Panel Regression with a Moderator Variable

[john matthews]

*Please forgive my lack of Stata/Statistics knowledge or basic questions, I'm fairly new to Stata and statistics in general. I've tried other avenues of learning and understanding it but it's all been a bit overwhelming thus far so I'm hoping someone can point me in the right direction*

To give a brief background, for my research I am assessing whether the presence of a Chief Sustainability Officer (CSO) will affect the Sustainability Performance of a firm. Also, I would like to test whether Degree of Firm Multinationality has a Moderator effect on this.

Data
32 Firms
(DV) Sustainability Performance: Measured as a score between 0 and 100. There are two measurements one in 2012 and 2017.
(IV) CSO Presence: Measured as a Yes/No in 2012 and 2017. Currently, all firms have a 'Yes' for both 2012 and 2017.
(Moderator IV) Degree of Firm Multinationality: Measured between 0 and 1 using the firm's (Foreign Tax/Total Tax) for 2012 and 2017.

Control Variables: Firm Size, Firm Age

I used a youtube tutorial to set up my data, import it to Stata and run a panel regression for the effect of CSOpresence on Sustainability Performance. But CSOpresence was omitted due to collinearity. Is there something I can do to solve this? Are there any other tips on how I can setup my data etc that will help me run this panel regression? Are there some other things I should do? I'm also unsure if a fixed effect or random effect regression is suited for this.

Literally any and all advice would be much appreciated!

Also, if there is a problem regarding the data, my research is flexible so I am able to change things such as years of measurements etc (as long as it isn't the variables themselves).

Sorry for the long post and quite basic questions, but I hope someone can help me out. Thanks in advance!

[Carlo Lazzaro]

John:
welcome to this forum.
Some comments about your post:
- you probably used -xtreg,fe- and, as such, the -fe- machinery wiped out the invariant predictor -CSOpresence-. As you can easily envisage yourself, the recommendation included among the FAQ to share (via CODE delimiters) what you typed and what Stata gives you back makes a lot of sense, because the suggested approach reduces the guess-work of interested listers, who may well be off target;
- -hausman- test is the usual tool to compare -fe- vs -re- specifications (as long as if you use default standard errors);
- as far as the set of your predictors is concerned, you should skim through the literature of your research field to be sure you gave a fair and true representation of the data generating process you're investigating;
- you do not tell us whether or not a postestimation routine has been performed (eg, did you check whether your regression model suffer from misspecification?);
- from the -xtreg- entry you can retrieve a list of helpful panel data econometrics textbooks. In addition, Stata listers dealing with this stuff are usually fond of https://www.stata.com/bookstore/micr...metrics-stata/