US Output-Gap estimation using state space

[Xavier]

I am trying to estimate the Output Gap for the US, using unobserved components obtained from a state space model, like the BIS did ("The global output gap: measurement issues and regional disparities", Quarterly review, June 2011).
The model split the observed GDP between two unobserved components: a Potential GDP and a Gap (GDP=POT+GAP). Then the state variable Potential GDP is defined as a time viriable AR(1) model (POT= GROWTH + POT(-1) + var where the time varying growth rate of Potential GDP is a random walk GROWTH = GROWTH(-1) + Var). The other state variable GAP is an AR(2) process (GAP=c(1)+c(2)*GAP(-1)+c(3)*GAP(-2)+var). var's are independently and normally distributed innovations.

I've written the following code for an EViews state space object:

@signal PIB = POT + GAP

@state growth = growth(-1) + [var = exp(c(6))]
@state POT = POT(-1) + growth + [var = exp(c(5))]
@state GAP = c(1) + c(2)*GAP(-1) + c(3)*GAP(-2) + [var = exp(c(4))]

param c(1) .0 c(2) .0 c(3) .0 c(4) .0 c(5) .0 c(6) .0

It doesn't work because Eviews doesn't accept the signal equation dependent AR(2) process GAP neither the lagged signal equation dependent variable POT(-1) in state equations.

I would appreciate explanation on how can I code this state space model in EViews coding.

[startz]

Code: Select all

@state GAPLAG = GAP(-1) @state GAP = c(1) + c(2)*GAP(-1) + c(3)*GAPLAG(-1) + [var = exp(c(4))]