Hausman test different outcomes

[Sydney Niezen]

Hello,

I am doing research about how CEOs with youth sports background influence exploration and exploitation behaviours. To test if I need fixed effects or random effects, I perform the Hausman test. However I get to different outcomes for my two hypothesis. The first is if CEOs with youth sports background are positively associated with exploration and the second with exploitation activities.

xtreg Explore_Uit_orig sport_background age gender ceo_tenure firm_size rd_intensity leverage roa i.industry i.year , fe
estimates store fe_Explore_Uit_orig

xtreg Explore_Uit_orig sport_background age gender ceo_tenure firm_size rd_intensity leverage roa i.industry i.year, re
est store re_Explore_Uit_orig

hausman fe_Explore_Uit_orig re_Explore_Uit_orig
hi2(13) = 33.90
prob> chi2 = 0.0012 so FE is preferred
** EXPLOITATION

xtreg Exploit_Uit_orig sport_background age gender ceo_tenure firm_size rd_intensity leverage roa i.industry i.year, fe
est store fe_Exploit_Uit_orig

xtreg Exploit_Uit_orig sport_background age gender ceo_tenure firm_size rd_intensity leverage roa i.industry i.year, re
est store re_Exploit_Uit_orig

hausman fe_Exploit_Uit_orig re_Exploit_Uit_orig
chi2(13) = 18.61
Prob > chi2 = 0.1359, RE is preferred.

Does anyone have an idea on what is the preferred effect?

[Jeff Wooldridge]

I’m wondering how you’re using FE in this setting because I’d think the variable measuring youth sports background wouldn’t vary across time for adults.

[Jeff Wooldridge]

I’m wondering how you’re using FE in this setting because I’d think the variable measuring youth sports background wouldn’t vary across time for adults.

[Carlo Lazzaro]

Sydney:
As Jeff pointed out, whatever panel-related background (as it refers to the past) is time-invariant.
But what remains to be seen is your research question, as the two dependent variables are pretty different.