Pay Equity Analysis: Step-by-Step Guide for HR Teams

Why pay equity analysis matters more in 2026 than ever before

The regulatory environment around pay equity has shifted faster in the past three years than in the previous three decades. State-level pay transparency laws now cover more than 40% of the US workforce, and the EEOC has stepped up enforcement activity around compensation discrimination under both the Equal Pay Act and Title VII. Meanwhile, the Office of Federal Contract Compliance Programs (OFCCP) continues to require federal contractors to evaluate compensation practices as part of affirmative action obligations.

But the pressure is not only regulatory. Employees now have more salary data than at any point in history. Glassdoor, Levels.fyi, Payscale, and the growing number of jurisdictions requiring salary ranges on job postings mean that pay disparities that once stayed hidden are increasingly visible to the people affected by them. When an employee discovers they are paid significantly less than a peer in a comparable role, the conversation that follows is never comfortable — and it often ends with either a retention-threatening morale problem or a lawsuit.

For HR teams, the question is no longer whether to conduct a pay equity analysis, but how to do it well — with the right methodology, the right data, and the right framework for acting on findings. A poorly executed analysis is arguably worse than none at all: it creates a document that demonstrates you knew about disparities without evidence that you addressed them.

The legal framework: what the law actually requires

Before designing your pay equity analysis, you need to understand what legal standards your organization is measured against. Three federal statutes form the foundation, supplemented by a growing patchwork of state and local laws.

The Equal Pay Act of 1963. This prohibits sex-based pay differences for employees performing "equal work" — defined as jobs requiring substantially equal skill, effort, and responsibility performed under similar working conditions. The law allows four affirmative defenses: seniority systems, merit systems, systems based on quantity or quality of production, and differentials based on "any other factor other than sex." Courts have interpreted this fourth defense narrowly, and an employer must demonstrate that the factor is actually used consistently, not merely available as a post-hoc justification.

Title VII of the Civil Rights Act of 1964. Title VII's prohibition on compensation discrimination is broader than the Equal Pay Act in two important ways: it covers all protected characteristics (race, color, religion, sex, national origin), not just sex, and it uses a "comparable" rather than "equal" work standard. The EEOC's guidance on compensation discrimination makes clear that Title VII analysis can compare employees across different job titles and families when the work is comparable in skill, effort, and responsibility.

Executive Order 11246 and OFCCP requirements. Federal contractors and subcontractors with contracts exceeding $50,000 and 50 or more employees must develop affirmative action programs that include compensation analysis. The OFCCP's Directive 2022-01 outlines the agency's approach to pay equity investigations, including the statistical methods it considers appropriate and the standards for identifying actionable disparities.

State and local laws. At least 20 states now have pay equity or pay transparency statutes that go beyond federal requirements. Colorado, California, New York, Washington, Illinois, and others now require salary ranges on job postings, ban salary history inquiries, or mandate pay data reporting. Several states — notably Massachusetts, Oregon, and California — have broader definitions of comparable work than the federal standard, making it easier for employees to establish pay equity claims. Your analysis methodology must account for the most stringent requirements in every jurisdiction where you employ people.

The 7-step pay equity analysis process

A credible pay equity analysis follows a structured, repeatable methodology. Skipping steps or conducting them out of order produces findings that are difficult to defend — either internally when presenting to leadership, or externally if challenged by regulators or in litigation. The following process reflects the methodology endorsed by the American Psychological Association for workplace statistical analysis and aligns with OFCCP expectations for federal contractor audits.

Step 1: Define scope and objectives

Before pulling any data, establish the boundaries of your analysis. Document answers to these questions:

• Which employee population? Typically all regular full-time and part-time employees. Temporary workers, contractors, and employees in countries with different compensation structures may require separate analyses.
• Which compensation elements? At minimum, base salary. A thorough analysis also examines total cash compensation (base plus bonus), equity grants where applicable, and starting salary patterns for recent hires.
• Which demographic dimensions? Gender is the starting point for most organizations. Race and ethnicity should be included wherever you have sufficient data. Age is relevant for age discrimination analysis. Intersectional analysis — examining outcomes for women of color versus white women versus men of color, for example — reveals disparities that single-dimension analysis misses.
• What is the analysis period? Point-in-time analysis using current compensation data is the baseline. Longitudinal analysis examining pay progression over multiple years reveals whether gaps are growing, shrinking, or stable.

