Announcement

Collapse
No announcement yet.
X
  • Filter
  • Time
  • Show
Clear All
new posts

  • reg or xtreg?

    Hello Stata-Users,

    I have a Panel-data and i would like to estimate the impact of a treatment on firms. I use the DID estimator.

    Here is my regression:

    y time_dummy group_dummy interaction_time_group controls

    In addition, i would like to add progressively several fixed effects. ( i.year ; i.location_of_the_firm ; i.industry_code )

    Which command is the most appropriate in my case? -reg- or -xtreg- with fe option?

    Thank you for your help,

    Pierre

  • #2
    It is hard to comment, as what is 'appropriate' depends (much) more on your research question and theory than on anything else. Why do you want to control for fixed-effects progressively?

    If implemented correctly, both the regress command with indicator variables (LSDV estimator) and xtreg with fe option (fixed-effcets model) will give the (same) within estimates. Since we are usually not interested in the unit-specific constants (or there are just to many of them), we apply the within-transformation by specifying the xtreg command with fe option.

    Note that the concept of DID is equivalent to the fixed-effects (within) estimator with controls for time-period. If you have more than two periods you might wan to consider fitting a first differences model, depending on the theoretical assumptions about the effect of the treatment.

    Best
    Daniel

    Comment

    Working...
    X