Seemingly Unrelated Regression Analysis: Dummy Variables

[Chris Timmerman]

I am estimating the relationship between X and Y, but am interested in how this relationship is affected by a binary variable. However, X and Y are endogenous. I am estimating the relationship now using SUR analysis where X and Y are dependent variables and the right hand side is identical for both equations. Where I get confused is, because I cannot use interactions as dependent variables I am unsure how to view/interpret the impact of the binary variable on the relationship. Could someone explain how I can estimate the equation to view the impact of the binary variable on the relationship of the dependent variables when using SUR?

[Phil Bromiley]

You normally get a better answer if you provide Stata code (using code delimiters), Stata output, and sample data (using dataex).

If both X and Y are endogenous then SUR does not solve the problem. Your estimates are inconsistent. Your model needs variables that differ between the two equations to use normal simultaneous equation estimators like ivreg and reg3.