|
|
|
|
Early Warning and Early Response, by Susanne Schmeidl and Howard Adelman (eds.)
Dane Rowlands
1
The Norman Paterson School of International Affairs
Carleton University
Abstract
Increased financial integration has been accompanied by greater volatility of capital flows. For many less developed countries (LDCs) this volatility can have serious economic repercussions. The International Monetary Fund (IMF) has been assigned an important role in dealing with the problem through emergency lending and enhanced surveillance duties. Predicting these crises, however, is difficult. The paper begins by outlining the task which the IMF has been directed to undertake, and why “early warning” is essential to fulfilling its mandate. A brief review of the literature on the use of early warning models to predict financial crises is then provided. The paper then provides some preliminary evidence on the relative value of systemic and country-specific analyses in managing IMF resources.
1. Introduction
The Mexican peso crisis in 1994 provides stark evidence of the volatility of international capital flows, and the potential consequences of this volatility for less developed countries (LDCs). 2 In this example the United States led the international effort to stabilize the crisis by creating a sixty billion dollar (U.S.) package to bail out the government of newly-elected Mexican President Zedillo. A large part of this money (roughly one-third) was from the International Monetary Fund (IMF), the international financial institution (IFI) with primary responsibility for short-term stabilization. In this instance, however, the IMF was a follower of the U.S. lead rather than an equal. Furthermore, the assistance provided to Mexico was reactive rather than preemptive. By allowing the crisis to develop, the international financial community contributed to the severe economic hardships which followed. While the Mexican population bore the brunt of this hardship, adverse repercussions were felt in several other emerging markets, particularly in Latin America.
In the wake of the crisis, the IMF’s role as the guardian of international financial stability was expanded. Two new policy recommendations emerged: the creation of a rapid-disbursement facility, and enhanced surveillance (Williamson, 1995). The first policy requires an augmentation of IMF resources as well as an adjustment in lending procedures to speed up administrative processes. Implicit in resource augmentation, however, is the need to manage finances more carefully and with greater forethought. The deployment of the rapid disbursement facility will ease the hardship of countries which suffer from rapid capital flight, and preserve the IMF’s preeminence in financial stabilization.
Enhanced surveillance is designed to prevent crises from developing by keeping the market informed of key economic developments in IMF member states. With timely information, the reaction (or over-reaction) to new data will tend to be less dramatic. In addition, this surveillance is a prerequisite for the management of the IMF’s own resources. The IMF’s failure to predict the number and magnitude of future economic disruptions may result in the use of IMF funds for less deserving cases, leaving it unable to respond adequately to new crises. Alternatively, the risk of having too few resources may cause the IMF to become too parsimonious in its normal activities in order to conserve funds for potential future crises. Thus it is important for the Fund to have an accurate model for predicting crises.
This paper examines the problem of forecasting economic crises and IMF resource use. The next section provides a review of past IMF disbursements and outlines the arguments regarding the IMF’s need for enhanced “early warning” models. Section three then reviews the literature on predicting financial crises and IMF credit disbursements. The subsequent section provides some preliminary empirical evidence on the use of systemic aggregate data as a predictive tool versus the use of country-specific data. Section five presents the conclusions, extensions and policy implications arising from the analysis.
2. Resource Disbursements by the IMF
The IMF was created at the end of the Second World War in an attempt by the victorious powers to restore international financial stability and avoid the crises which characterized the interwar period. Although many of the IMF’s original functions have changed, for example it no longer seeks to manage an international fixed exchange rate system, its primary raison d’etre of promoting a smoothly operating global financial system remains.
