Year and industry dummy

[Larissa MM]

Hi, I want to run a fixed effect regression model with both year and industry fixed effect.

I would like to know if I am doing it correctly.

 xtreg mtd prof size tang growth liq dc1 dc2 dc3 dc4 i.year, fe

Fixed-effects (within) regression               Number of obs     =      4,820
Group variable: induscode                       Number of groups  =         10

R-sq:                                           Obs per group:
     within  = 0.2796                                         min =         80
     between = 0.5962                                         avg =      482.0
     overall = 0.2991                                         max =      1,630

                                                F(18,4792)        =     103.34
corr(u_i, Xb)  = 0.1301                         Prob > F          =     0.0000

------------------------------------------------------------------------------
         mtd |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
        prof |  -.3832766   .0236579   -16.20   0.000    -.4296569   -.3368963
        size |   .0361722   .0020567    17.59   0.000     .0321402    .0402042
        tang |   .1248675   .0139474     8.95   0.000     .0975241    .1522108
      growth |  -.0083867   .0007413   -11.31   0.000    -.0098399   -.0069335
         liq |  -.0102962   .0004226   -24.37   0.000    -.0111247   -.0094678
         dc1 |   .0615475   .0095531     6.44   0.000     .0428191     .080276
         dc2 |   .0420181   .0074701     5.62   0.000     .0273732     .056663
         dc3 |   .0213741   .0079806     2.68   0.007     .0057284    .0370199
         dc4 |   .0781161   .0110077     7.10   0.000      .056536    .0996963
             |
        year |
       2009  |  -.0585273   .0120194    -4.87   0.000    -.0820909   -.0349637
       2010  |  -.0834637   .0120274    -6.94   0.000     -.107043   -.0598844
       2011  |  -.0741264   .0120149    -6.17   0.000    -.0976812   -.0505716
       2012  |  -.0744085   .0120271    -6.19   0.000    -.0979871   -.0508299
       2013  |  -.1149802    .012032    -9.56   0.000    -.1385683    -.091392
       2014  |  -.1233646   .0120518   -10.24   0.000    -.1469917   -.0997376
       2015  |  -.1275209   .0120579   -10.58   0.000    -.1511599    -.103882
       2016  |  -.1244362   .0120844   -10.30   0.000    -.1481271   -.1007453
       2017  |   -.132976   .0121247   -10.97   0.000    -.1567459   -.1092061
             |
       _cons |   .0220694   .0284766     0.78   0.438    -.0337578    .0778966
-------------+----------------------------------------------------------------
     sigma_u |  .07936114
     sigma_e |  .18634853
         rho |  .15352465   (fraction of variance due to u_i)
------------------------------------------------------------------------------
F test that all u_i=0: F(9, 4792) = 46.85                    Prob > F = 0.0000

[Carlo Lazzaro]

Melissa:
technically speaking, you simply included a time categorical predictor in the right-hand side of your -fe- regression model.
This predictor will show, when adjusted for the remaining predictors, variation of the regressand within the same panel as time goes by.
This might be exactly what you want.
Conversely, if you want a double fixed effects model, you shoud consider the user-written command -reghdfe- (type -search reghdfe- to spot and install it).

[Phil Bromiley]

However, I strongly suspect that your model is identical to to what you will get from reghdfe. Regression with dummies for panel generally gives identical results to xtreg,fe.

I ran
use http://www.stata-press.com/data/r15/nlswork, clear
xtreg ln_wage grade age ttl_exp i.year ,fe
reghdfe ln_wage grade age ttl_exp ,absorb(year idcode)

The parameters on age and tt1_exp are identical in the two estimates. The only difference is that while xtreg drops grade because it is colinear, reghdfe estimates a parameter on grade but with an enormous (e+07) standard error.