Sentence Variability in a Mathematical Sentencing Framework: A Statistical Analysis of Brazilian Court Data

This article presents the findings of a quantitative study on sentencing practices in Brazil, focusing on the presence of numerical patterns and “penal clustering” in judicial decisions. Drawing on a dataset of criminal sentences from São Paulo—the country's most populous and active judiciary—the research statistically investigates whether Brazilian judges display preferences for certain sentence lengths, such as even numbers, “round numbers,” and multiples of 6 months. The study also examines the role of numerical fractions in shaping sentence variability. Despite Brazil's civil law tradition and the absence of formal sentencing guidelines, the results reveal a significant degree of penal clustering, with judges consistently relying on a narrow set of sentencing values. These patterns suggest the influence of cognitive heuristics, particularly “anchoring effects,” in sentencing decisions. While the use of numerical fractions may enhance consistency and predictability, it may also constrain individualization—raising concerns about mechanical decision‐making and reduced responsiveness to case‐specific circumstances. The article discusses the implications of these findings for judicial training, institutional reform, and public policy. It argues for a more data‐informed and psychologically aware approach to sentencing, capable of balancing the dual demands of consistency and individualization within both common law and civil law traditions.