Since its inception, the IMF has come to rely more heavily on its provision of “temporary” financial assistance to member countries as a means of avoiding or ameliorating international payments problems. This emphasis on lending has become particularly pronounced since the demise of the fixed exchange rate system in the early 1970s. In essence the IMF assists its member states by providing them with the hard currency they require to finance their balance of payments deficit, though with this money come policy commitments designed to address the external imbalance. In economic terms, the provision of this short-term liquidity theoretically allows a country to maintain imports of consumer and capital goods in the face of a temporary shock, thereby removing the necessity of imposing severe restructuring requirements on the economy. In political terms, IMF financing allows a government to maintain the provision of certain goods and services to its public. In practice, however, countries generally turn to the IMF in the face of longer-term structural problems. In these cases the money provided by the IMF is supposed to ease the transition to a more realistic economic structure. Whether the difficulties are self-inflicted (eg. excessive government intervention or expenditure) or the result of forces beyond the country’s control (eg. adverse terms-of-trade or interest rate shocks), it has become commonplace to turn to the IMF for financial assistance and technical advice whenever an economic crisis strikes.
The assistance provided by the IMF is not free. Borrowers are expected to repay loans with interest (though this rate can sometimes be concessional relative to that charged to the country by the market) and is generally associated with policy conditions which the borrower must fulfil. These policy conditions have been a major irritant in IMF–LDC relations, as they typically entail restrictive fiscal and monetary policy and require the state to minimize its role in the economy. The stress on narrowly defined economic efficiency at the expense of distributional concerns has led many countries to avoid the IMF as long as possible. The IMF’s profile and lending portfolio, therefore, are generally larger during periods of intense international economic crisis.
Members acquire IMF resources in two ways. The first is the traditional method in which a country applies to borrow some portion or multiple of its quota in the form of a hard currency. This method is called a “purchase” in IMF parlance (a repayment is called a “repurchase”), and accounts for over 86% of outstanding IMF credit. The second method is referred to as a more traditional loan, and typically refers to the concessional financing provided under the Structural Adjustment Fund (SAF) or Enhanced Structural Adjustment Fund (ESAF). These loans are targeted towards the poorest members of the IMF.
The IMF has three primary means of financing its activities. The first is the subscriptions (“quotas”) paid to it by its member states. Only a portion of these subscriptions is usable, however, as the majority of a nation’s subscription is paid in the form of its own domestic currency. These currencies are generally not desired by countries unless they are “hard” convertible currencies of the type issued by developed market countries. The quotas paid in the form of hard currencies make up the bulk of the IMF’s usable resources. While periodically revised, quota changes are subject to negotiations between Fund members. Increasing quotas to meet the immediate financing requirements of a crisis, therefore, is not possible.
The IMF’s second means of acquiring resources is to simply increase the issue of its own currency, the Special Drawing Right (SDR). As a mechanism for short-term financing, however, SDR allocations are inadequate. Negotiating an SDR increases is a long and arduous process which makes it unsuitable as a means of generating resources to meet the needs of an immediate or imminent crisis. In recognition of the recent heavy demands on its resources, however, the IMF is currently pursuing these negotiations with its member states.
The third method of increasing its resources is to borrow money from capital markets (though this capability has not been exercised as a means of augmenting its basic lending resources) or from individual states either bilaterally or through the General Arrangements to Borrow (GAB). These arrangements have recently been supplemented with the New Arrangements to Borrow (NAB). The capacity to borrow greatly increases the ability of the IMF to deal with the extraordinary financing requirements of a major crisis, though there are some important restrictions. In 1996, for the first time since the mid-1970s, the IMF has paid back its debts arising from previously borrowed funds.
The GAB was first negotiated in the early 1960s, and was motivated by the realization that the Fund did not have sufficient resources to meet the borrowing needs of its wealthier industrialized members. These states, however, have not borrowed from the IMF since the mid-1970s. The GAB currently allows the IMF to borrow up to 17 billion SDRs 3 from major industrialized countries. A similar agreement commits Saudi Arabia to make available a further 1.5 billion SDRs. The resources available under the GAB, however, are not guaranteed. A country that has made resource commitments under its auspices may refuse to lend to the IMF if it is suffering from balance of payments or liquidity problems. The NAB is similar to the GAB, but is twice as large in terms of resources. (IMF Annual Report 1997: 125).
