of reinvestment that the distressed firm can finance with equality, and solving ^n^^ which reduced the program must be modified accordingly: (PI') combining is its now by ^ A„A„. for is given by 6'^: yn^th - doj + -J^f^nj equal to the leverage ratio times the firm's debt capacity. Thus, as long as, Ari?tfct-rfoj>(l-^)A;„, holds, the distressed firm is able to finance all of its (11) We refer to this constraint as the needs. domestic collateral constraint. Unhke the international collateral constraint, which is purely aggregate and hence affects units through equilibrium prices exclusively, the domestic one affects microeconomic imits directly through quantity constraints. that are behind the externalities and excess volatility that Regions. Each lowered. Region we It is discuss below. of the regions in the previous section gives rise to 1 in the latter constraints two new regions as A„ the previous section, where the international constraint 19 is is slack, gives and rise to regions I II. In the former A„ while in the latter the opposite is is is high so the domestic constraint true. Similarly, region 2, is slack as well, where the international constraint and IV, sorted again by whether the domestic constraint binding, yields regions III is slack or binding, respectively. Figure 4 provides examples of equilibria within each of these regions. in demand and supply panel (a) represent the unconstrained (solid lines) is the unconstrained the case where A„ = 1. demand But when A„ < can be pledged, which gives rise to purchase where A„ all L'^ = i 1 or, it L= A - it it when needs, even corresponds to the demands depicted is The needs. the latter > i ~^\ is in available at is \ e L= 1 it cannot afford minimum all price. In V -A 1 dashed hnes. The zero. I \ in short dashes portray a situation L "'^ demand curve high enough that a distressed firm can equivalently, the international spread e The upper envelope imable to fully reinvest since is solid lines only a fraction of domestic output that is where A„ i L= there the international collateral the international collateral both cases I, small enough that the firm is 1, funds the constrained long-dashes represent a case in region still for of Dia- The \ \ \ \ \ '\ 1 '•.. \ 1 1 Panel (a) : Regions I and Panel (b) II Figure 4: Equilibrium when : Regions \< III and IV 1 Panel (b) has a binding international constraint. The long dashes represent a situation in region III, where A„ is large enough so all rents can 20 still be transferred to collateral supphers as in the case with perfect domestic markets. The short dashes, on the other hand, represent small enough to limit the ability to fully transfer the rents to a case in region IV, where A„ is collateral-suppliers, but high enough to fully aggregate international collateral (unlike region still In both cases L'^ II). IV both domestic and is hence there 1, now also a domestic spread, s^, vestors.^^ When as in regions II We the domestic market go; smnmarize the regions at date sufficiently is underdeveloped, this spread When A„ < 1)). spreads are zero and there 1 is full In this case, economy remain fails to aggregate all > downsized, s^ ^''Part of this economy can positive, one of four he in 0,5^ = 0,6^;^ < of zero, and the domestic and domestic both international project completion, binding. International spreads as the the collateral constraint (7) Region II occurs when the international constraint is is 1. collateral constraint (11) are slack. • able to promise other domestic in- is results of this sub-section as follows: Region I occurs when both the international • in a distressed and IV.^^ Proposition 2 (Regions (A„ < We which we define as and the return that the distressed firm 1, a positive international spread is between the internal return on investment reflects the difference firm at date a positive international spread. In region Besides the issue of whether there d This spread is international collateral constraints bind. Domestic Spreads. or not, there > is = s'^ s'i = 0,6':^ = 1. slack but the domestic constraint however domestic spreads are positive its collateral some resulting in projects being 1. wedge may be arbitraged by a resource-consuming bank, as in Holmstrom and Tirole (1997). revisit this issue in section 4. ^*In fact, emerging economies are likely to spend most of the time between these two regions. Region corresponding to normal times, while region IV represents deep crises. In region ^— 1. as the In region IV, this spread starts to economy moves collateral rises. during crises, At, rather While at first glance this behavior of domestic spreads than just the rise as well, it is is and IV rather than true in region IV. We II economy's will A as may seem II is any better, but because at on all domestic odds with the data crisis shock that reduces international collateral both A„ and may be one capture this feature below by letting the economy fluctuate and IV. In region More endogenously, collateral hinders the is important to point out that during severe latter, are likely to fall. that also reduces domestic collateral. I not because the domestic market the domestic spread into a crisis the international spread rises as the required interest rate where they often between regions fall, II, we develop I, no collateral constraint binds, while the opposite later in the paper, ability to aggregate collateral feeds through as a shock to domestic collateral as well. 21 an external shock to international by compromising the banking sector. This Region III occurs when the international constraint • straint is slack. International spreads The economy aggregates at zero. aggregate resources and IV • Region occurs when some jump of all ^— while domestic spreads remain 1, collateral, its projects are downsized, however there = s'^ both the international constraint are binding. Both, international spreads projects are downsized, to binding but the domestic con- is s'^ > 0,s^ > 0,s^ and < < 0,9'^ 1. The What are the costs of having an undeveloped domestic financial market? Fixing Externality: Internationcd Collatercd Undervaluation main economy cost to the resulting in lower investment is that in region and output. But II collateral is /c„,/cf and doj do not remain fixed as A„ over-exposed to the crises brought about by external shocks. externality in this subsection The Date To sharpen the and its At date Problem. focus, we assume 1 that the good state leaves the bad state bad becomes binding as an entrepreneur /firm — e(a;)^/^ = 1 State: u) in In either good or from production less debt maxltt*^ s.t. is ct+Cn/e{u})=W2.} linear in wealth, production and financing is exactly reflect. =G this case collateral is, 2. solves, In the good state, neither collateral constraint binds. That in region exclusively to are taken to maximize the expected value of wealth - which what (PI) and (P2) Good is where well.^^ Since the value function over date 2 wealth decisions at date economy I enough so that the domestic at date 2 with wealth (output repayments) of W2 in units of T-goods in region Although the negative shock problem we work backwards from date arrive at the date state, economy leaves the international collateral, equilibrium asset prices decline sharply To describe the the economy can be in either a good or a bad state. collateral constraints bind. collateral constraint We result, the vulnerability consequences in the next one. neither of the collateral constraints bind, while the IV where both and not aggregated properly, with underdeveloped financial markets undervalues international collateral As a is fc„, kt problem, we show that the private sector in an economy changes. Focusing on the date economy and some are positive 3.2 do J, the 1. the domestic constraint and domestic spreads 0,6'^ > insufficient is is fully aggregated and that all Proposition 1 dictates that in N-productions units are saved. The only our scenario a decline in the availabihty of international financing leads to a decline in domestic "intermediation" as well. 22 difference between an intact and a distressed the fraction of distressed firms on state G firm, is that the latter must reinvest kn- Since the expected (as of date 0) date 2 wealth conditional is p, is: W2{G) Market clearing = Rtkt + ^-doj-pkn. at date 2 (condition (5)), yields a real ^ ' (12) exchange rate of RtKt - Doj - pKn' Cp while both spreads -domestic and foreign- are equal to zero at date Bad State: 1. =B uj In the bad state, both collateral constraints bind. Investment constrained and both is the foreign as well as the domestic spread are positive. The binding domestic