Hi there,
I am looking at the relationship between corruption distance and bilateral FDI flows. I have chosen to use a two stage method as this eliminates the problem of many missing observations in my FDI data.
I have completed a first stage in which I determine the decision to invest or not using the following regression specification: Dij,t = α + βlogCorDistij,t−1 + γXij,t−1 + Trend + εij,t , where D is decision to invest, logCorrDist is the logged difference in the level of corruption between the host and source country, X is my control variables which also includes my exclusion variables (these will not be inlcuded in the second stage of the method), a time trend and then my error term.
I now want to generate an inverse mills ratio to use in my second stage of the model.
My second stage of the model is: log(FDIij,t) = α + βlogCorDistij,t−1 + γXij,t−1 + λMillsij,t+εij,t
So far I have been given a basic outline of code to generate an inverse mills ratio
probit ___ ____
predict ___, xb
g mills1 = exp(-.5*___^2)/(sqrt(2*_pi)*normprob(___))
Where the underscores are variables that I need to put in for it to work on my dataset.
The variables I am considering putting in here are FDI, logFDI, logCorrDist, but I may be completely wrong and have to include one or more of my control variables (GDP, GDP growth, GDP per capita, inflation, unemployment, trade orientation, political stability for all host and source countries). I have attached a sample of my dataset to give you an idea.
Could anyone please provide me some guidance of where to go from here.
Thank you very much,
Zak
I am looking at the relationship between corruption distance and bilateral FDI flows. I have chosen to use a two stage method as this eliminates the problem of many missing observations in my FDI data.
I have completed a first stage in which I determine the decision to invest or not using the following regression specification: Dij,t = α + βlogCorDistij,t−1 + γXij,t−1 + Trend + εij,t , where D is decision to invest, logCorrDist is the logged difference in the level of corruption between the host and source country, X is my control variables which also includes my exclusion variables (these will not be inlcuded in the second stage of the method), a time trend and then my error term.
I now want to generate an inverse mills ratio to use in my second stage of the model.
My second stage of the model is: log(FDIij,t) = α + βlogCorDistij,t−1 + γXij,t−1 + λMillsij,t+εij,t
So far I have been given a basic outline of code to generate an inverse mills ratio
probit ___ ____
predict ___, xb
g mills1 = exp(-.5*___^2)/(sqrt(2*_pi)*normprob(___))
Where the underscores are variables that I need to put in for it to work on my dataset.
The variables I am considering putting in here are FDI, logFDI, logCorrDist, but I may be completely wrong and have to include one or more of my control variables (GDP, GDP growth, GDP per capita, inflation, unemployment, trade orientation, political stability for all host and source countries). I have attached a sample of my dataset to give you an idea.
Could anyone please provide me some guidance of where to go from here.
Thank you very much,
Zak
