The UK government has promised to give their views and proposals for introducing Ethnicity Pay Gap (EPG) reporting by the end of 2020. A year ago, I pointed out statistical, data & ethical issues with EPG and listed 5 possible ways EPG could be introduced but I have not yet focused on what employers should be reporting. I have now concluded that Ethnicity Pay Fingerprints are vastly superior to Ethnicity Pay Gaps and my new recommendation is that all employers with 500 or more employees should be required to report their Ethnicity Pay Fingerprint (EPF) instead of their Ethnicity Pay Gap. If EPF is widely adopted and found to be beneficial then I would recommend that reporting of other protected characteristics such as gender & disability should be reported using Pay Fingerprints instead of Pay Gaps.
When I wrote my first article on EPG “Should the UK introduce Ethnicity Pay Gap Reporting?“, it followed my appearance at the Treasury Select Committee in Parliament to give evidence on “Effective of Gender Pay Gap Reporting“. In both the article and my evidence I made the following point.
“I completely support the idea that pay gaps for ethnicity and other protected characteristics should be measured. My concern is about how this is done since the gender pay gap reporting process is not fit for purpose when measuring minorities and an alternative process is needed.“
My biggest concern with EPG was that most employers would end up with really small sample sizes in some categories of ethnicity that would render any analysis of ethnicity pay gaps meaningless unless the employer was very large or in an area with high ethnic diversity. In my second article on EPG “How could the UK introduce Ethnicity Pay Gap Reporting?” I listed 5 options for introducing EPG and whilst I stated a preference for the 5th option, I foresaw a number of problems even with that. I now realise that I had not focused on what data an employer should be reporting around ethnicity and having pondered that, I have concluded that all employers with 500 or more employees should be required to report only their Ethnicity Pay Fingerprint. By focusing on EPFs only, I am of the opinion that EPF are more robust to some of the statistical and data issues I have commented on at length in my previous two articles
What is a Pay Fingerprint? It is a single chart which shows the breakdown by a protected characteristic within each of the 4 Income (or Pay) Quarters for an employer. If the breakdown is by ethnicity, I call the Ethnicity Pay Fingerprint or EPF. I now believe that this is the only data that an employer needs to report as it more or less tells the entire story.
Ethnicity Pay Fingerprints tell your story in 1 chart
The EPFs of six fictional employers are shown in the graphic below along with a statement in the brown label of how many employees they have e.g. employer A has 8,000 employees. Each income (or pay) quarter is then broken down into 6 broad ethnic categories White, Asian, Black, Mixed, Other & N/A for those who chose not to disclose their ethnicity.
Why not take a moment to look at each employer in turn and note what you see in each chart? I’m talking about the overall story that jumps out of the chart to you, not the detailed numbers.
6 Ethnicity Pay Fingerprints, 6 very different stories
Here are the stories that each EPF tells me.
Employer A (n=8,000) – This employer is broadly 50:50 white:non-white and non-white employees are of diverse ethnicities. Around 10% of employees have chosen not to disclose their ethnicity which is typical. Most significantly, there is no pattern of ethnic category by pay quarter. Both high and low paid roles have ethnically diverse employees and the overall picture is one of an employer where all ethnicities are represented in all pay scales. Whilst there may be a statistical pay gap by ethnicity, visually any such gap is not obvious from the EPF and would not be regarded as meaningful.
**Note** if you are not familiar with the mathematical relationship between income (pay) quarters and the median pay gap, please read either of these articles “Gender Pay Fingerprints are superior to Gender Pay Gaps” or “How to spot an incorrect median pay gap“.
Employer B (n=1,000) – Non-white employees are more or less invisible here! With only a 1,000 employees it is obvious from the EPF that there are only a handful of non-white employees. Consequently any statistical pay gap is completely meaningless since the real question here is why is this employer overwhelmingly white? If the employer is based in a location that is 95%+ white then the answer is obvious but if ethnic minorities make up 10% or more of the locality or candidate pool, then this employer needs to be asking themselves why do they have so few ethnic minorities? This is one of the reasons I like EPFs, it is obvious when sample sizes for an ethnic category are too small for analysis. If you can’t see them in the chart, that is the story!
