Consider the following model for a regression : $y = \beta_0 + \beta_1 x_1 + \beta_2 x_2+ \beta_3 x_1 x_2 + u$.

Suppose $x_1$ is exogenous and $x_2$ is endogenous. Endogeneity precludes me from analyzing the effect of $x_2$ on $y$. However, I still want to interpret the effect of a change of $x_1$ on $y$. The fact that $x_2$ is endogenous wouldn't bother me too much normally, but here, because of the interaction, the effect of $x_1$ is affected by the value of $x_2$. Not only is $x_2$ potentially correlated with the error term, but it is likely correlated with $x_1$ ($x_1$ and $x_2$ are not independent variables). Can I still interpret the marginal effect ($\beta_1 + \beta_3 x_2$) as the effect of $x_1$ on $y$, holding all else constant?

In my problem $x_2$ is endogenous because it is jointly determined with $y$. I study international agreements and these variables are characteristics which are decided at the same time by the parties.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.