I am quantifying the relationship between FDI (as dependent variable) and corruption ( as an independent variable) in addition to other repressors (inflation, political stability). I am hypothesizing that corruption is endogenous to FDI and thus the estimation by OLS will produce inconsistent and biased results. To account for this I propose a system of simultaneous equation to account for the joint determination of corruption and FDI. How can I do 2sls using eviews? How to define a system ( see the attached picture)
Dear Niematallah,
You need to build a system of equations, using some apriori knowledge. The old way to specify constants is to use C(1), C(2) etc. For a system with two equations, you may write: (may compress the variables to Co for corruption, PE for Policing effort)
FDI = c(1) +c(2)*Corruption ..................(1)
Corruption = c(3) + c(4)*PolicingEffort ......(2)
Then you have to specify instrumental variable(s). Basically, these are variables that will affect the whole system. For example, you can use PoliticalStability and Inflation that you already have. Also, if you have a time series, the you can make a variable call Time and you can use that as well as an instrument. Others instrumental variables can be Gdp, growth rates etc. So your first try of instrumental variables may be:
PoliticalStability,
Inflation
Time
Eviews allows you to correct for Auto Regressvie (AR) and Moving Averages (MA), and differincing (I). Don't worry about this up front. First build the system, and if the statistics are not good, then start working on AR and MA and I. Note that you can specify AR as AR(1), AR(2) etc; similarly for MA. Or ARIMA(1,0,1) for instance.
Now the two equations may not give good statistical fit. Then you have to find other variables. But before you do that, try transforming the variables. The easiest tranformation is to take log values, but be careful with dummy variable transformation. (Log of 1 is zero, and log of zero does not exist)
These are instructions for beginners. A next step will be to do 3SLQ. This is because the two equation belongs to one economy or sector or industry, so their error terms may be correlated. Maybe I have said enough for now. But you should be able to start your project with these minimal information. Good luck in your project.
2sls, basically employe in simultaneous equation or system equation. 2 stage least square mean that by applying two times OLS method. But recent problem shows this method potentially endogeneity bias. When the equation includes the intercept, if it doesn't satisfy othogonal condition...then the regressor is correlated with error term. So I suggest first try GLS method if it doesn't fix the problem second you may try another method like GMM.
2SLS is implemented separately for each equation so you don't need a system approach for 2SLS. Eviews calls it TSLS and lists this as one of the estimation methods available for an equation object.
Select the variable (multiple selection). Right click + open as Equation. Then in the method optison choose 2SLS. See the picture. Good luck!
To estimate an equation using Two-stage Least Squares, open the equation specification box by choosing Object/New Object.../Equation… or Quick/Estimate Equation… Choose TSLS from the Method: dropdown menu and the dialog will change to include an edit window where you will list the instruments.
Niematallah Elamin With the reference of answer by Lall B. Ramrattan i also suggest you to check the GMM or Lag augmented 3SLS model.