collateral constraint implies that in equilibrium: e^kn = ^ [xfRth - Doj + ^f^) < ^"' (13) with _ 1 ^ Ar,A„. ' LBe{B) while the (aggregate) international collateral constraint implies that: P Combining the aggregate coimterpart of (13) and (14) we obtain the date 1 prices: A„ T;tiZb \ l-p\e{B) .e(S)^e(B)/' A„ f where ry The ratio Z _ PKn XfRtKt - Doj indicates the economy's exposure to crises. availability of collateral per unit of distressed capital. As Z rises, there is less aggregate Other things equal, this means that the discount of N-claims, L^, rises as well. The real exchange rate at date 2 ^ ' is determined as before: = CI ^ (^-pAn(l-OK._ Cf RtKt-p9^Kn-Doj 23 ^ ' . Given these equilibrium prices and B date 2 wealth conditional on state W2{B) = availability of reinvestment resources, the is: + -^{R^ - pAr^) - Rtkt e{B) "^ ^^ i^W) + The first line by date of the RHS /^ „ _L ^" ^[W)~ N and T " (16) do,f ^^^"^ ~ ^^^^'^' tb\ ^nRnjOn) ) e{B)LB in (16) reflects the investment in the ' "^''^^ ,^ ^"- expected wealth, in the bad state, generated sectors, net of foreign debt contracted at date 0. In particular, this line corresponds to the expected wealth assuming that production units are allowed to in the next With lines. probability collateral On two The second [1 — p) and lends fail. the firm The intact in which case if the firm distressed, then is units, directly earning the internal return of its it distressed is captured having international collateral. borrows against the international to the distressed firm, thereby earning the domestic interest rate the other hand, last line reflects all value of saving these production units line reflects the return to is expected ^y borrows to save it against the return to having domestic collateral. domestic collateral at the domestic interest rate of its If its oiL^ own production international collateral. a firm L^ and is distressed, it The mortgages uses these proceeds to save production units, earning ^rgyFollowing the logic of the previous paragraph, we can rewrite (16) directly in terms of spreads as, W2{B) It is = Rtkt + -^{R^ - pAn) e[B) + (^sf + ^''^ doj (17) + ps^){X^Rtkt-doj) e{B)LB instructive at this stage to construct two ^-- one different versions of tW2(-B); dividual firms and the other for a central planner. To construct the former, for in- we simply substitute in (17) the expression for reinvestment implied by the microeconomic domestic collateral constraint, (13). To construct the central planner's version, which W2{B), we substitute instead the equilibrium collateral constraint, (14). we denote by While separately these expressions are obscure, their difference clearly highlights the basic externality: W2{B) where tp is pip <l - w*2{B) = sf (^p^j^^^kn - follows from equations (13) and (14). (1 - p^){\fRtkt - doj)^ (18) 30 the leverage ratio applied to collateral for an individual distressed firm. That 24 , is, it is the multipher At a given equilibrium in region IV, W2{B) and 1^2 (-B) must be equal. But it is apparent that individuals and the central planner value a marginal unit of international collateral and domestic collateral quite We differently. note three things in particular: domestic collateral more than the central planner. • Individual firms value • Individual firms value international collateral less The divergence • At the aggregate in these valuations are proportional to the B domestic spread, s^ the central planner only cares about creating and protecting level, international collateral. This is because international collateral However attract foreign investment. and international than the central planner. microeconomic at the level, can be used to secure financing. collateral that can be used to is all both domestic This is collateral the basic tension between the individual and the central planner's problem. A positive domestic spread heightens this tension by placing a wedge between the social and private valuation of collateral. A unit of international collateral generates asocial return of ^y at date 1, while a unit of domestic collateral generates a social return of one at date At the individual one. On the other hand, difference an intact level, for between social domestic collateral generates a return of firm, a unit of a unit of international collateral only generates a retinrn ofL^. The and private valuation of s^ = ^y — L^ The domestic spread prevents of international collateral. firms) 1. from transferring the results in undervaluation collateral demanders (distressed marginal value of collateral to the collateral providers (intact full firms).^^ Underdeveloped financial markets (low values of Ajj) the systematic mis-valuation of collateral. have described is not of the traditional Bardhan-Harberger type, where individual borrowers the aggregate collateral constraint. by distressed is p, and important to realize that the externality we It is on collateral that determines how much a firm can borrow. distressed firms result in high domestic spreads then frtp X If A simple way to see that each distressed firm uses a multiple oi (Value of Domestic -I- International Collateral) firms. In aggregate, however, the country tp is can only borrow against /^j/) and < the total its 1 is to think about in total the fraction of amount borrowed international collateral. Thus, we must have that, P'^X (Value of Domestic or, fnp + International Collateral) < Value of International Collateral <1. Overvaluation of domestic collateral collateral. At the individual firm a return of ^t^- On the necessary counterpart of undervaluation of international a distressed firm, a unit of international collateral the other hand, a unit of domestic collateral and generates output of an excess return of level, for is zero. ^rgt, providing a return of sf At the individual firm of domestic collateral, resulting in too > 0. level, this much domestic At the is still generates discounted at the interest rate of mis-pricing causes firms to overvalue the A„r„A:„ collateral relative to international collateral. 25 L* social level, domestic collateral generates do not internalize the fact that the country faces an upward slopping supply of foreign loan.'^^ Indeed, in our setup the economy does face a very steep international funds supply, for the country rationed after is some point, but the externality arises only markets are underdeveloped as The constrained nature well. collateral limits the price that collateral users arises, which reflects what they can can afford. when domestic of the domestic demand for In so doing, a domestic spread the gap between what an extra unit of collateral Collateral suppliers anticipate such spread, afford. financial is worth for and therefore them and see their provision incentives depressed. now establish this result on the undervaluation of international collateral and its dependence on the underdevelopment of domestic financial markets more formally. We do Let us this in two steps. First, making date constraints. we write decisions of {Kn, down Kt,Dof) subject We arrive at the constrained in particular, that these solutions planner's is the problem of a central planner Pareto to the domestic who has freedom in and international collateral the problem and show, efficient decisions of do not depend on A„. Second, we contrast the central and the decentralized solution and show that the market equilibrium with A„ < 1 constrained Pareto inefficient. Consider the problem of a central planner Central planner's problem. equally weighted sum of utilities of agents in the who maximizes economy by choosing Kn,Kt and Doj directly at date 0, subject to the respective collateral constraints. In the that the solution to this problem is max;^„,K.,Do,/ appendix we show equivalent to the solution of the following program - which resembles that of the individual (P4) the firm, with the exception of the expression foru>2(.B): Ea;e{G,B}'r'^e(w)2ti;*(a;) wl{B) s.t. w*2iG) = = RtKt + RtKt + - Doj + - Doj - pK^ ^^~4§)"^^" ^ (sf + sf ) {X^RtKt - Doj) Doj < mm^{Xr RtKt + A„^} Kn + Kt = Wo + Doj Propositions (Constrained Optimality) (a) Let {K*,K*,DQf) These solutions are the constrained Pareto optimal date {K*,K*,DQf) (P4). , Once we show that It is choices of the economy. (6) are independent of the strength of domestic financial links, A„. Part (a) see the appendix. Part Proof: be solutions to (P4)- it (6) can be done by simple inspection of program does not depend on the value of A„, its solution can't either. apparent that none of the terms in (P4) depends on A„ but for the constraint on foreign leverage. ^^ Which is However, we will show by contradiction that the inequality not to say that such mechanism ^'indeed, note that {sf + sf ) = is not important in reahty. A„/e(S). 