Compensation, Equity, Diversity and Inclusion Analytics

Introduction

Compensation and diversity, equity and inclusion (DEI) are often treated as separate HR domains. In practice, they are tightly connected. Pay structures influence attraction, retention, motivation, perceived fairness and organizational reputation. Representation affects access to opportunities, innovation and decision quality. Promotion and inclusion outcomes reveal whether formal equality is translated into comparable employee experiences.

HR analytics helps leaders move from general claims—such as “our pay is competitive” or “promotion is merit-based”—to evidence-based assessment. The analyst’s task is not merely to calculate differences. It is to identify meaningful patterns, distinguish association from causation, test alternative explanations, protect privacy and support fair action.

This chapter examines compensation structures, market competitiveness, pay equity, representation, promotion and inclusion analytics. It also explains adjusted pay-gap analysis, its assumptions and limitations, and the responsible use of data in rewards and DEI decisions.

Learning Objectives

By the end of this chapter, students should be able to:

1. Explain the relationship between job architecture, pay structures and market competitiveness.
2. Calculate and interpret core compensation, representation, promotion and inclusion metrics.
3. Distinguish unadjusted pay gaps from adjusted pay differences.
4. Evaluate regression-based pay-equity analyses and their assumptions.
5. Diagnose demographic representation and career-progression outcomes.
6. Design evidence-based compensation and DEI interventions.
7. Identify ethical, legal, privacy and governance risks in HR analytics.

1. The Strategic Architecture of Compensation Analytics

A compensation system normally contains four linked elements:

• Job architecture: job families, levels, grades and career paths.
• Internal structure: relationships among jobs based on contribution, complexity, skills and responsibility.
• External competitiveness: comparison with relevant labor-market rates.
• Individual positioning: an employee’s pay within a grade or range, reflecting experience, performance, skills and other legitimate factors.

A compensation philosophy states how the organization intends to position itself. For example, an employer may target the 50th percentile of the market for base pay while providing stronger variable pay or benefits. The selected market must be relevant by occupation, geography, industry, organization size and talent scarcity. Comparing a technology firm with a national all-industry average can produce a technically precise but strategically misleading conclusion.

Pay structure terminology

For employee i, the compa-ratio is:

where P_i is the employee’s base pay and M_i is the midpoint of the applicable range. Both must use the same currency and time basis. A value of 1.00 means pay equals the midpoint; 0.90 means pay is 90% of midpoint. Compa-ratio is descriptive, not proof of fairness. Employees may legitimately differ because of experience, scarce skills, performance or time in role.

Range penetration is:

where Min_i and Max_i are the range boundaries. It is meaningful only when the range is valid, pay is within or appropriately compared with the range, and the same pay component and currency are used.

2. Market Competitiveness and Pay Compression

Market competitiveness is assessed by comparing organizational pay with external reference data. A market index can be expressed as:

If an organization’s typical pay is 72,000 currency units and the relevant market reference is 75,000, the market index is 0.96. This indicates a position approximately 4% below the reference, not necessarily a problem. Sampling error, job matching, geography, total rewards and data age must be examined.

Pay compression occurs when pay differences between employees with materially different experience, tenure or responsibility become unusually small. It can arise when new-hire salaries rise rapidly while incumbent pay does not. Compression analytics should compare similarly situated employees and examine hiring rates, promotion increases, merit budgets and turnover.

A useful diagnostic is cohort comparison: compare employees hired in different periods while controlling for job, location, level and relevant experience. A simple time trend may reveal compression but cannot establish why it occurred.

3. Pay Equity: Concepts and Measurement

Pay equity concerns whether employees receive comparable pay for comparable work after considering legitimate, job-related factors. It differs from pay equality, which may describe identical pay regardless of job or contribution. Equity does not mean every employee receives the same amount; it means differences are explainable, consistently applied and not based on protected characteristics or their proxies.

Unadjusted pay gap

The unadjusted group pay gap compares average or median pay between groups. For groups A and B, the percentage gap relative to group A is:

If group A’s average pay is 80,000 and group B’s is 72,000, the gap is 10%. Median pay is often preferred because it is less sensitive to highly paid executives, although it does not solve all distributional problems. The reference group must be stated explicitly; reversing the denominator changes the percentage.

The unadjusted gap is valuable because it captures the overall outcome experienced by groups. However, it combines multiple mechanisms: occupational sorting, level distribution, part-time status, tenure, location, performance, hiring, promotion and possible discrimination. It should therefore be treated as a signal requiring investigation, not as a complete causal estimate.

