Statistical Evidence in Discrimination

Introduction

Statistical evidence in discrimination cases refers to the use of numerical data to demonstrate that a particular group of employees, applicants, or workers is treated differently from another group.

Statistics can be particularly important in employment discrimination because discriminatory treatment may not always be expressed openly. Instead, it may appear through patterns in:

• recruitment;
• hiring and promotion;
• pay;
• termination;
• disciplinary action;
• performance ratings;
• allocation of opportunities;
• transfers;
• layoffs; and
• access to benefits.

Statistical evidence generally does not by itself prove discriminatory intent. Courts usually consider it together with the employer's policies, individual circumstances, explanations, comparator evidence and other relevant evidence.

1. What Is Statistical Evidence?

Statistical evidence involves comparing outcomes between groups.

For example, suppose an organisation has:

• 1,000 employees;
• 500 employees from Group A;
• 500 employees from Group B.

If 90% of senior-management positions are occupied by Group A despite the two groups having comparable qualifications and experience, the disparity may be statistically relevant.

Similarly, if employees belonging to one protected category are terminated at substantially higher rates than comparable employees outside that category, statistics may support an allegation of discriminatory treatment.

The important question is:

Is the observed difference likely to be explained by legitimate factors, or does the pattern suggest unequal treatment?

2. Types of Statistical Evidence

A. Workforce-composition statistics

These compare the composition of the workforce with the composition of the relevant labour pool.

Example:

Such a disparity may justify further investigation.

B. Promotion statistics

Statistics can compare promotion rates.

For example:

• Group A: 20% promoted;
• Group B: 8% promoted.

The figures may become more meaningful when employees have comparable:

• qualifications;
• tenure;
• performance ratings;
• job classifications; and
• experience.

C. Pay statistics

Statistical analysis can identify differences in compensation between groups.

A simple comparison of average salary, however, can be misleading because salary may depend upon:

• seniority;
• qualifications;
• job level;
• location;
• experience;
• performance;
• working hours; and
• responsibilities.

Therefore, more sophisticated statistical analysis may attempt to control for these factors.

D. Termination and disciplinary statistics

An employer may have records showing that one group is disciplined or dismissed disproportionately.

For example:

Out of 100 disciplinary dismissals, 80 involve employees from Group A even though Group A represents only 40% of the workforce.

Such statistics can become relevant evidence, particularly when combined with comparator evidence and inconsistent disciplinary treatment.

3. Statistical Significance

One important concept is statistical significance.

A disparity can arise by chance. Statistical analysis therefore attempts to determine whether the observed difference is sufficiently unusual to make random chance an unlikely explanation.

Common techniques include:

• statistical significance testing;
• standard deviations;
• regression analysis;
• probability analysis;
• correlation analysis; and
• adverse-impact ratios.

However, statistical significance is not the same as legal discrimination.

A statistically significant disparity may still have a legitimate explanation.

Conversely, a legally significant disparity might not satisfy a conventional statistical-significance test because the dataset is too small.

4. Practical Significance vs Statistical Significance

Courts and decision-makers should distinguish between:

Statistical significance

Whether the observed disparity is unlikely to have arisen merely through random variation.

Practical significance

Whether the disparity is sufficiently large to have meaningful real-world consequences.

A difference may be statistically significant but practically insignificant.

Conversely, a substantial difference may be practically important but difficult to establish statistically where the sample is small.

5. Disparate Treatment and Statistical Evidence

Statistical evidence can support a claim of disparate treatment, where individuals belonging to a protected group are allegedly treated less favourably because of their protected characteristic.

For example:

If employees from one group consistently receive lower performance ratings despite similar performance indicators, statistics may support an inference requiring further examination.

But statistics normally need to be connected to the particular decision-making process.

6. Disparate Impact

Statistical evidence is especially important in disparate-impact discrimination.

Here, the issue may not be whether the employer intentionally discriminated.

Instead, the question can be whether a facially neutral rule or practice disproportionately disadvantages a protected group.

Examples include:

• recruitment tests;
• physical requirements;
• promotion criteria;
• scheduling rules;
• attendance policies; and
• redundancy-selection criteria.

The employer may then have to justify the practice according to the applicable legal framework.

7. Griggs v. Duke Power Co.

Griggs v. Duke Power Co., 401 U.S. 424 (1971) is a leading United States Supreme Court decision concerning employment discrimination and disparate impact.

The employer used educational and testing requirements that disproportionately excluded Black applicants from certain positions.

The Supreme Court held that employment practices could violate federal employment-discrimination law even without proof of discriminatory motive where they disproportionately excluded a protected group and were not sufficiently related to job performance.

