Journal Issue

IMF Economic Forum: Early warning systems: fad or reality?

International Monetary Fund. External Relations Dept.
Published Date:
January 2001
  • ShareShare
Show Summary Details

Since the eruption of financial crises in Mexico and Asia in the mid- to late 1990s, there has been a growing demand within the international community for a system that could help predict the onset of such crises. But how realistic is this goal? Do existing models have a good track record? Even if we develop a system that looks as if it would have been successful in the past, can we be confident that it will work in the future?

To tackle these questions, the IMF hosted an Economic Forum on early warning systems on November 1. The panelists were Peter Garber, Global Strategist with Deutsche Bank; Kristin Forbes, Professor, Massachusetts Institute of Technology; and Eduardo Borensztein, Division Chief in the IMF’s Research Department. Carmen Reinhart, Senior Policy Advisor in the IMF’s Research Department, served as moderator for the panel.

The general consensus among participants was that, although such systems still need to be further refined, they can provide a useful starting point for anticipating the onset of financial crises. However, some speakers were more enthusiastic than others.

Crystal ball?

Garber opened the discussion with a perspective from the private sector, where early warning systems offer the hope of higher profits and lower risks. He said investment banks are busily devising in-house models that attempt to predict currency crises to help clients craft effective foreign currency trading strategies or assess values and risks in emerging market currencies.

How are these models doing? He said that one of the newer economic models—Deutsche Bank’s “Alarm Clock” (DBAC)—estimates simultaneously the probabilities of exchange rate and local interest rate “events” in 19 emerging markets on a monthly basis. The DBAC defines separate exchange rate and interest rate events as depreciations greater than a certain size (estimated separately for levels ranging from 5 to 25 percent) and increases in the money market interest rates of more than 25 percent in a month. The methodology for the model jointly estimates the probability of these two types of events, allowing the probability of a simultaneous increase in interest rates to influence the likelihood of an exchange rate crisis and the probability of a depreciation to affect the predictions of an interest rate crisis. Summarizing the results of some 360 observations over a 20-month period, he said the model’s overall performance was “reasonably good.”

Good for what?

But Forbes, taking an academic perspective, had some reservations about the work being done on early warning systems. First, she asked, were the findings of the early warning systems consistent? To answer this question, she examined three private sector models, including the DBAC, covering predictions for 16 countries, including the least and most vulnerable ones, over a 10-month period (January–October 2001).

The result: a troubling absence of consistency in the findings. Using one measure, for example, she found that while there was, within each model, some consistency in prediction, there was virtually complete disagreement across the three models for any given month. “They are not sending a consistent message,” said Forbes, noting that the alarm clock did not always ring at the same time.

Second, she asked, did the models predict relative vulnerabilities for given countries? Here, the answer was more positive. One might have assumed that vulnerability had increased over this period. But in fact, the reverse occurred in a number of potentially vulnerable economies, including Argentina, Brazil, Mexico, and Taiwan Province of China, and the models captured this. The reason, she said, was that the models correctly predicted a series of substantial currency devaluations that subsequently took place.

“For what they were designed to do—predict currency depreciations—these models are not bad,” she acknowledged. But was predicting exchange rate movements the most important function of early warning systems? Not in her opinion. Proponents of such systems, she suggested, should direct their energies instead to addressing more pressing, policy-laden issues, such as external financing difficulties and financial system vulnerabilities.

Ultimate value

For the IMF’s Borensztein, however, the advantages of these models were already compelling for a number of reasons.

  • They are objective and mechanical and thus avoid the biases that may sometimes cloud analysts’ views. For example, the impressive economic growth record of Korea over the previous decades led many observers to overlook the country’s short-term external vulnerability in 1997.

  • The models are capable of processing several vulnerability indicators (such as the short-term ratio of debt to reserves and the current account balance) into a single figure that measures the probability of crises over some future time period. Watching a large number of indicators may prove inconclusive when they are not all moving in the same direction.

While far from perfect, the models seem far superior to some of the alternative indicators often used by markets and analysts, including bond spreads and sovereign rates issued by the major agencies. Despite many false alarms, the models helped anticipate potentially dangerous pressures at work in foreign exchange markets that did not develop into a full-blown crisis, thanks to appropriate policy responses or good luck.

Thus, while the use of early warning systems has a compelling logic for both policymakers and the private sector, research on crisis definition and model specification and more experience in “real time” application of the models are needed to fully develop this tool for crisis prevention.

Photo credits: Royal Canadian Mounted Police, page 341; Denio Zara, Padraic Hughes, Pedro Marquez, and Michael Spilotro for the IMF, pages 341, 343, 345–47, 350, 354, and 356; and Tony Ranze for AFP, page 355.

Other Resources Citing This Publication