how to control for individual fixed effects in Stata?

[Yao Zhao]

I know one method can be:

regress y x1 x2 i.id, vce(robust)

But the problem is in my data set, I have a lot of id. And it kills my Stata.

[Andrew Musau]

For panel data, you want to cluster your standard errors at the individual level. The -robust- option is thus valid for xtreg but not regress.

regress y x1 x2, absorb(id) cluster(id)
xtset id time
xtreg y x1 x2, fe cluster(id)

[Yao Zhao]

I don't find absorb option in -help regress-. I only find absorb option in -help areg-.

[Carlo Lazzaro]

Yao:
surely Andrew meant -areg- instead of -regress-.

[David HajHooj]

Hello Yao,
ssc install reghdfe
For regression:

reghdfe x y1 y2 y3, a(id) vce(cluster id)

[Carlo Lazzaro]

If Yao has many -panelid- but a fixed effect only, -areg-, -xtreg- and the community-contributed programme -reghdfe- give back similar results:

. use "https://www.stata-press.com/data/r16/nlswork.dta"
(National Longitudinal Survey.  Young Women 14-26 years of age in 1968)

. areg ln_wage age, absorb(idcode) vce(cluster idcode)

Linear regression, absorbing indicators         Number of obs     =     28,510
Absorbed variable: idcode                       No. of categories =      4,710
                                                F(   1,   4709)   =     738.02
                                                Prob > F          =     0.0000
                                                R-squared         =     0.6636
                                                Adj R-squared     =     0.5970
                                                Root MSE          =     0.3035

                             (Std. Err. adjusted for 4,710 clusters in idcode)
------------------------------------------------------------------------------
             |               Robust
     ln_wage |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
         age |   .0181349   .0006675    27.17   0.000     .0168262    .0194436
       _cons |   1.148214   .0193889    59.22   0.000     1.110202    1.186225
------------------------------------------------------------------------------

. xtset idcode year
       panel variable:  idcode (unbalanced)
        time variable:  year, 68 to 88, but with gaps
                delta:  1 unit

. xtreg ln_wage age, fe vce(cluster idcode)

Fixed-effects (within) regression               Number of obs     =     28,510
Group variable: idcode                          Number of groups  =      4,710

R-sq:                                           Obs per group:
     within  = 0.1026                                         min =          1
     between = 0.0877                                         avg =        6.1
     overall = 0.0774                                         max =         15

                                                F(1,4709)         =     884.05
corr(u_i, Xb)  = 0.0314                         Prob > F          =     0.0000

                             (Std. Err. adjusted for 4,710 clusters in idcode)
------------------------------------------------------------------------------
             |               Robust
     ln_wage |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
         age |   .0181349   .0006099    29.73   0.000     .0169392    .0193306
       _cons |   1.148214   .0177153    64.81   0.000     1.113483    1.182944
-------------+----------------------------------------------------------------
     sigma_u |  .40635023
     sigma_e |  .30349389
         rho |  .64192015   (fraction of variance due to u_i)
------------------------------------------------------------------------------


. reghdfe ln_wage age, absorb(idcode) vce(cluster idcode)
(dropped 551 singleton observations)
(converged in 1 iterations)

HDFE Linear regression                            Number of obs   =     27,959
Absorbing 1 HDFE group                            F(   1,   4158) =     884.06
Statistics robust to heteroskedasticity           Prob > F        =     0.0000
                                                  R-squared       =     0.6540
                                                  Adj R-squared   =     0.5936
                                                  Within R-sq.    =     0.1026
Number of clusters (idcode)  =      4,159         Root MSE        =     0.3035

                             (Std. Err. adjusted for 4,159 clusters in idcode)
------------------------------------------------------------------------------
             |               Robust
     ln_wage |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
         age |   .0181349   .0006099    29.73   0.000     .0169391    .0193307
------------------------------------------------------------------------------

Absorbed degrees of freedom:
---------------------------------------------------------------+
 Absorbed FE |  Num. Coefs.  =   Categories  -   Redundant     |
-------------+-------------------------------------------------|
      idcode |            0            4159           4159 *   |
---------------------------------------------------------------+
* = fixed effect nested within cluster; treated as redundant for DoF computation

[Andrew Musau]

Carlo Lazzaro, the -absorb- option in regress is undocumented. I suspect that the estimator implements areg if you specify this option.

. use "https://www.stata-press.com/data/r16/nlswork.dta"
(National Longitudinal Survey.  Young Women 14-26 years of age in 1968)

.
. regress ln_wage age, absorb(idcode) vce(cluster idcode)

Linear regression, absorbing indicators         Number of obs     =     28,510
                                                F(0, 4709)        =          .
                                                Prob > F          =          .
                                                R-squared         =     0.6636
                                                Adj R-squared     =     0.5970
                                                Root MSE          =     .30349

                             (Std. Err. adjusted for 4,710 clusters in idcode)
------------------------------------------------------------------------------
             |               Robust
     ln_wage |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
         age |   .0181349   .0006675    27.17   0.000     .0168262    .0194436
       _cons |   1.148214   .0193889    59.22   0.000     1.110202    1.186225
------------------------------------------------------------------------------

[Carlo Lazzaro]

Andrew Musau :
many thanks for pointing this out.
Admittedly, I was not aware of the availability of the -absorb()- option for -regress-, too.
As a consequence, my previous reply #4 reveals my lack of knowledge about this option and is not helpful for the original poster.

[Andrew Musau]

Your example in #7 is very helpful as it illustrates that you get the same results whether you use areg, xtreg or reghdfe. Your remark in #4 may very well be true for earlier versions of Stata. I have only checked this in Stata 16.

[Carlo Lazzaro]

Andrew Musau:
very interesting thread.
Thanks, as usual, for your contributions.

[Yao Zhao]

whether you use areg, xtreg or reghdfe doesn't affect your point estimate. But the standard error still have a little bit difference.

[Carlo Lazzaro]

Yao:
yes, I meant that point estimates are the same. As expected, standard errors differ.