Difference-in-difference and fixed effects

[Dominik Koeppl]

I estimate two models on the impact of conflict on education.

1. The first model is based on a single post-conflict household survey and a comparison of a young (affected) and an old (unaffected) cohort.
2. The second model is based on pre- and post-conflict household surveys (cross-section) and a comparison of a young cohort before (unaffected) and a young cohort after (affected) the conflict.

For both models I also exploit the geographical variation of conflict and that some areas were affected while others were not ('conflict').

For Model 1: I start with following regression:

regress $dependent young conflict did $controls_ind $controls_hh

Then I add area and age fixed effects. As expected, Stata drops 'young' and 'conflict' due to multicollinearity:

regress $dependent did $controls_ind $controls_hh i.area i.age

For Model 2: I start with following regression:

regress $dependent post conflict did $controls_ind $controls_hh

I again add area and age fixed effects. Stata drops 'conflict', but not 'post'.

regress $dependent did post $controls_ind $controls_hh i.area i.age

My questions are:

1. Why does Stata not drop 'post' in the fourth equation? I would have expected to be dropped similiar to 'cohort' in my third equation?
2. Should I keep it in the model or remove it?

Thank you very much for your time!

[Phil Bromiley]

You didn't get a quick response. You'll increase your chances of a useful response by following the FAQ on asking questions - provide Stata does in code delimiters, readable Stata output, and sample data using dataex. We have no idea what is in all your macros for example. I'm also not clear on your data structure. Also, try to simplify your posting to the minimum needed to demonstrate the problem.

Often is is easier to do such models with xtreg or areg - they're built for this kind of data. Your questions depend on data and substantive issues on which we have no information.

You can start by regressing post on the other iv's and see what you get.