In constructing an indicator of financial fragility, the choice of which filter (or transformation) to apply to the data series that appear to trend in sample is often considered a technicality, but in fact turns out to matter a great deal. The fundamental assumption about the likely nature of observed trends in the data, for example, the ratio of credit to GDP, has direct effects on the measured gap or vulnerability. We discuss shortcomings of the most common filters used in the literature and policy circle, and propose a fairly simple and intuitive alternative – the local level filter. To the extent that validation will always be a challenge when the number of observed financial crises (in the US) is small, we conduct a simulation exercise to make the case. We also conduct a cross country analysis to show how qualitatively different the estimated credit gaps were as of 2017, and hence their policy implications in 29 countries. Finally, we construct an indicator of financial fragility for the US economy based on the view that systemic fragility stems mainly from high level of debts (among households and corporations) associated with high valuations for collateral assets (real estate, stocks). An indicator based on the local level filter signals elevated financial fragility in the US financial system currently, whereas the HP filter and the ten-year moving average provide much more benign readings.
About the Authors
Simon Kwan is a senior research advisor in the Economic Research Department of the Federal Reserve Bank of San Francisco. Learn more about Simon Kwan