The IMF can also borrow—from individual country governments or institutions such as the Bank for International Settlements (BIS)—without activating the GAB. It is this method which the Fund has resorted to most frequently; borrowing under the GAB has not occurred since the late 1970s. In times of intense need, therefore, the IMF can call on these countries to provide additional resources, though once again the provision of loans is clearly voluntary, not mandatory. Since 1976, IMF loans have been made primarily to LDCs. In terms of current SDRs, the total credit and loans outstanding have grown from just over one billion SDRs in 1973 to over forty-one billion SDRs in 1995. The pattern of outstanding IMF credit is illustrated in figure 1.
The oil shock in 1973 led to first big crisis which demanded IMF attention. Net IMF credit outstanding rose from one billion SDRs in 1973 to thirteen billion SDRs in 1977. This level declined slowly until the second oil shock in 1979, which led to a new round of net borrowing in 1980. In the early 1980s, the tail end of the oil price increase coincided with the continued increase in interest rates arising out of the contractionary monetary policy in the United States. This policy also led to an appreciation of the U.S. dollar, the primary unit of account for international debt. As a consequence of the conjunction of these unfortunate circumstances, many LDCs were forced into default on their foreign debt. Reliance on the IMF as the key supervising IFI overseeing the debt rescheduling process also meant greater pressure on IMF resources. The total credit outstanding owed to the IMF rose from 9.3 billion SDRs in 1979 to 37.7 billion SDRs in 1984. This level declined gradually until 1990. Credit outstanding started to increase gradually again in 1991, reflecting in part the need for financing by the post-communist economies in transition as well as continuing LDC financial problems arising from the recession in the developed economy markets. The gradual increase became a dramatic rise from 30.3 billion SDRs in 1994 to 41.6 billion SDRs in 1995. The primary cause of this rapid acceleration in lending was the Mexican peso crisis, though a new lending program for Russia was a contributing factor.
Figure 1
Figure 1 clearly demonstrates the volatility of IMF resource use. Predicting the events which account for this volatility represents a challenge for forecasters: none of these events were clearly foreseen. The IMF’s procedures and structure do permit a high degree of adaptability, however, and so they have been able to respond to crises with an expansion of credit. There are three facets of the IMF which are particularly useful in meeting the challenges of an increasingly volatile international financial system. The first is the early warning approach which is built in to the IMF’s Article IV consultations with each member country. Secondly, the IMF can increase resources on short notice through the provisions of the GAB and other borrowing mechanisms. Finally, there is considerable discretion exercised by the Fund with respect to which countries are allowed to borrow, and how much credit they may be offered. In this regard the use of conditionality can be used as a means of rationing credit.
Despite these instruments, however, the IMF frequently expresses concerns over its resource constraints. Recently the Fund has calculated an index of its “liquidity”, i.e. the level of usable cash balances relative to short-term liabilities. Figure 2 illustrates how this liquidity ratio has fluctuated from 1982–1996. Furthermore, there is empirical evidence to suggest that the ease of acquiring IMF credit varies over time (Bird 1994), and is more difficult to obtain during periods of crisis (Rowlands 1994). Such variance suggests that despite its flexibility, the IMF’s ability to carry out its mandate of financing balance of payments difficulties is being compromised by inadequate resources.
Figure 2
4
do provide valuable information on developments in the world economy. The resulting reports on country performance, however, have rarely predicted the crises which have precipitated a high demand for IMF credit. In the 1994 report on Mexico prior to the 1994 peso crisis, for example, the concern expressed over the current account deficit was relatively mild, and overall the report was optimistic in light of the government’s fiscal position (IMF Annual Report 1994).
The provisions of the GAB have occasionally been used to augment IMF resources. It should be noted, however, that the current borrowing arrangements amount to only 18.5 billion SDRs, which is roughly the increase in outstanding credit from 1990–1995, and is only slightly higher than the IMF’s commitment of resources to the financial package put together for Mexico after 1994. Bilateral borrowing associated with the enlarged access to resources amounted to less than 16 billion SDRs. In addition, borrowed resources are costly to the IMF, and there may be a reluctance to use them in order to avoid these costs. Finally, as the recent peso crisis demonstrated, the IMF’s ability to extract solid commitments of funds on short notice is difficult; the commitment of countries to provide $10 (U.S.) through the BIS was not met, and the IMF had to provide the missing funds itself.
