Kalman filter HP equivalence

[jarry]

Hi all,

I am trying to replicate results of HP filter (with lambda=1600) with Kalman filter. Following Boone(2000) I have rewritten the system into state-space form as follows:
(the objective is to estimate NAIRU given the unemployment rate (U))

@ename e1
@ename e3

@state NAIRU=NAIRU(-1)+g(-1)
@state g=g(-1)+e3
@signal U=NAIRU+e1

@evar var(e1)=exp(c(1))
@evar var(e3)=exp(c(1))/1600

However, the results between the two methods differ substantionally (the Kalman follows unemployment series very closely compared to HP filtered series)
Does anyone have an idea whats wrong with the model?

Thanks,
jarry

[trubador]

It is difficult to locate the problem without seeing the actual workfile, but:
1) Make sure that you have extracted smoothed (not filtered) state variables.
2) Make sure that the estimation is actually converged.