Pay-equity audits, end to end

60-Second Summary

A pay-equity audit is a regression that controls for legitimate pay drivers (level, tenure, geography, performance) and isolates the unexplained gap by protected class.
Run it under attorney-client privilege — otherwise the analysis itself becomes discoverable evidence in litigation.
Budget the remediation before running the audit. An audit that finds a gap and can't fund the fix is a legal liability, not an equity win.
EU Pay Transparency Directive (effective 2026–27) makes audits mandatory above thresholds; the US patchwork (CA, NY, CO, WA) is already mandatory for many employers.

Most 'pay equity' work in companies under 1,000 people is a spreadsheet that averages comp by gender and notes a gap. That isn't a pay-equity audit — it's a pay-gap measurement, and the two are routinely confused to the detriment of both legal defensibility and actual remediation.

What an audit actually is

A pay-equity audit is a multivariate regression that estimates how much of the raw pay gap between groups is explained by legitimate, job-related factors (level, function, tenure, geography, performance rating) and how much is unexplained. The unexplained portion is the legally and ethically actionable number. The raw gap is a media headline.

Raw gap vs. adjusted gap

Glassdoor's 2024 analysis: the raw US gender pay gap is ~17%. After controlling for occupation, industry, experience, location, and education, the adjusted gap is ~4.6%. Both numbers are real; they answer different questions. Equity work targets the adjusted gap; representation work targets the raw gap.

The regression model

Variables that go into a defensible model

1
Outcome variable
Log of total cash compensation (base + target bonus), or log of base salary if equity is handled separately.
2
Legitimate controls
Job level, function, tenure, geographic differential, performance rating, education where job-relevant. These are the variables Title VII and EU directive recognize as legal pay drivers.
3
Protected-class variables
Gender, race/ethnicity (US), age (40+ in US), disability status. Run separate models per protected class — interaction effects matter.
4
Output
Coefficient and confidence interval on the protected-class variable, expressed as percent of comp. A 95% CI that excludes zero means the gap is statistically significant.
5
Cohort drill-down
Re-run within job family and level. Aggregate gaps often hide larger pocket-specific gaps that aggregate away.

Running it under privilege

Under US law, the audit analysis itself is generally not privileged unless it was conducted at the direction of counsel for the purpose of providing legal advice. Companies that run audits internally and informally have, in multiple cases, had the audit results subpoenaed and used as evidence of knowing pay discrimination.

Engagement letter from outside or in-house counsel commissioning the audit for legal-advice purposes
All analyst communications routed through or cc'ing counsel
Audit memo marked 'Attorney-Client Privileged / Attorney Work Product'
Remediation decisions documented as legal-advice-driven, not just policy-driven
Pre-determined remediation budget — fix what you find or don't look

The EU exception

Under the EU Pay Transparency Directive, the joint pay assessment triggered by a >5% unexplained gap is mandatory and the results are reportable to employee representatives. Privilege structures common in US audits do not shield EU audits from these obligations.

Remediation mechanics

Approach	How it works	Tradeoff
Targeted raises	Identify under-paid individuals from regression residuals; raise to band midpoint or model-predicted pay	Cleanest legally; expensive in year 1
Cohort floor adjustment	Raise everyone in an affected cohort to a band floor regardless of individual residual	Politically easier; may over-correct
Off-cycle correction + next-cycle structural fix	Immediate residual correction + change to comp process to prevent recurrence (e.g., remove salary history, enforce band discipline)	Best practice — fix the symptom and the system

Regulatory landscape

EU Pay Transparency Directive 2023/970 — reporting from 2027 (1,000+), 2031 (100+); mandatory joint pay assessment if unexplained gap >5%
California SB 1162 — pay scale disclosure on job postings; annual pay data reporting for 100+ employees
New York Pay Transparency Law (2023) — pay range on all job postings
Colorado Equal Pay for Equal Work Act — pay range plus promotion-opportunity posting requirements
Iceland Equal Pay Standard (since 2018) — certified equal-pay management system required for 25+ employees

Frequently asked questions

How often should we run a pay-equity audit?

Annually at minimum, with a refresh after any major comp event (annual cycle, acquisition, large RIF). EU directive requires triennial reporting but most companies move to annual once the muscle exists.

What's a 'normal' unexplained gap?

Best-in-class companies report sub-1% adjusted gaps. The market median for companies running rigorous audits sits around 1–3%. Anything above 5% is both a remediation priority and, in the EU context, a regulatory trigger.

Should we publish our pay-gap numbers?

Mandatory in UK, Ireland, France, Spain, and shortly the EU broadly. Voluntary publication elsewhere is a transparency signal that helps recruiting and retention — but only if you're publishing alongside a remediation plan. Publishing a gap with no plan reads as resignation.

Can salary bands alone solve pay equity?

Bands prevent new gaps from opening (entry pay), but don't fix accumulated gaps from years of pre-band promotions and merit decisions. You need both: bands prospectively, audit-and-remediate retrospectively.