Financial Crises: Contagion and Market Volatility Report of a Conference held in London on 8-9 May 1998

The opening paper was "Volatility and Contagion in a Financially-Integrated World: Lessons from East Asia’s Recent Experience", presented by Stijn Claessens (World Bank) and written with Amar Bhattacharya, Swati Ghosh, Pedro Alba (all of the World Bank) and Leonardo Hernandez (Central Bank of Chile). This paper provided an overview of the build-up of financial vulnerability in East Asia, showing that structural inadequacies, more than inadequate macroeconomic fundamentals, played a major part in the build-up to crisis. In particular, insufficient supervision of financial sectors and lack of transparency in reporting, were aggravated by undisciplined and uncontrolled foreign lending, leading to an unexpectedly large accumulation of short-term unhedged debt. These structural failings served to enhance the severity of the crisis, when it was eventually triggered, resulting in an impact on each individual economy that far exceeded the effect of the Mexican peso crisis of 1994-95.

The actual trigger appears to have been increasing pressure from external factors, when investors began increasingly to focus on the misalignment of the Thai exchange rate. An inadequate response by the domestic authorities eventually precipitated the first crisis, which spread further afield by means of bank-run type behaviour in other economies in the region.

The nature of the contagion effect is difficult to distinguish. Asset prices show high correlation across the region over certain periods, but there is no clear evidence of whether these were the ultimate result of pure spillover or region-wide similarity in changes in fundamentals.

The authors find that the policy prescriptions resulting from this study are clear: structural reform, together with diversified sources of external financing, is needed to prevent the occurrence of crisis. Should a crisis be triggered, such measures will also help to speed up the process of recovery.

Michael Dooley (University of California, Santa Cruz), discussing the paper, pointed out that most stories of the genesis of crisis were untestable and therefore irrefutable. Unrecorded financial flows were a historic feature of economies such as that of Indonesia and substantial expatriation of capital led to vulnerability to crisis. Japan, on the other hand, is a high-saving high-growth economy whose government is unwilling to recapitalise. A question that must be addressed is whether these economies are capable of recovering on their own, and whether substantial infusions of foreign capital are required to mend them. Sule Ozler (Koç University and UCLA) reiterated the lack of testable models of crisis and added that the current debate recapitulated that following the debt crisis of the 1980s. The main difference between now and then is that the present situation involves large numbers of private borrowers and lenders rather than the government borrowing seen in the 1980s, leading to increased difficulty in information monitoring and coordination of action. More theoretical and empirical work was needed on the nature and behaviour of crisis triggers.

The second paper, on "Determinants of Emerging Market Currency Crises and Contagion Effects" was presented by Manmohan Kumar (Credit Suisse First Boston) and written with William Perraudin (Birkbeck College and CEPR) and Uma Moorthy (Birkbeck College). The paper undertook an empirical analysis of monthly data on 32 emerging markets over the period January 1985-March 1998, seeking to determine the extent to which the probability of the occurrence of crisis can be predicted by the behaviour of macroeconomic fundamentals. Most existing empirical studies are either case-studies of single crisis episodes, "signalling" models where the behaviour of a fundamental such as the real effective exchange rate or the ratio of debt to GDP is seen as a signal of entry into a crisis state, or qualitative-response models using annual data.

A logistic regression model was used to model the probability of crisis. Two distinct specifications were used to generate the dichotomous dependent variable, in order to evaluate the robustness of the model to definition of crisis as sudden sharp depreciation against the definition of "unanticipated" depreciation. The regressors were chosen from among macroeconomic fundamentals, financial variables comprising both stocks and flows, policy indicator dummies and regional dummies. The paper explicitly modelled contagion effects both as correlated crisis effects and as regional spillover effects. The model was evaluated by calculating profits accruing as a result of trading on markets for which a high probability of crisis was forecast.

