#### What’s changed?

We’re improving the way we think about pay gaps.

Breaking down your overall pay gap into two gaps – the **unadjusted gap **and the **adjusted gap **– allows you to better understand the root causes of pay disparity.

Some pay disparity is fair and justifiable – for example, it is fair that someone with 10 years experience is paid more than someone with 2 years experience. Looking at **adjusted gaps **enables you to see what your numbers look like when you **remove the factors which might justify pay differences.**

In essence, you are now able to look at your unadjusted pay gap – the same pay gap that you were seeing in Gapsquare before – and your adjusted pay gap – which helps you look at **unexplainable differences **in pay.

This provides you with a better understanding of why pay gaps occur and which areas you need to investigate further.

#### NB: To access these features you will have to finalise your dataset. Learn how to do this here >

### Learn more with our video demos

###### Updated Dashboard & Unadjusted Pay Gap

###### The Adjusted Pay Gap

#### Defining and Understanding

**Unadjusted “Raw” Pay Gaps **

The unadjusted pay gap is the difference between the average pay of a non-baseline group to the average pay of the baseline group. This gives an overall position on equity but is often related to structural issues within the organisation, rather than individual cases of discrimination.

In this example, the mean average hourly rate of a female in the organisation is 10 per hour. The mean average hourly rate of a male in the organisation is 20 per hour.

The unadjusted pay gap i.e. the difference between the two, is 10 per hour.

**Explaining Unadjusted “Raw” Pay Gaps **

The unadjusted pay gap can be broken down further. It is made of an explained and an unexplained proportion.

The explained proportion accounts for factors which may impact the pay of an individual, like tenure, job level and education. These can be called pay determining characteristics.

The unexplained proportion is what is left, and should be investigated further. This unexplained proportion is also called the adjusted gap, as it is the gap which accounts for pay determining characteristics.

#### A New Approach to Pay Gap Analysis

##### This blog piece will help you to:

- Understand the difference between unadjusted and adjusted pay gaps
- Demonstrate how this can help you
- Show you how real organisations are looking at their unadjusted & adjusted pay gaps

#### FAQ

Regression is a statistical method that determines the impact of a dependent variable on other independent variables. In this case, we’re looking at how Pay Determining Characteristics impact your pay gap(s).

Pay Determining Characteristics are factors that should impact someone’s pay. For example, tenure is a Pay Determining Characteristic – the longer someone works in a company or industry, the higher their pay typically is. Protected characteristics like ethnicity or gender aren’t Pay Determining Characteristics, because employers cannot pay someone more (or less) than others based on their protected characteristics.

The Adjusted Pay Gap is the pay gap after adjusting for factors that influence the pay gap – the Pay Determining Characteristics.

Each Pay Determining Characteristic is treated as a separate variable that will impact pay. Likewise, different Pay Determining Characteristics will impact your Unadjusted Pay Gap at different magnitudes. Your Unadjusted Pay Gap could be largely explained by Job Level but not explained as much by Education Level. For instance, Job Level could contribute 60 percentage points towards explaining your Unadjusted Pay Gap, 30 percentage points by Education Level, and the rest of it can be explained by less significant characteristics.

Each contribution is then subtracted from the Unadjusted Pay Gap to give the Adjusted Pay Gap. The Adjusted Pay Gap can therefore be interpreted as the gap that is unexplained by these variables.

The variables that contribute to explaining your Unadjusted Pay Gap can be further understood through consultancy, using a deeper dive of your data. Other areas of interest that are not provided in the data can also be determined.

The first thing to look for is the R-squared and Adjusted R-squared. The R-squared describes how well the dataset fits the regression curve.

The Adjuster R-squared shows how well the dataset fits the regression model, taking into account the actual number of data points considered by the model. Therefore, if you add insignificant data points into the regression, the Adjuster R-squared will decrease as more and more data points are unused by the model.

For both R-squared and Adjusted R-squared, we describe 80-100% as high accuracy, 70-80% as fair accuracy, 50-70% as low accuracy. We tend not to use results with R-squared or Adjusted R-squared lower than 50%.

If it is within the +/-5% boundary, then it can be considered a “good” Adjusted Pay Gap. Generally, a negative Adjusted Pay Gap would imply that some women are being paid more than men for reasons that the data cannot explain. This doesn’t necessarily mean that “positive discrimination” is occurring; it may mean that we require more data or more granular analysis to further justify the pay gap.

