Simpson’s Paradox in Canadian Police Clearance Rates

Abstract

Police clearance rates and other forms of aggregated criminal justice data can be susceptible to statistical artefacts such as Simpson’s paradox (Yule-Simpson effect). Simpson’s paradox occurs when a trend apparent in separate data groups reverses itself once the groups are combined. Within 15 years of clearance rate data from the 50 largest Canadian police jurisdictions, 210 instances of Simpson’s paradox were discovered (annual mean = 14.0). These instances included four cases in which a reversal occurred simultaneously in all crime categories and subcategories. This finding suggests the need for caution when using clearance rates as a comparative measure of police performance, particularly between jurisdictions or time periods with different crime mixes. Criminal justice researchers, policy makers, and crime analysts should be aware of Simpson’s paradox and its potential effect on aggregated data. Finally, the possibility of a double reversal must be considered when attempting to resolve Simpson’s paradox.

Desempenho de organizações policiais: revisão sistemática da literatura

“…Neste contexto, "chamar os infratores para prestar contas" foi a terceira dimensão mais utilizada para avaliar o desempenho policial, com 16 Desempenho de organizações policiais: revisão sistemática da literatura | 280 Revista do Serviço Público (RSP), Brasília 72 (1) 262-291 jan/mar 2021 pesquisas. O número de policiais (Weisburd et al, 2010;Doerner;Doerner, 2012;Asif et al, 2018;Agarwal et al, 2019;Sinuany-Stern;Alper, 2019) e o tipo de crimes praticados (Weisburd et al, 2010;Doerner;Doerner, 2012;Hur, 2013;Demers;Rossmo, 2015;Kuo;Shih, 2018;Vidal;Kirchmaier, 2018;Scott et al, 2019) foram as variáveis mais empregadas para avaliar essa dimensão de desempenho. Na maioria das pesquisas foi constatado que o maior número de policiais implica mais crimes apurados.…”

Demonstration of Simpson’s Paradox in PISA 2015 Data: Confusing Differences between Boys and Girls

“…Blythe neglected that another British statistician Udny Yule wrote about same paradox already in 1903 (Yule, 1903). To acknowledge this some authors nowadays also call it Yule-Simpson effect (Demers & Rossmo, 2015). We will use a shorter name throughout the paper.…”

“…Judging from the amount of research literature the occurrence is certainly not uncommon. If we focus only on the recent research literature it can be found in different areas of science and life in general: medicine (Baker & Kramer, 2001;Rücker & Schumacher, 2008), administration (Demers & Rossmo, 2015) and even sports (Wright, 2012). In this paper we will explore its occurrence in large scale assessments in education.…”