The final means of influencing credit, the severity of rationing, is also problematic. While the evidence that conditionality is an effective rationing tool is mixed (Williamson 1982, Bird 1994), there are other means by which the Fund can limit credit. The variability of credit availability is undesirable, as it reinforces the cyclical processes of the international economy. Access to credit should not be tightened during those periods when more countries need it to finance adjustment. The terms of IMF credit need to be relatively constant if it is to fulfil its role as a stabilizing influence on the global monetary system.
Finally, it is arguable that with the increased integration of financial markets, and their increased volatility, the potential for extraordinary demands on IMF resources has been multiplied. The Mexican peso crisis had serious repercussions for other currencies, and the recent currency crises in South-east Asia clearly demonstrate the potential for contagion effects which may bring many countries to the IMF at the same time.
In order to provide credit in a more consistent fashion, therefore, the IMF needs to supplement its inherent flexibility with a forward-looking resource management process. Specifically, the IMF needs a reliable means of predicting those crises which have led to significant changes in resource use in the past. The following section examines the literature on predicting these crises.
3. Literature Review
The debt crisis in the 1980s provided a strong motivation for the development of models which could predict financial crises. The relatively small literature which existed prior to 1982 became quite extensive by the end of the 1980s. Two other literatures also developed which are relevant to this paper: models for predicting which countries receive IMF credit, and models for estimating IMF resource use by countries. A brief review of these literatures is provided here.
Sovereign debt theory is still an area of considerable debate. 5 Despite these theoretical problems, work has proceeded on the empirical analysis of international lending to less developed countries (LDCs). Several early papers explore the factors which contribute to debt servicing difficulties, and are reviewed by Saini and Bates (1984), while Bhatt (1991) provides a later review of the literature. As Saini and Bates point out, commercial banks were the first to construct lists of indicators to assess country risk, which “were often used as an early warning system in conjunction with qualitative country evaluation” (Saini and Bates 1984: 343).
Frank and Cline (1971) started off the more formal statistical approaches to sovereign risk assessment by applying discriminant analysis to cases of debt rescheduling. Debt reschedulings became the dependent variable of choice for the literature that followed, with important contributions being made by Feder and Just (1977), Mayo and Barrett (1977), and Feder, Just and Ross (1981). The contribution of Mayo and Barrett is particularly noteworthy in its attempt to explain debt servicing difficulty five years into the future.
After 1982, more effort was put into statistical analyses of debt financing problems. Important post-debt crisis contributions to the “country risk” literature were made by Edwards (1984), Schmidt (1984), McFadden, Eckaus, Feder, Hajivassiliou, and O’Connell (1985), Hajivassiliou (1987), Taffler and Abassi (1987), Nunnenkampt and Picht (1989), Kutty (1990), Brewer and Rivoli (1990), and Savvides (1991). These papers vary widely in their samples and estimation techniques. Not surprisingly, their main conclusions suggest that variables such as debt-service ratios, foreign reserves-to-imports ratios, per capita GDP, export growth rates, GDP growth rates and other financial variables were all helpful in explaining instances of financial difficulty such as rescheduling of debt. Brewer and Rivoli (1990) deserve mention for their focus on political determinants of country creditworthiness as measured by investment reports. 6
Although considerable insight may be gained from the preceding literature, the question of IMF resource use is only indirectly addressed. The focus on reschedulings, while clearly correlated with IMF involvement institutionally and statistically (Rowlands 1994), does not provide a direct measurement of the demand for IMF resources. Secondly the focus of these studies is to estimate the behaviour of individual countries, not aggregate behaviour. Finally, only a small number of the studies explicitly attempt an early warning approach, the majority of studies being focused instead on simply identifying the factors associated with financial difficulties.
The second strand of literature is much smaller, and deals with the identification of IMF agreement recipients. Contributions to this literature include Joyce (1986), Rowlands (1994), Conway (1994) and Knight and Santaella (1994). The last three are the most extensive papers, using larger samples and more sophisticated techniques to draw inferences on a number of issues. The use of a binary dependent variable to indicate the presence or absence of an IMF agreement in a particular country, however, limits the applicability of these papers in terms of examining IMF resource use in aggregate. Nonetheless, the elements of this approach can be adapted to provide an early warning model of IMF credit demand.