Paolo Pesenti (Princeton University and NBER), as the first discussant, described the paper as a useful contribution to the empirical literature. The model took an "agnostic" view of crisis but confirmed a key result: that careful examination of macroeconomic fundamentals can help forecast financial instability. The paper’s contributions were in the use of high-frequency data, and in showing that economic growth and fiscal imbalance, such as budget deficit or public dis-saving were correlated with the occurrence of crisis. Andreas Fischer (Schweizerische Nazionalbank and CEPR) saw the paper more as an investigation of exchange rate risk management from the point of view of an US investor than as an explanation of contagion or crisis. The dataset was impressive, but the analysis could be extended by, for example, allowing the definition of crisis to be endogenously determined. The role of conditionality between crises could also usefully be examined. Finally, he commented on the strikingly positive nature of the results and suggested that parallels with other empirical work could be drawn more explicitly. Several further contributions to the discussion pointed out the need for examination of the time series properties of the flow variables and for more extensive out-of-sample estimation: for example, which crises did the model fail to predict?

Pierre-Richard Agénor (World Bank and IMF) and Joshua Aizenman (Dartmouth College) jointly presented their paper on "Volatility and the Welfare Costs of Market Integration". This paper examines the short-term costs incurred when increased foreign access to a country’s domestic financial system leads to increased vulnerability on the part of that country to financial market volatility. While such openness may be beneficial in the long run, welfare costs can be incurred in the short term through domestic market inefficiencies, notably inefficient intermediation, inadequate lending practices, large unhedged short-term foreign borrowing , inaccurate disclosure and ineffective supervision. In previous work, the authors demonstrated how such imperfections could magnify an initial exogeneous shock into a crisis in the domestic financial market. In this paper, they extend their model to consider the impact of volatility on financial market integration.

The basic framework employed is one in which risk-neutral banks provide credit intermediation services to domestic agents. The authors contrast the behaviour of borrowers and lenders under financial autarky, where domestic banks are the only source of credit, with their behaviour under financial openness. Greater openness is assumed to result in more intense competition among lenders; foreign lenders are also assumed to be more efficient than domestic credit sources as a result of greater experience in lending and of economies of scale. These efficiencies are reflected in lower costs of intermediation and lower mark-up.

However, opening up a market to capital inflow can magnify the welfare cost of domestic distortions, overwhelming the existing capital infrastructure and leading ultimately to collapse. This insight is illustrated by the recent Asian experience of crisis. An economy, which worked well while financially self-contained, may fail to adjust quickly enough to sudden inflows of foreign capital, and as a result of congestion externalities and domestic financial distortions, be precipitated into crisis by the shock of a relatively small capital inflow. It is therefore necessary to ensure that domestic distortions are dealt with before integration is sought

The first discussant, David Begg (Birkbeck College and CEPR), felt that the paper was a serious contribution to the banking and financial integration literature. But he argued that much of the theory was mistaken, in particular, the use of the backward bending interest rate/cost of credit curve. The paper should be extended to include consideration of incentive compatibility and participation constraints. The borrower’s and the lenders’ expectations should also be incorporated explicitly in the model. Anne Sibert (Birkbeck College and CEPR) added that domestic credit markets were characterised by asymmetric information, since an entrepreneur is likely to be better informed than the bank about the likelihood of default. Hence, moral hazard and adverse selection were also important considerations. The role of legal restrictions on borrowing and lending would be a useful addition to the model. The authors replied at this point, to say that the backward bending supply curve was not crucial to their argument. Their previous paper explicitly modelled credit rationing and enforcement and agency costs, and these were also present in the current model. Considerable discussion ensued, which Marcus Miller (University of Warwick and CEPR) summarised by saying that the authors presented serious arguments against financial liberalisation. The paper focussed on "second-best" arguments and opened up debate on their accuracy.

Ishac Diwan (World Bank) and Bernard Hoekman (World Bank and CEPR) presented their paper on "Competition, Complementarity and Contagion in East Asia". Their paper sought to examine the genesis of the series of crises affecting East Asian countries. These countries can be viewed either as competitors, by virtue of export similarity, or as complements, who exhibit similar but not necessarily synchronous patterns of economic development. These two factors act as transmission channels for the spread of crisis within the region. It is possible that both factors have operated simultaneously in East Asia where the increasing importance of China in the region, and the weakness in the Japanese economy have each played their part. It is necessary to disentangle them as the ensuing ripple effects and stabilising policy prescriptions are factor-dependent, however.

