H-P filter and natural rate of unemployment

[dani]

Hi,
i have some problems in estimating natural rate of unemployment with hp filter.
i've used annual data from 1980 to 2010, and i've got a graph that doesn't show a regular tendency.
the problem is the following: i need unemployment gap to estimate phillips cure with TVC, and i don't know if i should use for each year the difference between the observed rate of unemployment and the corresponding hp forecast for that year, or if i should use the same natural rate for all the sample (for example the mean of hp)..
i'll be very grateful to everyone will help me!
thanks in advance!!
dani

[trubador]

Assuming a constant natural unemployment rate (NUR) will be too restrictive. The ideal solution would be to estimate the natural rate along with the phillips curve. However, if you are planning to build a time varying coefficients model, then you'll end up with nonlinear state equations, which you'll need extended kalman filter to solve. You can try multivariate HP filtering or use another unobserved components model to obtain a more feasible time series of NUR.

Startz' paper, Measuring the NAIRU with Reduced Uncertainty: A Multiple Indicator-Common Cycle Approach, can also provide useful insights.

[dlccmn]

Dear users

I was trying to estimate the NAIRU from the following model

A dynamic Phillips curve

Dp=c(1)*Dp(-1)+c(2)*(-2)+c(3)*(U-N)+c(4)*D(U-N)+e

where
D= first difference
P= inflation
U=unemployment rate
N=Nairu which it is assumed to follow a random walk
e=error term

2) N=N(-1)+v
v=error term


I wrote the SSpace model in the following way

@signal dinf= c(1)*dinf(-1)+c(2)*dinf(-2)+c(3)*(un_rate-nairu)+c(4)*(un_rate-un_rate(-1)-nairu+nairu2)+[var =exp(c(5))]

@state nairu=nairu(-1)+[var =exp(c(6))]

@state nairu2=nairu(-1)

I got very unsatisfying results. Is this due to a very bad specification of the SSPACE moder, or there are other reasons?

Thanks in advance for your help,

Clemente De Lucia