The final literature does examine directly the issue of IMF resource use, though again the focus is on explaining country-specific, rather than aggregate, demand. Bird and Orme (1981), Cornelius (1987), and Bird and Bedford (1992) examine the amount of IMF credit drawn by countries. Bird and Orme (1981) and Cornelius (1987) both deal with rather limited samples, however. Bird and Bedford (1992) uses a more comprehensive data set, but they conclude that the results are rather disappointing: their model explains only 23% of the variance in borrowing from the IMF. In addition, many countries draw on the IMF for assistance when faced with natural diasters (such as droughts, hurricanes, earthquakes and floods) which are inherently difficult to predict (Killick and Malik 1992).
The three research areas identified in the previous literature are only indirectly related to the question of predicting aggregate IMF resource use. Nonetheless they point to three different approaches to the problem. In the first instance, estimates of which countries will face rescheduling or other financial difficulties can be used in conjunction with country size to generate a country-specific estimate of the likelihood and magnitude of financial problems. These measures can then be aggregated across countries to come up with a measure of potential financing needs, which in turn can form the basis of an estimate for IMF resource use. The second approach uses an identical aggregation method, but is based on predictions of IMF agreements rather than reschedulings. The third approach estimates each country’s use of IMF resources directly, and then aggregates over the complete set of countries.
These first three approaches all require the aggregation of country-specific estimates. This paper also uses a fourth method of estimating the demand for IMF resources. This fourth model focuses on the aggregate level of analysis only, and attempts to explain IMF resource use by referring only to global factors which may be contributing to subsequent credit use. The next section presents a preliminary comparison of the results based on these different approaches.
4. Early Warning and IMF Resource Use: Alternative Quantitative Approaches
In order to establish a basis for comparison across the techniques, all four approaches are used the to predict measures of IMF resource use for one, two, and three years into the future. While the lags used here are not as stringent as in some of the early warning models used by commercial banks, IMF lending portfolios are of considerably shorter maturity than private lenders. Most IMF credit agreements have a one-year duration, and repayment periods begin shortly thereafter, so the IMF can modify its resource use fairly substantially in the space of short period of time, at least in comparison with some other lenders.
The first model presented here uses a large panel data set to estimate the cross-country pattern of reschedulings and arrears. The dependent variable in the underlying model is binary, so probit estimations are used. The independent variables in the model include both economic and political factors, and is a slight variant of the one discussed in Rowlands (1994). The predicted probabilities of debt repayment difficulties are then multiplied by the IMF quotas for each country, and added across countries for each year. The result is a “prediction” about the annual demand for IMF resources. This basic model is used to construct these predictions on the basis of data which are lagged by one, two, and three years. Thus there are three sets of predictions generated by the model. These predictions are then used as the explanatory variable for actual IMF resource use in a simple OLS regression.
The second model replicates the process used in the first. The only difference is that instead of estimating debt payment difficulties, the probit estimations are used to predict actual IMF credit agreements. This model is also used to construct three different sets of predictions using different lag structures. Finally, the predicted values are used as the independent variable in a simple OLS regression with actual IMF resource use measures as the dependent variable.
The third model replicates the work of Bird and Orme (1981) by estimating the use of IMF resources directly on a country-by-country basis. The underlying estimation used here is a simple OLS regression 7 using similar explanatory variables as the first two models. The OLS estimates are then taken as predictions using the same one, two, and three year lags in the explanatory variables. Again these predictions are used in a regression on actual resource use.
These first three models roughly correspond to the approaches discussed in the previous section’s review of the literature. The final model uses OLS regressions on aggregate IMF resource use. Since the number of observations is much smaller for this model, only three explanatory variables are used: the average reserves-to-imports ratio for all LDCs, the average world real economic growth rate, and the aggregate current account balance of payments for all LDCs. Although a formal model of the expected relationship is not provided here, intuition clearly indicates that all three variables should be negatively related to the total demand for IMF resources. As before, these independent variables are lagged to come up with equivalent predictions based on information available one, two, and three years earlier.
