I am not completely convinced by the methodology employed (particularly by the way causality is ultimately dealt with) in this article from Scientometrics, but this is certainly a very interesting question (download PDF) and it is also true that not many have so far looked at the incentives to self-cite oneself. Here's the gist of it:
We report the results of a macro study of more than half a million citations to articles by Norwegian scientists that appeared in the Science Citation Index. We show that the more one cites oneself the more one is cited by other scholars. Controlling for numerous sources of variation in cumulative citations from others, our models suggest that each additional self-citation increases the number of citations from others by about one after one year, and by about three after five years. Moreover, there is no significant penalty for the most frequent self-citers – the effect of self-citation remains positive even for very high rates of self-citation.
So, if you believe this, start citing yourselves! Keep however in mind that you need to have published something 'citable', to start with :-).
The link to the article is via Tyler Cowen, on Marginal Revolution . Check out also the comments to that post; some of them raise very interesting issues: I tend to agree for instance with the fact that citing oneself is a way (a trial) to 'revive' (discussion around) older personal articles, much like bringing to attention older blog posts by now and then referring back to them in newer ones :-).