[Elizabete Lurie]

My supervisor was convinced that I should be able to use time fixed effects as there should be enough degrees of freedom, soI run the regression. I am just not exactly sure how to interpret the output. Are the relevant coefficients summarized above, and then what are the year specific coefficients indicating? Thank you in advance!

This is the regression I run and the output:

 xtreg InvWinsor i.CR_POM##i.Tech_Industry Size Profitability i.D, fe
note: 1.Tech_Industry omitted because of collinearity

Fixed-effects (within) regression               Number of obs      =      5715
Group variable: GlobalComp~y                    Number of groups   =       777

R-sq:  within  = 0.0302                         Obs per group: min =         1
       between = 0.0083                                        avg =       7.4
       overall = 0.0166                                        max =        29

                                                F(37,4901)         =      4.13
corr(u_i, Xb)  = -0.4124                        Prob > F           =    0.0000

--------------------------------------------------------------------------------------
           InvWinsor |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
---------------------+----------------------------------------------------------------
            1.CR_POM |   .0178671    .055042     0.32   0.745    -.0900398     .125774
     1.Tech_Industry |          0  (omitted)
                     |
CR_POM#Tech_Industry |
                1 1  |   .0604338   .1094199     0.55   0.581    -.1540783    .2749458
                     |
                Size |  -.6470393    .112147    -5.77   0.000    -.8668976    -.427181
       Profitability |   2.480885   .3963357     6.26   0.000      1.70389    3.257881
                     |
                   D |
               1982  |  -.5369541   1.960509    -0.27   0.784    -4.380429    3.306521
               1983  |   -.538144   1.960509    -0.27   0.784    -4.381621    3.305333
               1984  |  -.1877277   1.960527    -0.10   0.924    -4.031239    3.655784
               1985  |   .1194069   1.438708     0.08   0.934    -2.701105    2.939919
               1986  |   .1997108   1.436548     0.14   0.889    -2.616567    3.015989
               1987  |   .3664296   1.436989     0.25   0.799    -2.450712    3.183572
               1988  |    .178331   1.437071     0.12   0.901    -2.638972    2.995634
               1989  |   .1695821   1.437041     0.12   0.906    -2.647662    2.986826
               1990  |   .0375606   1.437186     0.03   0.979    -2.779968    2.855089
               1991  |   .3115575   1.437327     0.22   0.828    -2.506248    3.129363
               1992  |   .3470687   1.437634     0.24   0.809    -2.471339    3.165476
               1993  |   .3045631   1.437907     0.21   0.832    -2.514378    3.123504
               1994  |   .4245565   1.438361     0.30   0.768    -2.395277     3.24439
               1995  |   .4616731   1.438715     0.32   0.748    -2.358853    3.282199
               1996  |   .4550687   1.438986     0.32   0.752    -2.365988    3.276126
               1997  |   .6407996   1.438933     0.45   0.656    -2.180154    3.461753
               1998  |   .3358685   1.439193     0.23   0.815    -2.485595    3.157332
               1999  |   .4356521   1.439473     0.30   0.762    -2.386359    3.257664
               2000  |   .5243281   1.439975     0.36   0.716    -2.298668    3.347324
               2001  |   .2911237   1.439033     0.20   0.840    -2.530025    3.112273
               2002  |   .4913246    1.43894     0.34   0.733    -2.329642    3.312291
               2003  |   .5176196   1.439023     0.36   0.719     -2.30351     3.33875
               2004  |   .8726261   1.439405     0.61   0.544    -1.949254    3.694506
               2005  |   .7534948   1.439769     0.52   0.601    -2.069097    3.576087
               2006  |   .7115173   1.440175     0.49   0.621     -2.11187    3.534905
               2007  |   .7458274   1.440475     0.52   0.605    -2.078149    3.569804
               2008  |   .2385598   1.440833     0.17   0.869    -2.586119    3.063239
               2009  |   .4700265   1.441173     0.33   0.744    -2.355318    3.295371
               2010  |   .9485511   1.441537     0.66   0.511    -1.877507    3.774609
               2011  |   .6591929    1.44191     0.46   0.648    -2.167596    3.485982
               2012  |   .5360067    1.44201     0.37   0.710    -2.290979    3.362993
               2013  |   .8487168   1.442369     0.59   0.556    -1.978974    3.676407
               2014  |    .407778   1.500747     0.27   0.786    -2.534359    3.349915
                     |
               _cons |   5.723433    1.74418     3.28   0.001     2.304058    9.142807
---------------------+----------------------------------------------------------------
             sigma_u |  1.3131517
             sigma_e |  1.3862796
                 rho |  .47292974   (fraction of variance due to u_i)
--------------------------------------------------------------------------------------
F test that all u_i=0:     F(776, 4901) =     1.63           Prob > F = 0.0000

[Matt Roberts]

A fixed-effects model where a main effect is collinear with the fixed effect is almost the only situation in which it is OK to have an interaction term and omit one of the main effects.

Apologies for bringing this thread back to life but I have a similar problem and I was interested by this comment. Is there any literature that backs up this statement? I was of the understanding that even in a fixed effects model you need to include all main effects of an interaction term and if one was collinear (e.g. due to time invariance) then you needed to find another method.

[Carlo Lazzaro]

Elizabete:
i.D says that years have no significant effect in explaining the variation of yir dependent variable; hence, nothing change across the years (taking 1981 as the reference category).
However, this result might depend on how did you -xtset- your data.
As far as I can see that table output, the only relevant predictors in your model are profitability and size.
I do not know whether this is in line with your expectations and/or more relevant, with what the majority of the literature in your research field claims.