26 is always strict: K Doj < rmn{Xr RtK, + Suppose ^^^^ " }• that, D,J = min{XtRJU + Xr^^^^^^}, then, the amount of international collateral at date 1 in X^RtKt - Doj < However if this is the bad state is, 0. the case, then the market for exchanging domestic collateral for interna- tional collateral will necessarily break down. must return zero imits " Swapping ' z units of domestic collateral of international collateral since the supply of international collateral not positive. L*^ rises until such a swap does not take place, or until Undervalued international collateral Doj < mii\^{XfRtKt}. Let us contrast the central planners problem with the decentralized solution. Taking the case where the good state leaves us in region bad state in region IV, (P5) we can write the program maxfc„,fc,,do,/ s.t. is of a firm at date I and the as: Ec^G{G,B}7r'^e(w)5u;2M ^ - doj - pkn = Rth + = Rtkt + ^^^^^§f^-doj + (s] + pi,s^){\fRth-doj) W2iG) W2{B) doj kn < + Xnj^} m[ikj{X'^ Rth + kt = Wo + doj. Proposition 4 (Undervaluation of International Collateral) solutions to the program (P5) in the case when A„ < (kn,kt,doj), are constrained Pareto inefficient. provement by perturbing these decisions A 1 and s^ > 0. Let {kn,kt,doj) be Then the decisions, central planner can effect a Pareto im- to {k'^,k[,d'Q f), where at least two of the following inequalities are strict, doj < 0^0,/ kn S kn K ^ ^t Proof: see Appendix The proof follows closely from the logic laid out in the previous section. the only difference between (P4) and (P5) is 27 in the expressions for Note that W2{B) and W2{B). When A„ planner's Pareto = 1 the domestic spread, s^ and the decentralized = 0, and there is no difference between the central market equilibrium solution. In this case, the However, when domestic markets are undeveloped, s^ efficient. > is constrained and there is a divergence in the social and private value of collateral, resulting in the inefficiency. Excess Volatility and Policy 3.3 Proposition 4 has a straightforward but useful corollary: Corollary 5 and solutions {kn,kt,dQj) What is I and Z' denote {k'^,k[,d'Q f) the indices of exposure corresponding to and IV, as volatile we have Figure 5 illustrates this. — Z> Then, in proposition 4- the cost of a higher index of exposure? economy becomes more regions Z Respectively, let Z' . easy to see that as It is economy remains as long as the . Z rises the fluctuating between assumed.^'' The solid lines represent the supply of international collateral in the bad state and in the good state. In the good state the supply of international collateral is at X9RtKt — Do = Given our definition of Zg ''f- ^Y^- Similary in the bad state, since there constrained at a lower point. decisions, Z' ?^^ The dashed — C "— UQf Ajg HiKt , is less economy falls to C varies less across realization of G the supply curve turns vertical international collateral, the supply lines represent is supply at the Pareto improving at A during good states of the rather than B. There are two observations we can make from the L i While both economies share the equilibrium world, the less efficient since r. and B First, reading off the y-axis, figure. states at Z' than at Z, we can see that asset prices are less volatile. Second, reading of the x-axis, since equilibrium reinvestment varies less at Z' there is little thus, the than at Z, investment and output are also surplus of international collateral economy is left less volatile. by high, nearly defenseless in the event of a bad external shock. investment and output severely is even under the best of the scenarios; While we have stressed the impact of underdeveloped domestic more When Z volatility, it also follows crisis when A^ is financial markets on that since reinvestment options are curtailed lower, the country's stock market value suffers more as well. If the domestic constraints becomes so tight that, collateral regions, volatility e.g., the with respect to external shocks economy may finds itself at all times in the wasted disappear altogether, and be replaced by persistently low reinvestment rates. See our discussion of this issue in the conclusion. '*In this figure, we vary Z by changing Kt and Doj, while keeping Kn only illustrative, as such change in Z also changes the date 2 28 exchange constant. Moreover, the figure rate, and hence shifts demand. is L=A PVz", P*^ Figure Policy. The constrained Pareto to policy considerations. We 5: Excess and Costly Volatility inefficiency of more broadly discuss policy (Caballero and Krishnamurthy (1999)). Here naturally from the previous proposition: and domestic market equilibrium we limit When an open invitation companion working paper ourselves to pointing out collateral are warranted; they could take the form of capital inflows taxation during non-crisis periods (date collateral (or, more consuming -particularly requirements from tmdoing may it follows market equilibrium leaves the economy Measures to boost international investment in the T-sector what both international financial links are weak financial markets are imderdeveloped, excessively fragile. in a is 0); subsidies to generally, international collateral creators); or taxes dinring crises- activities. also work, although supervision through additional borrowing. 29 is Mandatory on collateral-provision needed to prevent the private sector Endogenous Breakdown 4 plification and Multiple Am- in Collateral Aggregation: Equilibria. Distressed firms access the international collateral of intact firms by borrowing from these firms against their N-collateral. this collateral aggregation is The done implicit assumption in the preceding analysis that is in a perfectly functioning public debt market. Distressed firms issue debt claims in the domestic market secured by N-collateral. Intact firms issue They then use debt claims to foreign investors secured by international collateral. the proceeds from this issue to purchase the debt claims of the distressed firms. In emerging markets, liquid domestic debt markets are the exception rather than the rule. Invariably, the financial system is bank based - for good reasons: lack of investor protection and the necessity of monitoring to alleviate asymmetric information places banks at the center of the financial system. Banks issue claims to foreign investors against balance sheets that are partially backed by international collateral - both of the distressed and the intact firms. These funds are then channeled to the distressed firms. Implicitly, the intact firm extends credit to the distressed firms against their domestic collateral. But unlike the immutable debt markets of the previous experience sharp economy's falls banks themselves run into trouble. ability to aggregate its scarce collateral in asset prices and Even worse, real activity. this perverse feedback process if fits conventional can lead to multiple wisdom well. every emerging market's external among many, tequila crisis, 4.1 it is crisis. And equilibria. its in its it of the early 80s, to The Banking System: Amplification and all reallocation plays a role in virtually extreme form of a worst equilibrium In order to highlight the role of banks, in this section in feedback into real activity and asset reminiscent of Indonesia in the recent Asian crisis crisis, name Mexico during the a few.^^ Financial Bottlenecks we suppress entirely domestic debt must be done via the banking system. need of funds take loans from banks. To attract these funds, banks market determined interest enter distress, the weakened, triggering further declines is Arguably, some form of and Chile during the debt markets so that As banks rate. Intact firms international collateral to foreign investors asset prices banks are undercapitalized at the outset, This endogenous disintermediation process and prices when sections, become the depositors. Distressed firms offer deposits at They mortgage and then deposit these funds in the a their banking See Gclos and Werner (1999) for a study of lending practices by Mexican banks during the postliberalization period. Among other things, they document the significant role played by banks in lending to manufacturing firms without access to foreign funds, this mix contributed significantly to the collapse of the 30 and their reliance on N-collateral. banking system during the 94-95 They argue crisis. that system. By collateral, intermediating these transactions, the banking system serves to aggregate which is the only role We Balance Sheets. it has in our model. unit measure). Suppose that at date and fund these loans by of the bank are a„ T-firms. ^^-^^ bank consider a representative 0, banks make loans to firms issuing debt to foreigners. this with d^ both and is N at face value of units of foreign debt (in units of f the value of capital (entering date 1) in the bank and T Foreigners extend credit to banks against the ''" date 2 debt on T the excess of assets over goods). Then liabilities, (19) same no-default constraint as before, +at-4j>0. and sheets of distressed (of sectors = T:^ + «t-</. 