**Note** if you want to answer the question as to whether the overall proportion of non-white employees (or indeed any category) differs significantly from what could reasonably be expected given the candidate pool, you must consult a statistician. I give an overview of how a statistician will answer this question in my article “Is all White alright?“
Employer C (n=5,000) – This employer clearly has a large ethnicity pay gap since the lowest paid quarter is ~70% white whilst the highest paid quarter is ~90% white. The exact size of the pay gap will depend on the pay scale of the employer. An employer with a relatively flat pay scale (sometimes measured by something called the Pay Ratio) will have a smaller pay gap than an employer with a very wide pay scale even if the two employers have exactly the same EPF. However, when it comes closing the gap, both employers will have to do the same thing, find ways to promote/recruit ethnic minorities up the pay scale in the same proportion.
On a separate point, I note that Asians make the main proportion of ethnic minorities here and again if this differs from what might be expected given the candidate pool or location that should prompt questions. Finally, N/As are more common among low paid employees than high paid employees which could point to a trust issue among the lower paid.
Employer D (n=300) – Like employer C, this employer clearly has a pay gap. This time though, there appears to be a difference between Black & Asian employees. Black employees are more common among low paid employees whereas Asian employees are equally represented across all pay quarters. That suggests the employer’s focus should be on why are black people less likely to be found in higher paying roles.
One point should be borne in mind though is that this employer only has 300 employees which means each pay quarter only has 75 employees. Therefore the small black bar in the upper middle quarter may only be 2 employees and that is something that could change very quickly. Recommendation #8 of the Royal Statistical Society’s “10 ways to improve gender pay gap reporting” states that any employer with less than 100 employees in a single category should be flagged as having hard to interpret. I take the same view with income (pay) quarters, there needs to be a minimum of 100 in each quarter to allow reasonably robust interpretation of the data. This explains why I recommend the minimum employer size for EPF reporting should be 500 as that guarantees at least 100 employees in each pay quarter.
**Note** – I explain the statistical background to “minimum of 100” in my article “Life on Mars” which demonstrates how sensitive pay gaps are to small sample sizes.
Employer E (n=400) – Given that white employees are slightly more likely to be found in higher paying roles than low paying roles, the employer will have a small ethnicity pay gap. More notable though is that this employer is minority white with White and Asian employees in roughly equal proportions together making up 80% of employees. Again we should note that there are only 400 employees in total so we have to be cautious about over interpreting any small differences.
A more interesting point is that this time, black employees are more likely to be found in higher paying roles though obviously the sample size is small. This might seem unusual but note that so far I am only using broad ethnic categories. The UK census breaks these categories down further with black typically split into Black-British, Black-Caribbean, Black-African and Black-Other. Available data (can be found at the Race Disparity Unit) shows that Black-African do much better economically and educationally than Black-Caribbean. So it would be worth looking at these employees in more depth to see if that explains what we see here.
Obviously, splitting the broad categories into more detailed categories comes at the cost of sample size as any such analysis could often involve only a handful of employees in the more detailed categories. There are only 2 ways to overcome this. First is to raise the minimum reporting threshold to 5,000 employees which was my 3rd option from my list of 5 in my 2nd article on EPG. The second is to strongly encourage any employer who wishes to do such detailed analysis to consult a professional statistician who can tell you which conclusions are valid and which are invalid. This option was in fact my 1st option from the list of 5 in the link just given. I have now concluded that the latter is preferable as that allows the minimum reporting threshold to be kept lower at 500 employees instead of 5,000.
Employer F (n=1,200) – If EPG was mandated and followed the GPG approach, this employer would in all probability be reporting no ethnicity pay gap. If instead they are mandated to publish their EPF chart instead, it is immediately obvious that the real question to be answered here is why did only 50% of employees declare their ethnicity? This is not unusual, I have seen non-response rates for ethnicity questions from many surveys over the years vary between 5% & 40%. But without the chart, the employer might be able to avoid having to make that point clear. I strongly believe a clear visual image is superior to a dry set of numbers and this is what EPF offers. The calculations are simple to do and the narrative from the chart is clear to all in a way that is not clear from a set of numbers.
Of course, the employer would have to comment on this and they may say one of two things. The first might be a workplace with some real trust issues among the staff and the employer would have to work hard to restore trust among staff. The alternative response might be that staff have genuine objections to their employer keeping a permanent record of their ethnicity on the employer’s IT system. After all, the lesson of history is not kind to those societies that have required people to be classified by race and I am strongly against employees being mandated to state their ethnicity. Employees must have the right not to state their ethnicity and it is for the employer to persuade them of the value of doing so.
My Recommendation for Ethnicity Pay Fingerprint Reporting
- Employers with a headcount of 500 employees or more should be required to submit EPF data. The equivalent limit for GPG reporting is 250 and I estimate about 50% of GPG reporters would also need to submit EPF data as well.
- They will need to allocate their employees to 4 Income (Pay) Quarters just as they currently do for gender pay gap reporting.
