In the golden age of particle physics (1940s to 1970s) progress both theoretical and experimental was so fast that people just couldn't wait for months for new research to be published in a journal in order for important discoveries to be disseminated. As a result, a culture of "preprints" developed, whereby researchers would widely circulate draft papers not only among their immediate colleagues, but worldwide. I recall visiting CERN in 1993 and seeing a large panel on the wall dedicated to displaying preprints, both local and from foreign institutes.

It is not a coincidence that Tim Berners Lee developed HTML while working at CERN: his first webpage (local only to CERN as there were no commercial browsers!) was intended to provide links to his own research. Again, a means to preprint dissemination.

With the advent of the World Wide Web, the Los Alamos National Laboratory in the USA came up with the idea of an "e-prints archive" and the arXiv was born. The original URL was xxx.lanl.gov and it later migrated to Cornell University when growth made maintaining the system unsustainable for LANL. The first papers on the theoretical high-energy physics repository hep-th date from August 1991.

For some time the high-energy physics community was the only one to use the arXiv, but the practice soon spread to the rest of physics and the arXiv diversified into thematic subarchives as we know it today. However, here's where cultural differences started to become apparent. Astrophysicists baulked at using the arXiv and at the concept of pre-prints itself. How could you go to a repository of non-peer-reviewed papers?

Other fields have been faster or slower in adopting the preprint culture, mathematics being one of the most eager adopters also because of the cross-pollination with theoretical physics. Mathematical finance and economics have embraced the model and they have their own preprint archive, SSRN (Social Science Research Network). For economics and finance this is also natural for two reasons: one is the existing culture of "working papers" published by non-academic economic policy institutions (and, lately, academic departments); the other is the time-sensitive nature of much quantitative economic/financial research. If you wait 3 months your paper on the money market or business cycle analysis will be obsolete.

It is because of the sheer weight of researcher expectations (and because researchers themselves are part of journal editorial boards) that physics and mathematics journals accept arXiv references for submission of papers, and do not demand that pre-prints be removed after publication of a paper. Other disciplines, not so much.

It's all about culture. Maybe I should have been an anthropologist.