9r The balance in banking sector In particular, suppose that the assets face value of date 2 debt of N-firms Banks fund in a competitive intact firms (equations (2) and (3))are modified slightly: Qi = At Rtk -at+ for the intact firm, while for the distressed firm, Qs Both equations That is, do,/. jjj^ lending from the banking sector. Notice that we have taken A„ to be one in both balance sheet any further in the absence of to the public debt market with A„ now -at^ i<-tf-t it is, are modified to reflect the date Capital Constraints. calculations. -\ doj ^^^ = 1 and hence frictions, the banking system collateral aggregation is is equivalent We frictionless. introduce capital constraints that limit the capacity of the banking system to aggregate collateral. At date 1, a bank takes in deposits from intact firms of d\ ^ and makes loans of x\ ^ to the distressed firms. Let respectively. If Lx Lq and no other constraint is represent the loans' and deposits' interest rates, present, both of these are made at the market de- termined interest rate of L, and the expected date 2 repayment to the bank from this At date 0, Thus issuance not all agents in the domestic economy are the same of domestic debt cannot be netted to zero as - there are both banks as well as firms. we have done in previous sections. Additionally, throughout the paper, our firms are conglomerates, each of them with an N-firm and a T-firm in it. We assume here that each of these mini-firms can contract debt independently. '*The purpose of this section is to highlight the potential date 1 bottleneck brought about by the deterioration in banks' balance sheets. There are several interesting date issues discussion but which would lengthen this section substantially. For example, one sign contingent contracts with borrowers and depositors. 31 which merit an extensive may ask why banks don't operation is {x\ d — d\ ^)L (in units of face value of T-goods at date We 2). assume that capital constraints restrict the size of this operation. In particular, > qt < a < where 1 is value of x\ Juj).^^ there axl^iuj), (20) the capital that banks must hold against making loans with present If this Lx constraint binds, equal to is still L L^ but to one since falls an excess supply of deposits but a scarcity of loanable funds. is Taking as given the market interest rate (on loans) of D^ and date 2 exchange rate of e'^, the problem of a bank at date (P6) s.t. - 4,/ - ^t/(^) + (^" - at (i) q^>axl^iuj) objective of the bank L^ and taking deposits capital constraint fS reflects profits from making loans of x\ ^ requires raising deposits against international collateral of at — d\i]^ it to the appendix. while in equilibrium V^ = 1 Banks is for behave exactly as in the previous sections. In the problem As scenario. is ^ 1 the interest rate (i) is simply the resource constraint: plus attracting funds investors. very similar to that of the previous section, funds much is less fact, as two regions, (i) as they can whenever L'^ Panel and > than loanable funds. Firms long as constraint entirely analogous to that in section 2.4. in that section, there are the date from intact firms of d\ will lend as when demand reflect (iii) from foreign Since the formal analysis of equilibrium relegate and (ii) J^lj) at L^. Constraint of d\ ^{uj) at the interest rate of on banks. Constraints making loans of x\ we ^)AM ' (L^d " l)^ld(^) 4»<d5_rf(a;)+4/(a;) d\j < at - d^j{u;) (iii) of + max,.^^(^),,.^^(^) iii) The 1 is, (ii), (i) is 1, will not binding, (a) in figure 6 depicts this which are distinguished by whether or not the international collateral constraint binds. In practice, capital adequacy requirements are based on both assets as well as BIS standards assign capital requirements to sorted (i.e. common stock, preferred stock) different assets held liabilities. For example, liabilities of the bank are and weighted to determine the amount of capital held by the bank and hence ensure that the capital requirement principles by banks. The is met. Capital requirements can be justified from on the basis of moral hazard within the banking example. 32 sector. first See Holmstrom and Tirole (1997), for t A i \ L= \ L= \ \ \ L= J L \ A L= 1 \ I I ' ' 1 „=1 Panel (a) : Regions (i) e„=i and Figure The new interesting the economy can demand from fall cases arise into regions low and output is high. then interest rates (iii) is that or (iv) depending on now if Regions capital requirement how line in figure 6 (b) equilibrium: loan Conversely, : (iii) and (iv) Equilibrium with Banks when the (iii) rise to ration ingredient to highlight 6: The dashed distressed firms. loans which yields a region Panel (b) (ii) demand loan supply is is (z) does bind. In this case tightly it binds relative to corresponds to a supply of fully satisfied, interest rates are constrained by capital requirements, the scarce supply (dotted line, region (iv)). The key the supply curve for loans bends backwards at the point where the capital requirement binds. This occurs because the value of banks' capital when interest rates rise (see equation (19)). The backward bending nature differentiates constraint there is it from region IV and a microeconomic wasted collateral, in of supply plays a significant role in region (iv), in section 3. aggregation. While in both cases an aggregate and collateral capital requirement bind, in the banking-intermediated case the sense that not to the distressed firms. Equilibrium full falls is all excess domestic resources are channeled A rather than B, where the latter reflects at point Constrained banks become a financial bottleneck, which bring about simultaneously the worst of each scenario described in the previous sections: sharp declines in asset prices coupled with collateral waste. 33 Multiple Equilibria 4.2 The domestic interest rates and rise in a cause of the crisis, it is also fire sale in We crisis. region (iv) is not only a The fall in domestic asset prices amplifies the impact of the crisis by deepening the credit crunch caused But distressed balance sheets."*" and it feasible intermediation brings and (B) in the figure represent is of the As we described above, illustrate this in figure 7. the contraction in loan supply causes the rise in interest rates. symptom by banks' easy to see that the feedback between asset prices about the possibility of multiple equilibria.'^ Points (A) two equilibria which are distinguished by low interest rates, non-binding capital requirements (A) and high interest rates, credit crunch (C). Point (B) also a possibility in is in The equilibrium in (C) < 1 Low and is Pareto inferior to that of (A) and is interest rate equilibria are mechanism ''"This amplification production units are saved, this all or a concave but not rectangular ranked. prices interest rates rise to constrain loan supply to exactly both (A) and (B) demand. While A„ which is demand more causing the constraint efficient than high rate ones. similar in spirit to Kiyotaki fall in is on the demand side is and Moore (1997) - constrained demand reduces and further constraining demand. However, asset prices, on the supply side - in that a fall in asset In this sense, our In Kiyotaki model rise, is model the key markets during closer to the amplification literature (see also disregarded and multiple Lamont macroeconomic equilibria that (1995)). is effects of the strength of firms' balance sheets, it is emphasize the balance sheet problems is in More in firms another source introduce a non-convexity in the investment possible to have multiple equilibria. We have simply chosen to banks where capital constraints are most natural. done after the Tequila crisis, probably exacerbates the problems we have described making these requirements by valuing assets at outdated bank runs can be is - Given the feedback between investment and specifically to banks, however, increasing capital liquidity ratios, as Argentina has latter debt-overhang problems the appropriate policy response in this case? To start, the pohcy suggestions in the previous section apply here as well. possibility of concern in international both empirically plausible and has been pointed out in the Debt-overhang problems of firms as a function of the strength of their balance sheets. "^What A crises. new public debt crises. model has of amplification financial causing bank capital to mechanism studied in Krishnamurthy (1998). ^'A related mechanism is "the disappearance of markets (and market-makers)" during often expressed by emerging markets' policy makers in their attempt to place and Moore the productivity of assets, in our constrained supply causes interest rates to be compromised, and further constraining supply. The with reinforced by a tightening financial constraint. However, one key diiTerence between our approach the key financial constraint and generally, function, equilibria can be strictly Pareto theirs lies in the identity of the agent with the crucial financial constraint. ''^Our not so in (C). is More (B)."