Step 2: Collect and validate data

The quality of your pay equity analysis is entirely dependent on the quality of your data. Garbage in, garbage out is not a cliche here — it is the single most common reason pay equity analyses produce misleading results.

Pull the following data for every employee in scope:

• Employee identifier (anonymized for the analysis)
• Current base salary and total cash compensation
• Job title, job family, and job level or grade
• Department and business unit
• Work location (city and state, at minimum)
• Hire date and tenure in current role
• Most recent performance rating (and ideally the prior two years)
• Highest education level and relevant certifications
• Gender, race/ethnicity, and age
• FLSA status (exempt vs. non-exempt)
• Any documented pay adjustments (retention offers, market corrections, critical-skill premiums)

Validate the data before analysis begins. Common problems include: employees with missing job levels (often new roles not yet graded), inconsistent job title mapping (the same role called different things in different departments), stale performance ratings, and missing demographic data. Each gap reduces the reliability of your findings. If more than 10% of records are missing a critical field, fix the data issue before proceeding.

Step 3: Group comparable roles

This is where the analysis either gains or loses credibility. The fundamental question in any pay equity audit is: who should be compared to whom? Comparator groups must be narrow enough to contain genuinely comparable roles, but broad enough to have a statistically meaningful sample size.

The most defensible approach uses a combination of job family and job level as the primary grouping variable. For example, all Software Engineers at Level 4 form one comparator group; all Marketing Managers at Grade 3 form another. Within each group, employees should be performing substantially similar work requiring comparable skill, effort, and responsibility — the standard courts apply in Equal Pay Act cases.

Common mistakes in grouping:

• Groups too broad: Comparing all "individual contributors" across engineering, marketing, and finance dilutes the analysis to the point of uselessness. A gap in that group tells you nothing about whether specific roles have equity problems.
• Groups too narrow: A comparator group with three people cannot produce statistically meaningful results. If a job family and level combination has fewer than five employees, consider rolling up to a broader level or combining with an adjacent level.
• Ignoring location: A software engineer in San Francisco and one in Tulsa may be in the same job family and level but operate in fundamentally different labor markets. Location-based pay differences are legitimate, but only if they are applied consistently.

Step 4: Select legitimate pay factors

Within each comparator group, identify the factors that legitimately explain pay variation. These are the "control variables" in your analysis — the reasons two employees in the same group might justifiably be paid different amounts.

Commonly accepted legitimate factors include:

• Tenure: Time in the organization and time in the current role. Longer-tenured employees have typically accumulated more merit increases.
• Performance: Documented, consistently applied performance ratings. The emphasis is on "consistently applied" — if performance ratings themselves show demographic bias (a separate analysis worth conducting), using them as a control variable launders that bias into the pay equity analysis.
• Education and credentials: Where directly relevant to the role. A JD for lawyers, a CPA for accountants. Be cautious about using education as a factor for roles where it does not materially affect job performance.
• Shift differential or hazard pay: Where employees in the same group work different schedules or conditions.
• Geographic location: Cost-of-labor adjustments applied through a documented geographic pay policy.

Factors that are NOT legitimate — even though they explain pay differences — include: prior salary (banned as a factor in many states, and using it perpetuates historical discrimination), negotiation outcomes (where evidence shows systematic demographic differences in negotiation behavior), and subjective manager assessments not captured in formal performance reviews.

Step 5: Run statistical analysis

With comparator groups defined and legitimate factors identified, you are ready to run the quantitative analysis. The choice of statistical method depends on your group sizes, data quality, and the level of rigor required for your compliance obligations. See the detailed breakdown of statistical methods below.

For each comparator group, the analysis should produce:

• The unadjusted gap: the raw difference in median pay between demographic groups before controlling for any factors.
• The adjusted gap: the difference remaining after controlling for all legitimate factors.
• Statistical significance: whether the adjusted gap is large enough and consistent enough to rule out random variation (typically using a p-value threshold of 0.05).
• Effect size: how large the gap is in practical terms (dollars and percentage of pay).

