"The proofing tools for marking up PDFs are not well designed ... marking by hand takes a lot less effort, even when it isn't faster. Proofreading in Word isn't so bad, but of course most people's final product isn't in Word... one packager request[ed] a list of corrections in Word rather than marking up the PDF. Now THAT took a hideous amount of time ... More recently I have been able to export a list of corrections from Acrobat into Word to satisfy this requirement. Still a bit tedious, but much much faster, as the page numbers go in automatically."
I have _utter and complete trust_ in ethelmay's accuracy. You should, too.
I've been trying to make the following assertion: workflow time savings mean Cox et al are able to pre-read or sort-of proofread at a rate much, much faster than proofreading for what Eisler terms "legacy publishers". (At this point, I'm sort of hoping that Ms. Cox or someone doing this work will weigh in and tell me I'm totally full of it, completely right, or somewhere in between, ideally with backing details. If I don't hear from someone soon spontaneously, I might even contact someone and just _ask_ them.)
First, two definitions. Workflow is the sequence of tasks that something goes through in order to become a finished product. In this case, I'm describing the "manuscript" as submitted by the author as the starting point, the "proofreader" as supplying a service that definitely includes noticing when you wrote elicit but meant illicit or their when you meant they're and probably includes some number of other observations but probably does not include insisting that you change dialect in dialog to be Standard English. It also includes noticing that some part of your technical process lost or mangled text.
As ethelmay observed, proofreading is an activity that can be assessed in a variety of different ways (other than output): how "easy" it is and how "fast" you can do it. "Easy" is a product of factors like: how many errors are there, how does the customer want the resulting output formatted, how familiar is the proofreader with the systems involved. "Fast" depends on related but not identical factors. If proofreading is easy (not very many errors, inline corrections to a simply formatted file in a word processor or other piece of software with which one is very familiar) and can be done fast (because you read quickly, have a knack for spotting errors, and have done a lot of work of exactly this nature before), a proofreader can charge a lower per word or per page rate and still make a living wage per hour. If proofreading is hard and/or slow (many errors, corrections must be made in a separate document including descriptive references back to the original document, any or all of the files are in software with which one is not particularly familiar and/or is buggy or difficult to use even for an expert), the per word or per page rate must be much higher.
Another factor, which I am _purposely_ ignoring, is the issue of proofreading non-text or non-flowing text -- proofreading poetry or tables or anything like that is an entirely different matter but one which does not typically come up among fiction authors proposing to selfpub in order to get a larger fraction of the book revenue for themselves.
Finally, the output of some proofreaders will be "better" than the output of other proofreaders, for any given definition of "better". That is, some proofreaders will miss two ands in a row, or let a too/two slide on by, while other proofreaders would never, ever, ever allow that to happen. To some degree, the more attention is paid, the slower the process (some people are faster at a given rate of accuracy than others, just like typists). It is reasonable to expect that a more careful and thorough proofreader would expect a higher per word or per page rate to reflect their slower production and/or higher quality, in much the same way that Hondas command a higher resale value and thus a higher initial sales price because they break down less frequently than many other makes. It is _also_ reasonable to expect that some customers will decide they're okay with the lower reliability, and buy a Kia instead, and thus also pay someone to do a worse job proofreading if they're a lot cheaper and "good enough". (Notice me not getting into why anyone pays a ton of money for a highly unreliable vehicle. Me, not going there.)
Eisler's term "legacy publisher" is very evocative in this context. Offset presses which can produce paper books cheaply are not the kind of thing you update every year or two -- but word processors have been. Workflow which _must_ terminate in something acceptable to an offset press from the late 1980s (say) may -start- in Word, but have to go through a lot of conversions on the way to the press. Each conversion costs money, and risks disrupting the design of the book. Proofreading the results of these conversions often involves dealing with an ancient and/or awkward interface.
Judging by this:
at least some selfpublished authors are consciously composing in their destination format. Conversions which happen after submission to Smashwords aren't really available to the selfpubber to adjust, other than whatever you might choose to do to a mobi file that you then submit on Amazon. That's not to say that these things can't totally destroy an ebook, just that you cannot pay a proofreader to catch those problems and then try to fix them.
Derrico also notes that rather than using Word, you can create an HTML file and thus eliminate some of the conversion steps that happen after submission and which allow you to create a better looking/better functioning ebook. There's nothing stopping an author from composing their book in vi or emacs or some other editor with HTML (or, to be fair, using much more appropriate tools for this purpose) and then uploading that to Amazon or wherever (<-- I know this because I've done it).
Ever since I did some looking around to understand how academic publishers were adjusting to ebooks (answer: .pdf, Ingram and/or really smart people who can shepherd something through two parallel workflows), I've been aware of the legacy aspects of publishers workflow. Never mind shipping the damn paper around and warehousing it. As long as workflow is pointed at the inputs for those presses, it will be a twisted, demonic mess resistant to every effort to simplify it and make it more efficient and effective.
To sum up: "proofreading" services aimed at selfpubbers on Smashwords and/or kindle who have composed a Word document or HTML document benefit from three things. (1) Simplification of workflow. (2) A familiar destination format that is easily modified "in place" (Word, HTML, etc.). (3) Assuming you've selected your clients carefully, a product which does not need extensive correction. The result of these three things means a "proofreading" service can charge a dramatically lower per word/per page rate and still make a per hour rate comparable to someone suffering with legacy workflows. If they hit the speed/quality sweet spot, they'll have repeat customers. I do not think that this is a controversial statement (altho it might be wrong in some key element(s)) and I don't think that as a business this can be reasonably regarded as exploitative.
I also don't expect the impact on "legacy publishers" to be that significant, but I'll get into that later.