I ran an ordered logit difference-in-difference estimation and got stuck in terms of interpretation. The dependent variable takes on values 1 2 and 3 and I am trying to obtain coefficients for the variables for each of the three categories as I see in papers but I am not sure how to go about it.

  • 1
    $\begingroup$ Please provide a specific simplified example of your desired result. $\endgroup$ – Frank Harrell Aug 14 '14 at 14:36
  • $\begingroup$ say I am trying to estimate the effect of a policy on marital status-single, married and divorced. $\endgroup$ – user54087 Aug 14 '14 at 15:01
  • $\begingroup$ That sounds like a standard 2 degree of freedom test comparing 3 groups, which can be done by removing marital status from the model or by using a 2 d.f. contrast to get a Wald $\chi^2$ statistic. A double difference is not implied by the example question you provided. $\endgroup$ – Frank Harrell Aug 14 '14 at 15:05
  • $\begingroup$ when I run the ordered logit, I get treatment effect for the dependent variable as a whole, but i am trying to get the treatment effect for each of the categories. $\endgroup$ – user54087 Aug 14 '14 at 15:07
  • $\begingroup$ Get those from the coefficients (which represent difference from the reference cell in log odds scale) or by taking differences in coefficients to get other comparisons. $\endgroup$ – Frank Harrell Aug 14 '14 at 16:05

If you have specified the difference in differences as the interaction of your time and group variable then you need to be aware that interactions in non-linear models (like ordered logit) have a different interpretation. This relates to a similar question here which asked how difference in differences works in the context of logit models. In this reply there are alternative models and links to articles that explain the intuition and interpretation of difference in differences in the context of nonlinear models. Have a look and then see whether ordered logit is actually what you want to do. Particularly relevant for your case would be Karaca-Mandic et al (2012). Note though that what you are trying to do is not very well explored in the econometric literature which is mostly for the fact that difference in differences in nonlinear models is far from trivial (for an overview of the problem see Athey and Imbens, 2006).

| cite | improve this answer | |

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.