VECM with user specified ECT

[mob7]

Hi,

Is it possible to estimate a VECM and get the estimation output in first differences and impulse responses generated in levels while specifying the error correction term as a specific variable in the dataset and not the residuals from the long-run relationship that is estimated before setting up the VECM?

For example, the standard way of doing a 2 variable, 4-lag VECM is:

lin(define=model_ecm) x / resids
# constant y

system(model=model_1)
var x y
lags 1 to 5
det constant
ect model_ecm
end(system)
estimate(outsigma=s,residuals=resid)
errors(impulses,model=model_1) 2 24 s


However I want to provide a separate error correction term (var_ecm) and estimate:
system(model=model_2)
var x y
lags 1 to 5
det constant var_ecm{1}
end(system)
estimate(outsigma=s,residuals=resid)
errors(impulses,model=model_2) 2 24 s

Is it possible to do this and get the estimation output in first differences as well as the impulse responses in levels?

[TomDoan]

You can do that, but the IRF can't take into account how VAR_ECM is related to X and Y since VAR_ECM is "data" not a definition. How is it computed if not from X and Y?

You can put a blank ECT instruction, i.e.

ECT

in the system to get the estimates in differences. And as with a system with an actual ECT, it will give you the IMPULSES in levels.