The results of random effect and fix effects are very similar, the only difference lies in the analysis of each particular effect, i.e. time effect, or panel effect, or population effect.

2个解释变量的P-value小于0.1， 其他4个都大于0.1小于1. is obviously not a good model.

The best thing you have to do at this stage is to identify why the result is crampy: is it because of missing variable bias? is it heteroscedasticity? How about the specification errors? Do you understand your data well? i.e. data size, causality relationship between variables.

It is dangerous to apply the model directly without testing the assumptions. Try to start by testing the assumptions, i.e. (xi,yi) i.i.d. , error term to be white noise. Though these are the basic things, it might lead you to the conclusion that your OLS produces the best results.

Modeling is tedious and arduous, but you will learn a lot by asking yourself questions along the way toward a satisfactory model. In my observation, no matter how crampy the data is, there is definitely a model there. Please take your time.

Best of the luck.