*^'^^ meet offset slightly more prices, as in Japan). with improved credit more comphcated but lines, in this section may by adequacy requirements help avoiding bank runs but raising a. At the very procyclical should be pondered (which The perverse impact this may often done have on the possibility of interbank markets, and lender of last resort certainly not impossible in "currency board" scenarios. 34 is it least the mechanisms. L j L= L= A 1 1 1 Figure 5 FinEil 1 7: Multiple Equilibria with Banks Remarks Since models are stylized representations of complex realities, terpret them narrowly when extracting their lessons it is always unwise to in- and relevance. This applies both to assmnptions and implications. In the context of our model, there are several implications and assumptions that deserve further discussion in order to facilitate their broader inter- pretation and empirical assessment. Our first central implication is the presence of fire sales. substantial expected excess returns in the midst of a the form of a sharp rise in crisis. That is the emergence of In the model, fire sales take domestic interest rates, by which we understand the discount applied to assets that are not good international collateral. There are two related questions that naturally follow this abstract description of fire sales: Do they include the frequently observed collapse in the exchange rate? More generally, which observed rates and prices are the empirical counterparts of the model's rates Our model is and prices? designed to highlight collateral and balance sheets we have eliminated a series of traditional ingredients 35 effects, and in so doing which are needed to pin down exchange rates during crises, such as [date and nontradeables, ables Ours is consumption decisions involving a choice between trade- monetary decisions that pin down a nominal exchange or than on rising risk possible is Indeed, our emphasis on is premia - although connecting the if any combination of these two, are is decline in stocks domestic asset prices. This implication seems highly consistent with the there one robust is In many situations debt less than prime fire sale, The banks' instead. which is which is assets. In such cases interest rate surges underestimate the avail, is also banks in the post-Russian crisis is a fairly well established fact. is that implication, fire sales many high marginal product sharp decline in investment and output. This The second is projects are foregone, resulting of view of policy, its markets —due to the is sector, which leaves insufficient monopsony argument, but development of domestic finan- — that renders inadequate the compensation received by collateral providers. There are several questions that immediately come to mind with respect to What the undervalua- international collateral. This un- dervaluation stems not from the traditional Bardhan-Harberger cial occur, the dual of this and accumulation by the private the economy unfit to withstand significant shocks to from the domestic credit constraint The behavior also a central feature of actual crises. and main one from the point tion of international collateral creation when a good example. Regardless of the specific mechanism by which the in the and quantity constraints reluctance to lend in the midst of a capital flow reversal, even decline in asset prices it is the practitioners' catch-all term for partially reflected in rationing they have plenty of liquidity at their of Argentine As we argued and equity markets are nearly non-existent and most financing occurs through the banking sector. extent of the facts. and other fact of crises associated to financial factors, precisely the sharp decline in "risk appetite," a higher required return on all possibili- a shift in the "collateral/risk premium" term. Internal arbitrage must cause a corresponding if risk interpre- given to incentive constraints.''* is which are consistent with the model as long as there in the introduction, and from an exchange rate overshooting, interest parity condition results or from a sharp rise in domestic rates, or collateral supply shortages rather collateral shortage a more probabilistic interpretation But whether the new ties rate. a model of a shift in the interest parity condition brought about by collateral short- ages that hinder arbitrage. tations 1] this implication. are the observable counterparts of the model's domestic financial underdevelopment? Is there evidence on the connection between crises and such underdevelopment? the specific mechanism behind the externality? And, does The model's mean is have empirical content? indicator of the degree of underdevelopment of domestic markets domestic spread, by which we Where it What is the the gap between the marginal product of distressed the probabilistic term must have an exogenous or endogenous aggregate component. Also, see Holmstrom and Tirole (1998b) for a model of asset pricing based 36 on liquidity rather than risk premia. firms and the domestic directly has always been problematic involve at least one unobservable. financial markets time. Measuring such gaps interest rate -in the sense described above. for What supporters of credit rationing models since they is reasonably measurable, however, and the bulk of the intermediation that takes place With these measures mind, the evidence in is in is the size of an economy over overwhelmingly in favor of the view that domestic financial markets in emerging economies are extremely underdeveloped compared to those of we OECD's. Interestingly, most of the very few exceptions reside in Asia, a point return to below.''^ More outsiders generally, our basic insight is that a limited ability to transfer internal return to —the genesis and essence of a market to suppliers since financial constraint effectively reduces its size. it — reduces the appeal of that Our model limits these suppliers to domestic producers of T-goods, but in reality international market-makers are also included, and these To a typically dislike markets that do not support significant volumes. this extension follows the same logical steps and conclusions of our first order, stylized analysis. Although we have emphasized the implications of underdeveloped financial markets crises, an interesting dimension of our model is that extremely underdeveloped financial mar- kets can lead to lower volatility with respect to external financial shocks. when equilibrium falls in for the wasted collateral region (region II), is The chief factor the underdevelopment of domestic financial markets leading to a failure of domestic aggregation of resources rather than to the scarcity of external resources.'*'^ This non-monotonic relationship between do- —low development mestic financial development and vulnerability development and very high volatility, and and high development and low volatility, volatility middle — connects our view with two pieces of supportive evidence. The one corresponds to the low to middle development range, where correlation first between development and crises is the rapid increase in of rushed and One volatility is positive. of the domestic intermediation. This careless lending, or the result of some main empirical predictors is of often interpreted as evidence moral hazard As institution. Some of that is midoubtedly true, but our framework offers a different angle: the country moves from region wasted collateral region) into region IV, where external factors may lead to crises and II (the fire sales. financial development rises, Moreover, since financial underdevelopment means that private agents undervalue international collateral, they do not take enough precautions against these crises. Within a range, their increased financing do not come hand-on-hand with "The large spreads for problems with domestic banks in collateral. full emerging economies between deposit and lending rates These spreads are partly accounted significant role in determining these spreads as well al. possibilities incentives to take the adequate precautions required incurred by banks to overcome firms' contractual problems. '^See Aghion et and aggregation — e.g. (1999) for similar implication in a 37 for may by the monitoring and also reflect legal costs Although undoubtedly other factors play a inflation, oligopolistic model of endogenous banking structures, credit cycles. etc. by higher leverage."