Three kinds of effects are identified as possible channels for crisis transmission: an income effect and an investment effect, which are most affected by complementarity, and a substitution effect which is most affected by competition. The authors then explore competitive and complementary behaviour empirically, looking at correlations in the growth of fundamentals such as GDP, investment, consumption and exports in East Asia with Japan and China and the role of foreign direct investment in the region, particularly its sources and composition.

The first discussant, Jenny Corbett (St Andrews’ College, Oxford and CEPR), said that the paper was important in emphasising the role of trade as a channel for contagion. She questioned the evidence of similarity between Asian economies, pointing out that the group of countries which underwent the first crisis were distinct from the second wave. Trade shocks originating in China had considerable impact on competitor countries, but real-side shocks were also important. The paper could usefully explore further channels of contagion as suggested by other contributions to this conference and perform further empirical analysis of the role of correlations and rank correlations. David Vines (Institute of Economics and Statistics, Oxford and CEPR) pointed out that falls in growth seemed to occur as a result of declining exports, and suggested that the empirical features discussed in the paper were reduced from results, and he proposed a structural model from which they might be thought to come. Gabriel Galati (Bank for International Settlements) suggested the use of more disaggregated data. Reuven Glick (Federal Reserve Bank of San Francisco) mentioned the difficulties of distinguishing between Chinese and Hongkong exports as a result of trans-shipment. Paolo Pesenti (Princeton University and NBER) pointed out that the nature of trade between Japan, Korea and the United States was very different from that between China, Indonesia, Malaysia, and the United States.

Giancarlo Corsetti (Yale University and Universita di Roma Tre) presented "What Caused the Asian Currency and Financial Crisis", a paper in two parts written with Paolo Pesenti (Princeton University and NBER) and Nouriel Roubini (New York University, CEPR and NBER). The first part of the paper, subtitled "A macroeconomic overview", comprised a survey of East Asian indicators taking in fundamentals such as current account imbalances, GDP growth, investment, saving and reserves as well as the role of the banking system and moral hazard, and the role of political instability and uncertainty about economic policy.

The second part of the paper, subtitled "Interpreting and modelling the crisis", begins by analysing the period 1995-1996 in the lead-up to the crisis. A detailed discussion of the 1997 crisis period, encompassing the initial financial distress, spillover and contagion effects, the role of Japan and the state of current debate about the crisis follows. Even though current explanations of crisis, the "first-" and "second-generation models", might not appear to be able to explain the events in East Asia, it can be argued that these theories, if appropriately reformulated, still have explanatory power. The authors provide two extensions to these models: one looking at moral hazard in the context of bail-out promises and the behaviour of banks, the second focussing on currency crisis and contagion in a region comprising several countries. They find the roots of crisis in the combination of inconsistent government policy and an underdeveloped financial sector. These resulted in unsustainable patterns of investment, and ensuing government mismanagement of a worsening situation did the rest.

The first discussant was Richard Portes (London Business School and CEPR). He said that financial crisis, which is generally marked by interactions between the foreign exchange market, bank failures and debt default, is distinct from currency crisis. <N>The effects of the latter have already been extensively modelled, especially in the case of speculative attacks. He asked how the models presented could be extended to identify and measure the effects of moral hazard and to model hyperinflation. What was the effect of pegging an exchange rate and what were the safe limits on the exchange rate and the real exchange rate? He also noted that not even the IMF had correctly predicted this episode of crisis, and that danger signs had been ambiguous, to say the least. Jacques Olivier (HEC School of Management, Jouy-en-Josas and CEPR) felt that the paper was very comprehensive in its scope, but that clarity was needed in the definition of the model: for example, the authors propose a deterioration in fundamentals as a mechanism for crisis, but also invoke multiple equilibria via a shift in expectations. The question of (un)sustainability also needed investigation. He was not convinced by the figures presented in the paper that the deterioration of macro-fundamentals actually occurred. Finally, with respect to the second model presented in the paper, the assumption of perfect sustitutability between exports from countries affected by the first and second waves of crisis and the absence of trade links within the region were not credible. Joshua Aizenman (Dartmouth College) and Stijn Claessens (World Bank) pointed out that there were substantial differences in the economies and the nature of financing in Japan, Korea and Thailand, which led to differences in the types of crisis endured by each country. Paul Masson (IMF) mentioned that a study of IMF advice to SE Asian countries was shortly to be published.