Flag any comparator group where the adjusted gap is statistically significant and exceeds a materiality threshold you define in advance — typically 2% to 5% of median pay, depending on your organization's risk tolerance.

Step 6: Identify outliers and investigate

Statistical analysis identifies patterns; it does not explain them. Every flagged gap requires individual case review to determine whether legitimate explanatory factors exist that were not captured in the quantitative model.

For each employee contributing to a flagged gap, review:

• Hiring circumstances: Was there a documented reason for a higher or lower starting salary?
• Promotion history: Did the employee enter the current role through promotion (potentially carrying a lower salary from the prior level) or external hire?
• Retention adjustments: Were market or counter-offer adjustments made that explain a higher salary?
• Role-specific skills: Does the employee possess a specialized skill set that commands a market premium within the broader role family?

Document the outcome of each individual review. The result for each flagged employee falls into one of three categories: (1) the gap is explained by a legitimate, documented factor; (2) the gap is partially explained but a residual unexplained portion remains; (3) the gap has no legitimate explanation. Categories 2 and 3 feed into the remediation plan.

Step 7: Build the remediation plan

The remediation plan converts your findings into specific salary adjustments with dollar amounts, effective dates, and approval chains. This is where the analysis transitions from a diagnostic exercise to an operational project with budget implications. See the detailed section on building a remediation budget below.

Pay equity analysis process: summary table

Statistical methods explained for HR teams

You do not need a Ph.D. in statistics to run a credible pay equity analysis, but you do need to understand what each method does, when to use it, and how to interpret the output. Here are the three most commonly used approaches in compensation analysis.

Multiple regression analysis

Multiple regression is the gold standard for pay equity analysis and the method most courts and regulatory agencies consider appropriate. It works by modeling pay as a function of multiple variables simultaneously — job level, tenure, performance, education, location — and then testing whether demographic variables (gender, race) have a statistically significant effect on pay after those legitimate factors are accounted for.

In practical terms, you build a regression model where the dependent variable is log-transformed base salary (log transformation is used because pay increases tend to be percentage-based, not flat-dollar) and the independent variables are your legitimate pay factors plus the demographic variable you are testing. If the coefficient on gender is negative and statistically significant (p < 0.05), that means women in your organization are paid less than men even after accounting for all the legitimate factors in your model, and the difference is unlikely to be due to chance.

The coefficient itself tells you the size of the gap. A coefficient of -0.06 on the gender variable in a log-salary model means women are paid approximately 6% less than men after controlling for the included factors. Whether that number represents a problem depends on the sample size, the confidence interval, and whether additional factors not in the model might explain the difference.

When to use it: When you have at least 30 employees per comparator group and complete data on the control variables. Regression with small samples produces unstable results.

Compa-ratio analysis

Compa-ratio analysis is simpler than regression and works well for organizations with established salary bands. The compa-ratio for each employee is their salary divided by the midpoint of their pay band. A compa-ratio of 1.0 means the employee is at the midpoint; 0.90 means 10% below; 1.10 means 10% above.

Once you calculate compa-ratios for every employee, you compare the average (or median) compa-ratio across demographic groups within each comparator group. If women in the Software Engineer Level 4 group have an average compa-ratio of 0.94 and men have 1.03, that 9-point gap tells you that women in this group are systematically paid below the midpoint while men are above it — a clear signal that warrants investigation.

Compa-ratio analysis has the advantage of normalizing for pay band differences across roles, making it easier to aggregate results across the organization. Its limitation is that it does not control for factors like tenure or performance within the group — it tells you the gap exists but does not tell you how much of it is explained by legitimate factors.

When to use it: As a first-pass screening tool, especially when you have well-defined pay bands. Follow up with regression for groups that show significant compa-ratio gaps.

Cohort analysis

Cohort analysis groups employees by hire date and tracks their pay progression over time. This method is particularly effective at revealing whether certain demographic groups fall behind in pay progression — even if their starting salaries were equitable.