The case demonstrates why numerical disparities can be important even where discriminatory intent is difficult to prove.

8. Hazelwood School District v. United States

Hazelwood School District v. United States, 433 U.S. 299 (1977) is another important authority concerning statistical evidence.

The U.S. Supreme Court considered statistics comparing the racial composition of teachers with the relevant labour market.

The Court recognised that statistical disparities can provide important evidence in discrimination litigation but stressed that the appropriate comparison population matters.

This is an important methodological principle:

The relevant comparison group must actually represent the population from which the employer could reasonably have recruited.

9. International Brotherhood of Teamsters v. United States

International Brotherhood of Teamsters v. United States, 431 U.S. 324 (1977) is a leading U.S. authority concerning statistical evidence in employment discrimination.

The Supreme Court recognised that statistics showing a significant disparity in employment opportunities could be powerful evidence of a discriminatory pattern or practice.

The Court nevertheless emphasised that statistics must be interpreted in their factual context.

The case illustrates how statistical evidence can help demonstrate a systematic pattern, rather than merely an isolated discriminatory decision.

10. Bazemore v. Friday

Bazemore v. Friday, 478 U.S. 385 (1986) concerned statistical analysis of pay differences.

The U.S. Supreme Court rejected the idea that statistical analysis should automatically be disregarded simply because some potentially relevant variables were not included.

The case is significant for the principle that statistical models should be evaluated carefully rather than rejected merely because they are not perfect.

At the same time, omitted variables can affect the weight and reliability of statistical conclusions.

11. Wal-Mart Stores, Inc. v. Dukes

Wal-Mart Stores, Inc. v. Dukes, 564 U.S. 338 (2011) involved a proposed nationwide class action alleging gender discrimination.

Statistical evidence concerning pay and promotion disparities was presented.

The Supreme Court concluded that the plaintiffs had not demonstrated the required commonality for the proposed nationwide class in the manner required by the applicable procedural rules.

The case demonstrates an important limitation:

A broad statistical disparity does not automatically establish that all individual employment decisions resulted from a common discriminatory policy.

12. McCleskey v. Kemp

McCleskey v. Kemp, 481 U.S. 279 (1987) concerned statistical evidence in the criminal-justice context rather than ordinary employment discrimination.

The petitioner relied upon statistical evidence showing racial disparities in capital sentencing.

The Supreme Court held that the statistical evidence presented was insufficient, by itself, to establish the constitutional claim concerning the particular defendant's sentence.

Although it is not an employment case, it illustrates an important evidentiary principle:

Group-level statistical disparities do not automatically establish discriminatory treatment in an individual case.

13. Statistical Evidence in Indian Discrimination Law

Indian employment-discrimination law has a somewhat different legal framework from the U.S. Title VII system.

The Constitution provides important equality guarantees through:

• Article 14 – equality before law and equal protection;
• Article 15 – prohibition of discrimination on specified grounds;
• Article 16 – equality of opportunity in public employment.

Statistical evidence may therefore become relevant where a claimant alleges that government employment policies or decisions produce systematic unequal treatment.

However, statistics must be considered alongside the particular constitutional or statutory provision involved.

14. E.P. Royappa v. State of Tamil Nadu

E.P. Royappa v. State of Tamil Nadu, (1974) 4 SCC 3 is a foundational Indian equality decision.

The Supreme Court moved away from treating equality merely as a formal comparison between two individuals and recognised the broader relationship between arbitrariness and equality.

Although the case was not principally a statistical-discrimination case, its reasoning provides an important constitutional background for examining patterns of unequal treatment.

Statistical evidence may be relevant where it helps demonstrate that a classification or governmental practice produces systematically unequal outcomes.

15. Ajay Hasia v. Khalid Mujib Sehravardi

Ajay Hasia v. Khalid Mujib Sehravardi, (1981) 1 SCC 722 further developed Article 14 principles concerning arbitrariness and discriminatory treatment.

The Supreme Court emphasised that State action must satisfy constitutional equality requirements.

Where statistical evidence demonstrates a systematic pattern, it may assist a court in examining whether apparently neutral criteria are operating in an arbitrary or discriminatory manner.

However, statistics must still be connected to the legal classification or governmental action being challenged.

16. Indra Sawhney v. Union of India

Indra Sawhney v. Union of India, 1992 Supp (3) SCC 217 is highly significant concerning constitutional equality, reservations and identification of socially and educationally backward classes.

The case demonstrates the importance of empirical material and data when governments formulate classifications and reservation policies.

The Supreme Court examined constitutional limitations concerning reservations and the identification of backward classes.