Holger Wolf (Stern School of Business, New York University and NBER) presented "The Spatial Properties of Capital Flows is Location Destiny?" written with Swati Ghosh (World Bank). Rational investor decisions have been invoked as determinants of the direction and magnitude of financial flows. This paper seeks to augment this set of predictors with geographical and spatial factors. It is noted that, even though the volume of new private flows has increased enormously in the last decade, their destinations are highly concentrated within SE Asia and Latin America. Sub-Saharan Africa attracts very little.

The relevant spatial factors appear to be cultural familiarity, distance and existing trade links. The result would be an alternative description of regional contagion effects, in that the negative dependence of flows on distance will affect all countries in a region in similar fashion. A cross-sectional probit model is estimated, using as the dependent variable an access score defined as a ranking based on a weighted average of access to bank lending, access to short-term finance and access to capital markets. A second measure of access is a ranking based on the average absolute net inflow of private capital as a percentage of GDP over the period 1990-95.

The explanatory variables used are relative population size, relative market size, real development stage and growth. Additional regressors are openness (export to GDP ratio for 1989), distance to the nearest G7 economy, GDP-weighted average distance to the closest G7 economy and continental dummies for Africa and Latin America. A series of probit regressions demonstrate that, even though location factors are correlated with access, the real development stage is the most important determinant of access; if this is controlled for, location effects do not add significantly to the model.

A classification tree approach is used to test for the presence of threshold effects, showing that location is a good predictor of access, so that Africa and Latin America have much lower access to world financial markets. Income per capita provides a (slightly) better indicator of access than does distance to the G7 economies. The authors point out that the causal links between income per capita and location are far from clear. It is noted that previous work in the empirical growth literature has shown significant negative continental effects on per capita growth.

Finally, the sources of capital flows are examined in an attempt to determine the characteristics that define the choice of country of investment. The presence of "home bias" is tested using regression in a modified gravity model framework. The elasticities of exports, FDI, loans, debt and equity are all negative with respect to distance and positive with respect to remoteness, market size of the recipient and development level. The cultural affinity variables are generally positive in effect but not as influential as the above variables.

The discussant, Ian Marsh (University of Strathclyde and CEPR), said that the link between investment and information that underlies this paper was undoubtedly correct, and that barriers to investment arising out of little or no information had been well documented, such as the role of "pioneers" versus "fast followers". The authors are also correct in specifying location rather than geographical distance as a determinant of access, but the actual role of location is unclear. Related questions: why did Latin America have access in the nineteenth Century but no longer? what role does proxy investing play? does the model provide a mechanism for the spread of "Asian flu" from the ASEAN countries to the newly industrialised countries? Sule Ozler (Koç University and UCLA), contributing to the discussion, pointed out that former patterns of colonisation were probably important determinants of access and were well documented in the development literature.

The second day of the conference opened with the paper "Contagion and Trade: Why are Currency Crises Regional?" which was presented by Reuven Glick (Federal Reserve Bank of San Francisco) and written with Andrew Rose (University of California, Berkeley and CEPR). Currency crises tend to be regional. Neither the "first generation" models nor "second generation" models explain why crises tend to occur in regional clusters. Since trade patterns are regional, trade linkages are an important natural source of explanations for contagion in currency crises. The authors restrict their attention by ignoring the possibility of simultaneous regional shocks and the timing of crises. They substantiate their hypothesis by examining the histories and ensuing contagion effects of three recent major crises: the 1992 EMS crisis, the 1994 Mexican peso crisis and the Asian crisis, which began in 1997.

The authors assume, first, a speculative attack on a country and, second, the presence of contagion.

The data originate from five different episodes of currency instability:

1. the breakdown of Bretton Woods in 1971

2. the collapse of the Smithsonian Agreement in 1973

3. the EMS crisis of 1992-3

4. the Mexican crisis of 1994-5

5. the Asian crisis of 1997-8

Annual data from 161 countries is used, of which many were involved in none of the crisis episodes.

