Why did we choose our methodology: Rake?

The Rake method is designed to ensure that men and women are compared on an equal footing by rebalancing the distribution of background variables (such as years of experience and job level). Every individual is assigned a weight so that, across these variables, the male and female groups become statistically comparable. See an example below where we assign everyone a weight based on the variable job level. You can see that the distribution of this variable is the same for the male and female group when using the weights.

With those weights, we can calculate the weighted salaries and derive the equal pay gap using a definition similar to the legally required definition for the gender pay gap:

(Average weighted male earnings – Average weighted female earnings) / Average weighted male earnings

That brings us to the first advantage of using the Rake method: it aligns the gender pay gap (also called nominal pay gap) with the equal pay gap (also called corrected pay gap).

Other advantages are:

• The model does not have to obey assumptions:
  • It works with nonlinear and correlated variables.
  • No complex data transformations are needed to make it work.
    
    
• The result is accurate and transparent:
  • It offers transparency in how the data is adjusted as this is reflected by the weights.
  • Other than regression, there is no error-term included in the equation, which makes its results more accurate.
  • The results can be reproduced by hand (even though we are more than happy with the computational support), which makes it easier to explain.
    
    
• It gives more than an equal pay gap and the impact of background variables.
  • It allows reporting the results both in the absolute pay difference (in EUR) and the relative one (as a percentage). This makes it easier for stakeholders to grasp the real-world implications.
  • Besides the difference in average (mean) pay, it is also possible to explore the difference in midpoint (median) pay. The median often gives a more balanced picture as it is not affected by extreme salaries (a few very high or low earners can distort the average).

However, like any approach, our Rake method also comes with its own set of challenges:

• Group definitions matter:
  • To equalize the distribution of variables, background factors like age or years of experience must be categorized into groups. These groupings significantly influence the results. Too narrow, and there will not be enough employees in the group to compare or get reliable results; too broad, and you may mask important differences.
    
    
• Requires robust data:
  • The method requires a sufficient number of individuals in each subgroup to generate reliable outcomes. If certain groups (e.g., young females in a specific role) are too small, the weights can become extreme, and the results less dependable as it will then be highly impacted by a small set of employees.
  • There cannot be unknown values; everyone needs to belong to a group.
    
    

In contrast, regression-based methods will still produce a result with very small groups or when the values of individuals are unknown. However, the quality of that result may be questionable.

When we weigh up (pun intended) the pros and cons of all methods, we prefer our Rake method as it aligns with reporting requirements, the results are accurate and transparent, and the results offer many relevant insights.

So far in this blog series, we have looked at why Equal Pay methodologies matter, what methods are available, and the strengths and weaknesses of each. In the final blog of the series, we will tackle the most important question of all:

👉 What’s next?

Because, while identifying a corrected pay gap is a vital step, closing it is what truly counts.