0
$\begingroup$

Should we include some control variables in regressions when we test a moderation effect of a variable?

$\endgroup$
  • $\begingroup$ In general, control variables are a good idea. Please give an example of the specific regression you would like to know about. $\endgroup$ – Student Nov 15 '19 at 4:42
  • $\begingroup$ Thanks a lot! when I added relevant control variables to my regressions, the moderator variable "become not significant" but the interaction term remain significant..Should we say that there is a moderation effect?..Without control variables, both were significant.. $\endgroup$ – Elkwather Nov 18 '19 at 15:22
1
$\begingroup$

Without additional information, it is hard to answer your question but I would say that a moderation model is like any econometric model (and in fact, it is an econometric model).

Therefore, I would tend to agree with @Student, adding control variables is a good idea but it must represent your theoretical considerations (is there an omitted variable bias, a confounding variable, etc ?). So you should only add relevant variables to your model in order to increase its precision.

Just adding random/useless variables will unnecessarily increase the complexity of your model; and as my econometrics professor used to say: "why use a bazooka to kill a fly when you can use a fly swatter ?"

Edit: I (scholar) googled "moderation effect economics" and I randomly chose a paper published in a good journal, you can see that they also use relevant control variables (p. 512).

|improve this answer|||||
$\endgroup$
  • $\begingroup$ Thanks a lot! I added relevant control variables to my regressions. For some regressions, the moderator variable "is not significant" but the interaction term is significant..Should we say that there is a moderation effect? $\endgroup$ – Elkwather Nov 18 '19 at 15:20

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