Noise in judicial decision‐making: A research note

Abstract Researchers suspect large unsystematic variation (noise) in criminal sentencing, but past attempts to quantify it have used short hypothetical vignettes administered in low‐stakes settings to small, heterogeneous samples of judges. Such vignettes are deficient in detail and ecological validity. Sample heterogeneity also makes it difficult to separate noise from systematic variation (e.g., experience effects). In the current study, these concerns were mitigated with a naturally occurring noise audit: an early‐career judicial exam in Poland, which included a 390‐min, high‐stakes sentencing task, wherein performance affected career prospects. Based on a uniform court file of over 100 pages, each judicial candidate ( N = 232) composed a sentencing judgement in a criminal case and then justified their decision at length. The proposed sentences exhibited substantial disparities in the type and amount of punishment; different judges gave the same defendant a small fine or a long prison term. Continuous sentencing outcomes (e.g., fine amounts and imprisonment terms) were widely dispersed. The implication is that statutory sentencing ranges, which are intended to allow adjustments for aggravating and mitigating circumstances, also give judges room to treat identical defendants differently.