Analysing both the unadjusted and adjusted gaps is useful to pinpoint areas of improvement in your workforce. **For example, the unadjusted gap reflects under- or over – representation of the reference or comparator group e.g. if the unadjusted mean gender pay gap is 20%, this means that on average, men as a group get paid 20% more than women as a group. **This reflects the lack of female representation in the higher paid roles and signifies to the company to question the distribution of their female workforce and the accompanying salary distribution. As a whole, this is a very useful insight to improve representation of a certain group of employees.

On the other hand, looking at the adjusted pay gap – which enables you to see what your numbers look like when you remove the factors which might justify pay differences – allows you to see if groups are being paid equally for the same work with the same experiences and backgrounds.

Any positive (or negative) adjusted gap means that the data cannot fully explain differences in pay among groups and further investigation is required. This can be because not enough data was analysed, there are unobserved variables (e.g. ability) or that there is some form of discrimination present e.g. unconscious bias.

If the dataset is too small, the regression results will be statistically insignificant. We typically require at least 25 employees of each gender to run regression, but this number can vary by +/- 10 employees depending on how your data is distributed and what Pay Determining Characteristics you provide in the data.

The unadjusted pay gap is the difference between the average pay of a non-baseline group to the average pay of the baseline group. This gives an overall impression of your pay gap, although outliers can impact the figure. However, this does not yet account for factors that can explain why the pay gap is there.

The difference between the Unadjusted and Adjusted pay gaps is the explained Unadjusted Pay Gap – it represents how much of your pay gap has been explained by Pay Determining Characteristics.

For example, if you have a positive Unadjusted Pay Gap, but more men have longer tenures than women, then tenure will explain why these men are paid higher. Therefore, tenure will explain part of your Unadjusted Pay Gap.

Each Pay Determining Characteristic (provided by you in the data) will have a Contribution by Variable towards explaining your Unadjusted Pay Gap.

A positive pay gap is when a group (such as women or BAME) is paid less than their comparator group (such as men or White people). Conversely, a negative pay gap is when the comparator group is paid less. For both positive and negative pay gaps, a negative Contribution by Variable will make your Adjusted Pay Gap increase; subtracting a negative number is the same as adding a positive number.

There are a few ways to interpret negative contributions. The first is that based on the variable(s) with negative contribution, a pay gap should be higher than it is. Taking the gender pay gap as an example, if the only Pay Determining Characteristic in the data is Education, and women are paid more than men for the same level of Education, then the regression results would tell us that men need to be paid more. However, this is not conclusive, as other factors – such as Job Level or Performance – have not yet been provided in the data. Your actual Adjusted Pay Gap will be influenced by the sum of all variables’ contribution, so even if some variable contributions are negative, there could be larger positive values.

A second interpretation depends on the magnitude of the negative value. If a value is fairly close to zero – for example, -0.0001 – then the variable in question is not as significant.

For the UK, EU and US, we’d expect a good Adjusted Pay Gap to be below +/- 5%. Other regions in the world are benchmarked on a case-by-case basis.

Companies are able to set their own standards on our app to categorise their pay gap as either Red, Amber or Green – high pay gap, medium pay gap, and low pay gap respectively.

We either require more data that you have not provided within the app, or it can be a case of current data not justifying your initial Unadjusted Pay Gap. If it’s the latter, then an equal pay audit and scenario planning can be done to find the tailored actions you can take to close your pay gap.

The adjusted pay gap can be larger than the unadjusted pay gap when the total contribution by variable is negative, since: Adjusted pay gap = unadjusted pay gap- (totally contribution by variable). A total negative contribution by variable tells you that for the **pay determining characteristics in your data, **women are paid more than men. But overall, in your organisation, men are paid more on average than women. For example, if Job Level is the only pay determining characteristic mapped, and women are paid more than men in the same job level, you get a negative contribution by variable. However, your raw difference in pay still tells you across the whole company, men are paid more than women- which means that the adjusted pay gap increases. This doesn’t mean discrimination is necessarily happening- it could mean that you need to add more pay determining characteristics in the data.

#### Want to know how it works?

This technical whitepaper looks at our regression analysis method, with limitations, equations and easy explainers