Estimation is performed by means of a cross-sectional binary probit model whose dependent variable is defined to be unity in the event that the country underwent a given crisis episode. In each crisis episode, the first country to be hit (the "ground-zero" country) is identified. The regressors are a set of macroeconomic control variables and a "total" trade index, which quantifies the importance of the trade links between this country and the rest of the world, relative to its trade with the "ground-zero" country. A "direct" trade measure was also constructed to measure trade flows between each country and the "ground-zero" country.

The macroeconomic control variables are chosen among the usual first generation model indicators, together with a measure of the degree of currency under-valuation. The last is an index constructed as the export-share weighted sum of bilateral real exchange rates in relation to the currencies of all trading partners. Initial analysis shows that the difference in trade linkages between crisis and non-crisis countries is statistically significant, but that this is not true of the macroeconomic controls.

The analysis is extended using a multivariate probit model, whose results reinforce those already obtained. Model checking was performed by means of varying the controls used as regressors, different measures of trade linkage, and different dependent variables, but in each case, the results obtained bore out the results of the initial model specification.

Finally, two continuous measures of crisis were calculated to examine the determinants of exchange market pressure. The first measure was defined as the cumulative percentage change in the nominal devaluation rate with respect to the ground zero currency for three, six and nine months following a crisis. The second measure was a weighted average of the devaluation rate and the percentage decline in international reserves over the same time period. The conclusions drawn from regression of these measures on the same regressors as before support the hypothesis that trade is a significant factor in the explanation of the incidence of crisis. The policy implications clearly provide support for some form of international monitoring of trade.

The first discussant, Mark Taylor (University College, Oxford and CEPR), said that the extent to which two countries competed in a third market is clearly a significant predictor of crisis. The policy implication that surveillance should incorporate a regional factor was also clear. While the econometric analysis in the paper was well executed, it would benefit from the following extensions: inclusion of other macroeconomic and financial variables, and the explicit modelling of moral hazard and asset bubbles, as well as poor corporate governance and other development and cultural factors. There were also problems of identification: how could one distinguish between a common shock such as the ERM crisis, in which countries within a customs union were affected, and contagion? Gravity models could also provide useful modelling insights. A final point: why was the Australian dollar not affected by the Asian crisis, given the importance of Australian trade linkage with Asia. He was followed by Javier Suarez (CEMFI, Madrid and CEPR), who said that he had enjoyed reading the paper because of its clarity of exposition. The empirical challenge faced by the paper, of identifying a "trade channel" for contagion could not, however, be based on the distinction between trade linkages and similar macroeconomic fundamentals. Strong fundamentals might serve as a defence against attack, but were not necessarily the channel for the spread of crisis. The empirical identification of the contribution for these fundamentals to crisis is a different question from whether regional crises are caused or amplified by the operation of a contagion channel. Alternative possibilities to a trade channel are financial channels, due to cross-border borrowing and lending, and information channels, based on observation by agents of the behaviour of comparable countries. Indeed, the trade effects might be proxying regional dummies, which were left out of the present analysis.

The proposed empirical model was difficult to interpret. Some kind of loss function approach might be useful in imposing a structural context on the model. The chronology of crisis is fundamental to this model leading to further problems with interpretation in the light of simultaneity and omitted variable bias. The paper makes stimulating evidence for the contribution of trade linkages, and would further benefit from more direct reference to theoretical and structural models.

The second paper of the day was "A Theory of the Onset of Currency Attacks" presented by Hyun Song Shin (Nuffield College, Oxford and CEPR) and written with Stephen Morris (University of Pennsylvania). Multiple-equilibrium models of crisis incorporate as given the self-fulfilling nature of belief in the imminence of attack. They do not seek to explain why switching occurs between equilibria, nor do they predict the timing of attack. The model presented in this paper seeks to deal with these lacunae. It relaxes the assumption of perfect observation of macroeconomic fundamentals by speculators, substituting differential information in its place. Fundamentals evolve according to a Brownian motion process, which speculators observe with noise, which is itself normally distributed.

The current state of the economy is observed perfectly only by the monetary authority. A continuum of speculators observes this as a noisy signal, but have perfect information about the past state of the economy. The willingness of the authorities to defend a currency peg is proportional to the state of the economy: a healthy economy is likely to be defended more strongly than an ailing one. The ferocity of the speculative attack needed to induce authorities to abandon the peg is therefore directly proportional to the strength of the economy. If the peg is removed, the currency depreciates by a known amount and never reverts to its original value.

