The lack of meaningful boundary differences between journal impact factor quartiles undermines their independent use in research evaluation

Journal impact factor (JIF) quartiles are often used as a convenient means of conducting research evaluation, abstracting the underlying JIF values. We highlight and investigate an intrinsic problem associated with this approach: the differences between quartile boundary JIF values are usually very small and often so small that journals in different quartiles cannot be considered meaningfully different with respect to impact. By systematically investigating JIF values in recent editions of the Journal Citation Reports (JCR) we determine it is typical to see between 10 and 30% poorly differentiated journals in the JCR categories. Social sciences are more affected than science categories. However, this global result conceals important variation and we also provide a detailed account of poor quartile boundary differentiation by constructing in-depth local quartile similarity profiles for each JCR category. Further systematic analyses show that poor quartile boundary differentiation tends to follow poor overall differentiation which naturally varies by field. In addition, in most categories the journals that experience a quartile shift are the same journals that are poorly differentiated. Our work provides sui generis documentation of the continuing phenomenon of impact factor inflation and also explains and reinforces some recent findings on the ranking stability of journals and on the JIF-based comparison of papers. Conceptually there is a fundamental problem in the fact that JIF quartile classes artificially magnify underlying differences that can be insignificant. We in fact argue that the singular use of JIF quartiles is a second order ecological fallacy. We recommend the abandonment of the quartiles reification as an independent method for the research assessment of individual scholars.