This paper attempts to test the natural rate hypothesis by leaking at
historical unemployment data for France, the UK and the US. We find th
at the unemployment series can be described as stationary around an in
frequently changing mean. Moreover, that the speed of convergence towa
rds mean unemployment is slower when unemployment is high and differs
across the three countries: the two European countries having more per
sistence. We do not find any evidence that the persistence of a change
in unemployment is a function of the size of the change. We conclude
that the data are consistent with multiple equilibria models where lar
ge shocks bring the economy from one equilibrium to another, and also
with models with a moving natural rate. (C) 1997 Elsevier Science B.V.