In this case it's better to survey a sample that is skewed but in a known way (you can then address these biases in the paper) as opposed to skewed in an unknow

In this case it's better to survey a sample that is skewed but in a known way (you can then address these biases in the paper) as opposed to skewed in an unknown way via mechanical turk (how many are pretending to be part of the demographic you targeted just because western academic surveys pay more?).