Several specifications of state-space models are used to obtain estimates o
f the NAIRU for the G7 except Japan. plus Australia, over the past 28 years
. A Phillips curve-type regression is shown to deliver estimates that do no
t mimic low-frequency movements in unemployment rates, even when a drift is
included in the specification of the NAIRU. Standard errors around the est
imates are extremely large. Using information about the behavior of unemplo
yment, in addition to inflation, alleviates both these shortcomings.