This study analyzes 58,033 retractions from the Retraction Watch Database via Crossref to identify trends in retraction causes across time.
A dual-date framework distinguishes between the year of retraction to detect patterns and the publication year to trace origins. Statistical analyses include slope calculations for misconduct trends and a structural break test to examine the impact of potential policy or practice shifts, such as retraction awareness campaigns.
A shift is observed from classical misconduct, e.g. “Falsification” (slope = −0.006, r2 = 0.51), to procedural concerns like “Fake Peer Review” (slope = 0.004, r2 = 0.39), especially post-2010. A statistically significant structural break is detected in 2014 (F = 4.24, p = 0.021), aligning with policy updates and the maturation of Retraction Watch. The mean retraction lag of 2.31 years (SD = 3.43) shows varied detection timelines.
The study introduces a dual-date analysis approach and visualizes shifts in misconduct typologies across decades, highlighting retractions as evolving reflections of scholarly oversight. It proposes implications for editorial transparency and peer review reform in the face of global publication expansion.
