Several authors have questioned the evidence claimed by Galí and Gertler (1999) and Galí et al. (2001) that a hybrid version of the new Keynesian Phillips curve approximates European and US inflation dynamics quite well. We re-examine the evidence using the vector autoregressive framework and likelihood based methods, paying particular attention to the stationary and nonstationary, and possibly cointegrated, nature of variables involved. Our results show that the exact as well as the inexact form of the hybrid NKPC are clearly at odds with the European data. On the other hand, Galí and Gertler (1999) finding that the inexact hybrid NKPC is a "good first approximation" to the US inflation dynamics seems less at odds with the data. However, the assumption of the model that the stochastic term forms a sequence of innovations may be problematic as we find indication of autocorrelation in the estimated residuals. The exact form of the hybrid NKPC is firmly rejected by the US data.