random effects or OLS

I am really sorry but I am quite confused.

I have panel data (200 firms over the time period of 2002 - 2014). I ran a test whether I should use fe or re; and I think I got the result of random effects, which makes sense since my sample is drawn from different industries. The result made me think I need to use random effects.

code:

. hausman fe re

             ---- Coefficients ----
                (b)          (B)            (b-B)     sqrt(diag(V_b-V_B))
                 fe           re         Difference          S.E.

zchangeinROA  -.0078418    -.0094914        .0016495        .0002656
Narcissimi~    .004696    -.0003595        .0050556        .0030671
interaction   -.0008962    -.0005818       -.0003144        .0003695
zlntotal_a~    .0674206    .0044034        .0630172        .0163112
zlnrevenue    -.0052717    -.001084       -.0041878        .0122064
zrole       -.0097101    -.0096499       -.0000602         .001937

    b = consistent under Ho and Ha; obtained from xtreg
 B =    inconsistent under Ha, efficient under Ho; obtained from xtreg

Test:  Ho:  difference in coefficients not systematic

chi2(6) = (b-B)'[(V_b-V_B)^(-1)](b-B)
        =        3.92
Prob>chi2 =      0.6881
(V_b-V_B is not positive definite)

Now I did a Breusch and Pagan test to see if I could use pooled OLS instead of random effects. As I understood it, I reject the Null and thus need to do a random effects. Could you help me if I did the right decisions? This is for my thesis, so any help or explanatio is appreciated.

Breusch and Pagan Lagrangian multiplier test for random effects

workforcechange[firm_ID,t] = Xb + u[firm_ID] + e[firm_ID,t]

Estimated results:
| Var sd = sqrt(Var)
---------+-----------------------------
workfor~e | .0253792 .1593084
e | .0232228 .1523903
u | .0019666 .0443464

Test: Var(u) = 0
chibar2(01) = 88.06
Prob > chibar2 = 0.0000

I posted the same question here; but have not gotten a reply yet.

http://ift.tt/293lf0L