Estimating State Space Model with exogenous variables

[purdued]

Hi all,

I'm trying to estimate the following specification:

y(t) = trend + cycle
trend(t) = 2*trend(t-1) - trend(t-2) + e_t(t)
cycle = C(2)*cycle(-1) + c(3)*x(t) + c(4)*z(t-1) + e_c(t)

I'm essentially breaking gdp into trend and cycle with the cycle being described by financial variables x (house prices) and z (interest rate). I have tried to formulate that specification in Eviews as follows:

@signal gdp = trend + cycle
@state trend=2*trend(-1)-dtrend(-1)+[var = exp(c(1))]
@state dtrend=trend(-1)
@state cycle = c(2)*cycle(-1) + c(3)*gph+c(4)*r(-1)+[var = exp(c(5))]

But the answers I get are not correct (coeffs all non-significant). I'm looking to replicate a paper's results which I know to work out so I'm worried its the specification I'm getting wrong.

Any thoughts?

[startz]

That looks right. Try setting the initial values of the c vector to the final values reported in the paper.