The predictions and regressions based on these four models are used to explain three different measures of IMF resource use. There are three different dependent variables used at the final stage of the regression. These variables are: total IMF credit outstanding (a stock), the change in total IMF credit outstanding (a flow), and gross new credit issues in a given year (a flow). The estimations are all run using a constant in addition to the other explanatory variable or, in the case of model 4, variables. Furthermore, for the dependent variables which are flows (net changes and new loans) the first difference of the explanatory variables is used. The simple R-squareds for the final stage regressions are presented in Table 1. Although using R-squareds to compare model performance is crude, the procedure serves the purposes of this paper due to the focus on explaining as much variance as possible and comparing the relative “explanatory” capacities of each approach. It should also be noted that the R-squareds over-estimate the “predictive” capacity of the model, since the expected values generated by the model are being compared to within-sample actual values, not out-of-sample values.
Table 1
The results at this preliminary stage of the research are intriguing. Despite its relative lack of sophistication, model 4 (using only three explanatory variables) performs relatively well. The within sample estimates have a higher explanatory power than any of the first three models in five out of the nine cases presented. In only one case is the reported R-squared extremely low (0.02, the next lowest being 0.29), and this is the only case where it has the lowest R-squared of to all the models for a particular cell. Furthermore, in three of the four cases where model 1 outperforms it, model 4 has only marginally lower R-squareds, and is a close second-best. Thus model 4, the systemic approach, does quite well relative to the models which use a country-specific approach, at least according to the simple standards adopted here.
As for the three country-specific models, model 1 dominates the other two approaches in all nine cases. Model 1 uses a probit procedure to estimate the probabilities of a country having repayment problems (arrears or reschedulings). 8 The superiority of this model is unexpected since, being based on a probit estimation of difficulties, it does not have actual IMF behaviour directly reflected in the initial dependent variable. Figures 3–6 present the actual and estimated values for total resource use and changes in total resource use for the two best models: model 1 and model 4.
Figure 3
Figure 4
Figure 5
Figure 6
In terms of the other patterns of the R-squareds, the effects of different lag structures for predictive purposes followed the anticipated pattern: the longer the lag, the lower the variance of IMF resource use explained by the models. Furthermore, all the models had higher R-squared statistics when estimating the total IMF resource use as opposed to changes in resource use or new IMF credits. Given that total resource use is a less volatile series than the other two measures of IMF resource flows, this result is not surprising. In terms of the two flow variables, models 2, 3 and 4 performed better when estimating the change in total IMF resource use than when estimating new credits issued. Model 1, however, was more accurate when new loans were being estimated.
Although these preliminary results are not overwhelming, there are grounds for some optimism. In some cases the amount of variance explained in the model with the two year lag is not much less than in the regressions based on information with a one year lag. Thus there is some evidence that information on IMF resource use is present in data two years prior to the credit actually being extended, a useful feature for early warning. Secondly, the explanatory power of the model based on aggregate resource use (model 4) is generally superior to the other models in explaining the flow variables measuring IMF resource use (change in credit outstanding and gross new loans). This model used only three simple explanatory variables which are chosen without reference to a formal theoretical model. By constructing a more rigorous foundation, this last model may have room for considerable improvement.
Other extensions to the present work may also prove to be valuable. Recent work has emphasized the forecasting of currency crises a literature that may be usefully integrated into the current work. Currency crises are also grounded in more formal theoretical modelling, and such an approach will add to the rigour and robustness of attempts to forecast IMF resource use. Finally, the current paper relies primarily on econometric forecasting rather than the more traditional early warning methods used in political science. By combining theory and empirical forecasting with more traditional early warning approaches, however, may provide the IMF with even better guidance about future resource demands.
5. Conclusions and extensions
This paper has attempted to motivate the need for early warning techniques to be applied by international financial institutions in order to manage their resources. The IMF is a prime candidate in this regard, as it is frequently called upon in times of economic crisis to make large quantities of money available on short notice. While the IMF obviously does make projections on resource use, the measures which currently give it flexibility have proven insufficient in terms of maintaining stable access to its resources in a manner consistent with the promotion of global financial stability.
