This question is somewhat related to my previous question, see here: https://stats.stackexchange.com/questions/463221/why-when-do-attitudinal-scales-introduce-endogeneity

I wonder why a scale as an explanatory variable will always make a binary-dependent model endogenous?

This is what an editor told me along with rejecting my paper, and as long as I don't change it/acknowledge the possible endogeneity I cannot resubmit the paper.

  • 1
    $\begingroup$ Would it be possible to share the exact wording of the concern? $\endgroup$ – 1muflon1 May 15 at 18:28
  • 1
    $\begingroup$ Absolutely. It was: including a self-reported scale will induce endogeneity. $\endgroup$ – canIchangethis May 19 at 9:14
  • $\begingroup$ Oh okay now it makes sense. $\endgroup$ – 1muflon1 May 19 at 9:34
  • $\begingroup$ Really,why? I cannot see the sense yet. All variables (except spatial ones) are self reported even age and sex is... $\endgroup$ – canIchangethis May 19 at 9:35

Based on the clarification in your comment this is not because you use scale variables but because the scale variable that you use is self-reported.

It is well known in literature that self-reporting leads to endogeneity. See for example: Lindeboom, M., & Kerkhofs, M. (2004). Subjective health measures, reporting errors and endogeneity in the relationship between health and work.

The reason why that is when people self-report some outcomes those are often affected by other unobservables that you do not include in your model which is a form of endogeneity. For example, studies that measure happiness show that the self reported happiness can vary depending on which time of the day/week/season you make the measurement.

Moreover, you can have also endogeneity in more narrow sense unemployed people can self-report that they cannot work as a rationalization of their unemployment status and hence you would get reversed causality where the unemployment actually causes the self-reported inability to work even though the person actually would be able to work as measured by objective measurable criteria (in a model where employment is regressed on self-reported ability).

So it’s not that there is any problem with having scale variables. For example, having some 1-10 index but rather the problem is that in your case the scale was made from self reported data.

| improve this answer | |
  • 2
    $\begingroup$ great answer! Please could you give the full reference of Lindeboom, M., & Kerkhofs, M. (2004)? There seems to be a typo in the first sentence. $\endgroup$ – emeryville May 19 at 9:59
  • 1
    $\begingroup$ @emeryville I added hyperlink to ungated version of the paper $\endgroup$ – 1muflon1 May 19 at 10:03
  • $\begingroup$ the link doesn't work for me. I would like to comment in detail, but I would like to read the paper beforehand. Thank you. $\endgroup$ – canIchangethis May 19 at 12:09
  • $\begingroup$ @canIchangethis I updated the link it should work now - sorry I was logged into SSRN when I copied the link so it probably affected the URL hope that now it works $\endgroup$ – 1muflon1 May 19 at 12:16
  • 1
    $\begingroup$ I just found it, too here poseidon01.ssrn.com/… $\endgroup$ – canIchangethis May 19 at 12:17

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.