PAGs are Pay Analysis Groups. Essentially, these are groups of employees who share similar characteristics or pay structures, such as Job Level, Department, or Job Function. This is the starting point of group-by-group equal pay audits, by grouping employees based on their Pay Analysis Group.
For equal pay, employees who do substantially similar work should be paid similarly.
It is important that equal pay audits are conducted for groups that have similar pay structures, as different departments may have different metrics for promotions and pay raises. PAGs will give you an overview of groups with comparable pay structures, such as departments or job functions. Within those groups, you can see whether there are pay gaps and employees to investigate for equal pay issues. PAGs will also flag if there is identified disparity in a group.
PAGs therefore allows you to prioritize which groups to start conducting equal pay audits for.
PAGs provides group-level analysis of the following metrics:
- Unadjusted and adjusted gaps
- Identified disparity
- Number of employees to investigate for equal pay
- Statistical significance of the results
The unadjusted pay gap is the raw difference between the average pay of your reference group and the average pay of your comparator group. This gives an overall overview of your pay gap, although outliers can impact the figure. The adjusted pay gap is the pay gap after adjusting for factors that influence the pay gap (i.e., Pay Determining Characteristics (PDCs)). The adjusted pay gap thus becomes more accurate when you map as many PDCs as possible in your data. An acceptable adjusted pay gap would be below <|+/-5|%, but ideally you would have an adjusted pay gap of 0%.
This is flagged when a protected characteristic is statistically found to be potentially contributing to your unadjusted pay gap, according to the Oaxaca-Blinder decomposition method. Your unadjusted pay gap can be explained by contributing factors, which could be Pay Determining Characteristics (PDCs) such as tenure, education or performance. However, in certain cases, it may be statistically explained by employees’ protected characteristics. Ideally, all pay differences should be explained by justifiable PDCs, and not based on employee gender, , disability or other protected characteristics.
These are employees whose salaries fall under their predicted pay based on a linear regression model, by at least 2 Standardized Differences. This means that these employees earn less than their predicted pay by a significant amount compared to other employees in the same PAG.
A result is only statistically significant when it has a P-value less than or equal to 0.05. A P-value measures the probability that a result could have occurred by random chance. Generally, a smaller P-value tells us the results are more reliable.
This classification splits your PAGs into groups with pay gaps more than 5% (red), between 2-5% (amber), and under 2% (green). These intervals are based on benchmarking in the EU and USA, but you may have different targets for your specific company. You can change this classification in your settings. For example, you may want to change the boundary for red groups to 3% if your organization generally has smaller gaps.
The calculation for an unadjusted pay gap can only be done for employee groups with >10 people. This is to meet the conditions such that a group can be split into four quartiles (as outliers are based on the interquartile range), and that the group must contain both a reference group and comparator group. For example, a Job Level with only men will not have an unadjusted pay gap because there are no women in the Job Level to compare to.
If the unadjusted pay gap cannot be calculated, the adjusted pay gap also cannot be calculated as the adjusted pay gap is calculated from the unadjusted pay gap.
The calculation for an adjusted pay gap uses multivariate regression. This requires at least 25 (+/- 10) employees for each employee characteristic to work (i.e., at least 25 men and 25 women). Additionally, datapoints such as data on Pay Determining Characteristics should ideally be distributed across the whole group.
If a PAG does not fit the requirements for multivariate regression, you can either look at the regression results for your entire workforce or investigate the individual PAG manually. The Pay Review feature may still produce results, even if your adjusted gap is not showing. For instance, it may still be able to produce a list of employees to investigate. If so, we recommend doing a manual check on each employee flagged, such as using compa-ratios as benchmarks.
PAG results can be useful for equal pay audits. It helps you identify areas in your workforce where there may be pay disparity issues. By clicking “Explore” on a PAG, you will be taken to an equal pay audit page. There, you can find a list of employees to investigate and do a comparison between them and other employees performing substantially similar work. On the same page, you can also look deeper into pay gaps by understanding the variables that contribute to your pay gap. For example, you may have a pay gap in a PAG due to women being in lower-seniority roles compared to men.
By clicking the “Explore” button, the Gapsquare™ tool provides suggested adjustments for employees to investigate equal pay issues. This can be used as guidance for changing these employees’ salaries. Additionally, we offer consulting services to help you plan scenarios that can close your pay gap and meet diversity and inclusion targets.