Eviews 5.1: panel unit root and hausman test

[lnp3]

dear forum,

I am currently using Eviews 5.1.. I am trying to decide between fixed and random effect estimator for a panel data analysis of bank profitability over the period 2000-2010

first, i am doing the panel unit root test (breitung, common root) for each variable in my estimation.. then I will conduct hausman test.

however, I am not sure about the order of tests. which test should I do first? panel or hausman?

1) I identified the variables with roots in levels. If unit roots exist, I will estimate with first (or second difference operator if necessary)

for example: this is going to be my estimation equation (without knowing fixed or random effect).

D((roa)) c d((eff)) d(lev) d(cris) d(liq) d((ta)) gdperc hhi d(mg) d((rsize)) mcapgdp d((nonixea))

2) However, when I am doing the Hausman test to decide which effect to use (fixed or random?), should I estimate it with levels only or first/second difference operators? In order words, should I filter the unit root effect in Hausman test as well?

Thank you so much in advance for your help.


Aysen
:?