I can say (but not visibly prove, unfortunately, NDAs) that large mechanical Turk surveys tend to get distinctly different results to mail and phone call survey

I can say (but not visibly prove, unfortunately, NDAs) that large mechanical Turk surveys tend to get distinctly different results to mail and phone call surveys. Enough so that the commissioners of the study were not interested in replacing existing surveys with mechanical Turk.