Adjusted pay analysis

An adjusted analysis estimates differences after controlling for variables considered relevant to pay. A common model is:

Here, ln(Pay_i) is the natural logarithm of pay, Group_i indicates group membership, the other variables are control variables, β_j are estimated coefficients and ε_i is the residual. The logarithm is often used because pay is positively skewed and coefficients can be interpreted approximately as proportional differences. For a binary group variable, the exact percentage interpretation of β₁ is approximately (e^β₁ − 1) × 100% when the model is otherwise appropriate.

For example, if β₁ = −0.06, the exact model-implied difference is approximately (e^−0.06 − 1) × 100 = −5.83%. This means the model estimates lower pay for the coded group conditional on included variables. It does not prove that discrimination caused the difference.

#### Assumptions and limitations

A credible adjusted analysis requires:

• an accurately defined comparison population;
• reliable, consistently measured pay and job variables;
• an appropriate functional form;
• sufficient observations within relevant groups;
• attention to missing data and measurement error;
• no severe multicollinearity that makes coefficients unstable;
• controls that are legitimate explanations rather than consequences of bias.

A particularly important issue is overcontrol. If promotion or performance ratings are themselves affected by unequal opportunity, controlling for them may conceal part of the inequity. Conversely, omitting legitimate job-related factors can exaggerate a residual difference. Analysts should therefore present multiple specifications, such as an organization-wide model, a job-family model and a tightly matched comparison, and explain how conclusions change.

Regression estimates are not automatically fair because they are statistical. Analysts should report confidence intervals, sample sizes and practical significance, not only p-values. Small gaps may be statistically significant in very large datasets, while important gaps may be uncertain in small groups.

Distributional and intersectional analysis

Mean or median gaps can conceal unequal outcomes at different points in the pay distribution. Analysts should examine quartiles, deciles, bonus receipt, bonus size and the probability of being below or above a pay threshold. Intersectional analysis examines combinations such as gender and race, but small cells create statistical instability and privacy risk. Suppression, aggregation and confidence intervals may be necessary.

4. Representation Analytics

Representation analysis asks who is present at each organizational stage. A basic representation rate is:

where g is a demographic group and s is a stage such as the workforce, a job family, a level or leadership. A representation rate of 30% in the total workforce and 12% in senior leadership signals a pipeline or progression issue, but the appropriate benchmark may be the qualified labor pool rather than the general population.

Representation should be examined across the employee lifecycle:

1. applicant pool;
2. shortlisted candidates;
3. hires;
4. probation or early tenure;
5. performance ratings;
6. development participation;
7. promotions;
8. critical assignments;
9. turnover;
10. leadership.

Selection rate and adverse-impact screening

For group g, the selection rate is:

A selection-rate ratio comparing group B with group A is:

A ratio below 1 indicates a lower selection rate for group B. Some jurisdictions use legal screening rules such as the “four-fifths” rule, under which a ratio below 0.80 may warrant investigation. This is a screening heuristic, not a universal legal test and not proof of discrimination. Legal requirements vary by jurisdiction and must be reviewed with qualified counsel.

Promotion and retention

Promotion rates should use a clearly defined risk set. The annual promotion rate for a group can be calculated as:

Eligibility might mean employees at a level who have completed a minimum period in role. If eligibility is not defined consistently, comparisons are unreliable. Survival analysis or discrete-time models may be useful for studying time to promotion or exit, particularly when employees have different observation periods.

Turnover analysis should distinguish voluntary and involuntary exits, regrettable and non-regrettable loss, and exits from critical roles. A high exit rate among a group may reflect pay, workload, manager behavior, career opportunity or external labor-market conditions. Analytics identifies patterns; listening, investigation and qualitative evidence help explain them.

5. Inclusion Analytics

Diversity measures composition; inclusion measures whether people experience respect, voice, access and belonging. Inclusion is therefore assessed through surveys, behavior indicators and institutional outcomes.

Useful survey dimensions include:

• psychological safety;
• respect and freedom from harassment;
• voice in decisions;
• access to information and development;
• confidence in speaking up;
• fairness of performance and promotion processes;
• belonging and intent to remain.

A favorable-score index may be calculated as:

The denominator should exclude invalid or deliberately missing responses according to a documented rule. Analysts should report response rate, item wording, scale design and confidence intervals where appropriate. Comparing small demographic groups without privacy safeguards can identify individuals indirectly.

