[Carlo Lazzaro]

Victor:
as per your low R.sq within and most of your predictors showing wide 95% Cis I wonder whether your model is correctly specified.
I would first investigate whether -fe- specification is the right one for your regression model. As you invoked non-default standard errors, -hausman- is out of debate there. You should rely on the glorious community-contributed command -xtoverid- which, in turn, being a bit old-fashioned, does not support -fvvarlist- notation. The usual fix is to prefix your -xtreg- code with -xi:-. If -xi:- does not do the trick, you should calculate interaction by hand (provided it makes sense to keep them).

[Victor Stuart-Smith]

 Model VCE    : Robust

Expression   : Linear prediction, predict()
dy/dx w.r.t. : lnta nim roaa assetg llrtgl depositg lev gdpg LCRpolicy

------------------------------------------------------------------------------
             |            Delta-method
             |      dy/dx   Std. Err.      z    P>|z|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
lnta         |
    emerging |
          0  |   .1985668   .2830566     0.70   0.483    -.3562139    .7533476
          1  |  -.1321813    .112635    -1.17   0.241    -.3529418    .0885792
-------------+----------------------------------------------------------------
nim          |
    emerging |
          0  |   .6450914    8.00381     0.08   0.936    -15.04209    16.33227
          1  |    .623683   6.878041     0.09   0.928    -12.85703     14.1044
-------------+----------------------------------------------------------------
roaa         |
    emerging |
          0  |   1.412726   1.977242     0.71   0.475    -2.462598     5.28805
          1  |  -4.543483   10.50964    -0.43   0.666      -25.142    16.05503
-------------+----------------------------------------------------------------
assetg       |
    emerging |
          0  |    .302561   .2780425     1.09   0.277    -.2423924    .8475144
          1  |  -.1158953   .2252908    -0.51   0.607    -.5574571    .3256665
-------------+----------------------------------------------------------------
llrtgl       |
    emerging |
          0  |   .3434655   .0809187     4.24   0.000     .1848678    .5020633
          1  |  -.3583853   1.031057    -0.35   0.728    -2.379219    1.662449
-------------+----------------------------------------------------------------
depositg     |
    emerging |
          0  |  -.1121131   .0264255    -4.24   0.000    -.1639062   -.0603201
          1  |    .139336   .2030365     0.69   0.493    -.2586082    .5372802
-------------+----------------------------------------------------------------
lev          |
    emerging |
          0  |    3.14852   1.674282     1.88   0.060    -.1330114    6.430052
          1  |   .3645413   .3024783     1.21   0.228    -.2283052    .9573879
-------------+----------------------------------------------------------------
gdpg         |
    emerging |
          0  |  -.0798069   .2180151    -0.37   0.714    -.5071087    .3474948
          1  |   4.558842   2.357175     1.93   0.053    -.0611364     9.17882
-------------+----------------------------------------------------------------
LCRpolicy    |
    emerging |
          0  |  -.0793828   .2034763    -0.39   0.696     -.478189    .3194233
          1  |  -.0793828   .2034763    -0.39   0.696     -.478189    .3194233
------------------------------------------------------------------------------

These are the results; What do you think? Thank you again it's been invaluable for me. (Ignore LCRpolicy I know what this is about)

[Victor Stuart-Smith]

Why is Hausman out for debate? The Hausman test said rejected the null; ie FE to be used.
The robust is just to control for heterscedasticity after the FE test was used. Im trying to not overcomplicate things.

Thanks

[Carlo Lazzaro]

Victor:
if you try running yourself -hausman- with non-default standard errors, your try will be unsuccessful:

. use "http://www.stata-press.com/data/r15/nlswork.dta"
(National Longitudinal Survey.  Young Women 14-26 years of age in 1968)

. quietly xtreg ln_wage age, fe vce(cluster idcode)

. estimates store fe

. quietly xtreg ln_wage age, re vce(cluster idcode)

. estimates store re

. hausman fe re
hausman cannot be used with vce(robust), vce(cluster cvar), or p-weighted data
r(198);

.

That said, performing -hausman- with default standard errors and then change them after -hausman- outcome, is methodologically wrong.
As previously replied, the community-contributed command -xtoverid- can do the trick (and, in this case, points the researcher to -fe- specification):

. xtoverid

Test of overidentifying restrictions: fixed vs random effects
Cross-section time-series model: xtreg re  robust cluster(idcode)
Sargan-Hansen statistic  14.529  Chi-sq(1)    P-value = 0.0001

Unfortunately, -xtoverid- being glorious but a bit old-fashioned, does no support -fvvarlist- notation.
The usual fix is to prefix your -xtreg- code with -xi:-. If -xi:- does not do the trick, you should calculate interaction by hand (provided it makes sense to keep them).