The hypertechnical reason to self-publish is to group-register periodical articles. Many article-writers figure that they'll be covered by the magazine's/newspaper's copyright registration. I'm afraid not. Under a probably incorrect line of decisions that began a couple of years ago in New York, a periodical's compilation copyright registration does not function as a registration for individual parts, including freelancers' articles (and photographs and other illustrations). One of the reasons that the particular decisions in question are probably incorrect is that this is a partially curable defect. Under the Copyright Act, one can only sue under a valid registration certificate, although as I've noted previously there are many things that an author can do short of filing suit to assert his or her copyright. See 17 U.S.C. § 411(a); see also Morris v. Business Concepts, Inc., 259 F.3d 65 (2d Cir. 2001). Timeliness is important; if the registration certificate precedes the beginning of the infringement—usually just by calendar date, but by definition in the Act a registration within 90 days of publication precedes the infringement with a few exceptions—the holder can recover attorney's fees in an appropriate case and possibly statutory damages. Untimely registration allows suit, but limits the recovery to actual damages (which are extraordinarily difficult to prove for articles).
So, then, must freelance authors spend $30 per article to register their works? Even the US government isn't that greedy. The Copyright Office provides a much-more-affordable alternative to individual registration: group registration. However, there remains an unstated problem here: Creating the deposit copy. One cannot just stuff clippings in an envelope to obtain a group registration! Instead, one should self-publish the deposit copies (two of them) and three or four backups, and in this modern age should probably maintain a PDF to make more copies when they become necessary. The Copyright Office's website has a scattered collection of materials that relate to this. The most important thing to keep in mind is that one need not (and, for litigation purposes later, should not) just retype the article or cut-and-paste it in from the original text file. Instead, one should do everything reasonable to maintain page breaks, and should on each page use both a page number from the publication source and a cardinal page number for the particular collection, and the table of contents should have complete bibliographic and publication information. This frequently means pasting clippings to letter-sized paper and photocopying (or scanning) the result, and adding the other information by hand (or computer if scanned) to each page, then taking it to the local copy shop for reproduction and binding.
This is a form of self-publishing. Now, instead of just a disparate group of articles, I have a "book" costing five or six bucks a pop that I can enter as an exhibit in a copyright action. I like to paste a copy of the registration certificate inside the front cover (photoreduced as necessary), but that's just me—because I've seen just how easy it is to misplace those things in an author's office and back files, to the occasional intense (if short-lived) annoyance of my clients' attorney…
In any event, this is one of the three rationales that justify self publishing. The other two are simple desire to see work in print without regard to financial success, and filling of a known (if not necessarily existing) market niche that is outside the commercial publishing industry's purview, often because the work has only a regional or hyperspecialized audience. More discussion on these coming up… and aren't you sorry you asked…