*^ The second one economies fairly why is when development refers to the other end, is high. of Asian in countries that could achieve such high levels of leverage -reflecting developed financial markets- were there such dramatic crises? one of "not enough development." long stretches of stability. economies The puzzle Financial development gave short of Still, G5 levels of Our perspective is them much growth and domestic market development, the themselves fragile and exposed to crises - albeit imlikely ex-ante - that left proved very damaging."*^ The latter discussion takes us to the externality ments of the model. In the latter, and probably agents do not worry about the nature and relevance beyond the confine- its in reality as well, domestic —domestic or international— of once the relative prices and market liquidity of these assets But the central planner does, savings calculations, At the it is just not in the instrument-mix that the central planner needs. it is which right prices, is what happens in our when microeconomic the central planner. But and taken into consideration.'*'^ the international collateral constraint model when domestic are well functioning, entrepreneurs align their portfolio prices collateral per-se, Individual entrepreneurs do their right microeconomic precautionary first. ^^ that binds since during crises is microeconomic interest rates do not and leverage selections with those of units are credit constrained in equilibrium, reflect fully the marginal value of resources. This the dual of a credit constrained equilibrium, and as such emerging economies as financial markets it is simply appears as a robust feature of well. Finally, the third implication worth expanding on is that the above factors can generate a serious financial bottleneck, with potential multiple equilibria consequences, due to the collapse of banks' balance sheets. Is this channel empirically relevant? And, does the underlying amplification mechanism extend beyond banks? In our model, rather than emphasizing the depositors-lenders maturity mismatches and the potential runs that come with '"'Eventually, however, incentives catch "^Having said collateral. A this, it we have chosen to emphasize the role of domestic up with financing opportunities probably the case that their is component big realization of this it, development continues. was beyond their international as financial levels of leverage of the crisis, especially right after Thailand, probably mismatch by international investors. had to do with the Traditional moral hazard arguments may have played a role here as well, although we continue to believe that the central problem of emerging markets too little — and too Note that volatile this claim is not inconsistent class of assets as collateral. is — not too much capital inflows. with our assumption that foreigners do not accept certain Their aversion to these assets comes from their perception that they have an informational and incentive disadvantage with respect to certain assets, and hence are priced out of those markets given their uncertainties. In a sense, during one that matters normal times our model operates with two "currencies," while during for total crises it is only reinvestment and activity, while the other one only decides internal transfers. 38 banks Banks borrow from in transforming domestic international collateral. Some lend to local firms that foreigners would not lend to directly. foreigners to of these firms are in Regardless of the nominal denomination of the domestic loans, banks are the N-sector. A naturally mismatched in their effective denominations. exchange rates reduces the banks' ability to intermediate foreign resources into the and much whether there is a crisis or not, but its real- economy extreme circumstances, the possibiUty of multiple equilibria at the worst of times. In to determine not so collapse in asset prices arises magnitude once embroiled in the crisis. This mechanism, as we have argued, plays a role And external crises. in its into the bank run type resources to crisis, name the early 80s, to crisis of Mexico during the tequila a few. Moreover, it crisis it reminds us and Chile during the complements naturally and feeds of multiple equilibria, as a reduced balance sheet means less residual This interaction probably had something to those depositors that run late. left emerging markets' extreme form of a worst equilibrium among many, of Indonesia in the recent Asian debt in virtually all do with the Argentine bank run of the mid-90s. But the mechanism we capture goes well beyond the banks themselves. In our model, banks are distinct from corporations in that the their T-component and their N-component. But a feature of reality. Once such sharp separation component this is is can separate incentive problem in clearly a modeling simpUfication not blurred, a collapse in the value of the N- of a corporation affects its international collateral as well, amplification mechanism may internal contagion the case, latter e.g., of our banks also arise is when and the feedback and extended directly to the corporate sectors are complementary sector. Similarly, in "production," as is with a government whose claims lose their rating, and via the sovereignty principle, causes a depreciation in the rating of domestic corporations as well.^^ All of our asset price and real implications arise from an abstract liquidity "shock." How should these shocks be interpreted in reality? They can arise as pure financial shocks, as often occurs in the so called "contagion" scenarios. Are these totally imrelated to fundamentals? Most short fist as given likely not, but by now most coimtries know quite well whether they are on the for a contagion induced shock. As such, taking our liquidity-shocks-abstraction and thinking about its impUcations - the domestic factors that amphfy them, and the generic type of policy responses that are most appropriate - But shocks need not be purely financial, as the latter shock to terms of trades, productivity, domestic wages, '"^Onc of the main payments trouble, it is Ukely to implement measures the private sector's repayments as well. sovereignty principle is — e.g. may be loss of justifications for the sovereignty principle is is a fruitful exercise. the reflection of a real competitiveness and a myriad that as the government finds itself in suspension of convertibility — that compromise Indeed, the very few broad exceptions allowed by reserved for highly dollarized economies, where 39 S&P convertibility is difficult to to the suspend. factors that Each weaken domestic corporations and a government's credit-quality perceptions. of these shocks has its own idiosyncrasies, but our point of view financial impact of these shocks, none of history has offered. in point. The them should have the real in isolation - at worst just a 2 or 3 percent issue when one in so doing compromise It fiscal stability, have omitted specific episode, output equivalent shock for a it becomes a the convertibility law, and with many it the private sector factors that play central roles in specific crises; currency attacks deficits, to crises name a few. Our goal has not been to explain a can be discussed. The standard intertemporal substitution driven, analysis of developed economies, alternative we offer is is for macroeconomic too distant from emerging markets' reality to be of a stylized framework that brings to the fore what much we view bone of emerging markets' structure: underdeveloped domestic asset markets coupled with weak international financial this structure as, for we order would be abated. deep markets, neoclassical model, which while useful as a benchmark as the back first but rather, to offer a useful framework within which the economics of emerging markets The a case seems safe to argue that absent imperfect financial markets considerations, and unsustainable budget use. is considers the uncertainty that this engenders: Will interest rates soar, and of this uncertainty We consequences that recent recent bout of uncertainty around Argentina's prospects country that has growth prospects well above OECD's. However, much that absent the Argentina's loss of competitiveness vis-a-vis Brazil seems a second order issue when considered at large? is links. Standard crises ingredients can be built on example, we are doing in Caballero and Krishnamurthy (1999) where discuss the virtues and pitfalls of different our "asset markets perspective." 