Poverty and Development Division
last updated : 27 April 2000
The provision of an early warning of a crisis is really the product of monitoring. One can make two conclusions after monitoring: either everything is on track or some things are going wrong. Once significant deviations from the desirable trend of a set of indicators are identified policy makers are alerted to the possibility of crisis. Early warning systems are models that attempt to predict systematically which countries are more likely to suffer from crises. Among these, models designed to predict currency crises are common (see box IV.1 for typologies of crises). The advent of the Asian crisis has stimulated work in this field further. The post-Asian crisis research can be mapped into the following two areas: the development of leading indicators or early warning systems to predict the onset of a crisis; and policy measures that can be taken to avoid similar problems in the future.
Here the concern is primarily with the first, including reviewing the various early warning systems that have been proposed. The second is of interest insofar as it can guide the construction of leading indicators. However, it should be noted that having an early warning system would be of great advantage if it were complemented by a reaction kit of policy response options whenever bad signals were flashing. If the authorities are required, on the basis of warning signals, to consider actions based on a set of policy options (similar to the prompt corrective action rules developed in the United States), then a speculative attack may be averted. It should be recalled that speculative attacks on a currency are significantly affected by uncertainty over the timing and quality of policy responses to disturbances. A reaction kit of policy options may well serve as a clearly articulated contingency plan for governments that could improve the confidence of market participants.
A fundamental issue confronting research in this area is whether crises are, by nature, predictable. If a crisis implies a correction of an unsustainable economic trend, presumably generated by an inconsistent set of policies, then crises should be predictable. This corresponds to the first case in the Radelet-Sachs classification (see box IV.1). To the extent that the monitoring or regulation of banks can spot problematic practices, moral hazard-based crises can also be predicted. It is also contended that some crises may be inherently unpredictable. This would be so in the case of contagion; Radelet and Sachs argue that the Asian crisis belongs more to this category. In their view, the macroeconomic fundamentals of most if not all of the affected countries did not justify the severity of the crisis.
Apart from such theoretical considerations, there are other issues that impinge on the practical side of predicting currency or other types of crises. First, there are doubts that crises are sufficiently similar across countries and over time to allow generalizations from past experience. There is also the question of whether adequate data on the signals of crisis are available, on time, with adequate accuracy and in a way that is interpretable by policy makers. There are difficult measurement problems for some vital indicators. For instance, much of the literature on the Asian crisis has identified underdeveloped banking supervision as a major factor in increasing the vulnerability of the banking system to a crisis. This variable, however, is difficult to quantify, and qualitative assessments may vary significantly. Lastly, even if data were available, the presence of "noise" may not guarantee that a correct interpretation is possible.
A second related issue is the difficulty of constructing an early warning system based on historical data to warn of a crisis that has origins in something new, that is, not covered by previous theories of crises. The world of finance is known for its rapid innovation and increasing complexity of instruments and so this situation is likely to arise. It is also true that policy makers, civil servants and technocrats tend to lag in their understanding of the innovations and their impact on policy.
Still another issue is how an early warning system could cope with a major structural change. For instance, currency crises erupt when an exchange rate peg is deemed unsustainable. In an era of freely floating exchange rates, would there be currency crises? If so, what is the transmission mechanism? Early warning systems constructed using data from a time when fixed exchange rates prevailed may not be useful or meaningful in predicting a crisis under a floating exchange rate regime.
If an early warning system was accurate and therefore credible, and authorities always acted on the basis of the warning, then policy action might avert a crisis. This is the intended effect. However, because of the action, technically the warning raised earlier becomes a false alarm as no crisis materializes - and the system may be seen as perverse. Thus an automatic reaction function of policy to crisis signals may ironically limit the predictability of crises.9
If the markets deem the early warning system to be credible, then a signal of an impending crisis may in fact also precipitate or trigger a crisis. This possibility, which works through the formation of expectations of the market players, creates a dilemma, and not just a conceptual problem, as market expectations have a tendency to overshoot and so may influence the economic situation of countries unnecessarily. Therefore, does having more information and warnings create crises, or lessen their severity? Does hiding information from the market avert crises or only postpone the day of reckoning and deepen the result?
In any case, if indeed leading indicators can be found, or early warning systems constructed, efforts in this direction are worthwhile in view of the large costs of a crisis. However, to anticipate a crisis, access to timely and accurate information and data is critical. Policy makers need to know the status of and changes in key economic variables. Surveillance, and the monitoring of key economic variables, are thus an important function of policy-making. Moreover, in a more open global economy, with greater economic links between countries via trade, finance and investment, this function needs to be fulfilled on the international level and not just nationally.
Andrew J. Filardo, "How reliable are recession prediction models?", Federal Reserve Bank of Kansas City Economic Review, vol. 84, No. 2 (Second Quarter 1999), pp. 35-56, available at
Please contact the webmaster with questions or comments about this web site.
For any queries concerning the substantive content of the page, please contact PDD homepage.