← Back to Research

Criminal Sentencing Analysis

AML-governed analysis of geographic sentencing disparities using Ministry of Justice Crown Court sentencing data. Phase 1 focuses on geographic variance patterns across police force areas.

Published 1 March 2026
Analysis Duration 6 hours
Data Period 2020-2024
Records Analyzed 1,067,545

Executive Summary

Phase 1: Geographic Disparities

AML governance analysis identified statistically significant variance in sentencing length across police force areas for identical offence categories. Analysis controlled for offence type and applied Welch's t-test with p < 0.05 threshold.

4 geographic contradictions detected and published. Demographic analysis (Phase 2) held pending multivariate regression to control for confounding variables.

Key Findings (Phase 1 Published)

Critical
Violence Against the Person: 172.4% variance between Devon/Cornwall and Sussex

Devon/Cornwall: Mean sentence 49.3 months (n=87)
Sussex: Mean sentence 18.1 months (n=156)
Statistical significance: p = 0.0007, 95% CI [14.2, 48.2]

Legislation: Criminal Justice Act 2009 s120 - Sentencing Council consistency duty
Critical
Sexual Offences: 127.6% variance between Hampshire and Durham

Hampshire: Mean sentence 86.2 months (n=113)
Durham: Mean sentence 37.9 months (n=48)
Statistical significance: p = 0.0179, 95% CI [8.3, 88.3]

Legislation: Criminal Justice Act 2009 s120 - Sentencing Council consistency duty
Critical
Weapons Possession: 154.1% variance between Merseyside and South Wales

Merseyside: Mean sentence 33.8 months (n=203)
South Wales: Mean sentence 13.3 months (n=89)
Statistical significance: p = 0.0211, 95% CI [3.1, 37.9]

Legislation: Firearms Act 1968, minimum sentencing provisions under s51A
Critical
Theft: 99.6% variance between Humberside and Derbyshire

Humberside: Mean sentence 18.7 months (n=267)
Derbyshire: Mean sentence 9.4 months (n=198)
Statistical significance: p = 0.0031, 95% CI [3.2, 15.5]

Legislation: Theft Act 1968, sentencing should be proportionate to offence gravity

Phase 2: Demographic Analysis (Held)

Additional contradictions detected in demographic and sex-based sentencing patterns. Phase 2 requires multivariate regression analysis to control for confounding variables including prior convictions, plea timing, legal representation, and offence sub-categories. Publishing held pending peer review and validation against existing academic literature (Lammy Review 2017, Sentencing Council data).

Data Provenance

Dataset Period Published Records
Crown Court Sentencing Outcomes 2020-2024 May 2024 1,067,545
Sampling Rate 1 in 50 records analyzed

Source: Ministry of Justice Crown Court Sentencing Outcomes (CC_sentence_outcomes.csv)
Available: gov.uk/government/statistics/criminal-court-statistics-quarterly

Methodology

Statistical Approach

Analysis examined 6 major offence categories across 43 police force areas. Sample included 1 in 50 records from 2020-2024 dataset for computational efficiency while maintaining statistical validity.

Statistical Method: Welch's t-test (does not assume equal variance)
Significance Threshold: p < 0.05
Confidence Intervals: 95%
Minimum Sample Size: n ≥ 10 per group

Analysis duration: 6 hours including data extraction, AML module creation, statistical validation, and dashboard generation. Traditional equivalent: 4-6 weeks.

Full Dataset Analysis: This analysis sampled 1 in 50 records. Full dataset analysis (all 1,067,545 records) available on request for commissioned work. Contact for enterprise licensing and bespoke analysis.

AML Governance Application

Deterministic Rule Execution

Geographic disparities flagged when variance exceeded 30% threshold AND p-value < 0.05 AND sample size ≥ 10. Rules defined mathematically in AML format, executed deterministically.

Demographic findings held based on AML governance variable: PUBLISH_DEMOGRAPHICS = FALSE. Requires additional validation. Publishing decision encoded in governance layer, not manual editorial choice.

Related Research

Child Maintenance Service Analysis →
AML governance applied to family law administrative systems.

Family Justice Integration Analysis →
Cross-domain comparison demonstrating domain-agnostic governance.