- Within each quarter, they should record and submit the % of employees falling in each of the 6 broad categories White, Asian, Black, Mixed, Other, N/A.
- If an employer wishes to publish analysis using more detailed ethnic categories, they should be strongly recommended to have such analyses audited or produced by a professional statistician especially when sample sizes of some categories are small.
- On their own website, they should post an EPF chart like the ones used in this article.
- The definition of pay should ideally be the total remuneration over a 12 month period. Currently gender pay gap reporting is based on a single pay period and splits hourly pay and bonus pay so I recognise that EPF may have to start with the same definition of pay as GPG. However, I am separately calling for GPG to change its definition of pay to total remuneration over 12 months and perhaps the introduction of EPF could be a catalyst for change to GPG at the same time.
- I am not recommending the mandatory publication of a calculated ethnicity pay gap. As I explained in the 6 examples, the EPF is there to prompt questions and actions in a way that can be missed by simply looking at pay gap statistics. Employers are free to calculate and publish pay gap figures as well but the only mandatory output should be the chart itself as I contend that tells the story.
Some further comments
I hope by now I have convinced you that a single Ethnicity Pay Fingerprint chart as used here is a simple and relatively robust way of bringing transparency to the ethnicity profile of the UK’s employers. By getting away from a dry set of pay gap numbers and instead taking in the story portrayed by the EPF chart, I believe it will be easier for the employer, employees and the general public to see what is going on and to ask the right questions. Unless the right questions are asked, employers run the danger of taking incorrect actions that change nothing or make things worse.
If I am right that an EPF chart is easier for a non-statistician to use in order to ask the right questions, I am now more relaxed about the sample size issues that were bothering me so much in my earlier articles. That was because back then, I was envisaging employers publishing a set of numbers just like they do for gender pay gap reporting and then agonising over the fact that their white-black pay gap had widened by 2.63%. I have seen this kind of error in GPG and I just knew it would be 10 times worse with EPG which would discredit the whole exercise. Interpreting the data correctly is so important and I think EPFs are vastly superior in this regard.
Of course, there is no reason why only ethnicity can be done using pay fingerprints. Other protected characteristics such as disability could be analysed in the same way. Indeed I would also like to see gender analysed with fingerprints since that could overcome a troubling gap in gender pay gap reporting. At present, anyone not identifying as male or female or who refuses to give their gender has to be excluded from gender pay gap reporting. With gender pay fingerprint reporting, they can be included and GPF (Gender Pay Fingerprint) chart would now have 4 broad categories of Male, Female, Other & N/A.
Finally for large employers, there is no reason why their analysis should be confined to income (pay) quarters. They have more than enough employees to break them down into eighths or tenths i.e. their EPF chart would have 8 or 10 bars. Indeed they would be able to undertake intersectional analysis where each quarter/eighth/tenth is split by gender and ethnicity and more detailed insights could be had. I plan to explore how EPFs can be taken to the next level in a future article.
** Update 30th November 2020 **
I submitted this article to the Commission for Ethnic Disparities & Inequalities in the UK who requested evidence on what could work to reduce such disparities. The commission posed 10 questions and I submitted a response to question 2 which covered the area of employment. This consultation has now closed and I will add any further relevant links below.
— Subscribe to my newsletter to receive more articles like this one! —-
If you would like to receive notifications from me of news, articles and offers relating to diversity & pay gaps, please click here to go to my Newsletter Subscription page and tick the Diversity category and other categories that may be of interest to you. You will be able to unsubscribe at anytime.
— Want to know more about pay gaps? —
You will find a full list of my pay gap & diversity related articles here which are grouped by theme.
— Need help with understanding your pay gap? —
I offer the following services. Please click on the headings for more details.
- Analysis – I can dig deep into your data to identify the key drivers of your pay gaps. I can build a model using a large number of variables such as pay band, seniority, job function, location, etc and use this to identify the priority areas for closing your gaps.
- Training – I run training courses in basic statistics which are designed for non-statisticians such as people working in HR. The courses will show you how to perform the relevant calculations in Microsoft Excel, how to interpret what they mean for you and how to incorporate these in an action plan to close your gaps.
- Expert Witness – Has your gender pay gap data uncovered an issue resulting in legal action? Need an expert independent statistician who can testify whether the data supports or contradicts a claim of discrimination? I have experience of acting as an expert witness for either plaintiff or defendant and I know how to testify and explain complex data in simple language that can be easily understood by non-statisticians.
If you would like to have a no-obligation discussion about how I can help you, please do contact me.