Inclusion analytics is stronger when survey results are linked, at an appropriately protected level, to outcomes such as absence, turnover, promotion and performance. Correlation does not establish that inclusion caused these outcomes. Reverse causality and common causes are possible: for example, a supportive manager may improve both survey scores and retention.

6. An Integrated DEI Analytics Framework

A practical framework links inputs, processes, outcomes and explanations.

Figure: An integrated framework for linking compensation and DEI evidence to action.

The framework discourages isolated dashboard metrics. For example, a leadership representation gap may originate in recruitment, early attrition, assignment allocation, promotion criteria or succession planning. Each mechanism requires different intervention.

7. Worked Example: Interpreting an Adjusted Gap

An organization compares annual base pay for 1,200 employees. The median pay is 64,000 for Group A and 59,520 for Group B.

The unadjusted gap, using Group A as the reference, is:

A regression controlling for job family, level, location, tenure and experience estimates β_GroupB = −0.035. The exact conditional difference is:

The analysis therefore shows a 7.0% descriptive median gap and an estimated 3.44% conditional difference under the specified model. It does not establish that the remaining difference is caused by discrimination, nor does it show that the controlled variables are unbiased. The next steps should include checking model diagnostics, reviewing hiring and promotion decisions, examining bonus and starting-pay outcomes, validating job matching and conducting structured case review. If the organization uses a remediation budget, it should prioritize employees whose pay is unsupported by documented, legitimate factors, while avoiding automatic equalization without role-level review.

8. Illustrative Case: Meridian Health Systems

This is a fictional illustrative case. Meridian Health Systems employs 4,800 people across hospitals, laboratories and corporate functions. Its annual dashboard reported that women represented 58% of employees but only 34% of senior managers. The unadjusted gender pay gap was 8.2%, while an adjusted model showed a 2.1% residual difference. Leaders initially concluded that the organization had “no material equity issue.”

Problem

Employee listening data showed lower confidence among women in access to high-visibility assignments and promotion transparency. The analytics team also discovered that the adjusted model controlled for current level and performance rating, although both could reflect unequal access to opportunities.

Analysis

The team created a lifecycle dataset linking applicant, hiring, pay, performance, development, promotion and exit records. It found:

• similar hiring rates for men and women overall, but lower female representation in high-paying specialist roles;
• lower access to international assignments and revenue-generating projects for women at middle-management levels;
• comparable promotion rates among employees nominated for promotion, but fewer nominations for women;
• higher voluntary turnover among women with strong performance ratings;
• wider pay differences in starting salary and variable pay than in base salary alone.

Approach and solution

Meridian introduced structured salary offers with documented approval for exceptions, quarterly pay-equity reviews, transparent promotion criteria, calibrated talent discussions, sponsorship for underrepresented managers and manager training on assignment allocation. It also revised the analytic model to present several specifications and explicitly flagged current performance and level as potentially affected by prior opportunity.

Outcome

After two review cycles, the organization reduced unexplained salary anomalies, increased women’s representation in the promotion pipeline and improved inclusion scores regarding career transparency. The organization did not claim that all disparities had disappeared. Instead, it established targets, owners, review dates and an escalation process.

Learning points

1. A small adjusted gap does not invalidate a large unadjusted outcome gap.
2. Opportunity allocation before promotion can be a central mechanism.
3. Pay equity requires analysis of starting pay, increases, bonuses and progression.
4. Multiple model specifications improve transparency.
5. Intervention should combine compensation correction with process redesign.

9. Governance, Ethics and Legal Responsibility

Compensation and DEI data are sensitive. Responsible practice requires:

• Purpose limitation: collect and use data for defined legitimate purposes.
• Data minimization: retain only what is necessary.
• Access control: restrict identifiable data and maintain audit logs.
• Small-cell protection: suppress or aggregate results where individuals could be inferred.
• Data quality controls: document definitions, missingness, revisions and source systems.
• Fairness in automation: test whether algorithms reproduce historical disadvantage.
• Human review: do not make consequential decisions solely from a model score.
• Transparency: explain measures, assumptions, uncertainty and remediation logic.
• Legal review: account for applicable equal-pay, anti-discrimination, privacy, labor and reporting laws.

Demographic data should not be collected casually or used to label individuals. Self-identification, voluntary participation, culturally appropriate categories and secure handling are essential. “Prefer not to say” responses should be respected and analyzed as missingness without penalizing employees.

