Panel Data CEO-firm fixed effects

[Amy Johnson]

Dear Statalists,

I am new to Stata. I have one inquiry on performing the CEO-firm fixed effects for an example of panel data as shown below:

Year	CEO Name	Firm	CEO firm_id
1996	John	A	1
1997	John	A	1
1998	John	A	1
1999	John	B	2
2000	John	B	2
1996	Mary	C	3
1997	Mary	C	3
1998	Mary	C	3
1999	Mary	C	3
1996	Tim	D	4
1997	Tim	D	4
1998	Tim	D	4
1999	Tim	E	5
2000	Tim	E	5
2001	Tim	F	6
2002	Tim	F	6

Year refers to the calendar year, Firm is the name (for example) of the company the CEO works. CEO firm_id is the identifier which changes if a CEO changes his/her firm. For example, Tim acts as CEO for firm D from 1996-1998, and firm E from 1999-2000, and firm F from 2001-2002. Thus, the corresponding CEO firm_id for Tim is 4 when he stays in firm D, 5 in firm E, and 6 in firm F. On the other hand, Mary, who stays in firm C and never changes her firm, then the identifier remains to be 3.

My inquiry is: if I "xtset CEO firm_id" and then run the "xtreg...., fe" , can the regression results be viewed as CEO-firm fixed effects? (I will not consider year fixed effects here).

Many thanks in advance.

Best,
Amy

[FernandoRios]

Dear Amy
If you want do run a CEO-Firm Fixed effect model, i would recommend you to look into "reghdfe".
xtreg will not do what you want, since it will only use CEO as fixed effect, while thinking the firm_id is the time component.
HTH
Fernando

[Phil Bromiley]

In addition to using reghdfe, you should probably xtset your data. You can use egen with group to create separate id numbers for each firm-CEO and then xtset the data with that id and year. This will let you do lags and use factor variable notation (available for most Stata estimators but don't think they work in reghdfe)