Missing compensation data should be treated as a data-quality issue before it becomes a statistical issue. Employers should first distinguish a true missing value from zero pay, not-applicable fields, delayed payroll data or failed joins between HR and payroll systems. Directive (EU) 2023/970 requires management to confirm the accuracy of Article 9 reporting, but it does not prescribe one universal imputation method for missing employer data. The safest approach is to quantify missingness, correct source records where possible, document every exclusion or transformation, avoid silent assumptions and run sensitivity checks when the treatment could materially affect the pay-gap result.
Jurisdiction: European Union
First Determine What the Missing Value Actually Means
A blank compensation field can represent several different situations. The worker may genuinely have no recorded value, the component may not apply, the correct amount may be zero, the payment may not yet have been processed or the value may have failed to join from another system. Those cases should not be treated identically. Before calculating a pay gap, analysts should classify the reason for the missing value and retain that classification in the working dataset. This prevents a blank cell from being silently interpreted as zero or from being excluded without explanation.
Quantify Missingness Before Choosing a Treatment
The next step is to measure how much data is missing and where the problem is concentrated. Analysts should calculate missing-record counts and percentages by field, worker category, sex, location and source system where relevant. A small random gap may have little effect, while missing values concentrated in one group can materially bias the result. This diagnostic work also helps identify system problems, such as one payroll feed failing for a particular country or one job family lacking working-time data. The treatment decision should be based on the pattern of missingness rather than on convenience.
Correct the Source Where Possible
If the missing value reflects an extraction, mapping or source-system error, the preferred solution is usually to correct the underlying data and rerun the analysis. Replacing missing values manually in the analytical file can solve the immediate calculation but create a repeatability problem later. Where a source correction is not possible, the analyst should create a documented transformation or exception field and preserve the original value. That approach keeps the data lineage intact and makes it possible to explain exactly how the reported or analytical result was produced.
Do Not Impute Compensation Silently
Imputation can be appropriate in some analytical contexts, but it changes the dataset and can introduce assumptions into a sensitive pay analysis. A missing salary should not automatically be replaced with the category average, midpoint or prior-year value without a clear methodological reason. The Directive does not prescribe one universal employer imputation method. If imputation is used for an internal analysis, the rule, affected records and rationale should be documented, and the employer should distinguish that analytical treatment from any official reporting method required by national law.
Check How Exclusions Affect the Result
Excluding incomplete records can also create bias. If missing values are concentrated among one sex, one category or one pay component, dropping those workers may make the apparent gap smaller or larger. Analysts should therefore compare the included and excluded populations and, where practical, run a sensitivity analysis using alternative defensible treatments. The purpose is not to manufacture a preferred result. It is to understand whether the conclusion depends heavily on a data-quality decision. Material sensitivity should be recorded in the methodology and escalated for review.
Tie Missing-Data Controls to Reporting Accuracy
Article 9 requires the accuracy of reported information to be confirmed by management after consultation with workers' representatives. Missing-data controls therefore belong in the reporting governance process, not only in the analyst's spreadsheet. A mature process should identify required fields, assign data owners, set validation thresholds, document unresolved exceptions and retain evidence of review. Where information is inaccurate or incomplete, the organisation should correct the issue and be prepared to explain the methodology used. This turns missing-data handling into a repeatable control rather than a last-minute cleanup exercise.
Frequently Asked Questions
Should missing pay values be treated as zero?
Only when zero is the true economic value. A blank field may mean unknown, not applicable, delayed or failed extraction, so the reason should be established first.
Does the EU Pay Transparency Directive prescribe a missing-data imputation method?
No. The Directive requires accurate reporting but does not prescribe one universal employer imputation rule. National reporting guidance should be checked for detailed treatment.
Can incomplete employee records simply be excluded?
They can sometimes be excluded for a documented reason, but the employer should assess whether the exclusions are concentrated in a group and whether removing them materially changes the result.
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.