Caution should be taken in interpreting these results since the individual characteristics of the companies cannot be identified by this estimation procedure (as can be done with the panel regressions in the previous specifications) and variables that are stable through time, such as size and the average annual premium, tend to gain significance due to possibly biased estimators.