while conduction panel data analysis, the outcome of unit root tests says that some of the variables are already stationary at level while other are non-stationary at level, but after first difference they became stationary. Question arises in my mind, that should i still proceed to conduct panel cointegration analysis? Is it mandatory that all the variables must be non-stationary at level but but stationary after the first difference in order to conduct panel cointegration analysis? Your suggestions and guidance are appreciated.
When order of integration is not same i.e. your series are stationary at I(0) and I(1) then you need to apply panel ardl to get long run and short run estimates.
Though panel ARDL can be suitable to use, extreme caution however needs to be exercised since application of unit root tests in the panel data context can be more complex relative to tests for unit root in single time-series. The greatest common complication in context of panel data emerges from the fact that test statistics may have distinct asymptotic distributions, depending on whether N is fixed and T tends to infinity (or vice versa) or both T and N increase concurrently in a static ratio. The problem of cross-sectional dependence in the errors across unit root testing regressions also needs to be dealt with quite carefully.
Dear Taimoor,
In a situation where you have different order of integration in panel data analysis, panel ARDL is applicable to derive a short run and long run eatimates of the analysis.
Regards
Since you have a mixture of I(0) and I(1) variables, the panel ardl will be ideal and the bounds test approach will determine the existence of long run relationship. Hope this helps.
- Badges
- Science topic
- Similar topics
- Psychology
- Cognitive Science
- Mind