Improved output gaps with financial cycle information? An application to G7 countries’ new Keynesian Phillips curves

ABSTRACT

In search for more robust cyclical imbalance indicators, recent research has highlighted the interactions between business and financial cycles. Output gap formulations increasingly take imbalances of the financial cycle into account, postulating finance-neutral output gaps (FNGAPs). To test their increased explanatory power in econometric models, we compare FNGAPs to univariate output gaps in their ability to explain inflation dynamics in hybrid new Keynesian Phillips curves. Results indicate FNGAPs to exercise (dis)inflationary pressure, but not to outperform traditional output gaps. Nonetheless, they have become increasingly significant in the course of the 2007/08 Global Financial Crisis.

KEYWORDS:

• Finance-neutral output gap
• financial cycle
• Phillips curve
• inflation dynamics

JEL CLASSIFICATION:

• E31
• E32
• E44

Acknowledgements

We would like to thank Mikael Juselius, Claudio Borio, and Piti Disyatat for providing us with details on their derivation of finance-neutral potential output.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

¹ In order to avoid cyclical means, financial cycle variables are demeaned using Cesàro means.

² Instrument lags between one and four are allowed depending on correlation statistics. Detailed variable descriptions and sources are available in Appendix 1.

³ The dynamic observation equation includes an autoregressive component of the output gap to replace the assumption of a purely white noise relationship. Variances of the state and observation equation are set to align cycle frequencies with that of the univariate HP filter. Lags k_rp;cr;r between zero and four depend on correlation statistics with the endogenous variable.

⁴ As coefficients are not always statistically different across countries, results provide an indication on the degree of forward- and backward-looking price setting and the role of the driving variable. Overall, significant cross-country differences exist and country-specific inflation dynamics are at the core of Phillips curve estimations so that alternative estimations, e.g. in a panel, are not taken into consideration.

⁵ To complement weak instrument tests, underidentification tests (Kleibergen and Paap 2006) are performed for both specifications. Kleibergen–Paap rk LM test statistics indicate that underidentification can formally not always be rejected. Nonetheless, p-values in the range of 0.05 to 0.10 demand a careful interpretation of results.

⁶ Setting the time break slightly earlier, e.g. in 2005, and repeating analyses with HP-filtered gaps yields similar results. Estimation results for both tests are available upon request.