Dear all,
Please can anyone clarify which is more appropriate to adopt when testing for structural breaks - is it on the dependent variable or on the main explanatory variable?
For instance, if I want to explain income inequality (using the Gini index) via the influence of trade reforms, do I construct the econometric specifications based on break points in the inequality series or the trade series? Although, I think it should be on the explanatory variable, I just need to be sure.
Kindly advise....thank you.
Dear Adeleye,
The break point shall be considered in the explanatory variable that trigger the shock. In your case, I think that the trade reforms may be taken as a shock variable and the break points should be considered in this variable.
Hope this is helpful.
Thanks Najib, but I have since modified the model. I now have just two variables explaining income inequality. They are domestic credit and per capita GDP. I'm thinking that since a series of financial reforms took place in Nigeria between 1986 and 2009, perhaps the domestic credit should be the break "variable" to consider. Or, I can use the Gregory-Hansen (1996) test result for the model to obtain the probable break dates. What do you think?
Please help if you can....gracias!
You should figure out which of the two candidates (domestic credit or per capita) are affected by the reforms. If the reform is focusing on the whole economy both of the variables may be affected. In the latter case, your break variable will be a dummy that take zero for before the reforms and 1 for after the reforms. The date of reform inaction (1986) may be the break point, which will define your dummy variable.
The procedures available for determining a break point such as Gregory-Hansen (1996) can be used. If your objective is testing the effects of this particular reforms, the date of reform of commencing the reforms is a good break point. If you just want to find out the effects of a structural break, and not sure about the break point, the Gregory-Hansen (1996) may be helpful. Lee-Stracizich (2003) unit root test can help you endogenously determine the break point, which is more reliable. I used the latter in a study.
I hope this will help you find your way.
Good Luck !
Thanks a lot Najib for the clear and concise explanation. I decided to go with your Gregory-Hansen suggestion...I will attempt the Lee-Stracizich (2003) unit root test approach if available with Stata with my next paper. For my unit root test, I used the Zivot-Andrews (1992) procedure...
thanks again, you are an angel!
- Badges
- Science topic
- Similar topics
- Filing