3
$\begingroup$

I only know of instrumental variables that causally influence the independent variable and, through the independent variable, are associated with the outcome (making it possible to assume the causal effect of the IV on the DV).

However, is it also possible to use an instrumental variable to investigate the causal role of a moderator? Do you know any examples of papers (preferably social science papers) that did something similar?

Illustrative Model

Thank you!

$\endgroup$
2
  • 1
    $\begingroup$ It will be possible. But I don't know of any examples (I thought about shift-share instruments but they seem different). So I guess that in this setting the moderator variable is included as an interaction term with the other explanatory variable? $\endgroup$ May 30 '20 at 12:15
  • 1
    $\begingroup$ Why would you want to? A moderator is, by definition, not a confounder. You don't have a backdoor path in your diagram. There are quite a few methods of analyzing moderation, especially in counterfactual analysis. I would recommend you look up Chapter 4 in Causal Inference in Statistics: A Primer, by Pearl, Glymour, and Jewell. $\endgroup$ May 30 '20 at 14:40
1
$\begingroup$

If you really want to do this, all you have to do is swap the viewpoint: treat the mediator as if it were the explanatory variable, and the explanatory variable as if it were a confounding variable. Here's an example causal diagram:

enter image description here

$X$ is the explanatory variable, $Y$ the effect, and $T$ the mediator. We have an "instrumental variable" $U.$ If you want the true causal effect of $T$ on $Y,$ then we must view $X$ as a confounder, because $T\leftarrow X\to Y$ is a backdoor path from $T$ to $Y.$ You can just use the backdoor adjustment formula, swapping the roles of the variables, thus: $$P(Y=y|\operatorname{do}(T=t))=\sum_x P(Y=y|T=t,X=x)\,P(X=x).$$ This would get the true causal effect of $T$ on $Y.$ $U$ is actually unnecessary in this diagram, if you want the true causal effect of $T$ on $Y.$

$\endgroup$

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.