Statistical and socio-economic material can therefore have an important role in demonstrating the factual basis for equality-related governmental policies.

17. M. Nagaraj v. Union of India

M. Nagaraj v. Union of India, (2006) 8 SCC 212 considered constitutional reservations in promotions.

The Supreme Court discussed the requirement of relevant constitutional and empirical considerations concerning reservation policies.

The case demonstrates that data and empirical assessment can be constitutionally relevant when the State seeks to justify certain affirmative-action measures.

The Court's reasoning also shows that statistics should not be treated as automatically conclusive; their relevance depends upon the constitutional question being examined.

18. B.K. Pavitra v. Union of India

B.K. Pavitra v. Union of India, (2017) 4 SCC 620 concerned reservation in promotions and consequential seniority in Karnataka.

The Supreme Court examined the evidentiary and constitutional requirements surrounding reservation policies.

The case is useful for understanding the role of quantifiable data and empirical material in equality-related employment policies.

It demonstrates that statistical evidence can be particularly important when the State relies upon empirical claims to justify differential treatment or affirmative action.

19. Problems with Statistical Evidence

Statistical evidence is powerful but has significant limitations.

A. Small sample size

If only 10 employees are being compared, one or two decisions can dramatically change the percentages.

B. Incorrect comparison group

Comparing the organisation's workforce with the entire population may produce misleading results if the relevant labour pool is substantially different.

C. Confounding variables

Differences may be caused by legitimate factors such as:

• experience;
• qualifications;
• job level;
• location;
• tenure;
• performance;
• working hours.

D. Selection effects

A disparity in promotions might originate at the recruitment stage rather than from discriminatory promotion decisions.

E. Data quality

Incorrect, incomplete or inconsistent HR records can undermine the reliability of statistical conclusions.

20. Employer's Defence to Statistical Evidence

An employer may respond to statistical evidence by showing that the disparity is explained by legitimate factors.

For example:

Employees from Group A have an average tenure of 12 years while Group B employees have an average tenure of 3 years.

If Group A therefore receives higher salaries, the raw salary disparity may not establish discrimination.

Statistical analysis should therefore ideally control for relevant variables.

21. Regression Analysis

In sophisticated employment-discrimination litigation, experts may use regression analysis.

A simplified model might examine whether salary is associated with:

• protected characteristic;
• experience;
• education;
• job level;
• location;
• tenure;
• performance; and
• other relevant factors.

If a significant difference remains after controlling for legitimate variables, it may provide stronger evidence of a relationship requiring explanation.

However, regression analysis still does not automatically establish unlawful discrimination.

22. Statistical Evidence and Individual Claims

A crucial distinction is between:

Group-level evidence

“Women in the organisation receive promotions at a lower rate.”

and

Individual-level evidence

“Employee X was denied promotion because she was a woman.”

The first may support the second, but it does not automatically prove it.

Courts may therefore examine:

• the decision-maker;
• comparator employees;
• stated reasons;
• performance records;
• communications;
• timing;
• deviations from normal procedure; and
• other circumstantial evidence.

23. Statistical Evidence in HR Investigations

Employers can also use statistics proactively.

An HR department may periodically analyse:

• hiring rates;
• promotion rates;
• pay;
• disciplinary actions;
• termination rates;
• performance ratings;
• bonuses;
• absenteeism-related discipline; and
• access to training.

This can identify potential disparities before they become litigation.

However, statistical monitoring should itself comply with applicable privacy and data-protection requirements.

24. Key Legal Principles

The major principles can be summarised as follows:

1. Statistics can reveal patterns that individual cases may not reveal.
2. A statistical disparity is evidence, not automatically proof of discrimination.
3. The correct comparison population is crucial.
4. Sample size affects reliability.
5. Relevant legitimate variables should be considered.
6. Statistical significance is different from legal significance.
7. Intentional discrimination and disparate impact require different analytical approaches.
8. Statistics are generally stronger when supported by individual comparator and documentary evidence.
9. Courts may reject statistics that are methodologically unreliable or unrelated to the relevant decision-making process.
10. Indian constitutional cases demonstrate the importance of empirical data in certain equality and affirmative-action contexts, although Indian law does not simply replicate the U.S. disparate-impact framework.

Important Case Laws at a Glance

Conclusion

Statistical evidence is an important tool for identifying and proving patterns of discrimination, particularly in employment decisions involving large groups of employees. Its evidentiary value depends on the quality of the data, the choice of comparison group, sample size, methodology and the connection between the statistical pattern and the challenged employment decision.

The central principle is:

A disparity may raise an inference of discrimination, but the legal conclusion depends on the complete evidentiary and statutory context—not on percentages alone.