Financial cycles: how long and how certain?

Abstract

Understanding and measuring financial cycles is important to academics and policy-makers, particularly those in charge of macroprudential policy. While business cycle periodic components have been largely estimated with different methods, empirical evidence related to financial cycles’ periodicities is scarce. Worse, ad hoc filters, are commonly used by policy-makers to calculate the credit gap and make important policy decisions based on the amplitude of this gap. To address the issue of depicting financial cycles, we estimate the credit gap and credit periodic components in 28 countries using Bayesian Structural Time Series Methods (STM) and Singular Spectrum Analysis (SSA). We find financial cycles of 13 to 20 years in most countries, but some present periodicities close to those of the business cycle. We also find discrepancies between our estimated cycles and the ad hoc filter calibration usually adopted by policy-makers to measure the credit gap, which conceal uncertainty and can be misleading.