1 proxy for 2 variables

[Iris Janssen]

Hi all,

I performed a test with the following command:

xtreg dependentvariable independentvariable1 independentvariable2 independentvariable3 controlvariable1 controlvariable2, fe

I actually want to test the effect of 4 independent variables on the dependent variable, but 2 of these variables were highly correlated. However, both proxies could actually proxy for both variables. Now my question is: can you have 1 proxy to explain the effect of 2 different independent variables?

Kind regards,

Iris

[Phil Bromiley]

The short answer is if you can't develop a good reason for a measure to be a measure of a specific variable, then there is no way to interpret the results.

By the way, proxy varialbes are not really explaining the true variable - they are simply attempts to measure or proxy for the true variable.