Publication cultures and the citation impact of open access

Regarding scholarly articles, two broad models of OA can be distinguished. Gold OA refers to scholarly journals that are financed not by subscription fees but by fees paid by the authors or their sponsors (so-called article-processing charges [APC]). Readers have free online access to such articles. A special case are hybrid OA (HOA) journals, where authors can either pay the APC and make their articles freely available, or not pay the APC, which means that readers will have to pay to access the full text. Issues of such journals typically comprise articles that are freely available as well as articles with a paywall.

The second broad model, green OA, refers to the posting of papers with a repository prior to or after their publication in a traditional journal. We may distinguish between institutional repositories, for example, on university websites, and subject-based repositories, such as PubMed Central for biomedical and life sciences, arXiv with a focus on physics, RePEc for economics, and SSRN for the social sciences. Moreover, some authors post versions of their articles on their individual websites.
The latter option, however, entails considerably larger search costs for potential readers compared to postings on institutional or subject-based websites. The rise of the different types of OA triggered a discussion as to whether OA will foster citation counts, or more precisely, do OA articles receive systematically more citations than comparable CA articles? The answer is important to the authors of scholarly articles, especially in the natural and social sciences, as their reputation strongly hinges on citation counts—to some extent on their individual citation counts, but in particular on the average citation counts of the journals they publish in, as a proxy of the reputation of those journals within the corresponding discipline. 

While early studies, relying on cross-sectional data, found a large OA citation advantage of up to 580%, in more recent studies, the OA status constitutes only one of several explanatory variables in the regression to explain citation counts. If additional control variables are taken into account, such as the number of authors, self-citations, quality of the article, time of first availability of content, citation windows, and secular trends in citations, the calculated OA citation advantage turns out to be much smaller than in the early studies. The results still vary widely: While many studies found an OA citation advantage of between 6% and more than 40%, others detected no OA citation advantage at all or even a negative one. These differences are partly due to the different objects of comparison. While some authors focus on OA and non-OA articles in the same HOA journal, others compare non-OA articles in traditional journals with pre- or post-prints in OA repositories, or they compare commercial online availability with OA. Moreover, there seem to be remarkable differences in OA citation advantages between different disciplines. Finally, different specifications of the econometric mode.