The retraction penalty: evidence from the Web of Science

Abstract

Scientific articles are retracted at increasing rates, with the highest rates among top journals. Here we show that a single retraction triggers citation losses through an author's prior body of work. Compared to closely-matched control papers, citations fall by an average of 6.9% per year for each prior publication. These chain reactions are sustained on authors' papers (a) published up to a decade earlier and (b) connected within the authors' own citation network by up to 4 degrees of separation from the retracted publication. Importantly, however, citation losses among prior work disappear when authors self-report the error. Our analyses and results span the range of scientific disciplines.

Figure 1. Retraction characteristics.

Of the 1,423 retractions indexed by the Web of Science, the percentage of total retractions is greatest in the sciences, with nearly half (49.1%) of all retractions occurring in biology & medicine journals (a). Normalizing by field publication rates, both biology & medicine and multidisciplinary sciences show the greatest retraction tendency (0.14 papers per 1000 publications) (b). The number of retractions issued in a given period has been rising rapidly since the year 2000 (c). A minority (21.9%) of retractions is due to authors' self-reporting errors to the publishing journal (d).

Figure 2. Effect of retraction on retracted papers themselves.

Citations losses, compared to control papers, are shown for (a) self-reported retractions and (b) non-self-reported retractions. Blue lines indicate mean citation losses and dashed lines present 95% confidence intervals. Compared to the control papers, citation losses are 86.2% (p < .0001) for self-reported retractions and 81.5% (p < .0001) for non-self-reported retractions, annually, five or more years after retraction.

Figure 3. Effect of retraction on authors' prior body of work.

Citations losses for prior work, compared to control papers, are presented after (a) self-reported retractions and (b) non-self-reported retractions. Orange lines indicate mean citation losses and dashed lines present 95% confidence intervals. After non-self-reported retractions, the authors' prior work loses 12.5% (p < .0001) of citations per year per prior publication five or more years after the retraction event, compared to control papers. By contrast, citation losses for the authors' prior body of work do not appear after self-reported retractions.

Figure 4. Effect on author's prior body of work by distance measures from retracted paper.

Considering the effect of non-self-reported retractions, citation losses are sustained on the authors' prior work published up to 10 years before the retraction event (a), with negative but statistically insignificant losses for still earlier work. Citation losses on the authors' prior work are also sustained up to 4 degrees of separation away from the retracted paper in the author's citation network (b), with negative but statistically insignificant effects on work at higher degree of separation. Red lines indicate mean citation losses, and dashed lines present 95% confidence intervals. After many years, publications tend to have few annual citations, limiting the capacity for change vis-à-vis matched control papers and resulting in noisier estimates at high distance.