cointegration with 2 structural breaks

[chiade]

Hi,

I am trying to incorporate a cointegration wih 2 structural breaks. I managed to utilize gregory-hansen to test conintegration in presence of 1 structural break. How do i do test cointegration with 2 structural breaks? is there a procedure for it? Pls advise.

Many thanks once again for your prompt replies.

Des

[TomDoan]

The number crunching in Kejriwal and Perron(2010), "Testing for Multiple Structural Changes in Cointegrated Regression Models", Journal of Business and Economic Statistics, vol 28, no 4, 503-522 can be done with the existing BaiPerron procedure. What the paper does is to analyze the behavior of the test statistics when applied to cointegrated models.

I didn't notice a paper which extended Gregory and Hansen to multiple breaks, but there is no conceptional reason it couldn't be done. The general procedure for searching for multiple breaks can be used for that.

[ege_man]

Dear Tom,
Is it possible to extend this code for unknown number of breaks like the following paper? As I understand this paper also uses the similar specifications with the Gregory and Hansen (1996)
Maki, D. (2012), “Tests For Cointegration Allowing For an Unknown Number of Breaks”, Economic Modelling, 29(5), 2011-2015.

[TomDoan]

The paper is behind a pay wall. But given that it's 4 pages, it can't be much more than a straight extension. As I mentioned in the earlier reply, the general process for handling multiple breaks is described in the User's Guide.

[elhampa]

Hello,

I am looking for this paper procedure: Kejriwal and Perron(2010),"Testing for Multiple Structural Changes in Cointegrated Regression Models", Journal of Business and Economic Statistics, vol 28, no 4, 503-522.
I read the discussion which it can be done by the Bai and Perron procedure but I do not know how to do that. I have downloaded the procedure for Bai and Perron. Do I need to make a change in that?

Regards
Elli

[TomDoan]

No. Just use the @BaiPerron procedure. Kejriwal and Perron uses the same calculations, but has critical values on breaks that are specific to cointegration.

[elhampa]

Hello Tom,

A quick question. Is the test proposed by Kejriwal and Perron "Testing for Multiple Structural Changes in Cointegrated Regression Models" Journal of Business and Economic Statistics 28(4):503-522 robust for a model suffers from heteroskedasticity and serial correlation?

Thank you for your support.

[TomDoan]

It depends upon what you mean by heteroscedasticity. The result allows for errors that would permit a functional CLT, which would mean mixing/ergodic with an unconditional long-run variance (typical "HAC" correctable errors). That would not allow for regime-based heteroscedasticity.

[elhampa]

Hello Tom,

I am trying to run test: Kejriwal and Perron "Testing for Multiple Structural Changes in Cointegrated Regression Models" Journal of Business and Economic Statistics 28(4):503-522, I have three I(1) and two I(0) time series that all include at least one break in trend.
I am using the Bai and Perron rats code and critical values of the above mentioned paper of Kejriwal and perron. The tables of critical values are only for qb=2 and pb=2 while I have 3 of q(b)s. Would you please advise me what I can do in such situation knowing that the model includes one break only.

Thank you for all your support.

Regards

[TomDoan]

Hello Tom,

I am trying to run test: Kejriwal and Perron "Testing for Multiple Structural Changes in Cointegrated Regression Models" Journal of Business and Economic Statistics 28(4):503-522, I have three I(1) and two I(0) time series that all include at least one break in trend.
I am using the Bai and Perron rats code and critical values of the above mentioned paper of Kejriwal and perron. The tables of critical values are only for qb=2 and pb=2 while I have 3 of q(b)s. Would you please advise me what I can do in such situation knowing that the model includes one break only.

Thank you for all your support.

Regards

You should probably contact Pierre Perron to see if they've extended the tables. However, it doesn't sound like you're understanding what this is testing. Kejriwal and Perron is testing for a break in the relationship connecting the series, not in the series themselves.

[elhampa]

Hello Tom,

Thank you for the clarification. Yes. I understand that the Bai and Perron is for break in relationship.
I tested the series for breaks using Lee and Strazicich.

Thanks again.

[TomDoan]

Please note that Lee-Strazicich (and similar tests) are not tests for breaks. They are tests for unit roots. See the discussion on Unit Roots and Breaks

[elhampa]

Hello Tom,

I am confused. Would you please correct me if I am wrong. I used Lee and Srazicich unit root test. Out of 5 independent variables, 3 of them are I(1) with breaks and two of them are I(0) with breaks.
Now if I want to use Kejriwal and Perron co-integration test I have three qb and 2 pb.
Am I right so far?

[TomDoan]

No. Again, you're confusing breaks in the variables with breaks in the specifications. The assumptions in K&P requires I(0) (stationary) and I(1) (accumulation of stationary) variables, not variables with breaks.

[elhampa]

Thanks Tom,
Yes I am not sure about pb, pf, qb and qf. I do understand that this test (K&P) is not for individual variables.
Have read the recommended pages.
So please explain what pb, pf, qb and qf are.