Citation Analysis for Collection Development

Generic scientometric citation analysis maps the global structure of a field, co-citation clusters, impact metrics, research fronts, using large bibliographic databases, and its goal is understanding science. Citation analysis for collection development is local and practical: it studies what your own community cites in their theses and publications to decide what to buy, keep, and cancel. The unit of interest is your collection, not the field's intellectual structure. The two share the basic act of counting citations but differ entirely in purpose, data source, and output, which is why this method is framed as a collection-evaluation tool rather than a mapping technique.

Theses and dissertations are an ideal citing sample for collection development because they are locally produced, abundant, and tied to specific programs, so their reference lists directly reflect the research needs the collection must support. They also tend to be thoroughly referenced and discipline-specific, making them a clean window into which formats, ages, languages, and titles a graduate community relies on. Faculty publications add another local perspective. Using locally generated citing works, rather than the published literature of the field at large, is what keeps the analysis focused on this library's community and therefore actionable for this library's collection.

Literature half-life is the age within which half of all the citations in a field fall, a measure of how quickly that field's literature stops being cited. It matters for collections because it informs retention and storage decisions: a discipline with a short half-life, where researchers cite mostly recent work, can have older material moved to storage or weeded with little impact, whereas a field with a long half-life, such as much of the humanities, needs deep retention of older monographs. Knievel and Kellsey's findings showed these patterns differ sharply by discipline, so half-life should be computed per field rather than assumed uniform across the collection.