1
$\begingroup$

I'm a bit confused with the definitions regarding causal inference. My question is whether we can call measured confounding an endogeneity problem?

$\endgroup$
1
$\begingroup$

I would say no. Confounding per se could be endogenous, or it might not be. But if the confounding variable is measured, then it's no longer endogeneity. Here's an example of endogenous confounding:

enter image description here

Here $U_Y$ is an exogenous variable that we will assume gives rise to the error term for $Y.$ $X$ is confounded with $U_Y,$ which is the definition of endogeneity. Typically, $U_Y$ is not measured. But you could have a similar-looking but quite different situation here:

enter image description here

Now $Z$ is a measured variable, and is a confounder for the effect of $X$ on $Y,$ the same as before, except that $Z$ is measured. Once the confounding variable is measured, you no longer include it in the error term; and once you no longer include a variable in the error term, it ceases being endogeneity.

$\endgroup$
1
  • 1
    $\begingroup$ Thank you! I agree with you. I started doubting because sometimes I hear people using confounding and endogeneity terms interchangeably. $\endgroup$ – Anita Apr 28 '20 at 8:25

Your Answer

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

Not the answer you're looking for? Browse other questions tagged or ask your own question.