The behaviour of speculators is modelled as an incomplete information game. First, it is shown that even a slight deviation from perfect common knowledge of the fundamentals, i.e. a small amount of noise, will result in outcomes which are substantially different from those arising out of perfect common knowledge. The authors demonstrate that multiple equilibria can arise only out of an extreme degree of common knowledge among speculators.

Harald Uhlig (CenTER, Tilburg University and CEPR), discussing the paper, said that it was clever, intriguing and bound to become a classic. The key result, contained in Theorem 1, shows that the addition of even a small amount of noise in speculators’ information to a "second-generation" model can lead to the collapse of multiple equilibria into a single equilibrium. This is a surprising and non-trivial result. The intuition underlying this result is as follows. If there is full knowledge, there may be multiple equilibria, so that, for a single level of fundamentals, there are two equilibria on which speculators might coordinate – either the crisis equilibrium or the no-crisis equilibrium. But, with asymmetric information, speculators lose common knowledge of the underlying fundamental uncertainty. If they have even a small amount of uncertainty as to how other speculators will behave, the paper shows that this can remove possibility of such multiple equilibria. It is possible that the multiplicity in equilibria could be restored by some other mechanism, such as sunspots or a crisis in a neighbouring country. There was a possible connection with Krugman (1996), who suggested that multiplicity could be ruled out with deteriorating fundamentals. Finally, the paper should be viewed as a perturbation analysis for small noise levels rather than as a stochastic theory of the onset of speculative attack. Paolo Vitale (LSE) continued the discussion, saying that the paper was a major contribution to the literature. Its objective was to clarify the link between differential information and speculative attack and to study the timing of attack. It was possible that one of the pieces of differential information that could emerge was the "type" of the monetary authority and its response to attack. The policy implications of the results were that authorities need to be less than clear about their objectives, and must also be ready to regulate the degree of uncertainty and differential information by means of signalling devices such as "cheap talk".

The third paper of the day, "The Portfolio Flows of International Investors I" was presented by Paul O’Connell (Emerging Markets Finance, LLC) and written with Kenneth Froot (Harvard University) and Mark Seasholes (Harvard University). This was an empirical paper which investigated the effects on prices when international investors are net buyers of equities in domestic markets. The existing literature suggests that the net impact of foreign investors on local markets is to push up prices. The data used to test this hypothesis comprised daily international portfolio flows of client institutions of a single major US custodian bank, covering 46 countries over the period 1994-1998.

The data are studied mainly via cross-country correlations. It is found that the data exhibit stronger positive correlation across regions than across individual countries, and that there is a considerable degree of trend-chasing behaviour. Differences are apparent in the behaviour of emerging and developed markets: in the former, inflows could forecast positive returns, whereas in the latter, returns could become negative.

Principal component analysis found that a regional factor, the "common-flow" component accounted for a considerable proportion of the co-movement, but that a country-specific effect was also at work. Variance ratios were used to examine persistence in flows. There appeared to be high persistence in flows in emerging markets, with shocks being associated with increased levels of future inflows over a considerable period of time. The US market, however, demonstrates little or no persistence.

Finally a two-equation vector autoregression was used to test for the value of returns in predicting flows in the future, over and above the contribution of lagged flows. There was little support overall for the hypothesis that emerging market inflows are due to an informational disadvantage among foreign investors.

The first discussant, Mark Salmon (City University and CEPR) began by commenting that the paper comprised a huge amount of work, but proved difficult to read because of its level of compression. The dataset was also very large, but only incorporated equity flows. The paper consisted mostly of descriptive statistics, and would benefit from the incorporation of conditional distributions of the data. The "heat map" used to summarise information on the correlations could benefit from the use of structure-seeking algorithms to elicit further patterns in the correlations. The "common-flow" component did not seem to account for enough of the correlation. It might be advisable to use the next one or two components as well. Cointegration analysis is a possible alternative to principal component analysis. The persistence of order flow is not an easy result to interpret. Once again, study of the integration structure of the data or of autocorrelation functions could provide clues. The VAR model is also heavily restricted, but evidence should be supplied on the justification for these restrictions. Manmohan Kumar (Credit Suisse First Boston) continued the discussion by saying that the paper was a useful investigation of the relationship between flows and returns. The dataset had considerable value, but should be extended to include information on debt as well as equity. Useful work had also been done by the authors on definitions of settlement. He asked about the proportion of the data which related to emerging markets and how the net flows were scaled by market capitalisation. It was possible, however, that generalisations made across emerging markets could lead to misleading results, and that more disaggregation might be desirable. The analysis should be extended to cover the portfolio aspects of investment and institutional constraints on investment.

