I'm working on a panel data and want to measure the causality of two variables. I doubt that granger causality can be used for panel data. Is there any better alternatives? Is there any software to conduct this kind of approach?
It depends on what kind of causality you want to capture. Yes you can do panel Granger causality, but you have bear in your mind that it is always "only" the causality in the Granger sense (with no exact direction and intensity...).
The standard method how to estimate causality in the panel data are Fixed-effects and Random-effects models. However, you have to be aware of the spurious regression problem. GMM is even little bit more sophisticated method (for example control for endogeneity), but at least you need to apply instrumental variables in this case.
When you want to measure long-run and short run causality between two cointegrated variables you can use panel Vector error correction model. For short-run causality between variables which are not cointegrated you can use panel VAR model.
This presentation about basic panel data regression methods could be perhaps usefull for you :
https://www.princeton.edu/~otorres/Panel101.pdf
Definitely you can use the Granger causality analysis to test causality between two variables. And, you can collect further details about E- views software from https://www.youtube.com/watch?v=qBYFd9fPA-k link.
Panel VAR will be the suitable method. You can consult Holtz_Eakin, Newey and Rosen, Econometrica, 1988 that deals with the theoretical issues pertaining to this.
- Badges
- Science topic
- Similar topics
- Philosophy
- Esthetics
- Interpretation