For example, if you compare men and women hired into the same role family in 2020, they should have received comparable merit increases, promotions, and market adjustments over the intervening years. If the 2020 female cohort's median salary has grown 18% while the male cohort has grown 27%, you have a progression disparity that the point-in-time analysis might attribute to "tenure" or "performance" but that actually reflects systematically different treatment over time.

Cohort analysis is powerful because it can reveal process problems that other methods obscure. If your regression analysis shows no significant gender gap after controlling for performance ratings, but cohort analysis reveals that women consistently receive lower performance ratings, the pay equity problem is in the performance management process, not the compensation process — but the financial impact on women is the same.

When to use it: As a supplementary method alongside regression, especially when you want to understand how gaps formed rather than just whether they exist today.

Interpreting results: explainable gaps vs. gaps that need fixing

Not every pay gap your analysis identifies represents a problem. The purpose of the analysis is to distinguish between differences that are justified and differences that are not — and to document that distinction rigorously.

Gaps that are typically explainable:

• A senior engineer hired from a FAANG company at a premium to secure a critical skill the organization lacked. The premium is documented, approved, and proportionate to the market data supporting it.
• A tenured employee whose salary has grown through 15 years of consistent merit increases, placing them above a recently hired peer in the same role. The difference reflects accumulated performance recognition.
• A geographic differential where employees in New York earn more than those in Nashville in the same role, consistent with a documented geographic pay policy.

Gaps that need investigation and likely remediation:

• A pattern where women's starting salaries are 5-8% lower than men's in the same role and level, without documented justification for the individual decisions.
• A comparator group where all employees of a particular ethnicity cluster in the bottom quartile of the pay band while all others are distributed across the full range.
• Merit increase patterns where employees with identical performance ratings receive systematically different increases correlated with a protected characteristic.
• A pay compression pattern where long-tenured employees from one demographic group are paid less than recently hired peers, while long-tenured employees from another group have kept pace with market rates.

The key test is consistency and documentation. A pay difference is defensible when the reason for it is documented at the time the decision was made, applied consistently across demographic groups, and proportionate to the factor it reflects. A difference is suspect when the explanation is constructed after the fact, when the same factor is applied inconsistently, or when a pattern emerges across multiple employees that correlates with a protected characteristic.

Building a remediation budget

Once you have identified gaps that cannot be explained by legitimate factors, you need to translate findings into a budget. This is where the conversation shifts from HR analytics to CFO-level financial planning, and the quality of your analysis determines whether the remediation gets funded or stalls.

Calculate the total remediation cost. For each employee with an unexplained gap, calculate the dollar adjustment needed to bring their pay to the level predicted by the regression model (or to the median compa-ratio for their group, if using compa-ratio analysis). Sum these individual adjustments to arrive at the total remediation cost. Industry data from compensation consulting firms indicates that first-time remediation typically runs between 1% and 3% of total payroll for the affected population.

Add the annualized burden cost. The salary adjustment itself is the base cost, but remember to include the downstream effects: increased employer payroll taxes (FICA), increased retirement plan contributions (if matching is percentage-based), and any benefits that scale with salary. For most US employers, the fully loaded cost of a salary increase is approximately 1.25x to 1.35x the base salary increase.

Phase the remediation if necessary. If the total cost exceeds what can be absorbed in a single budget cycle, propose a phased approach: close the largest individual gaps in Year 1 (these carry the most legal risk), address medium gaps in Year 2, and mop up remaining smaller gaps in Year 3. Three years is the maximum timeframe that courts and regulators generally consider reasonable for a good-faith remediation effort. Any longer and you risk the argument that the organization identified discrimination and chose not to fix it in a timely manner.

Budget for ongoing prevention, not just current correction. The remediation budget should include investment in the processes that will prevent gaps from recurring: updated hiring offer guidelines with tighter salary ranges, calibrated merit increase frameworks, manager training on equitable compensation decisions, and the annual analysis itself. Fixing today's gaps without fixing tomorrow's processes ensures you will be back in the same meeting next year with a new remediation budget.

Communicating results to leadership

The way you present pay equity findings to senior leadership determines whether remediation happens quickly, slowly, or not at all. Most HR teams make the mistake of leading with the methodology — regression coefficients, p-values, confidence intervals — which causes executive eyes to glaze over before you reach the part that matters.