The preliminary empirical results presented in the paper provide some basis for optimism in predicting IMF resource use. The two approaches used were to predict the demand for IMF credit for each country and then aggregate, and explaining directly the aggregate resource use. The aggregated approach of model 4 performed quite well relative to the models based on country-specific approaches. The implications of this result are potentially very important: economic problems may not be country-specific events, but may instead be manifestations of aggregate international economic disequilibrium.
The evidence also suggests that it may be possible to use basic international data to predict IMF resource two years in the future with some degree of reliability. There is considerable room for refining all of the models used, which may lead to improvements in the capacity of each model to make predictions. In all of the models there has been insufficient theoretical work to guide the structure of the estimating equations. Thus there is a need for the theoretical foundations to be developed. Furthermore, there is ample scope for modifying both the selection of independent variables as well as the deployment of more sophisticated estimation techniques in order to better capture the nuances of the data. One of the complications which clearly needs to be addressed is the potential bias introduced by the use of realized IMF credit use as the final stage dependent variable. Presumably this level is biased due to the fact that the IMF may be attempting to either expand or contract the use of resources depending on its own liquidity (Williamson 1982, Rowlands 1994).
In the end, however, the real need for the IFIs is to be able to predict the episodes of dramatic demand increase. Recent examples of these episodes have been the oil shocks, the debt crisis, and the peso crisis. The early warning—perhaps more accurately, forecasting—approaches used in this preliminary investigation have been calibrated to fit the smaller variations in IMF resource use as well. A more explicit focus on the dramatic crises may, in the end, prove a more useful direction for future inquiry. Until these periodic crises can be explained, the IMF and other key actors in the international financial system will have to rely on ex post crisis management rather than ex ante preemptive intervention.
References
Bhatt, P.R. (1991): “Determinants of International Bank Lending to Developing Countries: A Selected Survey of the Literature,” The Indian Economic Journal 38: 107–121.
Bird, Graham (1994): “The Myths and Realities of IMF Lending,” The World Economy 17(5): 759–775.
Bird, G. and D. Bedford (1993): “Why do countries borrow from the IMF? An analysis of the empirical evidence”. Department of Economics, University of Surrey Working Paper.
Bird, G., and T. Orme (1981): “An Analysis of Drawings on the International Monetary Fund by Developing Countries,” World Development 9: 563–568.
Conway, P. (1994): “IMF Lending Programs: Participation and Impact,” Journal of Development Economics 45: 365–391.
Cornelius, P., 1987. “The Demand for IMF credits by Sub-Saharan African Countries”, Economic Letters 23, 99–102.
Edwards, S. (1984): “LDC Foreign Borrowing and Default Risk: An Empirical Investigation, 1976–80,” American Economic Review 74: 726–734.
Feder, G. and R. Just (1977): “A Study of Debt Servicing Capacity Applying Logit Analysis,” Journal of Development Economics 4: 25–38.
Feder, G., R. Just and K. Ross (1981): “Projecting Debt Service Capacity of Developing Countries,” Journal of Financial and Quantitative Analysis 16: 651–669.
Frank, C. and W.R. Cline (1971): “Measurement of Debt Servicing Capacity: An Application of Discriminant Analysis,” Journal of International Economics 1: 327–344.
Hajivassiliou, V.A. (1987): “The External Debt Repayment Problems of LDC’S: An Econometric Model Based on Panel Data,” Journal of Econometrics 36: 205–230.
Joyce, J. (1992): “The Economic Characteristics of IMF Program Countries,” Economic Letters 38: 237–242.
Killick, T. and M. Malik (1992): “Country Experiences with IMF Programmes in the 1980s”, The World Economy 15(5): 599–632.
Knight, M. and J. Santaella (1994): “Economic Determinants of Fund Financial Arrangements,” IMF Working Paper WP/94/36.
Kutty, G. (1990): “Logistic Regression and Probability of Default of Developing Countries Debt,” Applied Economics 21: 1649–1660.
