The pooled estimator is simply linear regression that is run on all the data, without regard to cross section or time:
The error is assumed to be normally distributed with mean zero and a constant variance.