40 exchange rate systems within the context of Appendix 6 Incentive Constrcdnts 6.1 We have appealed to agency conflicts in order to justify the assumption that entrepreneurs One be required to retain a fraction of their shares. illiquidity assumption is possible formulation to justify this as follows. Consider a contract between a firm and a lender at date one unit at date [0, Now assume oo). can observe this that the realization oi Rt payment only by paying a = where E[Rt] yields date 2 flows of Rt is Suppose that a firm investing 0. and the support of Rt Rt, A private information of the firm. cost of c (in units of T-good). Then, is it is lender fairly standard to show that an optimal contract will be a debt contract (see Gale and Hellwig The (1985)). contract will specify a face of /; Rt ii >f and lenders pay the audit otherwise, the firm defaults the firm makes the repayment of /; and receive Rt cost Asstime that each firm has a continuum of such projects each with Each project is ^' This is = E[Rt - f\Rt /Ifl. < < /] define, , <' , f]Prob[Rt < f] + fProb[Rt > flProHR, < /I _ ^Prom < f] ^ ^^^^^^^ > f] , i _ ,, _ /""IR < ;] Rt suppressed the cost c, so that this /] Rt the component of firms that can be held by outsiders. In itself. Lenders receive is, Rt holds > return oi Rt. of, scaled by Rt this is f]Prob[Rt Rt E[Rt - c\Rt This > i.i.d. the share of each firm that the entrepreneur necessarily holds. expected flows m, - Then individually financed via a debt contract. — c. component is oiu: formulation we have simply the complement of what the firm However, this makes no qualitative difference to the problem. Derivation of the Central Planner's Program: Proposition 3 6.2 In this section we show how to arrive at the central planner's program, (P5) in the text. Consider a central planner that maximizes the weighted sum of utilities of agents in the economy. maxX:^:'^ [p«,.<.)^ It is + (1 -p)«,c-j5^ easy to show that this can be rewritten in terms of aggregate consumption. order condition for individual consimiption gives -S^ 41 = The first -^. This implies that aggregate consumption satisfies, -C^ <a< 1 P 1 The -a central planners objective can then be rewritten as, P«,s<s)' + (1 - pWuA^^ = (C„(a;)Ct(u;))^ or, maxX:7r-[KC<j'+(l-/^)«.<.)']=™^E^"(^»H^tM)^Consider next the aggregate consumption relations. In the good state there is full project completion, so that, In the bad state only 6nKn is Cn{G) = RnKn Ct{G) = RtKt-pKn completed, Cn{B) = kr,{Rn-pAr,{l-9^)) Ct{B) = ktRt-pe^kn-doj where, B '"= \?RtKt - Doj P Substituting in these relations for aggregate constraints, the central planners problem {PAl) maxfc„,fc„d„^^ S.t. is, Eu.e{G,B}7r'^(C»(w)Ct(w))^ = Rnkn Cp = Rth - doj - pkn C^ = k4R^-pAn{l-e^)) Cj = ktRt — pO^ kn — dgf C„ aB _ XfRtKt-Doj Doj < min^{Xt RtKt + Xn^^^S^ - e^K^} Kn + Kt = Wo + Doj A few remarks constraints. first. (PAl) incorporates both the domestic and international collateral Thus, the central planner is operating under the same constraints as agents 42 in the decentralized equilibrium. Prices do not appear in either objective or constraints except for the no-default debt constraint. Moreover, this constraint \n appears. As we showed earlier, this constraint will Let us next rewrite this program so that (P5), more easily. In doing this it is it is - the only one in which never bind. can be compared to the decentralized solution, easier to rewrite (PAl) and reintroduce thereby prices, getting (P4): {PA) Eu;G{G,B}7r^e(a;)2u;*(a;) max/f„,/c,,z)o,/ wl{B) s.t. w*2{G) = = RtKt RtKt + + ^^"4^)"^^" ^ - Do.f + (sf + sf) {X^RtKt - Doj) - Doj - pK^ Doj < min^{Xr RtKt + A^ ^ff^L""" } Kn + Kt = Wo + Doj (P4) arrives from substituting back in expressions for the exchange rate into the FOC for (PAl). First write (PAl) as follows, max^„,K.,Do,/,c-.,c-.,c5i.,.c^, s.t. + (1 - ^)«i<j2 pd^^, + (1 - /.)<, = cr p<,, + (l-p)c-, = C^n ~ Rn^n Cj = Rtkt — doj — pkn C^ = kn{Rn - PA„(1 - 9^)) Cf = ktRt - pO^kn - doj Ett'" [p{<,s(^,s)~^ ^n = p Doj < min^iXtRtKt + A^^^g^ - ^^i^n} Kn + Kt = Wo + Doj Let a = {Kn,Kt,Doj) be the vector maxa,c^^_,ej;^,<^,^,<^^ s.t. where F{a) reflects the -A^r(K. + date (1 of date choices, and rewrite the problem as, + (1 " P)i^,(^A)' pd^^, + (1 - p)c^, = Cr(a) pc^_, + (l-p)d;^,, = C-(Q) F(a) = E^r'^ constraints. [/'«.<s)^ Forming the Lagrangian, - pKs - Q"(a)) - <«,. + 43 (1 - pKs - c^i^))] - /^F^(") which gives the F.O.C.'s, & 1 IJ-t 2K« 1 I c ""' ' 2Vct. Notice that Which this is 6.3 is fJ-f = ]^ = e{u))2. Substituting this into the last constraint, just, the same F.O.C. as (P4) gives. Proof of Proposition 4 Consider the Lagrangian of (P5), jC= ^ TT'^e{uj)2W2{u;) - iJ.{kn + kt-wo-doj) w6{G,B} Which gives F.O.C.'s, 5do,/ ddoj ^c^o,/ Denote the solutions to these equations as {kn,kt,doj)- Next consider the F.O.C.'s to the Lagrangian of (P4), 44 We can evaluate the partial derivatives in L* at the choices dC* dC dkt dkt least s^ one > <7 < and 1, + n^e{B)h^{l-^)XfRt>0 we can conclude But by the budget 6.4 that welfare can be improved by choosing at of, < doj d'oj do as kt,doj). ddQj ddoj Since, (A;„, constraint, if kn — kn /C( > kt one of these holds strictly, at least one of the others must well. Contingent Debt Contracts In this section we rewrite the problem of sections 3 written contingent on the state of the world lj and 4 to include debt contracts that are £ {G,B}. The no-default constraint is now written as, do J The problem amount at date 1 is < \ ntkt + jj^ only altered by the fact that date of date 1 debt international collateral than decisions result in a different would be the case if debt contracts were not contingent. We can rewrite the date (P5C) program maxfc„,fc,_d^^ s.t. of (P5) as, E<^G{G,B}7r'^e(w)2'u;2(a;) W2{G) w,{B) d^j kn Allowing for economy by ^ - d^j - pkn = Rtkt + = R,h+ ^^-l^f- -dEj + (sf + p^s^){\^Rtkt-dEj) < X^Rtkt + Xnj^ + kt = wo + E^[d'^j]. contingency does increase the amount of insurance possible by the domestic offsetting some of the fall of A^. 45 However there will still be under-insurance relative the central planners problem as the basic externality the central planners problem for a comparison is unaffected. We can rewrite as, 1 (P4C) maxK„,/c,,D5'_^ E^e{G,B} ^ - D^j + (sf + 5^) {XfRtKt - D^j) = RtKt + - D^j - pKn w*2{G) = RtKt + D^j < X^RtKt + A„ ^ffj5" Kn + Kt = Wo + E4Doj] wl{B) s'.t. It is 7r"'e(tj)2w;*(w) ^ apparent that the expressions for W2{B) and W2 are different - again reflected in the domestic spread. For the same reasons as in the text, this will result in undervaluation of international collateral. Supporting Formulae 6.5 The solution to program (P6) for is Banking Section determined by the binding constraints. Constraint (i) requires that loans satisfy, a Given an amount of loans, the demand determined by the rate of for deposits l => be x'[^a{u,)<dl^iu;)+at-4j. Turning next to the decisions of firms, demand from intact firms follow from the same for loans logic as before (see by distressed firms and deposits PI and notational change with respect to sections 2 and the quantity of debt (supplied by distressed firms and of face value of date 2 T-goods. In 1 will interest, L^ = we make one and foreign borrowing by banks what 3. P2). For ease of exposition, demanded by follows, the quantity of debt present value of debt. Intact firms extend deposits to banks of Lf)> 1 ^ a:i_d(w) = X^Rth - doj - = 1 => xi,d(a;) < Lf) Distressed firms take out bank loans of ^> L"^ ^ %=L'^ ^ 0(1^^(0;) di,d(u;) XfRtkt - doj - a:i intact firms) in units is in ^(w), at at. to save production units, = max[kn - <j,0], di,rf(a;)<max[fc„-d^_/,0]. 46 we had denominated There, terms of the date Vis-a-vis the previous analysis, market clearing conditions need to be modified slightly The aggregate amount to take into account banks. must equal the of deposits in banks aggregate deposits of the intact firms, dl^iu) = = D^^iu,) X.