The standard derivation of a Phillips curve from a DSGE model requires that all variables are measured as deviations from their steady states. But in practice this is not done. The steady state for output is estimated by some statistical procedure, such as the HP filter, and the steady state for other variables, including inflation, is treated as a constant. This is inconsistent with the theory and raises econometric problems since inflation, for instance, is a very persistent series. We argue that the natural definition of the steady state is the long-horizon forecast and estimate these permanent components from a cointegrating VAR that takes account of global interactions. This estimate of the steady state will reflect any long-run theoretical relationships embodied in the cointegrating vectors. We then estimate Phillips Curves and other standard monetary transmission equations using deviations from the steady states on US data. This is both consistent with the theory and uses the relevant economic information about steady states.