Im trying to explain the intuition behind fixed effects to a group of people with no background in (formal) regression analysis. Explaining the intuition behind Diff-in-Diff estimation of the impact of a policy is straightforward (here is a good example). However, I am struggling to get fixed effects across. Here is what I have tried so far:

  • The Diff-in-Diff approach relies on measuring the outcome of interest for subjects who were treated and not treated, before and after. This is a 2x2 analysis
  • In practice, we would benefit from having more information, for example, by measuring the outcome for several treatment intensities and more than two time periods (SxT analysis).
  • In Diff-in-Diff, we are using the variation in outcome across subjects from different groups (S=2 for treatment and control groups) and across time (T=2 for before and after periods), to control for for time-invariant unobservable
  • Analogously, when T>2, we can control for a “subject fixed effect”: we observe outcomes for the same subject many times (not just once before and once after), so we can estimate what part of the outcome is due to the “fixed” characteristics of the subject (i.e. time-invariant unobservables of that subject).

Any ideas or resources?

  • 1
    $\begingroup$ That is IMO a pretty good description without equations. The equations one just needs to know algebra, so they can be presented pretty gently. $\endgroup$ – Andy W Feb 19 '15 at 13:43

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.