Control variables we added: GDP, GDP growth, trade openness, index for institutional quality, current account balance, lagged real effective exchange rate, education enrollment rate.
I never learned how to analyze panel data. Which method should we use (OLS?)? And which tests are necessary? Should we test for heteroskedasticity, multicollinearity etc? And should we use logharitmic values or not? Thanks in advance!
Dear Titia,
OLS would be biased in a panel case because of unobserved country effects (you have Zimbabwe and then Switzerland with completeley different levels), so I would recommend you have a look at the book of Wooldridge, 2002: Econometric Analysis of Cross Section and Panel Data. You will find Chapter 10 useful: Basic Linear Unobserved Effects Panel Data Models.
Here one may start with Fixed Effects of usind First Differences - I would recommend Stata - in Stata these are the "xtreg" commands. There are more complicated models like System GMM.
Another issue is that you have 19 years: you will definately need a White Noise test on your residuals after you run the panel regressions because there may be autocorrelation. N=13 is somewhat small.
When you have GDP and FDI together, I would think there will be some endogeneity issues. You can consider panel VAR approach. In this case, I would suggest you can use pvar ado in stata developed by Abrigo & Love (2015).
Abrigo, M. R. & Love, I., eds (2015), Estimation of Panel Vector Autoregression in
Stata: a Package of Programs.
Good idea, Saleh! One could also conduct panel cointegration tests - perhaps the proper model would be a panel VECM
- Badges
- Science topic