0
$\begingroup$

Is difference-in-difference based on group's internal trait -- socioeconomic status in this case -- appropriate?

I'd like to measure the difference in how people behave after a certain shock based on their SES. I have the data on both groups' behavior before the event and also after the event. But all the DID guides deal with cases that measure the effect of an external event -- such as a policy change. In my case, I'm trying to measure how an "internal" trait -- the socioeconomic status -- affects people's reaction to an external change.

Improve this question
$\endgroup$

1 Answer 1

Reset to default
0
$\begingroup$

As I understand it, you are still measuring the impact of a policy of an outcome variable. The additional terminology of "internal" vs "external" event/trait is a distraction. If you event is exogenous, you can compare the outcomes in one treatment SES with a control SES. With a DID, you will have to show us that the parallel trends assumption is not violated and that there is nothing else going on at the time the event happened. Please refer to http://econ.lse.ac.uk/staff/spischke/ec533/did.pdf for more.

Improve this answer
$\endgroup$
3
  • $\begingroup$ No, I'm not measuring the impact of a policy. I'm measuring the impact of the SES when the impact took place. i.e., The variable of interest I'm looking at is not the dummy variable used for policy change, but the income bracket of the people. $\endgroup$
    – Anon
    Jun 3, 2016 at 7:28
  • $\begingroup$ I am not sure what field you are in but in economics it's going to be very hard to defend that E(u|SEC) = 0 or SEC is exogenous. Without further details on your specific research question and design it is hard to give you any concrete advice. $\endgroup$
    – user26750
    Jun 3, 2016 at 7:32
  • $\begingroup$ I'm in health policy. I see researches along the lines of effects of the SES on cancer prevalence, etc. Don't they violate E(u|SEC) = 0 too then? $\endgroup$
    – Anon
    Jun 3, 2016 at 7:38

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.