I'm doing a diff-in-diffs type of model with xtabond, but would be grateful for help establishing if I'm doing it correctly.
In principle, I have app. 52 units, observed over, on average, 50 periods (but highly unbalanced, ranging from 9 to 99).
I previously did a panel fixed effects but would like to do Arellano-Bond with lagged dependent variable (one lag).
I also have one explaining variable that I believe affects only future periods (the explained variable is series episode viewership, and the explaining variable in question is episode rating).
And, finally, I want to include time trends (separate trend for the treatment group and separate for the control group), including a potential change in the trend post-treatment.
So, right now, this is what I have more or less:
Code:
xtabond Y POST_TREAT TREATED_POST_TREAT /// TIME TREATED_TIMETIME_POST_TREAT TREATED_TIME_POST_TREAT /// OTHERCONTROLS, lag(1) vce(robust) pre(RATING, lag(1,.))
The second line includes the trend, the trend for the treated group, and then both but only after the treatment.
Then I have some other controls (e.g. for the TV show season finale). I include the rating variable as a predetermined one. I understand that this is the right approach if I expect a shock in the variable to only have an effect on the future values of the dependent variable and not on the preceding ones.
I'd be very grateful for any comments - whether I should include something else as well, or if this is ok.
Thank you and kind regards,
WH
0 Response to Doing Diff-in-Diff with lagged dependent using xtabond
Post a Comment