Hi, I am running a FE regression to study the effect of diversification (GSDiv) on firm performance ( ROA_win05) amongst US stocks. Strangely, the diversification variable is getting dropped out due to collinearity. I find this strange, because diversification at a firm level is not identical across the years. Following is an output of my regression and also an extract of my data to show that diversification is not a time invariant value. Any insights or suggestions from this community will be very helpful for me in understanding the results and tweaking my model.

Output of my regression model:
Code:
. xtreg ROA_win05 LnAssets TDTE CoAge c.l1.GSDiv##c.l1.GSDiv##Crisis if SICInclude ==1 & Excl_by
> nd_delisting !=1 & Year !=7 & Year !=8 & Year !=9, fe
note: cL.GSDiv#cL.GSDiv omitted because of collinearity
note: 1.Crisis#cL.GSDiv#cL.GSDiv omitted because of collinearity

Fixed-effects (within) regression               Number of obs     =     28,257
Group variable: ID                              Number of groups  =      3,779

R-sq:                                           Obs per group:
     within  = 0.0832                                         min =          1
     between = 0.2222                                         avg =        7.5
     overall = 0.1183                                         max =         18

                                                F(6,24472)        =     370.38
corr(u_i, Xb)  = -0.3518                        Prob > F          =     0.0000

------------------------------------------------------------------------------------------
               ROA_win05 |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------------------+----------------------------------------------------------------
                LnAssets |  -2.357783   .0583975   -40.37   0.000    -2.472246    -2.24332
                    TDTE |   1.85e-07   5.84e-07     0.32   0.751    -9.59e-07    1.33e-06
                   CoAge |   .0062312   .0097091     0.64   0.521    -.0127991    .0252616
                         |
                   GSDiv |
                     L1. |  -4.02e-17   1.47e-16    -0.27   0.784    -3.28e-16    2.48e-16
                         |
       cL.GSDiv#cL.GSDiv |          0  (omitted)
                         |
                1.Crisis |    .417749   .1020776     4.09   0.000     .2176708    .6178273
                         |
         Crisis#cL.GSDiv |
                      1  |   1.13e-16   1.91e-16     0.59   0.556    -2.62e-16    4.88e-16
                         |
Crisis#cL.GSDiv#cL.GSDiv |
                      1  |          0  (omitted)
                         |
                   _cons |   23.62404   .4110291    57.48   0.000      22.8184    24.42968
-------------------------+----------------------------------------------------------------
                 sigma_u |   8.544612
                 sigma_e |  4.2479932
                     rho |     .80182   (fraction of variance due to u_i)
------------------------------------------------------------------------------------------
F test that all u_i=0: F(3778, 24472) = 16.15                Prob > F = 0.0000
Here is an extract of my data:
Code:
* Example generated by -dataex-. To install: ssc install dataex
clear
input int(ID Year) double ROA_win05 float LnRev byte SICInclude double GSDiv
2162 2001 4.66 5.306781 1  .3228801767137001
2162 2002 8.14 5.372961 1  .3380697720011363
2162 2003 8.67 5.403128 1  .3094243238644552
2162 2004 9.96 5.519459 1  .3094243238644552
2162 2005 9.13 5.627981 1 .28362709611685805
2162 2006 8.63 5.689683 1 .29554141367598213
2162 2007 9.59 5.792099 1  .3048453732817755
2162 2008 7.49 5.835688 1  .2763202052461048
2162 2009 10.9 5.933306 1 .25363894692169137
2162 2010 16.8 6.082675 1 .13124869920165286
2162 2011 15.7 6.241055 1 .12415397163303724
2162 2012 16.5 6.346338 1 .17110928343404325
2162 2013   17 6.499035 1  .1985152433458725
2162 2014 16.1 6.622071 1 .20436253747814215
2162 2015 15.5 6.688354 1 .23243549494150262
2162 2016 15.8 6.756467 1 .21863185219221729
2162 2017 14.9 6.805944 1 .23243549494150262
2162 2018 12.3 6.881103 1 .23243549494150262
2162 2019 6.55 6.899017 1 .23243549494150262
end