the regression model

[Huda Amin]

Dear STATA community,

I am working to investigate the impact of board gender diversity on firm performance in listed companies in UK. The gender diversity in boardroom code implemented in 2012 and based on David and Alexander reviews, the female proportion in boardroom should reach 25% by 2015. I want to investigate the impact before 2015 and after 2015, so can anyone suggest what is the best method to analyzing it (FE, DiD, GMM) and how can I split the years (note that ALL companies follow the code from 2012 but I need to know how to separate it to 2 groups before and after reaching the 25% target).

Can you suggest the suitable command, please?

I look forward to reading your recommendation.

[Carlo Lazzaro]

Huda:
if all companies follow the code, a control group (that is necessary for DID) is hard to identify.
As you're probably interesdet in within-panel variation from before/after 2015, you can go -fe-.

[Huda Amin]

Hi Carlo,
Thanks for your reply. I agree with you, and I am using a fixed effect now. Can I use (reghdfe) command? Can you recommend anything related to this command, particularly since the observation number is small and by using fixed effect, the R- square is always very low, so I am going to use reghdfe instead.

[Huda Amin]

Dear Carlo

Can I use DiD method to investigate the impact before 2012 (when start using the boardroom diversity code) and after implementation it for the same listed firms?

[Carlo Lazzaro]

Huda:
1) -reghdfe- will give you back the same within-Rsq as -xtreg,fe-;
2) the two requirements that support DID are: a treated and a a control group; a before/after-a-given-event time dimension. While the latter seems to be fulfilled in yur example, I'm a bit queasy about the former.