Announcement

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

  • Fixed Effect and Treatment Analysis

    Hello Dear All,

    I have a fixed effect model to measure the efffect of parental leave on motherhood wage gap.

    My data is merged cross-sectionals. Not a panel data, there is no individual observed two or more points in time.

    Here how I coded fixed effect model.

    Code:
    reg lnincome mother age education occupation i.posttreat i.surveyyear
    Do you think that model measures treatment effect because I dont have individual fixed effect ? Because my data is not a panel data I found this solution to apply fixed effect to cross-sectional.

    And most importantly I want to see the effect of treatment on mother variable?

    Code:
     reg lnincome  age education occupation i.posttreat##i.mother i.surveyyear
    Do you think that it will work ? I have treated mother non- treated mothers . But I don't have any non-mother that treated.

    What about if I wanna see the effect of treatment on different educational group ?

    Code:
    reg lnincome  mother age  occupation i.posttreat##i.education i.surveyyear
    margins educations , dydx(posttreat)
    Does this code provides me the info i am trying to get?

    Thanks in advance.

  • #2
    Hazal:
    are you sure that you're not dealing with a kind of diff-in-diff design?
    Kind regards,
    Carlo
    (Stata 18.0 SE)

    Comment


    • #3
      Originally posted by Carlo Lazzaro View Post
      Hazal:
      are you sure that you're not dealing with a kind of diff-in-diff design?
      Dear Carlo,

      I think you have a point. As DID model with fixed effect . Do you think that the code and design serves for its purpose?

      Thanks in advance.

      Comment


      • #4
        Hazal:
        usually a DID design has two groups (say, treated and untreated) and a pre/post whatever.
        The latter seems to be already present in your code, whereas I fail to get the former.
        Kind regards,
        Carlo
        (Stata 18.0 SE)

        Comment


        • #5
          Originally posted by Carlo Lazzaro View Post
          Hazal:
          usually a DID design has two groups (say, treated and untreated) and a pre/post whatever.
          The latter seems to be already present in your code, whereas I fail to get the former.
          To my knowledge there is DID design in two forms :

          1) Y=a0+a1*TREAT+a2*POST+a3*TREAT_POST+e
          2) Y=a0+a1*TREAT_POST+time fixed effects+firm fixed effects
          Code:
           
           reg lnincome mother age education occupation i.posttreat i.surveyyear
          Because I have merged cross sectional data. I have prefered the second version. In some websites, I have seen second version called as generalized DID design. But I have only survey year fixed effect and no other. The code that I employed is as above. So I wonder if ı am calling my model right and coding is accurate.

          Comment

          Working...
          X