Structural break ath the end of the sample

For econometric discussions not necessarily related to EViews.

Moderators: EViews Gareth, EViews Moderator

wolly77
Posts: 32
Joined: Mon Oct 05, 2009 4:28 am

Structural break ath the end of the sample

Postby wolly77 » Tue Oct 06, 2009 1:30 am

Hi guys, I have a question. I estimated a cointegrating relationship with a DOLS for the period 1952q1-2008q4. From the residual analysis emerges clearly that this long run relationship exists only until period 1952q1-2003q4. After this period the test on residual rejects the hyopthesys of stationarity (and then the existence of the long run relationship). I proved to insert in the estimation a regime shift dummy from 2004q1 until the end of period of estimation. This dummy is strongly significant and suggest to me that there is a structural break in the relationship in 2004q1. Some economics events confirm this hypothesys. The problem is that this structural break occurs at the end of the sample, and the Gregory-Hansen cointegration test is not able to ascertain it. There exists a statistical test able to capture the break ath the end of sample?

trubador
Did you use forum search?
Posts: 1518
Joined: Thu Nov 20, 2008 12:04 pm

Re: Structural break ath the end of the sample

Postby trubador » Tue Oct 06, 2009 3:33 am

For most of the econometric techniques, identification of a structural break works in an ex-post manner. This is also intuitive since one must have enough observations after the potential break to reach a valid and reliable conclusion that it is indeed "structural". Otherwise, it may simply be an outlier. Gregory-Hansen procedure is able to capture the structural shifts towards the end of sample, if you simply lower the trimming parameter. In that case, however, generating a trend dummy variable may not make sense at all.

wolly77
Posts: 32
Joined: Mon Oct 05, 2009 4:28 am

Re: Structural break ath the end of the sample

Postby wolly77 » Wed Oct 07, 2009 6:22 am

Dear trubador, can you help me interpret this result? I used DOLS technique (Stock-Watson (1993)) to estimate a long run relationship. To check the validity of a cointegration relationship estimated, I made the residual test based on the values of MacKinnon (1991). From the residual tests emerges that the cointegrating relationship breaks at the beginning of 2004. This can be viewed via a dynamic ADF test on residuals (that is, residual repeated every quisarter). From the graph of this test you can see that the ADF values are below critical vaules until 2004q1 when the realtionship breaks (from this point the ADF is above critical values). This result suggests me that a break occurs in 2004. But the Gregory-Hansen test suggests me that the break is in 2002. In your opinion which kind of test could detect this break? I thought to use the Hassler cointegration test (2004) for shifts in series at known date.
Attachments
dynamic_ADF.jpg
dynamic_ADF.jpg (28.82 KiB) Viewed 5284 times

trubador
Did you use forum search?
Posts: 1518
Joined: Thu Nov 20, 2008 12:04 pm

Re: Structural break ath the end of the sample

Postby trubador » Wed Oct 07, 2009 12:54 pm

Gregory-Hansen procedure may not be able to capture the structural break if it is located too close to the end of the sample. For more information on the issue, please refer to the original study. Besides that, different methods might yield different results and it is difficult to decide which method is best, since they are essentially data dependent and utilize different approaches to detect the location of structural break. Therefore, if you have a priori information or qualitative data to back the result obtained, then you can continue with the analysis more confidently.


Return to “Econometric Discussions”

Who is online

Users browsing this forum: No registered users and 13 guests