Why do some fields not use arXiv?

There are several reasons why a field might not use arXiv, or another pre-print service. Whether or not these are correct is left as an exercise to the reader:

• It will interfere with publication. This is a little circular, but if a field doesn't use pre-prints, journals might not recognize a pre-print being put in a depository as not prior publication. Whatever merits there are in putting something in a pre-print repository are offset if you can't get it into a journal.
• Pre-prints create version sprawl. If there's significant revision of pre-prints, there's now sort of two versions hanging out, the one on the pre-print server and the one that sees publication.
• Priority is not necessarily as important (not that it isn't important). In biomedicine, the ideal is for multiple studies to all independently estimate an effect, so being first isn't as critical.
• Pre-prints are not peer reviewed, but a pre-print of an RCT or something may still be used to influence medical decision-making etc.
• Pre-prints and double-blind peer review are basically mutually exclusive. This one is, in my mind, the biggest downside of pre-prints.

When it comes down to it though, much of it is just field convention - if preprints aren't used, there's very little utility to being a "pioneer" in that regard - it's work, and explaining, and some risk for very little benefit.

An important factor is how open are researchers in other fields to publish their manuscripts as a preprint. At least in biological sciences, we have biorxiv (inspired of course by arxiv) and several others preprint servers that are becoming popular. You can check some stats here: http://asapbio.org/preprint-info/biology-preprints-over-time.

This practice is also becoming extended to other fields. The Center for Open Science made available a platform for the creation of new preprint servers, e.g. https://osf.io/preprints/?_ga=2.149800182.1984918221.1511841280-890674891.1511841280

It has not been mentioned, but the preferred document format for arxiv is LaTeX, and naturally, fields heavily relying on typesetting equations are more prone to use LaTeX.

I am not sure that word documents uploaded ca. 2000 are perfectly compatible with todays version of Office.