VECM with mixed stationary and non-stationary series

[TomDoan]

You would have to show me the system instructions that you used, but if set up correctly, that would do the draws and responses correctly.

[ateeb]

linreg(define=ect1) logdebt
# logM1 logCPI

equation rect r
#

equation lsmect lsm
#

SYSTEM(MODEL=PK2MODEL)
VARIABLES logdebt logM1 logCPI logLSM r
LAGS 1
DET Constant du
ect ect1 lsmect rect
END(SYSTEM)
ESTIMATE 2003:07 2018:09
errors(model=PK2MODEL,results=errors,steps=48)

I await your response

[TomDoan]

In ECT form, the LAGS should be 1 2. Other than that, it's fine.

[ateeb]

The AIC has chosen 1 lag for the VAR therefore I used 1, so should i type 1 to 1?

Regards,

Thanks for quick responses

[ateeb]

a final question regarding this method, the document says that estimate will automatically difference all the series, however I do not want it to difference log of LSM which is Industrial Production as well as the interest rate, so given the way it is written, would estimate understand that they should not be differenced?

[TomDoan]

AIC in levels picks 1? Seems odd (it means there are no dynamics beyond the error corrections), but if that's what you got, then you would in fact do LAGS 1 only.

[TomDoan]

a final question regarding this method, the document says that estimate will automatically difference all the series, however I do not want it to difference log of LSM which is Industrial Production as well as the interest rate, so given the way it is written, would estimate understand that they should not be differenced?

For a stationary variable, differences on differences plus the lagged dependent variable is identical to level on lagged levels.