Self-citing is often frowned upon, being considered (and sometimes is) vanity, egotism or an attempt in self-advertising. However, everyone self-cite because sooner or later, everyone builds upon previous findings "Given the cumulative nature of the production of new knowledge, self-citations constitute a natural part of the communication process." (Costas et al., 2010).
The argument whether citation analysis should include self-citation has been going on since the early days of citation analysis and is still ongoing. Early citation studies tended to exclude self-citation, but today's Journal Impact Factor (JIF) includes them.
Self-citing, when and where?
The percentage of self-citings out of the overall citations to articles published in high-impact journals tend to be lower, not because these authors cite themselves less, but because these articles receive more external citations. The same is true about highly cited articles: the share of self-citation in their citations is lower than in less cited articles. Aksnes (2003) studied almost 40,000 publications by Norwegian authors and showed that the percentage of self-citations for articles cited less than 5 times was about 30%, but less than 20% for articles cited 46-50 times (Fig. 1). That is probably because there are only so many times an author can cite one publication.
The more authors an article has, the more self-citations it gets. Aksnes found, for example, that articles with one author receive 1.15 self-citations on average, but articles with 10 authors receive 6.7. Someone should really check the percentage of publication self-citations for high-energy physics, with their hundreds of authors per article.
The JIF and self-citing
The JIF is particularly vulnerable for self-citations, not only because of journals' attempts to "game" it, but because its citation window is two years, when the share of self-citations is at its highest level (Fig. 2).
Requests from editors to authors to cite more of their journal's articles are one of the many ways to manipulate said JIF. In a recent article, about 20% of the respondents reported being coerced by journals into citing more articles from the specific journal they were submitting to. The saddest part was that though 86% thought the practice unethical, 57% were willing to go along with it.
Thomson-Reuters, for their part, don't consider journal self-citation to be too much of a problem. They show (unfortunately, with data from 2002) that 82% of the JCR journals have self-citation rates of 20% or less. As for journals who overdo it, they are punished by being kicked out of the JCR (temporarily).
But really, excessive self-citation is for amateurs: the real thing is forming a "citation cartel" as Phil David from The Scholarly Kitchen puts it. In April this year, after receiving a "tip from a concerned scientist" Davis did some detective work using the JCR data and found that several journals published reviews citing an unusually high number of articles fitting the JIF window from other journals. In one case, the Medical Science Monitor published a 2010 review citing 490 articles, 445 of them were published in 2008-09 in the journal Cell Transplantation (44 of the other 45 were for articles from Medical Science Journal published in 2008-09 as well). Three of the authors were Cell Transplantation editors. This is even nastier than excessive journal self-citation because, as Davis points out, self-citing can be detected without too much effort, but citation cartels are trickier. In the good news, three out of the four journals Davis reported as suspects of cartel behavior were suspended from the JCR, and Retraction Watch reports that two of the manipulating articles have been retracted. This means TR will have to remove the sentence about citation circles who "call to mind the mythical unicorn" from their next white paper.
Self-citation and the author
We've already seen that self-citation can pay off at the journal level, but what about individual authors? Fowler and Aksnes (2007) did another research on the Norwegian database, but this time it was author rather than publication oriented. The percentage of author self-citation was rather low – 11% - but every self-citation yield, on average, 3.65 citations from others in 10 years. Fowler and Aksnes concluded that "self-citation advertises not only the article in question, but the authors in question." In practical terms, this could mean that the author who managed to fit in even one more self-citation per article will be much better off years down the road than her colleague who didn't. The effect turns negative only in very high levels of self-citations (40-50 a year). It seems that self-citation make authors more visible. It's like with SAT preparation courses: it's not that you'll do better than the others if you take them; it's that you'll do worse than the others if you don't take them. Because of this effect, Fowler and Aksnes suggest not only to remove self-citations from citation counts, but to penalize them as well. However, this could hurt older or more productive scientists, who cite more simply because they have more articles to cite and more publications where to cite them. Currently, every popular index has a version free of self-citations, but it seems that most people continue to factor them in.
Costas, R., van Leeuwen, T.N., & Bordons, M. (2010). Self-citations at the meso and individual levels:
effects of different calculation methods Scientometrics (82), 517-537 DOI: 10.1007/s11192-010-0187-7
Aksnes, D. W. (2003). A macro study of self-citation Scientometrics, 56 (2), 235-246
Fowler, J. H., & Aksnes, D. W. (2007). Does self-citation pay? Scientometrics, 72 (3), 427-437 DOI: 10.1007/s11192-007-1777-2
McVeigh, M. E. (0). Journal self-citation in the journal citation reports Thomson Reuters