3 SLS and R^2

[mbeckr18]

Dear Glenn or Gareth,

I am currently estimating a 3 SLS model using a system. Looking at the provided R^2 statistics my R^2 dropped significantly compared to the resutls of the simple OLS regression (from 89% to 29%). I guess this might be due to the (possibly weak) instrumental variables I used. Is that correct? Is there a way to test your IVs in E-Views 8? Also e-views provides me with the R^2 statistics for each equation. For the second equation my R^2 is negative. How is this possible?

Thank you very much for your help
Best Maren

[startz]

OLS always gives the best R-squared for a linear estimator.

A negative R-squared means the fit is worse than simply fitting to a constant.

There is nothing wrong with a low or negative R-squared when using an instrumental variable model.