I am doing a research on FDI , with six independent variables , my objective is to see long term relationship of IV with DV. FDI is stationary at I(1) , and other five IV are also at I(1) , but one IV is I(0) stationary at level. Can cointegration be applied in such case or some other test is used?
Yes you can apply cointegration test but since you have a mixed integrated series (i.e. I(1) and I(0)), you could consider utilizing the autoregressive distributed lag model (ARDL) of Pesaran et al. (2001). This method is appropriate for modelling long run relationships even if variables are stationary and non-stationary (I(1) not I(2)). Microfit software and EViews 9.0 can easily implement this procedure.
You can do cointegration analysis using Johansen as long as the variables are at most I(1). In fact, there is no need to decide on the order of integrating before doing the cointegration analysis. If one variable is stationary and I(0), it will be a cointegrating relationship in its own right. It will not enter other cointegrating relationships, though. From a practical point of view, how do you know that one variable is I(0)? Is the conclusion based on a unit root test?
There are many ways to handle the stationary and non stationary variables. You can do a cointegration test with all I(1) variables and include I(o) only in the VECM.
@NiklasAhlgren yes the conclusion is based on unit root test only, out of seven one variable is showing unit root at I(0)
The fact that a unit root test rejects does not mean that a series is stationary; the outcome of the unit root test may be false. You should think about whether there is any reason to believe the series is I(0) rather than I(1). If there is no such reason, you may proceed with the cointegration analysis on the assumption that all series are I(1).
I am doing a research on the impact of FDI on Total factor Productivity. Can I apply Bayer-Hanck cointegration test for a mixed integrated series.... I(0) and I(1)?
- Badges
- Science topic