As one sees the trends do more or less continue. The average number of submissions on the hep arXiv are shown in blue and include hep-th/hep-ph/hep-lat and hep-ex. Red is astro-ph, green is cond mat, violet is math with the pink addition being math-ph. The clear extensions are cross-links. The number of hep submission slightly increased again, so it seems the temporary drop was a fluctuation. It seems to stagnate around 730 papers/month.
I am still wondering how it comes though. Several people have argued this is to be expected and just suggests 100% participation from that community (this is also the explanation you find on the arXiv website). I don't think though this can be the full story. I would expect the average number of submissions to be roughly a product of
- the total number of people in the community
- their average productivity
- the fraction of them using the arxiv
First, it is plausible to me the fraction of people using the arxiv in the hep community is saturated in North America and Europe, but given that there are still new countries coming into the game and the arXiv is as global as can be I am not sure whether this actually holds. That is to say, globally seen I'd still expect that fraction to be increasing.
Second, since the world population generally is increasing, one should think this general trend underlies the statistics, unless the fraction of people in the hep-community relative to the total population decreases.
Think China, Japan, India. Then have a look at the submitter's affiliations and tell me one should not expect further growth. I mean, yeah, there are a lot of Germans, but we don't outweigh the rest of the world.
Taken together this means, if people stick to the arxiv either the number of people in the community doesn't grow as I'd expect, or they publish less. Another possibility, as Dylan mentioned, would be that people just submit papers to other arXivs. E.g. it might be that some stuff that used to end up on hep, now goes into physics general, or history of physics. Or, maybe now that there are more online journals that publish more or less immediately, people don't submit all their papers to the pre-print server? Anybody noticed something like this?
The other thing that I found surprising is the sheer number of math publications. Not the trend - I understand that probably that community is still not completely arxiv-ed - but gee, look at all those math papers! I always had the impression mathematicians publish only sparingly, seems I was somewhat mistaken there.
For more statistics see here.