Announcement

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

  • Panel data fixed effects with industry dummy

    Hi everyone,

    I am trying to replicate a study which is been made by Michaelas, Chittenden and Poutziouris in the paper “Financial Policy and Capital Structure Choice in U.K. SMEs: Empirical Evidence from Company Panel Data”.

    Here the link: https://link.springer.com/content/pd...8010724051.pdf

    In this paper the study is been conducted running a panel data fixed effect regression with time and industries dummies on a database of UK firm analysed for ten years.
    My database is composed by 5954 firm with data of ten years each (from 2008 to 2009). Since I extremely need both the values of time and industries dummies, I run this regression on Stata (See the two attachments).

    IND_* is the dummy variable that I created for identifying each industries.

    My question is:
    1. Since I already know why matematically the industries dummy is omitted due to collinearity (it is time invariat), why in the paper the authors can get the values from that dummies?
    2. Am I wrong in the way I am running the regression?
    3. If I am not wrong, Is there any chance to get those omitted values maintaing the panel fixed effect regression?
    Thank you very much for the answer!
    Attached Files

  • #2
    Giuseppe:
    In see nothing wrong in your regression.
    You may want to take a look at -search mundlak-.
    Kind regards,
    Carlo
    (Stata 18.0 SE)

    Comment


    • #3
      Carlo:
      First, I would like to thank you for your answer.
      Second, the -search mundlak- does not solve my problem. I need to estimate the coefficient of the time invariant dummy (the industry dummy “IND_*”) with the fixed effects regression. As you can see in the paper that I am referring to there is an estimation of the time invariant variable which is been made with the fixed effects.
      I still not able to figure it out they did that, so if you are other tips I will be happy to follow them.
      Thank you very much.

      Comment


      • #4
        Dear Giuseppe, It seems to me that the fixed effects in the paper are referred to industries rather than firms. As such, you should try:
        Code:
        reg y x1 x2 ... i.industry i.year, robust
        not xtreg.
        Ho-Chuan (River) Huang
        Stata 17.0, MP(4)

        Comment


        • #5
          River:
          If I use reg, than it will not be a fixed effect regression anymore and I need the fixed effects.

          Comment


          • #6
            Dear Giuseppe, Yes, it is a fixed-effect (in terms of industries, rather than, firms) regression. It is not uncommon in the finance literature.
            Ho-Chuan (River) Huang
            Stata 17.0, MP(4)

            Comment


            • #7
              River:
              Are you telling me that in the paper it’s been run a linear regression instead of a panel fixed effects?
              this thing make sense from the mathematical point of view (you cannot have coefficient for a time invariat variable) , but it is something that would surprise me a lot !

              Comment


              • #8
                Let me try to clarify. regress with panel dummies is precisely the same as xtreg, fe. You can't include i.ind in the regressors in xtreg because it is precisely colinear with the fixed effects. However, regress does not solve the colinearity problem that exists for variables that don't vary within panels. You can get the dummy parameters either in regress or by using predict after xtreg (see xtreg postestimation in documentation).

                Comment


                • #9
                  Dear Giuseppe, You can find many papers (finance field in particular) using the empirical strategy I mentioned above. (Please using Quote so that I will be informed)
                  Ho-Chuan (River) Huang
                  Stata 17.0, MP(4)

                  Comment

                  Working...
                  X