Publication

This paper addresses the discrepancy between the theoretical requirements and actual practice in creating Phillips curves. In order to derive a Phillips curve from a DSGE model, steady states for various variables need to be estimated rather than just held constant. The authors argue that the natural definition of the steady state is the long-horizon forecast and estimate these permanent components from a co-integrating VAR that takes account of global interactions. They estimate Phillips curves using deviations from the steady states on US data and argue that this is both consistent with the theory and uses the relevant information about steady states.