How to do endogeneity test using cmp

[dr liu]

Hello all,

I m using the cmp command to estimate an ordered probit model with endogenous variables. The problem is: how to determine whether a variable is endogenous or not. I do not know which part of the result is about the endogeneity test. Or I need to run another command to conduct the endogeneity test. I really appreciate any help about the problem.

Thanks

[David Roodman]

Look at whether the athanrho parameter for a variable is statistically difference from zero. For example, if you do

cmp (y = x1 x2 w) (x1 = z1 w) (x2 = z2 w) ....

Then the atanhro_12 parameter (also transformed and reported as rho_12) tells you about the correlation between the error terms of the first and second equations, meaning the y and z1 equations. If it is different from zero, then there are unobserved factors simultaneously affecting y and x1, meaning it is endogenous.

--David

[dr liu]

Look at whether the athanrho parameter for a variable is statistically difference from zero. For example, if you do

cmp (y = x1 x2 w) (x1 = z1 w) (x2 = z2 w) ....

Then the atanhro_12 parameter (also transformed and reported as rho_12) tells you about the correlation between the error terms of the first and second equations, meaning the y and z1 equations. If it is different from zero, then there are unobserved factors simultaneously affecting y and x1, meaning it is endogenous.

--David

Thanks a lot!!!!