I would say that the use of self-citations that you describe is very common, but whether it's dirty is not clear-cut and it's probably difficult to study. I can also see several different perspectives, some more malicious than others:
- Bias due to better knowledge of your own and related researches (friends) work. I find this rather benign, often unintential and natural.
- When providing a reference to support a specific point, everything else equal, most researchers will probably cite their own work before the work of others. This with the assumption that both papers provide equally good evidence/arguments, and that the author can only use a maximum number references for his paper (which is commonly the case). This behaviour is probably most often seen in soft/weak references of general points in e.g. introductions and discussions. I personally find this behaviour maybe not ideal, but rather human.
- The concious choice not to cite other researchers work that would be relevant. This can come in (at least) two forms: 1) ignoring researchers with a critical perspective on your work (~academic feud), 2) wanting to "claim" the field as your own. Both of these cases should ideally be spotted during the review process, but I think it is very common so still see traces of them in published papers. The problem is probably to prove bad intent, and it is also dependent on reviewers that know the litterature in the specific field extremely well and also bother to speak up about it. This form of behaviour is naturally the most problematic one.
Under point 3, the first form (ignoring criticism) has recieved most attention. However, reading between the lines, I feel that the second form is relatively common as well, where groups of researchers try to "claim" a specific topic (intentionally or unintentionally), so that their body of work is the go-to reference for this topic. A good way to achive this would be to preferentially cite work from within the group, especially if you are already a well established researcher. This can be particularly powerful if you are writing review papers of topics, since these can be used to define the core literature of a topic for future researchers that discover the field (probably related: Matthew effect).
I don't know about studies that try to quantify how common these specific behaviours are. However, the study of citation patterns and self-citations is common, and two papers that touch upon the things mentioned here are Hyland (2003) and Fowler & Aksnes (2007). Hyland (2003) looks at how self-citations can be used for different purposes, while Fowler & Aksnes (2007) tries to quantify the "value" of self-citations (which seems to be quite large).