In fl ation Dynamics and the New Keynesian Phillips Curve: a Bayesian Approach Andrea Carriero ∗ Università Bocconi First draft: October 2005 PRELIMINARY Abstract We propose a Bayesian approach to test the New Keynesian Phillips Curve (NKPC) both as an exact relation and as a noisy/potentially misspecified model. We do so in order to reconcile the contradicting evidence that the NKPC is rejected by a Wald test but nevertheless it provides a good approximation of inflation dy- namics. We apply the proposed approach to US and EMU data. When the NKPC is tested as an exact relation its rejection is confirmed. When modelled as a noisy or potentially misspecified relation the NKPC is supported by the data. Therefore, according to our test the NKPC does not hold exactly, but may hold on average. The degree of noise/misspecification needed to accept the theory is reasonably low. Importantly, all the results of the paper are valid for any admissible value of the parameters measuring the degree of price stickiness and the portion of backward looking firms in the economy. JEL Classification: C11, E31 Keywords: Bayesian VARs, Inflation, Phillips Curve 1 Introduction The New Keynesian Phillips Curve (NKPC) can be thought of as a set of restrictions on a VAR in inflation and marginal costs. When these restrictions are tested with a classical Wald test, they are strongly rejected, both in the US and in the EMU. This happens for any admissible value of the parameters measuring the degree of price stickiness and the ∗ Address: IEP Università Bocconi, Via Gobbi 5, 20136 Milano, Italy. E-mail: andrea.carriero@unibocconi.it. Home- page: www.igier.unibocconi.it/carriero. I thank, without implicating them, Filippo Altissimo, Giacomo Carboni, Stefano Gnocchi, Carlo Favero, and seminar participants at the ECB. 1