Hi!
I am examing the effect of the demographic transition in emerging Asia on my dependent varibale, the stock of FDI. According to the literature I used several control variables (GDP growth, trade openess, institutional quality, education level, real effective exchange rate (lag), total GDP and I am thinking about adding the lag of the FDI stock in to the independent variables as well.
So I am working with balanced panel data and if I understand correctly the Hausman test tells me to use fixed effects. Somewhere on the internet I found that I need to use cross section weights with the diagonal coëfficiënt covariance method since my t (=18 years) is larger than my N (12 countries). However, I have no idea if this is correct. Furthermore I expect my independent variables to be correlated. How should I deal with this? And with time series data I am used to performing tests for autocorrelation and heteroskedasticity, but how does this work for panel data?
So to sum my question: Am I using the right method of estimation (if not, which should I use and why). And which tests should I perform and how can I perform these tests in eviews 8 (I am thinking about test for autocorrelation, hetereoskedasticity, endogeneity etc.)
Thank you very much!!
PS: for convenience I uploaded the excel file with the data and the eviews file that I am working in. In this eviews file I am experimenting with different methods and different estimations (2 stage least square, EGLS, period fixed etc.)
I am examing the effect of the demographic transition in emerging Asia on my dependent varibale, the stock of FDI. According to the literature I used several control variables (GDP growth, trade openess, institutional quality, education level, real effective exchange rate (lag), total GDP and I am thinking about adding the lag of the FDI stock in to the independent variables as well.
So I am working with balanced panel data and if I understand correctly the Hausman test tells me to use fixed effects. Somewhere on the internet I found that I need to use cross section weights with the diagonal coëfficiënt covariance method since my t (=18 years) is larger than my N (12 countries). However, I have no idea if this is correct. Furthermore I expect my independent variables to be correlated. How should I deal with this? And with time series data I am used to performing tests for autocorrelation and heteroskedasticity, but how does this work for panel data?
So to sum my question: Am I using the right method of estimation (if not, which should I use and why). And which tests should I perform and how can I perform these tests in eviews 8 (I am thinking about test for autocorrelation, hetereoskedasticity, endogeneity etc.)
Thank you very much!!
PS: for convenience I uploaded the excel file with the data and the eviews file that I am working in. In this eviews file I am experimenting with different methods and different estimations (2 stage least square, EGLS, period fixed etc.)