The existence of a long-run trade-off between unemployment and inflation (the Phillips curve) was rebutted by Friedman and by Phelps in the late 1960s, paving the way for the `rational expectations revolution' in macroeconomics and the `Lucas critique' of the use of econometric models for policy evaluation. Lucas argued that optimal decision rules vary systematically with changes in the structure of variables relevant to optimizing decision-makers, so that any policy change will alter the structure of econometric models on which the policy change may have been based.
Yet the Lucas critique has largely been ignored by the majority of applied econometricians because of a lack of empirical evidence. In Discussion Paper No. 321, Research Fellow George Alogoskoufis and Ron Smith attempt such an empirical evaluation, using a version of the Phillips curve itself. Using historical time series for the United Kingdom over 1857-1987 and the United States over 1892-1987, Alogoskoufis and Smith show how shifts in the process generating consumer price inflation have generated shifts in parameters of estimated, expectations-augmented wage equations.
The authors find a dramatic upward shift in the persistence of consumer price inflation in both countries around World War I, very quickly reflected in the coefficient of lagged consumer price inflation in the Phillips curve. The estimated degree of persistence of consumer price inflation shifts from about 30% to about 70% for the United Kingdom, and from virtually zero to about 60% in the United States. The coefficient of lagged inflation in the Phillips curve shifts from 30% to 90% for the United Kingdom. For postwar data, such shifts also occur in the late 1960s. The corresponding shifts in the persistence of inflation are from about 30% to about 70% for both the United Kingdom and the United States, again accompanied by a shift in the coefficient of lagged price inflation in the Phillips curve.
The most likely hypothesis to explain these shifts, Alogoskoufis and Smith argue, is that they reflect the effects of changes in monetary regime on the persistence of inflation. The classical gold standard was abandoned at the start of World War I; attempts to restore it in the 1920s failed, and the gold exchange standard that prevailed for part of the 1920s was not the same regime. In the postwar period the discipline of the early years was provided by the Bretton Woods system, and the shift the authors find appears when the system began to break down at the end of the 1960s.
Alogoskoufis and Smith then assess the robustness of their results by examining instead the insider-outsider model of labour market behaviour, which predicts that lagged changes in unemployment, and not its level, should affect wage inflation. Estimates of this alternative model indicate that for the United Kingdom the fit of the equation is no worse than the Phillips curve; for the United States, however, the fit of this model is worse. As with the more traditional Phillips model, the results reveal large shifts in the coefficient of lagged inflation after 1914, and similar shifts after 1968

The Phillips Curve and the Lucas Critique: Some Historical Evidence
George Alogoskoufis and Ron Smith

Discussion Paper No. 321, June 1989 (IM/AM)