Interpretation of coefficients Fixed effect reg

[Abdullah Ijaz]

Hi,

I just need some help with the interpretation of the coefficients of panel fixed effects regression analysis as my output is shown below.

my dependent variable is log-transformed and it is ratio variable.

my main dependent variable is number of occurences annually.

Is the effect very low as you can see the coefficient value is quite low but still significant.

 xtreg tearate incidents procedureslog Gov GDPpercapitalog Credittoprivate FDI i.Year
> , fe

Fixed-effects (within) regression               Number of obs     =      1,201
Group variable: countrynum                      Number of groups  =        120

R-sq:                                           Obs per group:
     within  = 0.3002                                         min =          2
     between = 0.5829                                         avg =       10.0
     overall = 0.5590                                         max =         13

                                                F(18,1063)        =      25.33
corr(u_i, Xb)  = -0.5552                        Prob > F          =     0.0000

---------------------------------------------------------------------------------
        tearate |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
----------------+----------------------------------------------------------------
      incidents |  -.0003739   .0000709    -5.27   0.000    -.0005131   -.0002348
  procedureslog |  -.1283264   .0531598    -2.41   0.016    -.2326366   -.0240163
            Gov |    .450314   .1148559     3.92   0.000     .2249441     .675684
GDPpercapitalog |   .7937656    .155046     5.12   0.000     .4895346    1.097997
Credittoprivate |   .2573968   .0479212     5.37   0.000     .1633658    .3514278
            FDI |   .0084924   .0029515     2.88   0.004     .0027009    .0142839
                |
           Year |
          2007  |   .0667414   .0512528     1.30   0.193    -.0338269    .1673096
          2008  |   .0191358   .0514024     0.37   0.710    -.0817259    .1199976
          2009  |   .0231675   .0540206     0.43   0.668    -.0828316    .1291666
          2010  |   .0155554   .0524643     0.30   0.767    -.0873899    .1185008
          2011  |   .0552378   .0535749     1.03   0.303    -.0498868    .1603624
          2012  |   .1095555   .0550582     1.99   0.047     .0015204    .2175907
          2013  |   .1652532   .0555919     2.97   0.003     .0561709    .2743356
          2014  |   .1357808   .0563733     2.41   0.016     .0251653    .2463963
          2015  |     .11749   .0578124     2.03   0.042     .0040506    .2309294
          2016  |   .1796093   .0593656     3.03   0.003     .0631223    .2960963
          2017  |   .2224498   .0617384     3.60   0.000     .1013068    .3435927
          2018  |   .2581553   .0648637     3.98   0.000     .1308798    .3854308
                |
          _cons |  -7.593333   1.329511    -5.71   0.000     -10.2021   -4.984569
----------------+----------------------------------------------------------------
        sigma_u |  1.1740312
        sigma_e |  .31061184
            rho |  .93458244   (fraction of variance due to u_i)
---------------------------------------------------------------------------------
F test that all u_i=0: F(119, 1063) = 87.33                  Prob > F = 0.0000

[Carlo Lazzaro]

Abdullah:
if your regressand is continuous and logged, -incident- coefficient tells you that an 1-unit chage in -incident- reduces the regressand of:

. di  (exp(-.0003739)-1)*100
-.03738301

or about -0.037%.

[Abdullah Ijaz]

Carlo:

Many thanks for detailed response. Much appreciated.