Announcement

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

  • Triple Difference-in-Differences-in-Differences with interaction data only for treated group

    Hi StataList,

    I am new, but have read the guidelines so intend to abide by them. I am trying to conduct a triple Difference-in-Differences (DiD) where I have a matched sample of firms (treated and control firms) but only data for the triple interaction term for the treatment group of firms.

    Code:
    xtreg acquisition_spending i.treat##i.post##i.regulator_size i.cyear, fe
    First, does my overall code look okay?

    acquisition_spending = the amount of money firms spend on acquisitions by year
    treat = 1 for firms that were mentioned by a regulatory agency (regulator) in a court filing and 0 otherwise (for all matched control firms that did not experience a mention by a regulatory agency)
    post = 0 for pre-court filing dates and 1 thereafter
    regulator_size = the total number of employees of the regulator the filed the court filing against the treatment firm

    Second, my issue is that I only have data for the "regulator_size" variable for the treated firm, and no data for the control firms. This is because there is no regulatory agency in the first place to connect to the control (non targeted) firms. How do I solve this in my regression model?

    Thanks very much list!
    Eben K

  • #2
    I have been thinking more through this challenge. One solution I see is to do a sub-group analysis, where I only consider the treated firms with a HIGH level of regulator size compared to control firms, then do a separate model where I only consider the treated firms with a LOW level of regulator size compared to control firms? So i would have two regressions:

    Code:
    xtreg acquisition_spending i.treat##i.post i.cyear if regulator_size>50 | treat==0, fe
    xtreg acquisition_spending i.treat##i.post i.cyear if regulator_size<50 | treat==0, fe
    Then I could compare the coefficients of the main DID effect (interaction of treat X post) to see if the effect is larger/smaller in one case or the other. Does this work?

    Comment

    Working...
    X