following

Although this question may be better answered by econometricians, I have used unit root tests with structural breaks in my papers. The criteria for choosing an appropriate UR test is the behavior of your timeseries and the structural changes in the sector being studied. If there is a big change which affect the trend of your variable, you can use UR test with one structural break; In case you observe two big shifts in your timeseries, you can use UR test with two breaks... sometimes, variable is stationary with one break and in other times it is stationary with two breaks.

UR tests are also VAR or other kinds of models, and the way a break is considered may differ in different tests.

Hope the above explanation is helpful.

Hi,

When your data series shows structural break, conventional unit root tests are no more relevant (because SB alters the unit root tests results) and you have to use a unit root test that takes into account and reveals the SB. Perron (1989) asserted that unit root tests to be carried out without considering structural changes would produce erroneous results. Structural break unit root tests initiated by Perron (1989) have been studied by Zivot-Andrews (1992), Perron (1997), Ng-Perron (2001) and Lee-Strazicich (2003) using different algorithms. In these methods, one or two structural breaks are allowed in series. However, Bai and Perron (2003) and Carrion-i-Silvestre et al. (2009) have proposed tests that can determine multiple structural breaks (Göçer and Peker, 2014a, 2014b; Göçer, 2015).

However, if your goal is to identify SB (say you're working with returns rather than prices of stock markets so you are not worried about stationarity), there are specific tests without passing by unit root tests such as ICSS algorithm of Inclan and Tiao (1994) and the modified version of 2004, CUSM type tests, Nyblom Test, Hansen test, and many others.

Dear Hamid.

It is of the opinion that structural break can affect unit root process. According to Peron (1989), flunctuations are stationary around a break trend. So Zivot-Andrews test and others are used to indicate unit root with structural break in a model.

Regards

"Model Structural Break Test" is designed to detect the possibility of an unexpected change over time in the parameter estimates of the model. This is a test for structural stability of a model, that is the time-invariance of its parameter estimates. Structural breaks can lead to huge forecasting errors and cast doubt on the reliability of the model if not tested and fixed. On the other side, "Structural Break Unit Root Test" is a check of possible structural breaks when testing for unit root processes. Basically, this is a unit root test that takes into account possible structural breaks in the model being tested. Please read "Unit-root tests and structural breaks: a survey with applications" by John Glynn, Nelson Perera, and Reetu Verma, all from the University of Wollongong, Australia, Faculty of Commerce Papers (Archives), Faculty of Business and Law.

Dear Ako Doffou

Many thanks for your suggestion. How do you edit your initial model, and how do you estimate parameters, when your data/model has two or three structural breaks?

If you have break points there are two ways of estimation. The first and least advised way is to incorporate dummy temporal variables into your model corresponding to the dates of the structural break. The second and best option is to split your sample before and after the break which allows your parameters to change. You can use the switching models for that.

Dear Chiraz Karamti

Your suggestion is appreciated. Many thanks.

I understand the first four lines of Ako Doffou as referring to model invariance but thought Structural Break Unit Root Test refers to the stability of ith variable.