Better Confidence Intervals for the Population Mean by Using Trimmed Means and the Iterated Bootstrap?

In the quarterly Danish sample based wage statistics firm within a given industry are randomly selected with probabilities proportional to size. The hourly mean wage per employee is registered within each selected firm. Then the population mean per employee is estimated by the pps-estimator, i.e. the simple mean of the selected firm means per employee. I raise the question: Is it possible to produce better confidence intervals for the population mean by using trimmed means instead of the simple mean when the iterated bootstrap is used? Monte Carlo experiments indicate that it is the case