The Canadian Phillips Curve and Regime Shifting

Phillips curves are generally estimated under the assumption of linearity and parameter constancy. Linear models of inflation, however, have recently been criticized for their poor forecasting performance. The author investigates the linearity and constancy assumptions of a standard reduced-form Phillips curve for Canada using two different techniques: (i) the methodology proposed by Bai and Perron (1998), which allows for an unknown number of breaks at unknown dates, and (ii) a three-regimes Markov-switching regression model. Both methodologies strongly reject the linearity and parameter constancy assumptions. The author finds that the outputinflation relationship does not hold under the current monetary policy of inflation targeting, with its low and stable inflation. Since the inflation-control targets were adopted, inflation expectations appear to be more forward looking and well anchored at 2 per cent, the official target rate. Core inflation exhibits very low persistence and there do not appear to be significant asymmetries in the inflation response to output-gap shocks within regimes. Generalized impulse responses are computed to illustrate some properties of the Markov-switching Phillips curve model. JEL classification: C52, E31 Bank classification: Econometric and statistical methods; Inflation and prices