Announcement

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

  • Industry FE omitted due to collinearity

    Hi everyone,

    I have panel data with a list of company ID's, years 2010-2020 and the industry that each country is in. When I do i.industry to attempt to include industry FE, all of the industries are omitted due to collinearity. Does anyone know why this would happen? My code is:

    xtreg dv v1 v2 v3 v4 v5 i.year i.industry, fe

  • #2
    You don't say how you -xtset- your data, so I can't be sure what the -fe- option is doing. But I'm willing to guess that you did it with company id as the panel variable, so -fe- gives you company level fixed effects.

    Since any company remains in the same industry over time, industry is a time-invariant attribute of the company. It is a matter of linear algebra that in a fixed-effects regression it is impossible to identify effects of time-invariant attributes of the panel effect variable due to colinearity. If you would like to see the colinearity directly, run:
    Code:
    regress industry i.company
    and you will see that you get an R2 = 1.0 (or very close due to rounding errors) and the coefficient table will show you the exact linear relationship between industry id and company.

    So the question now is why do you want to include i.industry in your model. If you just wish to adjust for its effects, but have no need to estimate industry-level effects to answer your research question, then just forget about it--all time invariant attributes of company, including industry and also including time-invariant attributes you haven't measured or even thought of, are automatically adjusted for by the company fixed effects themselves. If, on the other hand, you actually need to estimate industry level effects, then you can't do that with a company-fixed-effects model and will have to use some other approach.

    Comment


    • #3
      Hi Clyde, thank you for your response.

      I use xtset CompanyID year. I do not need to estimate industry level effects, I am just trying to ensure that my regression is correct. Do you think I just ignore this as xtreg , fe will incorporate industry FE?

      Comment


      • #4
        Do you think I just ignore this as xtreg , fe will incorporate industry FE?
        Absolutely, yes.

        Comment

        Working...
        X