Lead with the business case, not the statistics. Frame the conversation around three numbers: the total unexplained pay gap in dollars, the remediation cost, and the potential cost of inaction (litigation, regulatory penalties, turnover of underpaid employees, reputational damage). A $1.2 million remediation budget is an easy approval when the alternative is a $5 million class action settlement plus the cost of the bad press.

Show the specific findings, not just the aggregate. Aggregate statistics ("women earn 4% less than men after controlling for relevant factors") are important for context, but leadership needs to see where the problems are concentrated. Which departments? Which job families? Which levels? This specificity makes the problem actionable and assigns implicit accountability to the leaders of affected areas.

Present the remediation plan with timelines and milestones. Leadership wants to know: how much, how fast, and how do we know it is working? Present a phased plan with quarterly milestones, interim metrics, and a clear definition of "done." The goal is not zero gap — some variation is normal and expected. The goal is that remaining variation is fully explained by legitimate, documented factors.

Address the communication strategy. Leadership will immediately ask: "Do we tell employees?" The answer depends on the size and nature of the gaps, the organization's culture, and applicable legal requirements. Some states require disclosure. Even where disclosure is not legally required, employees who receive equity adjustments will notice the change on their pay stubs. Have a communication plan ready that is honest without creating unnecessary alarm.

Ongoing monitoring: how often and what to track

A pay equity analysis is not a one-time project. Gaps recur because the processes that create them — hiring, promotion, merit increases — operate continuously. Effective monitoring requires both a full annual analysis and interim checks throughout the year.

Annual full analysis. Conduct the complete 7-step analysis annually, timed 2-3 months before your annual compensation review cycle. This timing serves two purposes: it gives you enough lead time to incorporate remediation adjustments into the planned salary review, and it ensures that the data reflects the most recent organizational changes (new hires, promotions, departures).

Quarterly interim checks. Between annual analyses, monitor leading indicators that signal emerging gaps:

• Starting salary ratios: Track the ratio of accepted offer salary to pay band midpoint by demographic group for every new hire. If women's offers consistently land at 92% of midpoint while men's land at 98%, a gap is forming in real time.
• Merit increase distribution: After each merit cycle, check whether average merit increases differ across demographic groups at the same performance level. A 0.5% difference in merit increases compounds to a meaningful pay gap within 3-4 years.
• Promotion velocity: Track time-to-promotion by demographic group. Slower promotions for one group effectively means slower pay growth, even if within-level pay is equitable.
• Voluntary turnover of recently adjusted employees: If employees who received equity adjustments are leaving at higher rates, the adjustments may have been insufficient or the organization may have other equity problems beyond compensation.

Document everything. Every analysis, every finding, every remediation action, every decision not to act on a finding (with the reasoning). This documentation serves as your evidence of good faith in the event of a future claim or audit. The organization that can demonstrate it has been analyzing, finding, and fixing pay disparities on a regular cadence is in a fundamentally different legal position than one that conducted a single analysis and never followed up.

Common pitfalls that undermine pay equity analyses

Having reviewed hundreds of pay equity audits across different industries, certain mistakes appear repeatedly. Avoiding these will save time, money, and credibility.

Using job title instead of job level as the comparator. Job titles are inconsistent, inflated, and often reflect organizational politics rather than actual work content. "Senior Vice President" at one company is equivalent to "Director" at another. Use a standardized job architecture with defined levels as the basis for comparator groups.

Failing to check whether control variables are themselves biased. If your performance review process systematically rates women lower than men for equivalent work (and research suggests this happens more often than most organizations believe), then controlling for performance in your regression model does not remove bias — it encodes it. Run a separate analysis on whether performance ratings differ by demographic group after controlling for objective output measures before using ratings as a control variable.

Ignoring total compensation. An analysis that looks only at base salary misses disparities in bonus payouts, equity grants, and other variable compensation. In many organizations, variable compensation is where the largest and least-documented pay gaps exist, because bonus and equity decisions often involve more manager discretion than base salary decisions.

Analyzing without a remediation commitment. Conducting the analysis, discovering problems, and then not acting on the findings is worse than not conducting the analysis at all. The analysis becomes a document proving that the organization knew about disparities and chose to tolerate them. If leadership is not prepared to fund remediation, do not start the analysis.

