From the Public Safety edition of October 19, 2026
Crime Science study using Chicago 2022–2024 records finds incomplete crime data can create or worsen ethnic disparities in predictive policing outputs, with the largest shifts for Black-majority communities
Mengqin K. Jiang, Federico Liberatore and Alun Preece (authors); Crime Science (Springer Nature) · Policing / predictive policing and data quality · Chicago, Illinois, United States (method transferable; single-city results)
- Publisher
- Crime Science (Springer Nature)
- Original publication
- Published: 14 October 2026; Crime Science volume 15, article 40
- Source retrieved
- 2026-10-20
What happened
Using Chicago crime records (2022–2024), census data and the HAGEN spatio-temporal model, the authors removed records at controlled rates (25–100% in low-diversity neighbourhoods; 50% or 100% in single-ethnicity-majority communities) and allocated prediction errors to groups as predicted over-policing (false positives) and under-policing (false negatives). Removing records from a group’s majority communities generally reduced predicted over-policing and increased predicted under-policing for that group, with Asian-majority scenarios an exception. The largest shifts came from removing data in Black-majority communities; in the appendix, predicted under-policing for Black residents rose by 58.36% from 0% to 100% reduction. The authors call for auditing missingness and its demographic distribution before deploying event-level predictive policing.
Why it matters
Agencies buying forecasting, hot-spot or resource-allocation analytics rarely audit where their records are missing. This study gives a concrete, testable reason to require data-completeness audits by neighbourhood before deployment, and it reinforces that model accuracy alone does not guarantee fair or effective outcomes.
Evidence and measured results
Opened the open-access article and read the abstract, methods, results, discussion and limitations. Outcomes are model-based proxies, not observed police behaviour; missingness was simulated by uniform random removal within targeted communities; one city, one period, one model. The article also cites a randomised trial in Montevideo finding no significant crime-reduction benefit from a predictive policing system over analysts.
Limitations and uncertainty
Simulation on one city and one model; recorded crime treated as proxy for crime; random within-area removal does not model real under-reporting; baseline data already incomplete; results may differ by season and crime type.
Put this evidence to work
Lighthouse Advisory interpretation, grounded in this source. Enriched 2026-10-20; this does not change the original publication date. Each role follows the same checklist so takeaways compare across sources.
Sales
Role takeaway
- Customer problem
- Agencies using predictive or allocation analytics cannot show whether data gaps skew outputs against particular communities.
- Stakeholders
- Crime analysis manager
- Deputy chief for operations
- Oversight board or inspector general
- City data officer
- Community advisory group
- Discovery questions
- Which analytics drive deployment decisions today?
- Is record completeness measured by neighbourhood?
- Are false positives and negatives reported by area or group?
- Has the vendor done disparity testing?
- Value hypothesis
- A completeness and disparity audit can identify where analytics outputs are unreliable before they shape deployment.
- Potential engagement
- 6-week data-completeness and disparity audit of one analytics product using the agency’s own records.
- Unsupported claims
- Effects on actual policing behaviour
- Crime reduction from predictive policing
- Results generalize beyond Chicago and HAGEN
Pre-sales engineering
Role takeaway
- Fit
- Agencies using spatio-temporal forecasting, hot-spot or resource-allocation analytics.
- Architecture and integration
- Add completeness metrics and controlled-missingness tests to the analytics pipeline, with neighbourhood-level error allocation using census data.
- Prerequisites
- Access to model outputs and training data
- Census or demographic data at neighbourhood level
- Analyst capacity
- Constraints
- Vendor models may be opaque
- Real missingness mechanisms unknown
- Small areas produce noisy estimates
- Security
- Handle incident-level data under existing CJIS and records rules; publish only aggregated results.
- Proof of value
- Proposed: replicate controlled data removal on the agency model
- Proposed: report predicted over/under-policing allocation by community before and after
Delivery
Role takeaway
- Work
- Build and run a completeness and disparity audit and embed it in model release and annual review.
- Dependencies
- Vendor cooperation
- Demographic data
- Oversight body engagement
- Ownership
- Crime analysis unit with inspector general or oversight review
- Skills and adoption
- Train analysts on missingness assessment and on communicating model limits to command staff.
- Governance checkpoints
- Pre-deployment audit
- Annual disparity review
- Public summary to oversight body
- Acceptance criteria
- Proposed: completeness measured for every neighbourhood
- Proposed: disparity thresholds defined and tested before release
- Risks
- Vendor refuses access to model
- Audit used to justify rather than question deployment
- Simulation assumptions misread as real-world effects
Implementation considerations
Lighthouse Advisory interpretation across the operating dimensions a public-sector buyer must settle before this evidence becomes a design. Each note answers the question under its heading for this specific source.
Architecture and integration
What must connect, and where does the AI sit in the workflow?
Analytics pipelines should report data completeness by geography and group, and test model outputs under plausible missingness before release.
Governance
Who approves, reviews and stays accountable for outcomes?
The authors frame missingness auditing as a prerequisite for deploying event-level predictive policing.
Security and privacy
What data, permissions and controls need testing?
Disparity risk arises from data gaps (under-reporting, uneven detection, recording practices), not only from model design.
Accessibility and workforce
Who is affected, and what skills or accommodations follow?
Crime analysts need skills to assess record completeness and explain model limits to commanders.
Procurement
What should contracts, pricing and exit terms secure?
Buyers should require vendors to provide disparity testing under data-incompleteness scenarios and neighbourhood-level error reporting.
Operating model
Which teams own the service once it runs?
Crime analysis units, not vendors alone, should own completeness audits and periodic disparity reviews.
Publication history
- 2026-10-19Public Safety · Issue 417 resources
Stable resource ID: crime-science-chicago-predictive-policing-data-incompleteness-20261014