AI can assist with anomaly detection, job-description analysis, survey-text classification and scenario modeling. However, AI systems can encode biased historical decisions, use proxy variables and produce opaque recommendations. Before deployment, organizations should validate performance across groups, conduct disparate-impact testing where appropriate, document model features, establish override procedures and monitor drift. Generative AI should not receive identifiable compensation records unless approved security, privacy and contractual controls are in place.

10. Implementation Blueprint

A mature analytics program can proceed through the following cycle:

1. Define the decision: clarify whether the question concerns pay correction, workforce planning, promotion, retention or inclusion.
2. Specify the population and unit of analysis: employee, job, application, promotion opportunity or survey response.
3. Build a data dictionary: define pay components, dates, levels, demographics, eligibility and exclusions.
4. Validate the data: reconcile payroll with HRIS, inspect missingness and test unusual values.
5. Describe outcomes: report distributions, medians, representation and lifecycle rates.
6. Diagnose mechanisms: examine job architecture, hiring, assignments, performance, promotion and exits.
7. Model responsibly: use suitable controls, sensitivity analysis and uncertainty estimates.
8. Engage stakeholders: include HR, legal, employee representatives, managers and affected groups.
9. Act: correct pay where justified, redesign processes and assign accountable owners.
10. Monitor: establish baselines, targets, review intervals and unintended-consequence checks.

A dashboard should not merely display a gap. It should show the reference population, time period, denominator, confidence or uncertainty information, data limitations, responsible owner and next action.

Challenges and Limitations

Several limitations recur. Demographic categories may be incomplete or culturally inappropriate. Job titles may not reflect actual work. Market surveys may use inconsistent job matching. Small samples reduce reliability. Pay data may omit equity, commissions, allowances or benefits. Cross-sectional analysis cannot establish career dynamics. Historical performance ratings can embed managerial bias. Simpson’s paradox may occur when aggregate results differ from subgroup results because workforce composition varies. Finally, measurement can create false confidence: a clean dashboard does not guarantee a fair organization.

The solution is not to abandon analytics but to combine quantitative evidence with job evaluation, process audits, employee voice, qualitative inquiry and independent review. Robust conclusions are usually cumulative rather than dependent on one metric.

Contemporary Developments and Best Practices

Current practice is moving toward pay-transparency ranges, skills-based job architecture, continuous rather than annual equity monitoring, intersectional analysis with privacy safeguards and outcome-based DEI accountability. Organizations increasingly analyze total rewards, not base pay alone, and use scenario models to estimate the effect of hiring, promotion or salary-budget choices.

Best practice is characterized by consistent definitions, relevant comparison groups, transparent assumptions, multiple measures, statistical and practical interpretation, protected data, employee participation and visible follow-through. The central managerial principle is simple: analytics should improve the fairness and effectiveness of decisions, not provide statistical decoration for decisions already made.

Chapter Summary

Compensation and DEI analytics connect organizational rewards with workforce opportunity and employee experience. Job architecture and market benchmarking establish the context for pay decisions, while compa-ratios, range penetration and market indices describe internal and external positioning. Unadjusted pay gaps reveal group-level outcomes; adjusted models explore differences after specified controls, but they do not prove causation and may conceal inequity when controls are themselves affected by unequal opportunity.

Representation, selection, promotion, turnover and inclusion measures should be examined across the employee lifecycle. A credible analysis defines denominators, uses relevant benchmarks, protects privacy and combines descriptive evidence with appropriate modeling and qualitative investigation. The Meridian case illustrates why pay correction alone cannot resolve pipeline and opportunity problems. Effective action requires governance, transparent assumptions, accountable intervention and continuous monitoring. AI can support analysis, but human judgment, legal compliance and ethical safeguards remain essential.

Key Terms

Adjusted pay gap: A group pay difference estimated after controlling for selected variables.

Adverse impact: A substantially different selection outcome affecting a protected group.

Benchmark job: A job with stable content and reliable external market data.

Compa-ratio: Employee pay divided by the midpoint of the applicable pay range.

Compensation philosophy: The organization’s principles for positioning and rewarding pay.

DEI: Diversity, equity and inclusion in workforce composition, opportunity and experience.

Intersectionality: Analysis of overlapping demographic and social identities.

Inclusion: The extent to which employees experience respect, voice, belonging and access.

Market index: Organizational pay compared with a relevant external market reference.

Pay compression: Unusually small pay differences between employees with materially different experience or responsibility.

Pay equity: Fair and explainable pay relationships for comparable work.

Pay gap: A difference in pay between defined groups, usually expressed as a percentage.

