Response

摘要

Davis accurately observes that our analysis of open access only accounts for scientific literature provided by publishers. We chose to analyze when journal volumes come online through publisher Web sites because publishers do not select specific articles for availability; instead, they select a batch of articles (one or more years' worth) based on duration since publication. The difficulty with analyzing the open-access effect for articles that authors have paid or taken pains to post freely is that authors likely select the best articles. Moreover, in author-pays and self-archiving analyses, articles cannot be compared with themselves over time-they must be compared with other articles that were not selected. As a result, they report large open-access effects-from 100% (7) to 286% (2). A reanalysis of one of these studies using instrumental variables to predict whether authors paid the open-access fee suggests that much of the purported open-access effect comes from author selection (5).