Beyond compliance: measuring the gender pay gap under the EU Pay Transparency Directive – a practical approach

The EU Pay Transparency Directive (2023/970) marks an important shift in how organisations approach pay equity. Companies will be required to report gender pay differences, demonstrating that compensation systems rely on objective and transparent criteria. This raises a practical question for many organisations: how can the gender pay gap be analysed in a way that is methodologically sound and at the same time useful for managerial decision-making?

We examines the mechanisms that shape pay gaps and the analytical tools that allow organisations to distinguish structural differences from unequal treatment.

Can we tell if the status quo provides equal pay for equal work?
Most organisations are comfortable with an implicit assumption: we reward merit. Presumably, pay reflects performance, responsibility and impact; whereas differences in salary signal differences in contribution. Managers know that the system isn’t perfect, but chose to believe that it’s broadly fair.

The EU Pay Transparency Directive unsettles that comfort. It is often treated as a compliance exercise, a reporting obligation to manage while limiting legal and reputational risk. Yet this framing misses the opportunity inherent in the directive: any company benefits from knowing whether ‘equal pay for equal work’ holds within its own walls. This question is not ideological; it is purely empirical. It reflects meritocracy and can only be answered with a thorough and skilful examination of the data. Understanding pay structures within the firm is an asset, and their correct management serves the firm’s performance.

Pay is the key incentive for workers to both contribute to the organisation and shape their career paths. Pay structure sends a forceful signal of what activities are valuable for the firm, where workers should excel in their efforts and which talents they should develop. Unequal pay for equal work distorts the signals, disincentivises effort, and disenfranchises the workers.

Unequal pay for equal work is not a sustainable business model. It develops over time.  Even initially perfectly meritocratic pay structures become archives of past negotiations, managerial discretion and changing market conditions. The resulting pay structure drifts away from the meritocratic logic it was meant to reflect.

This article examines the role of pay as an incentive. We explore how organisations might drift towards unequal pay, and how econometric tools can help firms distinguish structural differences from unequal treatment. We conclude on a vision of the pay transparency directive that exceeds compliance, in favour of an audit of the incentive architecture embedded in pay.

Pay structure as an incentive system
Pay is one of the organisation’s most powerful signalling devices. It speaks more credibly than words on which activities matter, where to invest effort, and how success is defined. Differences in salary are meant to reflect differences in responsibility, performance and contribution. If pay differences cannot be explained by transparent, objective criteria, the incentive mechanism weakens and is perceived as illegitimate. Effort and engagement decline when the link between contribution and reward appears loose.

The signalling effect of pay extends beyond the walls of the organisation. Even perceptions of unequal pay can discourage talent from entering or remaining in the firm. In this sense, distorted pay signals not only misallocate internal effort—they also affect the firm’s ability to attract and retain talent.

Equal pay analysis is therefore not only about fairness or legal compliance. It is a diagnostic test of incentive coherence: can the current pay structure be systematically justified in terms of productivity and value?

Why gaps emerge: unconscious bias and job architecture
Pay gaps rarely stem from explicit policy. More often, they arise from decentralised decisions operating within imperfect structure due to unconscious biases and job architecture.

Unconscious bias can enter through performance evaluation and promotion processes. Assessments of ‘potential’, leadership or readiness are inherently discretionary. Even when formal criteria exist, interpretation varies. Small differences in ratings, negotiation outcomes or promotion timing can translate into persistent wage divergence once linked to percentage-based increases and career ladders. These effects compound over time without requiring ill intent.

Structural features of job architecture create a second channel. Functions associated with coordination, support or care are often valued differently from revenue-generating or technical roles. If grading frameworks or market benchmarks reflect such patterns, disparities may arise before individual performance even enters considerations.

Equal pay analysis is therefore only as neutral as the underlying classification system. If comparable roles are placed in separate grades, they never enter the same comparison group. If groupings are too broad, meaningful comparisons become challenging. Defining responsibility, effort and value is not merely legal compliance; it is an organisational design choice that shapes pay outcomes. Without examining both evaluation processes and job architecture, salary comparisons risk overlooking the structural sources of divergence.

Measuring the gap: comparing the comparable
Measurement is essential, but it is far from mechanical. It requires defining what comparison is normatively relevant and statistically defensible. The difference in average earnings between men and women is transparent and required for reporting. Yet it reflects the contribution of multiple mechanisms: occupational segregation, hierarchical distribution, tenure, working-time patterns, among others. What we ideally need is to isolate (un)equal pay for comparable work.

To isolate the effect of gender we require to compare observationally equivalent workers: in similar roles, tenure, education, and all other outcome-relevant information that the organisation can determine. The aim is to approximate the counterfactual: what would a woman earn the same income as an observationally equivalent man?

Technically it is a bit harder: the analysis needs to adequately tackle the fact that the more precise the definition of comparable worker, the more likely it is that no such identically matched worker of the opposite gender actually works in the organization. One does not want to rely on extrapolation (an imaginary, statistical “comparable worker”) and thus sometimes trade-offs emerge between precision and coverage. Broad comparison groups risk overstating inequality by failing to account for relevant differences, whereas extremely narrow groups risk understating inequality by eliminating meaningful contrasts or excluding segments of the workforce. A thorough analysis necessitates explicit methodological choices and academic-level command of over 1000 statistical estimation methods for obtaining counterfactual wages to select a method adequate for the organisation.

Beyond averages, distributional and dynamic methods can reveal whether gaps concentrate at the top, at entry levels or in promotion probabilities. Dynamic analysis adds another dimension. Promotion hazard models and cohort tracking can identify whether men and women progress at different speeds through the organisational hierarchy. The objective may thus be to understand the mechanisms rather than to provide a single number. Measurement, properly designed, becomes a governance tool rather than a reporting formality.

