Why you should use the Hodrick-Prescott filter - at least to generate credit gaps

In a follow-up paper we show that when projection gaps are estimated using a panel with the same coefficients imposed for all economies, they perform marginally better than the credit-to-GDP gap in predicting financial crises. 

BIS Working Papers  |  No 744  | 
17 September 2018



Excessive credit growth is part and parcel of any financial boom and bust. But what is "excessive"? There is no clear cut answer to this question, so researchers use proxies. Typically, credit is divided, or "normalised", by some variable such as GDP so that it can be compared across time and between countries. The gap between the level of normalised credit and some "trend" level is then used as a proxy for excessive credit growth. 

The BIS was the first to develop a credit-to-GDP gap measure and show that this is a useful crisis early warning indicator. As such it was adopted by the Basel Committee on Banking Supervision as a guide for setting countercyclical capital buffers. It involves normalising credit by GDP and measuring the trend using a one-sided Hodrick-Prescott (HP) filter. 

However, this measure has been criticised because the HP filter has some undesirable statistical properties. Also, credit and GDP may move together. From a conceptual perspective, we agree with these criticisms. But, in the absence of clear theoretical foundations, any proposed gap measure should be treated only as a proxy.


We run a horse race between the baseline gap and alternatives that have been suggested to fix these problems. In particular, we assess how well the different gap measures predict financial crises up to three years ahead in a panel of 42 economies between 1970 and 2017. In addition to normalising credit with GDP, we also consider normalisations based on population. And we look at trends based not only on the HP filter but also based on five-year growth rates and local projections.


We find that no other gap measure outperforms the baseline credit-to-GDP gap. Across many forecast horizons and subsample specifications, the baseline gap is either the best or very close to (and not statistically significantly different from) the best predictor of crises using a standard measure to judge performance. Some other gaps do almost as well. But gaps based on linear projections perform poorly.



The credit gap, defined as the deviation of the credit-to-GPD ratio from a Hodrick-Prescott (HP) filtered trend, is a powerful early warning indicator for predicting crises. Basel III therefore suggests that policymakers should use it as part of their countercyclical capital buffer frameworks. Hamilton (2017), however, argues that you should never use an HP filter as it results in spurious dynamics, has end-point problems and its typical implementation is at odds with its statistical foundations. Instead he proposes the use of linear projections. Some have also criticised the normalisation by GDP, since gaps will be negatively correlated with output. We agree with these criticisms. Yet, in the absence of clear theoretical foundations, all proposed gaps are but indicators. It is therefore an empirical question which measure performs best as an early warning indicator for crises - the question we address in this paper. We run a horse race using quarterly data from 1970 to 2017 for 42 economies. We find that no other gap outperforms the baseline credit-to-GDP gap. By contrast, credit gaps based on linear projections in real time perform poorly.

JEL classification: E44, G01

Keywords: early warning indicators, credit gaps, HP filter