Quite a long time ago now, I reviewed a book for IEEE Spectrum (I think) on the “Reification problem” in formal methods of software development. (“The what?”, you ask? “Reification” is to make something less abstract. This was a book on how to convert beautiful proven programs on paper into ugly proletarian real computer code.)
The contents were normal for this kind of book – I’d just had a chapter published in a similar book, so I was familiar with the process – a collection of chapters which were provided in typeset form to the publisher by the authors and printed using photo-offset printing. The book itself was the cheapest sort of academic hardcover binding: sewn signatures in a cloth spine, hardboard covers and the sort of cheap plastic surface that would break and peel off in a semester’s hard use.
What struck me was what the book cost: over $200. I no longer remember the exact price and page count, but the number burned into my memory was that it was 62.5¢ per page. Making Xerox copies of the book would have cost 10¢ a page. The actual production costs of the book were as low as they could possibly have been in hardback, with no design, no typesetting, and cheap binding.
Now, there are some reasons why it would be expensive, but the reason they could get away with it was that thousands of research libraries were buying this sort of book for $200+ each so they could maintain the currency of their library.
Formal journals were — and are — even worse. You needed to produce your paper in typeset form to match the journal’s standards, after which the journal would charge “page fees” or a “publication fee”. Those fees are pretty significant now: up to $3000 in some journals. The journals are also published online, but if you want access to more than the abstract, you need to pay a fee — $30–$40 usually. (Here’s an example.)