Pay range: Minimum, midpoint and maximum rates associated with a grade or level.

Promotion rate: Promotions divided by the defined eligible population.

Representation: The proportion of a demographic group in a specified workforce population.

Selection rate: Selected applicants divided by applicants in a defined group.

Total rewards: Monetary and non-monetary value provided through employment.

Unadjusted gap: A descriptive group difference without statistical controls.

Review Questions

Short Answer Questions

1. What is the difference between pay equality and pay equity?
2. Define compa-ratio and explain what a value of 0.90 indicates.
3. Why is the unadjusted pay gap useful even though it does not establish causation?
4. What denominator should be specified when calculating a promotion rate?
5. Identify two privacy risks in intersectional DEI analytics.

Long Answer Questions

1. Explain how job architecture, market benchmarking and individual pay positioning interact in a compensation system.
2. Discuss the assumptions and limitations of regression-based adjusted pay-gap analysis.
3. Develop an employee-lifecycle framework for analyzing representation and progression.
4. Explain why inclusion analytics should combine survey evidence with behavioral and workforce outcomes.
5. Evaluate the governance requirements for using AI in compensation and DEI analytics.

Analytical / Application Questions

1. Group A has median pay of 80,000 and Group B has median pay of 74,400. Calculate the pay gap using Group A as the reference and explain what the result does and does not establish.
2. A promotion process has selection rates of 24% for Group A and 18% for Group B. Calculate the selection-rate ratio and identify appropriate next investigative steps.
3. An adjusted pay model controls for current level, performance rating and tenure. Explain how controlling for these variables might clarify inequity and how it might also conceal unequal opportunity.

Multiple Choice Questions with Answers

Question 1

What does a compa-ratio of 1.00 indicate?

A. Pay equals the range midpoint

B. Pay equals the range maximum

C. Pay is 10% below the range minimum

D. Pay is 100% above market pay

Correct Answer: A. Pay equals the range midpoint

Question 2

If Group A median pay is 80,000 and Group B median pay is 72,000, what is the gap using Group A as the reference?

A. 8%

B. 10%

C. 11.1%

D. 12.5%

Correct Answer: B. 10%

Question 3

Which statement best describes an adjusted pay-gap estimate?

A. It proves discrimination whenever the coefficient is negative

B. It describes a conditional difference under specified model assumptions

C. It eliminates the need for qualitative investigation

D. It is always more important than the unadjusted gap

Correct Answer: B. It describes a conditional difference under specified model assumptions

Question 4

A selection rate is 24% for Group A and 18% for Group B. What is Group B’s selection-rate ratio relative to Group A?

A. 0.50

B. 0.67

C. 0.75

D. 1.33

Correct Answer: C. 0.75

Question 5

Which is an example of overcontrol in pay-equity analysis?

A. Excluding all job-related variables

B. Controlling for a promotion outcome that may reflect unequal opportunity

C. Checking missing demographic data

D. Using a relevant geographic control

Correct Answer: B. Controlling for a promotion outcome that may reflect unequal opportunity

Question 6

Which measure most directly describes the proportion of a demographic group in senior leadership?

A. Range penetration

B. Representation rate

C. Compa-ratio

D. Market index

Correct Answer: B. Representation rate

Question 7

Why should analysts define the eligible population when calculating promotion rates?

A. To make every group’s rate identical

B. Because inconsistent eligibility makes comparisons unreliable

C. To remove all demographic information

D. Because promotion rates cannot use percentages

Correct Answer: B. Because inconsistent eligibility makes comparisons unreliable

Question 8

Which practice best protects privacy in small-group DEI reporting?

A. Publishing every individual result

B. Suppressing or aggregating small cells

C. Replacing all data with estimates

D. Reporting names with survey scores

Correct Answer: B. Suppressing or aggregating small cells

Question 9

What is a central limitation of an unadjusted pay gap?

A. It cannot describe overall group outcomes

B. It combines multiple possible mechanisms and does not establish causation

C. It requires a regression model

D. It can only be calculated for hourly employees

Correct Answer: B. It combines multiple possible mechanisms and does not establish causation

Question 10

Which is a responsible use of AI in compensation analytics?

A. Automatically reducing pay for employees flagged by a black-box model

B. Using anomaly detection with validation, human review and governance

C. Uploading identifiable payroll data to an unapproved public tool

D. Treating historical decisions as unbiased training labels

Correct Answer: B. Using anomaly detection with validation, human review and governance