Thursday, June 10, 2010

Early warning crisis indicators: truth in orthodoxy?

With the main drama of the world financial crisis behind us policy makers have lent increased urgency to coming up with ways to anticipate and thus prevent future crises. In other words, governments from around the world are looking for improved "early warning indicators" to keep this mess from ever happening again. Because, of course, NO ONE saw this crisis coming, right?

Whatever. The National Bureau of Economic Research has a new working paper out this week taking a fresh look at the early warnings literature. Unlike most papers studying the issue, which focus on particular variables specific to each past crisis in question, this paper approaches the 2008-09 crisis as if it didn't know anything about it before hand--prioritizing the variables most often identified in the past.

The authors, Jeff Frankel and George Saravelos from the Harvard Kennedy School, essentially ask the question: without the benefit of hindsight, would the standard crisis indicators have been able to successfully predict the 2008-09 crisis?

The paper starts with a very thorough review of past empirical studies to identify what indicators have most often been found to be statistically significant crisis predictors. Reviewing over 80 studies, the authors come up with these numbers, presented here courtesy of Maladjusted Charts:

[Number of times found to be statistically significant]
As can be seen above, there's no real shockers in this list as most of the top variables are the usual suspects. Foreign reserves has been found to be statistically significant by 50 different studies, followed by the real exchange rate (48), GDP (25), credit (22) and, of course, the current account (22).

Now, quite naturally, this on its own says more about the state of the economic literature than it does about what variables actually predict crises more accurately. The more orthodox variables have probably been studied for longer and hence more papers have found them significant. Similarly, variables like debt composition and capital flows only began to receive attention in the wake of the Asian crisis and the literature surrounding their role in financial crises has really only picked up in recent years.

So given this conceptual issue, a thorough empirical investigation--using innovative measures of crisis severity and incidence--might give less importance to the orthodox indicators, no? After all, a pattern in recent studies is that traditional indicators like foreign reserves to GDP or real exchange rate overvaluation do not predict crises as powerfully as previously thought.

Well, not so fast the authors of this paper say. After crunching the numbers they find that, without the benefit of hindsight and ignoring the particulars of the 2008-09 crisis, the two variables that predicted it's impact the best were foreign reserves and real exchange rate overvaluation.

Truth in orthodoxy? I wonder... That's really too bad though. I was kinda rooting for debt composition and capital flows.

1 comment: