An journal editor once told me about an experiment they did where for every article submitted to their journal during a particular period they tracked if and when the manuscript appeared as a preprint on the arXiv, and what the final fate of those submissions were. In their data they did find manuscripts that appeared on the arxiv before or during peer-review were more likely to be accepted by their journal.
Of course, correlation does not imply causation, i.e. it is not necessarily true that the simple act of making a preprint available raises the likelihood a particular manuscript gets accepted. More likely this is a selection effect, if authors are confident in their work they are more likely to make it public available as a preprint.