EVALUATING THE USEFULNESS OF EARLY WARNING INDICATORS FOR BANKING CRISIS: A PANEL VAR APPROACH

Abstract

In this paper, I examine the usefulness of various leading indicators in the prediction of banking crises in advanced economies. Using a panel vector autoregression methodology, I explore the dynamics between each leading indicator and the real costs of banking crises. The real costs include output and unemployment losses, fall in industrial production, and government deficits. My results based on the impulse-response analysis show that credit indicators, together with equity price growth and housing price growth, are useful in explaining real cost dynamics during a banking crisis. This highlights a set of relevant indicators policymakers can work with in predicting banking system failure. As an illustrative exercise, I formally test the early warning characteristics of the variables identified by applying a dynamic probit model on two specific countries. The same variables were again found useful in making probabilistic predictions on banking crises under a four quarters ahead early warning horizon.