^uj) Likewise, the total loan supply from banks = (1 - p)xrA^) must equal the new domestic debt issued by distressed firms. Taken together and combined with the capital constraint on banks \tRtkt-4f-do,f<— ''a XtRtkt-4f-doj>^ Consider the first inequahty. When =^ diX^) = => d,4{u) {1 - this implies that, p)xiAoj) 'a^ = {l-p)xt^ = the capital requirement on banks does not bind, banks aggregate domestic collateral perfectly - they channel the intact firms to the distressed firms. all of the excess debt capacity of The economy behaves exactly as if A„ = 1 in a public debt market (section 2.4). Asset prices and investment depend on the international collateral constraint, pkn<XtRtkt-doj-dlf pkn>XtRtkt-doj-d'oj When the capital requirement depending on how tightly L^ = 1 it (i) ^ ^ L^^n^^^l L^ = binds, the binds relative to L^j, = 0^ ^ 1 e^<l. economy can demand from = fall into regions (iii) In this case, distressed firms. and: p{kn-{XiRtkt-doj-At))<^ ^ L^ = p{k^-{XfR,kt-doj-At))>^ ^ B^ a 47 = l ^ et = or (iv) l e-<l. References [1] Aghion, Phillipe, Bachetta, Philippe, and Banerjee, Abhijit, "Capital Markets and the Instability of [2] Open Economies," MIT mimeo, January Aizenman, Joshua, "Country Risk, Incomplete Information and Taxes on International Borrowing," Economic Journal [3] 1999. 99, March 1989, pp. 147-161. Bardhan, Pranab K., "Optimum Foreign Borrowing," the Theory of in Karl Shell, MIT Optimal Economic Growth, Cambridge, Mass., ed., Essays on Press, 1967, pp. 117-128. [4] Bernanke, Ben, Gertler, Mark, and Gilchrist, Simon, "The Financial Accelerator in a Quantitative Business Cycle Framework," [5] NBER Working Paper 6455, March 1998. Bhattacharya, Sudipto, and Gale, Douglas M., "Preference Shocks, Liquidity, and Central to Bank Policy", Barnett, William A., ed.; Singleton, Kenneth J., ed.. New pproaches monetary economics: Proceedings of the Second International Symposium in Eco- nomic Theory and Econometrics. Econometrics Cambridge; series International Symposia in Economic Theory and New York and Melbourne: Cambridge University Press 1987, pp. 69- 88. [6] Blommestein, Hans rities [7] lADB "Institutional Investors, Pension WP # 359, November Reforms and Emerging Secu- 1997. Borensztein, Eduardo, and Atish R. Ghosh, "Foreign Borrowing and Export Promotion, [8] Markets," J., IMF Staff Papers, 36-4, December Bulow, Jeremy and Rogoff, Kenneth, 1989, p. 904-932. "A Constant Recontracting Model of Sovereign Debt", Journal of Political Economy 97-1, February 1989, pp. 155-78. [9] Caballero, Ricardo J. and Krishnamurthy, Arvind, "Capital Flow Dilemmas under Im- perfect Financial Markets: Macroeconomic Policy and Vulnerability," work in progress, October 1999. [10] Caballero, Ricardo J. and Krishnamurthy, Arvind, "Emerging Markets Assets Market Perspective," [11] mimeo MIT, November Crises: An 1998. Calvo, Guillermo A., "Balance of Payment Crises in Emerging Markets," mimeo, Maryland, January 1998. [12] Cavallo, Domingo, "A Comment," Latin America, ed. by R. in Volatile Capital Flows: Hausmann and L. Rojas-Suarez, pp. 45-48. 48 Taming Their Impact on IDB, Washington DC, 1996, [13] Chan-Lau, Jorge A., and Zhaohui Chen, "Financial Crisis and Credit Crunch as a Results of Inefficient Financial Intermediation-with Reference to the Asian Crisis," mimeo [14] Chang, Roberto and Andres Velasco, "The Asian Liquidity Diamond, Douglas W. and Dybvig, PhiUp Cole, Harold L., Debt [18] [19] Crisis," and Timothy Economy J. Report 210, April 1996. Staff and Alary HaUward-Driemeier, Ken and James "Crisis Adjustment, and Reform: Re- May 1998. Poterba, "Investor Diversification and International Equity 81, 1991, pp. 222-226. Fernandez- Arias and Davide Lombardo, "Private External Overborrowdng in Undistorted Economies: Market Failure and Optimal PoUcy," [21] Insurance, and Kehoe, "A Self-FulfiUing Model of Mexico's 1994-95 Markets," American Economic Review [20] "Bank Runs, Deposit H.. from the Thailand Industrial Survey," mimeo French, mimeo Atlanta 91-3, June 1983, pp. 401-19. Minneapohs FED, Research Dollar, David, sults Crisis," 1998b. Liquidity," Journal of Political [17] A Canon- Model," mimeo Atlanta FED, July 1998a. FED, July [16] August 1998. Chang, Roberto and Andres Velasco, "Financial Crisis in Emerging Markets: ical [15] IIVIF, mimeo March Gale, Douglas and Martin HeUwig, "Incentive Compatible 1998. Debt Contracts: The One- Period Problem" Review of Economic Studies 52(4), October 1985, pp. 647-63. [22] Gelos, R. straints, Gaston and Alejandro M. Werner, "Financial Liberalization, Credit Con- and Collateral: Investment in the Mexican Manufactuing Sector," lAlF \\T/99/25, March 1999. [23] Giavazzi, Francesco and Marco Pagano, "Confidence Crises and Pubhc Debt Manage- ment," in Public Debt Management: Theory and History, ed Rudiger Dornbusch and Mario Draghi. Cambridge Univesity Press, [24] Goldfajn, Han and Rodrigo O. Intermediation," [25] Valdes, "The IMF mimeo, August Harberger, Arnold, p. 94-118. Twin Crises and the Role of Financial 1998. "Lessons for Debtor-Country Managers and Policjonakers," J.Cuddington and G.Smith (eds.), in International Debt and the Developing Countries. Washington: World Bank, 1985. 49 [26] Holmstrom, Bengt and Tirole, Jean, "Financial Intermediation, Loanable Funds, and the Real Sector," Quarterly Journal of Economics 112-3, August 1997, pp 663-91. [27] Holmstrom, Bengt and Economy Journal of Political [28] Holmstrom, Bengt and Jean Tirole, "Private and Public Supply of Liquidity," (a), 106-1, February 1998, pp. 1-40. Tirole, Jean "LAPM: (b), Liquidity Asset Pricing Model," MIT mimeo, June 1998. [29] Charles Jacklin, J., "Demand Prescott, Edward C, ral trade. Minnesota Studies ed.; Deposits, Trading Restrictions, and Risk Sharing," Wallace, Neil, ed.. Contractual arrangements for intertempoin Macroeconomics series, vol. 1 Minneapolis: University of Minnesota Press 1987, pp. 26-47. [30] Monetary Fund, International Prospects, and Key [31] J. Mathieson, Garry Schinasi, and Bankim Chadha, 1998. Johnson, Simon, Peter Boone, Alasdair Breach, and Eric Friedman, "Corporate Governance in the Asian Financial [32] Developments, World Economic and Financial Surveys, directed by Policy," in Charles Adams, Donald Markets: Capital "International Crisis, 1997-98," Kang, Jun-Koo and Rene M. Stulz, Portfolio Equity Ownership in "Why Is MIT mimeo, September There Home Bias? An 21, 1998. Anlysis of Foreign Japan," Journal of Financial Economics 46-1, October 1997, pp. 3-28. [33] Kiyotaki, Nobuhiro and Moore, John, "Credit Cycles," Journal of Political Economy, 105-2, April 1997, pp. 211-48. [34] Kletzer, Kenneth, "External Borrowing sues." In Schultz, [35] The State of Development Economics New A ed. Survey of Some Theoretical Is- By Gustav Ranis and Paul T. York: Basil Blackwell, 1988. Krishnamurthy, Arvind, "Collateral Constraints and the Amplification Mechanism," mimeo Northwestern [36] by LDCs: Krugman, Paul, University, November "International J.Cuddington and G.Smith (eds.). Debt 1998. Strategies in an Uncertain World," in International Debt and the Developing Countries. Washington: World Bank, 1985a. [37] Krugman, Paul, "Prospects and Financial Issues for [38] for International Debt Reform," the Developing Countries, Geneva: Krugman, Paul, "What Happened to Asia?" 50 MIT in International UNCTAD., mimeo, 1998. Monetary 1985b. [39] La Porta, Rafael, Florencio Lopez-de-Silanes, Andrei Shleifer and Finance," Journal of [40] Political 85, December 1995, pp. 1106-1117. and Growth," American Eco- December 1994, 1310-29. 40, April 1996, 1037-47. Radelet, Steven and Jeffrey Sachs, "The Onset of the East Asian Financial Crisis," mimeo HIID, February [44] 1998, pp. 1113-1155. Obstfeld, Maurice, "Models of Currency Crises with Self-fulfilling Features," European Economic Review [43] December Obstfeld, Maurice, "Risk Taking, Global Diversification, nomic Review [42] 106-6, Lamont, Owen, "Corporate Debt Overhang and Macroeconomic Expectations," American Economic Review, 85-5, [41] Economy and Robert Vishny, "Law 1998. Sachs, Jeffrey, Theoretical Issues in International Borrowing. Princeton Studies in In- ternational Finance, #54, Princeton University, July 1984. [45] Simonsen, Mario H., "The Developing-country Debt Problem," 1985. [46] Shleifer, Andrei and Robert Vishny, "Liquidation Values and Debt Capacity: A Market Equilibrium Approach," Journal of Finance 47-4 September 1992. [47] Shleifer, 52-1, [48] Andrei and Robert Vishny, "The Limits of Arbitrage," Journal of Finance March 1997, pp 35-55. World Economic Outlook, IMF, May 1998. [49] The World Bank, East [50] Zhou, Chunseng, "Dynamic Portfolio Choice and Asset Pricing with Differential Infor- Asia: The Road to Recovery, September 1998. mation," Journal of Dynamics and Control, July 1998. 51 •ju =w= onf|)ate Due Lib-26-67 MIT LIBRARIES 3 9080 01917 6541