Paul Masson ( IMF) presented his paper on "Contagion: Monsoonal Effects, Spillovers and Jumps between Multiple Equilibria" in which several concepts of contagion were defined and distinguished one from the other. First, policies undertaken by industrialised countries may have similar effects across several emerging markets. These are "monsoonal effects". Second, a crisis in one emerging market could affect others via changes in macroeconomic fundamentals e.g. devaluation could trigger changes in competitiveness. These are "spillovers". Finally, a crisis in one country may trigger a crisis in another, for reasons which are not explainable by fundamentals. This is "true" contagion, which requires the existence of self-fulfilling expectations and multiple equilibria. Most previous work on multiple equilibria or sunspots considers countries in isolation. Here a simple model of two emerging markets is developed to illustrate the role of multiple equilibria and contagion in the transmission of crisis. Given simple assumptions, conditions for the existence of multiple equilibria in the first ("home") country are derived which depend on expectations of the future levels of the trade balance, external debt and reserves. Jumps between these equilibria are stochastic, with a simple Markov probability structure. Contagion is defined as an increased probability of moving to a "bad" equilibrium in the "home" country following a crisis in the other country. This definition of contagion would permit the explanation of a move between equilibria to be drawn from microeconomic reasons such as the revision of expectations or the investment behaviour of financial institutions. The model can be extended to demonstrate "monsoonal" and "spillover" effects as well by the introduction of trade interactions.

The paper continues by considering the empirical evidence of recent crisis episodes and suggests extensions to the basic model, in particular, the role of rollover risk, banking sector distortions and risk-averse investors. The model could be useful in constructing early-warning indicators of balance of payment crises and to identify countries that are vulnerable to multiple equilibria. Issues connected with the stochastic nature of jumps and with jump probabilities are subjects for future research.

The first discussant, Axel Weber (Universität Bonn and CEPR), said that the model was simple but rich in possibilities, opening up the possibility of time-series approaches to the study of speculative attack. Another innovative feature was the Bayesian approach used to model the evolution of probabilities. The model would be easy to generalise to three or more countries. More clarity was needed about whether shocks are permanent or transitory in nature, however. Other possibilities for extension included investigation of the parts played by foreign direct investment and the capital account together with intertemporal aspects of the model. Marcus Miller (University of Warwick and CEPR) distinguished between crises which were contemporary in effect and contagion effects and the mechanisms generating them. The appeal of multiple equilibrium models lies in the fact that they can account for crises not anticipated by the market. This model resembles Obstfeld’s (1994) model of a "bad" market, but is more appropriate to emerging market economies. He remarked on the role of interest rates as a crisis trigger and on the relationship with bank-run models. Peter Brandner (Oesterreichische Nationalbank) pointed out that there could be problems in distinguishing between the three kinds of effects defined in the paper. Bad fundamentals are essential to the occurrence of crisis, in addition to monsoonal and spillover effects. The link to the trade balance and issues of interest-rate parity could also usefully be explored. The choice of equilibrium and memory effects on this choice could be of interest: for example, was there asymmetric movement between equilibria. Modelling of transition probabilities could be performed by means of Markov-switching methods.

