Page 1 of 1

Test of Parameter Stability of a Cointegrating Vector

Posted: Thu Mar 29, 2012 10:27 am
by cecedi
Hi, I have 2 questions:

1. After estimating a Cointegrating Vector by the procedure DOLS of Stock and Watson (by the procedure @swdols in RATS), how can I test for the Stability of Parameters of this cointegrating vector? Is there a code in RATS to do this kind of stability test?

2. When I use CATS to estimate a cointegrating vector, I can use the "Recursive Estimation Procedure" to test for the "Beta Constancy" by graphics. However, it seems a stability test for all estimated cointegrating vector by CATS but not for one only chosen cointegrating vector. For example, if I set the rank = 1, the test for Beta constancy will be made for the first cointegrating vector; but if I set the rank = 3, the test for Beta constancy will be made for all of three cointegrating vector.
So, if I like to take the first cointegrating vector, it's OK. But if I have to take the second or the third cointegrating vector, I don't know how to make the stability test for the chosen cointegrating vector?

Many thanks
Best regards

Re: Test of Parameter Stability of a Cointegrating Vector

Posted: Thu Mar 29, 2012 12:03 pm
by TomDoan
cecedi wrote:Hi, I have 2 questions:

1. After estimating a Cointegrating Vector by the procedure DOLS of Stock and Watson (by the procedure @swdols in RATS), how can I test for the Stability of Parameters of this cointegrating vector? Is there a code in RATS to do this kind of stability test?
Not that I'm aware.
cecedi wrote: 2. When I use CATS to estimate a cointegrating vector, I can use the "Recursive Estimation Procedure" to test for the "Beta Constancy" by graphics. However, it seems a stability test for all estimated cointegrating vector by CATS but not for one only chosen cointegrating vector. For example, if I set the rank = 1, the test for Beta constancy will be made for the first cointegrating vector; but if I set the rank = 3, the test for Beta constancy will be made for all of three cointegrating vector.
So, if I like to take the first cointegrating vector, it's OK. But if I have to take the second or the third cointegrating vector, I don't know how to make the stability test for the chosen cointegrating vector?
See section 4.2.7 for recursive estimation of an identified system.