Announcement

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

  • Industry Dummy Variables and independent variables

    I am conducting a cross-sectional study on UK firms and aim to find (if any) a relationship between performance measures e.g ROA and board diversity while controlling for firm and board characteristics by industry. I have created dummy variables for each industry category. I want to know how I can regress my performance measures against the diversity of each industry category. Would this require creating new variables for the average of each diversity measure for each industry?

    I am also unsure of what to infer from the coefficients in the following regression
    Code:
    reg roa div fsize bsize age i.industry, robust
    Where fsize bsize and age are my control variables, roa is my dependent variable, and div is the independent variable with the coefficient of interest.

    I am unsure as to what exactly the coefficients represent in the attached photo below
    Click image for larger version

Name:	Screenshot 2019-05-10 at 20.33.52.png
Views:	1
Size:	196.7 KB
ID:	1497789

    Last edited by Daniel Kosoko; 10 May 2019, 13:59. Reason: unclear about variables

  • #2
    What do you mean by the diversity of each industry category? Do the variables fsize, bsize, and age represent that concept, or are there other variables you haven't mentioned that characterize it?

    Comment


    • #3
      diversity is a measure of the proportion of directors belonging to a demographic. (fsize bsize age) are my control variables. I want to regress roa on the average diversity of the firms in each industry category.

      Comment


      • #4
        Then, yes, you do have to calculate average values for those diversity variables. -egen, mean()- will do that for you. (Read -help egen- and then scroll down to the -mean- function for details if this is not familiar to you.)

        Code:
        by industry, sort: egen mean_diversity = mean(diversity)
        
        regress roa div fsize bsize age mean_diversity, vce(robust)
        Do not try to also include i.industry in this: you will have a colinearity between the mean_diversity values and i.industry if you try, and some of those variables will got omitted. Since you are interested in the diversity variable specifically, it is better to just omit i.industry rather than have Stata arbitrarily pick something to omit (as it might well choose to omit the diversity variable!)

        Comment

        Working...
        X