Mayo, A.L. and A.G. Barrett (1978): “An Early-Warning Model for Assessing Developing-Country Risk,” in S.H. Goodman, ed.: Financing and Risk in Developing Countries (New York: Praeger) 81–87.
McFadden, D., R. Eckaus, G. Feder, V. Hajivassiliou, and S. O’Connell (1985): “Is there life after debt: An econometric analysis of the creditworthiness of developing countries,” in G.W. Smith and J.T Cuddington, eds. International Debt and the Developing Countries (Washington D.C.: The World Bank) 179–209.
Nunnenkampt, P. and H. Picht (1989): “Willful Default by Developing Countries in the 1980s: A Cross Country Analysis of Major Determinants,” Weltwirtschaftliches Archiv 125: 681–702.
“The perils of prediction,” The Economist, August 1, 1998: 61–62.
Rowlands, Dane (1993): “Constitutional Rules, Reputation, and Sovereign Debt”, Journal of International Economics 35(3/4): 335–350.
Rowlands, Dane (1994): “Political and Economic Determinants of IMF Conditional Credit Agreements, 1973–1989”, The Norman Paterson School of International Affairs Development Studies Working Paper 6.
Saini, K.G. and P.B. Bates (1984): “A Survey of the Quantitative Approaches to Country Risk Analysis,” Journal of Banking and Finance 8: 341–356.
Savvides, A. (1991): “LDC Creditworthiness and Foreign Capital Inflows: 1980–86,” Journal of Development Economics 34: 309–327.
Schmidt, R. (1984): “Early Warning of Debt Rescheduling” Journal of Banking and Finance 8: 357–369.
Taffler, R.J. and B. Abassi (1987): “Country Risk: A Model For Predicting Debt Servicing Problems in Developing Countries,” in Z. Res and S. Motamen, eds.: International Debt and Central banking in the 1980s (New York: St. Martin’s Press) 191–223.
Williamson, John (1982): The Lending Policies of the International Monetary Fund (Washington: Institute for International Economics).
Williamson, John (1995): “Reform of the International Financial Institutions,” Canadian Foreign Policy III(1): 15–22.
Endotes
Note 1: The author would like to thank Carleton University and the Social Sciences and Humanities Research Council for their financial support through a GR5 research grant. Thanks are also due to the organizers and participants of the “Synergy in Early Warning” conference organized by the Prevention/Early Warning Unit of York University’s Centre for International and Strategic Studies. Back.
Note 2: This paper was written in the Spring before the Asian financial crisis in 1997. The commitments made by the IMF in the wake of this new crisis, coupled with the continuing assistance for Russia, dramatically highlight the importance of the issues discussed in this paper. Back.
Note 3: Currently, 1 SDR is approximately equal to $1.50 U.S. Back.
Note 4: These Executive Board reviews were recently increased in frequency in response to the Mexican peso crisis (IMF Annual Report 1996). Back.
Note 5: Rowlands (1993) provides a review of some of the theoretical issues. Back.
Note 6: There is a growing literature on predicting currency crises, events with a clear relationship to financial crises generally. According to a recent article (“The perils of prediction,” The Economist, August 1, 1998: 61), a forthcoming IMF working paper by Berg and Patillo reviews three models in this literature. They conclude that none would have been useful in predicting the current crises that have emerged in Asia. Back.
Note 7: Clearly the use of a TOBIT estimation procedure would be preferable. However, distilling the predicted values from a TOBIT model is computationally difficult. Furthermore, for the purposes of this paper, a continuous variable approach to the estimation may be preferable due to the need to measure total resource demand, not just resource demand above a certain threshold. Back.
Note 8: It should be noted that there were preliminary estimations which did not use first differences of the explanatory variables in the prediction of the dependent variables which were flows. In these cases, the model based on estimating actual IMF resource use on a country-by-country basis—model 3—actually had better results than model 1 in the two and three year lagged estimations of changes in total IMF resource use. Theory suggests, however, that the first difference approach is the most appropriate for estimating the flow, however, so these unusual (though encouraging) results were rejected. Back.
Email this citation