Hi everyone,
When there is non-stationarity in time series data due to structural breaks( more than 2 structrural breaks )
1) then how to conduct Cointegration analysis and
2) will use of VECM model still be meaningful in this case?(if found cointegrated)
Thank you
When there is non stationary of data due to structural break, use structural break unit root test. Then, the same procedure for cointegration would proceed as usual.
Testing for panel cointegration with structural break is used. There is GAUSS code for this.
Article A Simple Test for Cointegration in Dependent Panels with Str...
Article Testing for Panel Cointegration With Multiple Structural Breaks
Dear Mwoya Byaro
There are tests of Gregory and Hansen (1996a) and Hatemi-J (2008) when the cointegration relationship has one or two breaks respectively. And Maki (2012) cointegration test allows for an unknown number of breaks. But these research papers also does not give clear cut indication that Vector Error Correction Model (VECM) would be meaningful or not here.
And can we use dummy variables for these structural breaks and use the same Johansen cointegration test and VECM Methodology again here.
Or we should split the sample into sub-samples and do the cointegration, VECM test for each sub-samples seperately. Then how to interpret it.
Thank you.
1) Go for set of non-linear cointegration tests - ARDL/NARDL/QARDL/QNARDL/GH-HJ-Maki tests.
2) No VECM model may not be dependable even if the found cointegrated.
Dear Devesh Kumar Pant
If you are using time-series, you can identify the break years first. For this, you can use the Maki cointegration. After that you can apply the ARDL method where you can externally incorporate the break years to capture the effect of SB in your final model.
Brajesh Mishra Sir, can I use NARDL when the data have structural break ? If so then which unit root test should I be using ?
Dear @Aditya, you may apply NARDL test incase non-linear unit root test establishes structural breaks in majority of time series.
1. Bai and Perron test
2. KSS test
Thank you sir. I went for ZA unit root test. Found in many of the research papers like mine. And sir what about Threshold NARDL sir ?
Dear @Aditya,
It is always advisable to use multiple unit root tests (also from the perspective of learning new techniques) or cointegration tests. Now if you see NARDL, TARDL, QNARDL, QARDL, etc are extension of ARDL models. Yes you may explore TARDL as well.
In the presence of breaks, it is advisable you employ Gregory-Hensen cointegration test, run your vecm under constant and trend. You can call me on +2348060962048
- Badges
- Science topic
- Similar topics
- Statistics
- Time Series