Independent research study — not client work

Verified

RS-03Policing and accountability • checked 4 September 2026

Predictive policing: what the oversight bodies found when they audited it

Both audit reports were downloaded and read. The study reports what the oversight bodies themselves concluded, and does not extend those conclusions into a general claim that predictive policing does or does not work.

Section 01

The question and the answer

Question

What did official oversight bodies conclude about predictive-policing programmes they were asked to audit?

Answer as published

Two city oversight offices audited person-based predictive policing and both found the programmes poorly controlled rather than proven. Chicago's Inspector General found the police department's risk scores unreliable and the programme was decommissioned on 1 November 2019; the Los Angeles Inspector General found significant inconsistencies in how the Chronic Offender programme was run and cautioned against drawing strong conclusions from the crime statistics.

Section 02

Scope

What this study covers, stated before the findings so the limits are not read as an afterthought.
  • Two city programmes in the United States: Chicago and Los Angeles.
  • Findings are the auditors' findings, at the date of their reports, about the programmes as then operated.
  • Disparate-impact claims are not asserted here: neither report was used to establish one.

Section 03

Method as actually run

  1. 01

    Download each report as published by the oversight body and read the executive summary and findings sections in the PDF.

  2. 02

    Quote the auditors' own characterisations rather than campaign summaries of them.

  3. 03

    Keep separate what the auditors found from what they merely recommended.

Section 04

Findings, each mapped to a source

F1Verified

Chicago's programme was decommissioned on 1 November 2019 after federal grant funding of $3.8 million.

The Chicago Police Department used risk models known as the Strategic Subject List and the Crime and Victimization Risk Model, designed to predict the likelihood a person would become a 'party to violence'. CPD told the Inspector General in August 2019 that it intended to decommission the programme and did so on 1 November 2019.

F2Verified

The Inspector General found the scores unreliable and the underlying data poor in places.

Scores were not updated regularly — Strategic Subject List scores had not been updated since August 2016 until the successor model launched on 9 January 2019 — and the office observed missing age-at-latest-arrest values, criminal activity recorded after death dates, and conflicting weapons-offence data.

F3Verified

Officers using the scores did not consistently understand what they meant.

Of the 398,684 individuals with a score in version 5 of the model, 16.3% (64,947) were identified as confirmed gang members, yet one district commander estimated about 95% were gang members or associates and an intelligence officer stated everyone with a score was a gang member. Training material had not been updated past version 3.

F4Verified

In Los Angeles the Inspector General found the person-based programme inconsistently administered.

The review of Operation LASER's Chronic Offender component identified significant inconsistencies in selection and documentation practices from Area to Area, related to a lack of centralised oversight and formalised protocols, and found that the majority of people designated as Chronic Offenders had few, if any, actual contacts with police.

F5Verified

The Los Angeles auditors expressly cautioned against reading the crime figures as proof the programmes worked.

The report notes that overall targeted crime decreased as police presence increased, but that results broken down by quarter and Area were more mixed, and that given the difficulty of isolating the impact of these programmes the office cautions against drawing strong conclusions from the available statistics.

Section 05

Sources opened

Every source below was reached at the date shown. Where a page could not be fetched directly, that is stated.
S1Verified

Advisory Concerning the Chicago Police Department's Predictive Risk Models

City of Chicago Office of Inspector General (OIG File #18-0106)23 January 2020

PDF downloaded and read; figures in findings F1–F3 are taken from its background and findings sections.

Open source
S2Verified

Review of Selected Los Angeles Police Department Data-Driven Policing Strategies

Los Angeles Police Commission, Office of the Inspector General (BPC #19-0072)Transmitted 8 March 2019; report dated 12 March 2019

PDF downloaded and read; findings F4 and F5 come from its introduction and executive summary.

Open source
S3Partially verified

NJCM and others v The Netherlands (SyRI), ECLI:NL:RBDHA:2020:1878 — related welfare risk-profiling judgment

The Hague District Court5 February 2020

Listed for context only. The case record was confirmed reachable but the judgment was not read in full, so no finding on this page rests on it.

Open source

Section 06

Search log

QueryEngineLoggedOutcome
Chicago Office of Inspector General Strategic Subject List advisory 2020Web search4 September 2026Located the advisory PDF and its publication page dated 23 January 2020.
LAPD Inspector General review data-driven policing LASER PredPol 2019Web search4 September 2026Located BPC #19-0072 on the Police Commission's own site.
Download and read both audit PDFsDirect fetch4 September 2026Both files retrieved and converted to text; every figure quoted on this page was read in the report itself.

Section 07

What this study does not establish

Stated plainly, because a short study answering a narrow question is only useful if its boundary is visible.
  • Whether either programme disproportionately targeted any particular community — neither audit was used to establish that, and no disparate-impact analysis was conducted here.
  • What replaced either programme after decommissioning or reform.
  • Anything about predictive policing outside these two cities.
  • Whether the recommendations in either report were implemented.