Financial crises, macroprudential policy and the reliability of credit-to-GDP gaps

The Basel III regulation explicitly prescribes the use of Hodrick-Prescott filters to estimate credit cycles and calibrate countercyclical capital buffers. However, the filter has been found to suffer from large ex-post revisions, raising concerns on its fitness for policy use. To investigate this problem we study credit cycles in a panel of 26 countries between 1971 and 2018. We reach two conclusions. The bad news is that the limitations of the one-side HP filter are serious and pervasive. The good news is that they can be easily mitigated. The filtering errors are persistent and hence predictable. This can be exploited to construct real-time estimates of the cycle that are less subject to ex-post revisions, forecast financial crises more reliably, and stimulate the build-up of bank capital before a crisis.