Marcus Miller (University of Warwick and CEPR) presented "Asset Bubbles, Domino Effects and "Lifeboats": Effects of the East Asian Crisis" written with Hali Edison (Federal Reserve, Washington DC) and Pongsak Luangaram (University of Bristol). This paper began by describing the historical and economic background to the crisis in Korea, Indonesia and Thailand, showing that the crisis was triggered by speculative attack on over-valued currencies and extended into a downward spiral in other financial markets. This paper uses the framework suggested by Kiyotaki and Moore (1997) to analyse the way in which a scramble for liquidity in credit-constrained markets can move a financial system from boom to bust. Following Krugman (1998), the role of financial intermediaries and the price of land and other assets are seen as crucial to the understanding of the Asian crisis. The Kiyotaki-Moore model comprises two sectors: a credit-constrained, highly-leveraged small business sector and an unconstrained big business sector. Small businesses rely on land for collateral against borrowing; big businesses do not allow the total amount of debt to exceed the value of land. The equilibrium in this model is very fragile, so that the small business sector is wiped out when land prices drop fractionally.

This framework is used to examine land allocation and land prices after the bursting of an asset price bubble which originates in under-regulated financial institutions, and after an unanticipated devaluation. The result of both of these phenomena proves to be a downward spiral in land prices.

It may be possible to stabilise this fragile equilibrium by temporary financing designed to ensure the survival of borrowers after the shock and to avoid the "domino effects" arising when the failure of "imprudent" firms can trigger the failure of their "prudent" counterparts. Such domino effects can operate cross-sectorally as well as across frontiers, resulting in a possible channel for contagion.

The policy implications of the model are all concerned with ways of preventing the negative externality of wholesale collapse by keeping threatened firms afloat. The possibilities suggested include temporary financing, allowing "prudent" firms to take over their failing counterparts, transfers or debt write-offs, nationalisation and a temporary freeze or suspension of operation which acts as a "circuit breaker".

Joshua Aizenman said that this was a nice paper but there is an as yet unresolved question as to whether a bubble is really a good way to discuss the issue of a boom-bust cycle. It is really important to resolve this issue. The "black box" of financial intermediation needed to be modelled as well as openness in an economy. The case of Korea, e.g. the existence of chaebols and domestic bank cartels was an interesting example of the interaction between political and economic factors. In general, political economy could offer many useful insights into the understanding and modelling of financial behaviour. Rebecca Emerson (Barclays Capital) agreed that the paper was praiseworthy. She said that very high levels of gearing and overinvestment had been features of the economies studied, but disagreed that the economies concerned were only quasi-open. The model could be extended to make the probability of a bubble bursting endogenously determined, and the ability to borrow could be related to the discounted expected cashflows. Marcus Miller replied by agreeing with the need for conditionality in temporary financing as moral hazard was a particularly severe problem. The role of bubbles in the paper was to demonstrate the effects of a major shock. Indeed, moral hazard may be the actual generating mechanism for the creation or inflation of a bubble. The case of Japan, which has not taken decisive action to support firms wounded by the bursting of a massive asset price bubble, would repay further study.

The chair invited Mike Dooley (University of California, Santa Cruz) to sum up. He began by saying that the first paper by Stijn Claessens et al. summarised situations, such as foreign exchange crises, banking crises or insurance crises, where a government ran out of money. A government which has become insolvent has only one way to respond to crisis: by closing down banks and insolvent financial institutions. There were many, possibly too many, ways of modelling this simple fact.

The absence of clear thinking on this issue, an the failure to develop fundamentals-based models which illuminate it, had led to the growth of a plethora of multiple equilibrium models, of which there are too many, none of which are properly testable, not least because they do not "model" the data. A return to fundamentals-based models really is advisable, partly in order to recheck whether any model exists which will actually fit the data. The modelling challenge now is to try to construct a new generation of "first generation" fundamentals-based models which will meet this test. Multiple equilibrium models may be mathematically interesting. However, they are almost certainly unnecessary.

Taking up the issue of propagation of crisis and policy implications, it is likely that the problem in East Asia is of a similar magnitude to that in the United States in the 1930s. It is also likely to have long-lasting effects. The resolution of this problem cannot be handled by the private sector alone. Higher levels of intervention are required to plug the huge unallocated economic loss within the affected economies. There is no market mechanism in place that can perform this task, as evidenced by the Latin American crisis. Furthermore, the private sector will avoid any attempt to shoulder the burden.

In summary, we must admit that the nature of crisis still eludes us. Governments sink into bankruptcy for any number of reasons. It is not easy to identify culprits or to allocate blame, and it may be unwise to try. We should now move on from asking "what happened" to "what to do" as the crisis is still happening.