GREGORYHANSEN—cointegration test with breaks

[abi]

Dear Tom,

If the estimated results from the Johansen procedure shows that there is no cointegration vector but Gregory-Hansen cointegration test supports the existence of one cointegration vector, do you think @GregoryHansen can estimate cointegrating vector and then use that to estimate a VECM model? What should be done for such case?

Thanks,

[TomDoan]

No. Gregory-Hansen is a testing procedure, not an estimation procedure, and it's testing based upon a structural break in the time sequence.