Does your empirical result survive after trim by 2.5% ?