Can we rely solely on job valuation?
Job valuation defines the relative importance of roles and aligns the pay bands accordingly. It is not sufficient to determine equal pay for equal work. First, especially in small organisations, point scales will be insufficiently granular and pay bands may be excessively wide, to make comparisons based on job valuations meaningful. Second, pay is supposed to reflect performance, an inherently individual measure, whereas valuation refers to jobs. Third, job valuations invariably reflect some degree of unconscious bias by the evaluators: undervaluing of organisational and support tasks that are essential for firm performance. For example, assessments can undervalue tutoring and mentoring of new workers and overvaluing sales and technical roles. An excess zeal on metrics can lead to overvaluation of activities that are easy to objectively measure, and an omission of tasks that are crucial but hard to measure. If these tend to align by gender (e.g. ‘pink’ jobs concept), job valuation becomes a tool for justification rather than correction of pay structures. Cementing existing inequity, job valuation may rationalise status quo. Even if it meets formal requirements, the underlying distortions embedded in an unequal pay structure remain unalleviated.

Finally, an organisation runs the risk of cementing inequality even if it executes a perfectly gender-neutral job valuation. This will occur if pay bands across roles – which are benchmarked to market – reflect bias aggregated across other employers. In fact, benchmarking to market can act as a convergence mechanism: firms with more equal structures may drift toward less equal norms, while more unequal firms may improve. The result may be convergence – but not necessarily toward meritocratic outcome.

Job valuations and gender wage gaps measurement can be synergetic. A thorough measurement can serve to cross-validate job valuations, whereas thorough job valuations can in turn contribute to validating the identification of comparable workers.

Practical implications for organisations – how to approach measuring the pay gap
The discussion above shows that measuring the pay gap requires both statistical rigour and a clear understanding of how compensation decisions are made within the organisation.

Statistical expertise plays a key role. It serves to isolate the role of gender in effective pay structures. Such adjustment can be tailored to the organisation thanks to the availability of multiple econometric methods. Expertise helps to design appropriate model specifications, explain differences transparently and produce an intelligible report. The objective is not merely to produce a single number but to understand how the pay system functions. Measurement thus becomes a diagnostic tool, revealing whether the pay structure operates as intended and where corrections may be required.

In the approach applied by Arthur Hunt Consulting in cooperation with the research centre GRAPE (Group for Research in Applied Economics), practical HR and advisory experience is combined with academic expertise and established econometric methods used in labour market research.

Importantly, a formal job evaluation system is not a prerequisite for measuring the pay gap. While job evaluation may strengthen comparisons between roles, the essential requirement is a solid understanding of the data organisations already possess and the ability to use it effectively. This is good news for companies that have not yet implemented job evaluation frameworks and are concerned about meeting the new regulatory requirements.

In most organisations the necessary data already exists within HR and payroll systems. These datasets typically include base salaries, bonuses, job levels, tenure, working time, location and promotion history. In many cases they also contain information about organisational units, functions, career paths or performance ratings. The challenge therefore lies not in collecting new data, but in structuring and integrating existing information so that meaningful comparisons and reliable econometric analysis become possible.

The analysis typically begins with understanding the organisational context and defining the objectives of the study. This includes examining the company’s structure, compensation philosophy and HR processes that influence pay decisions. Based on this understanding, an analytical model can be constructed that separates differences explained by objective factors — such as experience, role, seniority or working time — from unexplained differences that may represent the actual pay gap.

Once the dataset has been structured, econometric methods can be applied to estimate the pay gap while controlling for relevant characteristics of employees and roles. The objective is not only to calculate an indicator but also to identify the mechanisms that generate differences in pay, such as structural differences between functions, promotion dynamics or historical compensation decisions embedded in the organisation’s pay structure.

An important stage of the process is the interpretation of results with HR and leadership teams. Analytical findings are typically discussed during workshops where statistical results are reviewed alongside organisational realities, helping translate analytical insights into an understanding of how pay structures function in practice.

Organisations may also conduct simulations of potential policy scenarios, modelling how changes in compensation policies, promotion rules or salary band structures could influence the pay gap before implementing any changes in practice. Such simulations help identify which actions may be most effective in reducing the pay gap and allow organisations to assess the likely impact of different policy options.

At the same time, the analytical framework enables organisations to prepare the data required for regulatory reporting under the EU Pay Transparency Directive as well as ESG and CSRD frameworks.

Once established, this analytical framework becomes more than a reporting exercise. It evolves into a management tool enabling organisations to monitor pay structures over time, evaluate HR decisions and guide future policy adjustments based on evidence.

Meritocracy needs to be demonstrated, not merely declared
There is a good business case to strive for meritocracy: it allows pay to play its role motivate workers to achieve organisation’s objectives. There are also good reasons why pay structures can in reality depart from this meritocratic ideal. Previous decisions cast a long shadow over the present, and what could be once an optimal call might have now detrimental effects.

The EU pay transparency directive helps to shake up the status quo. It provides organisations with an opportunity to understand their own structures, identify distortions, and realign incentives. The EU directive offers a real opportunity to develop measuring tools that identify departures from meritocracy. That knowledge, once developed, becomes a strategic asset.

A structured analytical approach — combining robust data analysis, clear interpretation of results and scenario simulations — allows organisations to move from assumptions about pay fairness to evidence-based understanding of their compensation systems, without making formal job valuation a necessary starting point for measuring the gender pay gap.