Panel Estimation and Model

[kafazi]

hello, i have a problem with panel estimation. the result of estimation is not good and seem bias ( like fix and random).
before i tried this fix or random, i used 3 step test to choose good model.

1. chow test
2. hausman test
3. Lm test

then, the result said random. but constant value of N-Cross same like FIX. according my database, IPM Banda aceh is greater than another regency. however, the estimation show another.
my lecturer give another option, is using Pool (Pool least square), but R2 = -137 and no one significant
how to estimate and choose model this data??

here i attach my work file.
N = 23
T = 4 (2010-2013)

thank you.
sorry for my bad language.