Small sample sizes make pay-gap analysis harder because a single worker can materially change the mean, median or apparent gap. They also create confidentiality concerns because published or shared figures may make individual pay easier to infer. Directive (EU) 2023/970 does not establish one universal minimum sample-size threshold that employers can use to ignore a category. Employers should document the group size, avoid false precision, consider descriptive and individual-level review, and check national rules on confidentiality or suppression. Categories should not be merged merely to create a larger sample if doing so would conflict with the Directive's same-work or work-of-equal-value framework.
Jurisdiction: European Union
Small Groups Can Produce Volatile Pay Metrics
In a small category, one high-paid or low-paid worker can move the average substantially. The median can also change sharply when only a few observations determine the middle of the distribution. This does not make the figures meaningless, but it changes how confidently they should be interpreted. Analysts should always display or record the number of workers behind a result and avoid presenting a precise percentage as though it had the same stability as a result based on a much larger population. The smaller the group, the more important it is to inspect the underlying records and distribution.
There Is No Universal Minimum Sample Size in the Directive
Directive (EU) 2023/970 requires defined reporting and category-level analysis, but it does not set one employer-wide minimum sample-size rule for every calculation. An organisation should therefore avoid inventing a threshold and then treating smaller groups as legally irrelevant. National reporting guidance may introduce confidentiality or suppression rules, and those rules should be followed where applicable. The legal question and the statistical question remain separate: a category may still be relevant to equal-pay analysis even when it is too small for a sophisticated statistical model.
Use Individual Review When Statistical Inference Is Weak
Where the group is very small, individual-level review can be more informative than formal inference. Analysts can compare job content, grade, pay components, working time, tenure, location and documented pay-setting criteria for the workers concerned. This approach can reveal whether the observed difference follows an objective rule or whether an unexplained disparity remains. The review should still be structured and documented rather than informal. A small sample is not a reason to skip analysis; it is a reason to choose a method that fits the available evidence.
Do Not Merge Categories Solely to Increase Sample Size
Combining groups can improve numerical stability, but only if the combined jobs remain comparable under the same-work or work-of-equal-value framework. Merging unrelated roles merely to reach a preferred sample size can hide meaningful differences and weaken the legal relevance of the analysis. If broader grouping is considered, the employer should document why the jobs belong together using objective gender-neutral criteria. Where that case cannot be made, it is better to retain the smaller category and use a more cautious analytical method.
Small Samples Increase Confidentiality Risk
When only a few workers are represented, a category-level figure may allow colleagues or managers to infer individual pay. Employers should therefore consider applicable national data-protection and confidentiality rules before publishing or widely sharing small-group results. The Directive itself may require information to be provided to workers, representatives or authorities, so confidentiality controls should be designed to support lawful access rather than suppress legitimate rights. Controlled disclosure, aggregation or other national-law mechanisms may be relevant depending on the context.
Flag Uncertainty Instead of Hiding It
A strong report should state when a result is based on a small number of workers and explain how that affected the method. It can record that formal significance testing was not appropriate, that the median was highly sensitive to one record or that individual review was used instead. This is more defensible than presenting a fragile result without qualification. Documenting uncertainty also helps future comparisons because the organisation can see whether the category has grown, whether the same pattern persists and whether a different analytical approach becomes appropriate later.
Frequently Asked Questions
What is the minimum sample size for a pay equity analysis?
The Directive does not set one universal minimum sample size for all employer analyses. The appropriate method depends on the question, data structure and any applicable national guidance.
Can a small worker category be excluded from equal-pay review?
A small sample may limit statistical techniques, but it does not automatically make the category legally irrelevant. Individual and descriptive review may still be appropriate.
Should small categories be merged to make analysis easier?
Only where the broader grouping remains defensible under objective same-work or work-of-equal-value criteria. Statistical convenience alone is not enough.
Related Guides
Official Sources
Requirements and practices differ by jurisdiction and organisation. Check current local law, official guidance and professional advice for a specific situation.