These losses are estimate from double-blind randomized controlled trials – both in the lab (using treadmills) but also in the field (picking tea, moving dirt for road-building, tapping rubber).  See Horton and Ross 2003 for references