On the long-run calibration of the credit-to-GDP gap as a banking crisis predictor

Abstract

The credit-to-GDP gap is a widely used early warning indicator of banking crises. It has become standard to calculate this trend deviation with a one-sided Hodrick-Prescott filter that uses a much larger value for the smoothing parameter λ than commonly applied in most business-cycle studies. We recalibrate the smoothing parameter with panel data covering almost one-and-a-half centuries of data. As a result, the 2008 crisis does not dominate the results and sample length helps contain filter initialization problems, i.e. most observations are preceded by decades of data. The optimal λ is found to be much lower than previously suggested.