Since the mid-1980s, Phillips curve forecasts of US inflation have been inferior to those of a conventional causal autoregression. However, little change in forecast accuracy is detected against the benchmark of a noncausal autoregression, more accurately characterizing US inflation dynamics.