To include time trend

[cjcostaj]

I am working with a model which will need to include a variable that represents a time trend as:
y = a + b * t
a and b are parameters and t is time trend (t = 1,2,3 ...)

How do I include it in the model? And How do dymare see this?

[StephaneAdjemian]

Hi,

All depends on what you are trying to do. General answer is that there is no interface to specify trends in the model block.

If you want to compute a stochastic simulation, with stoch_simul, it is not possible to include a time trend in the model. With a time trend it would not be possible to define the steady state level (of y in your example) which is needed to compute the local approximation of the model. The traditional approach is to write and solve the detrended version of the model... And to put back the trend on the reduced form solution (to simulate the endogenous variables with growth). If you want to solve a perfect foresight model, you can put a trend in the shocks.

Best,
Stéphane

[cjcostaj]

Hello Stéphane,

It is in shock that I'm thinking to use the temporal trend. But in doing so, will the model turn out to be non stationary?

Thank you very much.

Celso

[jpfeifer]

Hi,
it depends. In the deterministic/perfect foresight context (simul), this does not matter. You don't need stationarity there. For stoch_simul, non-stationary shocks result in a non-stationary solution (no finite second moments). In that case, you have to do as Stéphane says: detrend the model, solve it in detrended form and bring it back to non-stationary form.

[cjcostaj]

Thank you, Stéphane and Johannes.