Strong lensing statistics in large, z 0.2 surveys: bias in the lens galaxy population