Treating the analysis as an HR-only project. Pay equity analysis requires collaboration between HR, legal, finance, and IT. HR owns the methodology and findings. Legal advises on privilege, risk, and compliance. Finance approves the remediation budget. IT provides the data infrastructure. An HR team working alone will produce an analysis that legal cannot defend, that finance has not budgeted for, and that uses data IT cannot validate.

Frequently asked questions

What is a pay equity analysis and why does it matter?

A pay equity analysis is a data-driven examination of compensation across your workforce to determine whether employees in comparable roles are paid fairly regardless of gender, race, ethnicity, or other protected characteristics. It matters because unexplained pay gaps create legal liability under the Equal Pay Act and Title VII, erode employee trust when discovered, and increasingly trigger regulatory scrutiny as pay transparency laws expand across the United States and Europe.

How often should we conduct a pay equity audit?

Most employment attorneys and compensation experts recommend at least an annual pay equity audit, timed to coincide with your annual compensation review cycle. This lets you fold remediation adjustments into planned salary changes rather than making conspicuous out-of-cycle corrections. Organizations subject to OFCCP audits or operating in states with active pay transparency laws may want to run quarterly reviews of key metrics between full annual analyses.

What statistical methods are used in pay equity analysis?

The three most common methods are multiple regression analysis, compa-ratio analysis, and cohort analysis. Multiple regression controls for legitimate pay factors (tenure, performance, education, location) simultaneously and isolates the effect of demographic variables on pay. Compa-ratio analysis compares each employee's salary to the midpoint of their pay band and examines whether compa-ratios differ systematically across demographic groups. Cohort analysis groups employees by hire date and tracks pay progression to identify whether certain groups fall behind over time.

What is the difference between an adjusted and unadjusted pay gap?

The unadjusted (or raw) pay gap is the simple difference in median or average pay between two groups — for example, women earn 82 cents for every dollar men earn. The adjusted pay gap controls for legitimate factors like job level, tenure, performance, education, and location. The adjusted gap is typically smaller and represents the portion of the pay difference that cannot be explained by those factors. Both numbers matter: the unadjusted gap reveals structural representation problems, while the adjusted gap reveals potential discrimination within comparable roles.

Do we need outside counsel involved in a pay equity audit?

It depends on your risk profile. Conducting the analysis under attorney-client privilege — meaning outside counsel directs the analysis and the findings are treated as privileged legal communications — provides protection if the results are later subpoenaed in litigation. Organizations with known pay disparities, pending EEOC charges, or OFCCP audit obligations should strongly consider privilege protection. Smaller organizations with no pending legal issues may conduct the analysis internally, but should still consult employment counsel on methodology and remediation planning.

How much does pay equity remediation typically cost?

Remediation costs vary widely depending on the size and severity of identified gaps. Industry data suggests that organizations typically spend between 1% and 3% of total payroll on initial remediation when conducting their first pay equity analysis. A company with a $50 million payroll might budget $500,000 to $1.5 million for first-year corrections. Ongoing annual remediation costs are usually lower — around 0.5% to 1% of payroll — because the largest gaps have already been addressed and new gaps are caught earlier.

What data do we need to run a pay equity analysis?

At minimum, you need: employee identifier, current base salary, job title and job family, job level or grade, department, hire date, tenure in current role, most recent performance rating, highest education level, work location, and demographic data (gender, race/ethnicity, age). Additional useful data includes bonus and variable compensation, prior salary history within the company, market salary benchmarks for each role, and any documented reasons for pay deviations such as retention adjustments or critical-skill premiums.

Can we use pay equity analysis results in EEOC or OFCCP compliance reporting?

Yes, but carefully. The EEOC and OFCCP look favorably on organizations that proactively analyze and address pay disparities. However, the analysis itself can become evidence against you if it reveals problems you did not remediate. This is why many organizations conduct the analysis under attorney-client privilege and develop a documented remediation plan with timelines before sharing results broadly. If you are preparing for an OFCCP audit, your analysis methodology should align with the agency's standards for statistical significance and comparator group construction.