Average time alloted for manuscript review across disciplines

For mathematics, here's the latest survey done by American Mathematical Society:

http://www.ams.org/notices/201410/rnoti-p1268.pdf

You can find median times from submission/acceptance to acceptance/print/electronic print for various math journals as well as other statistics like the current estimate of waiting time for each journal. Not surprisingly, the median time from submission to acceptance in 2013 varies greatly; some take only a few months and others nearly a year and a half.

They do this survey every year. For instance, here's the one published 2 years ago:

http://www.ams.org/notices/201210/rtx121001473p.pdf

The worst offender that took the longest between submission and acceptance that year was Annals of Mathematics (which happens to be among the most prestigious math journals), and the median was 24 months.

I don't know if there is data for other fields. But American Physical Society occasionally makes statistics for their journals available, e.g., pages 14-17 of this PDF slides:

http://www.phys.nthu.edu.tw/~colloquium/2009F/T2.pdf

From my own experience as an author and reviewer as well as from what I hear, it appears that math journals typically take longer than physics journals. But probably things are drastically different across subfields even within one discipline (e.g., theoretical vs. experimental).

In any case, as the fact that they publish statistics indicates, both disciplines seem to take turnaround time very seriously. But for some reason, it appears that review tends to take more time if a journal publishes more mathematical papers. This seems to hold true for electrical engineering, too; the journal I have published in most frequently belongs to electrical engineering but is known for being heavily mathematical, and, lo and be hold, it's notorious for its slow, slow, slooow review...

In biology, things are much faster (although it can vary for different journals). From my experience, reviewers usually get a deadline of 2-3 weeks to submit their review, although many reviewers fail to meet this deadline. So the typical time of getting back the first review is around 1-2 months, depending on the journal.

Since speed is often an important factor, some journals make a point of advertising their typical review/decision times. For example the journal Genome Research states their average turnaround time for review is 30 days. The journal eLife gives the following median times: 3 days for initial decision (editorial decision whether to send the paper to review), 29 days to post-review decision, 90 days submission to acceptance.

How this affects the quality of reviews, especially those of interdisciplinary work (e.g. combining math and biology), is a completely different matter...

However, one way to explain the difference in speed from a theoretical field like math, is since biology is an experiment-based discipline, the reviewers basically evaluate what the authors report about their experiment - they cannot, and are not expected to, reproduce these experiments on their own. Once the paper is published, other labs will eventually try to reproduce its results.

As a linguist and former editor, I would say that your problems are on the outer edges of experience, which may have something to do with the journals you are submitting to, but not beyond belief. No names, but there are a couple of well-regarded theory journals that are known to have this problem. The main explanation is that reviewers take forever, and editors only have a limited power and willingness to nag: they are willing to wait another 4 months rather than drop the reviewer and get someone else. This problem is somewhat ameloriated by submission software which is now popular that automatically nags reviewers. The ultimate solution, I'm afraid, is that authors need to email the editor with a status query with a week after the supposed deadline (if the journal says when they expect to return a decision). A significant contributing problem, IMO, is that there is nearly no infrastructural support for the editor.

Handbooks and the like are another story. Expect a 4 year delay from invitation to appearance.