> Agent view (Markdown) of https://www.nycbiasaudit.com/resources/bias-audit-methodology-technical-guide
> Publisher: Holistic AI (https://www.holisticai.com). Site index for agents: https://www.nycbiasaudit.com/llms.txt · Everything in one file: https://www.nycbiasaudit.com/llms-full.txt

# Bias audit methodology: a technical guide to Local Law 144 calculations

Research paper · 19 pages · Published 2026-10-21 · By Holistic AI

Local Law 144 defines the minimum bias audit calculations precisely, but the choices around them decide what gets published. This paper implements every required table end to end on a seeded synthetic dataset, with the code, and shows seven implementation errors that move results across the 0.80 reference line. It closes with uncertainty and sensitivity methods, the publishing requirements and an audit readiness checklist.

## Key numbers

- **14**: Intersectional categories in a full EEO-1 analysis
- **7**: Implementation errors demonstrated on the same data
- **574**: Candidates tied at the median in our synthetic sample
- **2%**: Threshold below which a category may be excluded from impact ratios

## What is inside

- The statute and DCWP rule, mapped to what the data team must build
- Python implementations of selection, scoring and intersectional tables
- Unknown categories, the 2% exclusion and unstable reference groups
- Bootstrap intervals, significance tests and cut off sensitivity
- Seven implementation errors, each demonstrated, plus a readiness checklist

## Who it is for

- Data scientists and people analytics teams
- Bias auditors and model risk reviewers
- HR technology product and engineering teams

The full PDF is free. Request it with the form at https://www.nycbiasaudit.com/resources/bias-audit-methodology-technical-guide.

---

Holistic AI carries out independent Local Law 144 bias audits. What the service includes: https://www.nycbiasaudit.com/services.md. To commission a bias audit or ask a question, book a call at https://www.nycbiasaudit.com/demo or email we@holisticai.com.