I have yet to see the same results for Pooled OLS and FE, as the variables are different and the setup is different. For Pooled OLS, in my experience, you don't want to define a panel (that's why you are using "time" indicators in a traditional OLS framework, to account for that). Once you set up your panel, you wouldn't use all of those indicators in your estimation. Also, the RE may be better because you have little to no within-group variation (variation used for FE).

That said, I would suggest your defining the panel for FE and RE current, then ensure any time varying variables are either excluded or included (depends on the model - no time-invariant variables for FE, but allowed in RE if certain conditions hold/are assumed. As for Pooled OLS, make sure you are treating it as you would any OLS model with the panel indicators. As I said, in my experience, you should be getting different results.

Feel free to email or message if you have questions.

Jason

Thanks a lot.

Can I go through this situation? Is there any way to explain the reason and choose RE model?