Many alternative measures of core, or underlying, inflation have been proposed that are based on stripping out some unwanted or excessively volatile elements from the headline rate. A potential drawback of such measures is that they are necessarily atheoretic - based largely on purely statistical procedures. This paper proposes an alternative method of measuring core inflation utilising an explicit economic definition. It defines core inflation as that part of measured inflation that has no medium or long term impact on real output - a notion that is consistent with the vertical long-run Phillips curve. This definition captures the commonly held view that moderate movements in inflation can have no impact on the real economy once financial and wage contracts have been written taking it into account. Using this definition the paper estimates a measure of core inflation using the VAR identification technique developed by Blanchard and Quah. The estimated measure indicates that core inflation was higher than measured inflation in the early 80s suggesting that measured inflation was depressed by beneficial supply shocks. The opposite effect occurred in the late 80s. Currently, core inflation is above measured inflation.