Not all disciplines use arXiv -- far from it. Especially in the humanities and social sciences, there isn't really anything comparable to arXiv.
But you are right, preprints are not all that different from published papers, as this recent analysis confirmed: http://arxiv.org/abs/1604.05363
[edited to add:] I suppose you could look at the data (https://blog.datadryad.org/2016/04/28/sci-hub-stories/) and see what proportion of requested DOIs have a preprint version in the arXiv. I suspect it's rather small.