Difference in difference estimator for forward looking outcome variables that span multiple years

[Sam Wu]

Hello Stata experts,

I am currently trying to do a difference in difference estimator for a model where the outcome variable Y_it is forward looking and spans multiple years i.e. variance over five years. I would like to ask whether a difference in difference estimator is suitable for such a model?

I ask this because given the exogenous shock to some dependent variable X_it occurs in year t, one would expect that at t-4 the outcome variable would already partially capture the effect of the treatment, which means that the trend between the treatment and control group should already differ prior to the treatment. Is there a way to address this?

Thank you very much.

Best,
Sam

[Phil Bromiley]

You'll increase your chances of a useful answer if you follow the FAQ on asking questions - provide Stata code in code delmiters, Stata output, and sample data using dateex. Try to simplify the code and example to what you need to illustrate your problem.

I don't understand your logic. If you have an exogenous shock in t, it can only influence outcomes from t on. The shock might not be exogenous meaning you'd want a treatment model that lets the selection/treatment and outcome equation errors to correlate - look at the treatment model documentation. Such models take care of differences between treated and untreated groups. However, how an exogenous shock in t would influence t-4 is beyond me.

[Sam Wu]

Sorry about that Phil. I will try to follow this in the future.

I agree with you but I'm talking about it in terms of my model specification around the outcome variable. Since my outcome variable Y_i,t is a forward looking measure which includes values from t to t+4, the outcome variable at Y_i,t-4 would include values from t-4 to t.

My idea is that if an exogenous shock at t occurs to the dependent variable X_i,t, this could already influence the outcome variable at t-4, because the outcome variable at t-4 includes the value at t as well.