1
$\begingroup$

My question is about economic significance of dummy variable in regression analysis. To me, it is not intuitive to calculate percentage change relative to standard deviation(as usual for continuous variable) of dummy variable as economic significance because it is binary. I personally think that it is best to interpret coefficient compared to the case with zero value of the dummy if we want economic interpretation.

However, I am wondering if my understanding is correct. If you know published paper in economics that uses standard deviation of dummy variable for economic significance, please list that paper. Thank you so much for your help in advance.

$\endgroup$
3
  • $\begingroup$ Can you please tell what is "Economic significance"? $\endgroup$
    – Deep North
    Jun 26, 2017 at 2:16
  • $\begingroup$ It's nearly impossible for me to tell what it is you are asking. Perhaps you can provide an example and clarify your exact question? $\endgroup$ Jun 26, 2017 at 2:22
  • $\begingroup$ "Economic significance" likely means significance in the subject-matter sense, which is based on the effect size and subject-matter interpretation of it. E.g. some effects are statistically significant but too small to be interesting from the subject-matter point of view; such effects would be "not economically significant". Meanwhile, others may not be statistically significant but have large effect sizes to the extent that these are interesting from the subject-matter point of view; these would be "economically significant". $\endgroup$ Jun 26, 2017 at 12:38

1 Answer 1

1
$\begingroup$

Economic significance just means that an effect is substantively important. To determine that you need to substantively interpret your variables and your effects. If your variables have a meaningful scale (e.g. age in years, income in euros, etc.) then you do not want to standardize that variable in order to determine the economic significance, as that will make your effects harder to interpret. Standardization can play a role when you have a variable with non-interpretable scale (e.g. psychological test scores, though they are typically standardized already).

Indicator variables have a known scale, so you should not standardize it in order to determine the size of the effect.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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