Announcement

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

  • Multilevel fixed effects (-reghdfe-)

    Hello Statalisters,


    I Have a strongly balanced short panel (t=3) data and over 3000 observations in each panel. I am trying to run a village level fixed effects instead of a household fixed effect (I want the mean deviation transformation of the variables to be based on village means not household means but I also want to control for the household fixed effects). Here is my syntax:

    HTML Code:
       reghdfe y1 x1  [pweight=pw], a(village household) cluster(village)


    Am I doing it right? I cannot help it but think that I am not.

    I am also interested in the between villages effects estimation. I have mean collapsed my data by village and run OLS. Is there a built-in way of accessing the between effects for -eghdfe-?



    Any help is much appreciated.

    Yonatan

  • #2
    Do households move between villages? If not, the village fixed effect is redundant. You can obtain the fixed effects from reghdfe (see -help reghdfe-), but I think these are not consistently estimated, so whether you should is a big question.

    Comment


    • #3
      Thanks a lot Jesse. No. The households are nested within the villages.
      I am not interested in the fixed effect. I am rather interested in the ratio of the coefficient on one of my explanatory variables obtained from a within (fe) and a between effects regressions.

      would this ratio be consistent?

      Thanks again! Helped much!

      Comment


      • #4
        You can always replace the absorb part from (household) to (time), which effectively gives you the between regression and then calculate the ratio yourself. Whether this ratio is meaningful, I have no idea, but there might be literature about this?

        Comment


        • #5
          Thank you Jesse!
          You saved me a lot of for and gave me clarity.

          There is actually. It’s a contrast estimator for peer effects. Boozer & Cacciola (2001)

          Comment

          Working...
          X