When I submitted my evidence to the Commission on Racial and Ethnic Disparities (CRED) in November 2020, I recommended that ethnicity pay gap reporting (EPGR) using the Big 5 ethnicities of White, Asian, Black, Mixed & Other be made mandatory. In the CRED report published in March 2021, recommendation 24 called for analysis of ethnicity to use the Office of National Statistics (ONS) 18+1 ethnicities as much as possible and it is my opinion this is one of the reasons why CRED stopped short of recommending mandatory EPGR in recommendation 9. In June 2021, I published 7 + 5 recommendations for improving gender pay gap reporting (GPGR) and I made it clear that all 12 of my recommendations would also apply to any EPGR system introduced by the government. However, I stated my 2nd recommendation on number of categories to be reported needed more details before it could be applied to EPGR and this article will fill in those details whilst also recording my response to recommendations 9, 10, 23 & 24 of the CRED report.
This post was originally published as a draft on 2nd August 2021. What is published here is the final version with a number of changes from the draft version.
As this is a longer blog post than normal, here is a list of the section headings for guidance.
- What should the minimum number of employees for a reported category be?
- Should all employers be compelled to use the same ethnic categories?
- How can an employer decide how many categories to report?
- CRED Recommendations 23 & 24 – My Opinions
- How many employers will be able to report more than 2 categories?
- CRED Recommendations 9 & 10 – My Opinions
- My conclusions
I cover a lot of ground in this article which will ultimately end up as a list of 12 bullet points in the conclusions which I would like you to take away.
1 – What should the minimum number of employees for a reported category be?
I want to make it clear that what follows in this section is a discussion on the minimum category size for reporting data in the public domain i.e. published numbers that can be commented on, attacked, praised, criticised, etc. Such minima does not prevent employers from analysing their own data privately with categories smaller than the recommended minima. Indeed all employers regardless of their size should be analysing their diversity data regularly and small sample sizes will be impossible to avoid. The purpose of a minimum category size is to ensure that any data put in the public domain is reasonably robust and allows the public to draw reasonable conclusions that are not unfair to the employer.
I have already suggested 2 numbers in previous blog posts. Just to be clear, I am talking about the minimum number per ethnic category for the whole employer here, not per pay quarter.
- 20 – this was stated in recommendation 2 of my 7+5 recommendations for improving pay gap reporting. This was derived on the grounds of anonymity i.e. that it should not be possible to identify individuals from published data. This is common practice in the market research field and in my experience such minimums vary between 5 and 20. I personally use 20 when working for my clients but I would like the Information Commissioner to offer her view on this question as well before the government publishes any recommendations.
- 100 – this was recommendation 8 of 10 recommendations for improving pay gap reporting published by the Royal Statistical Society in 2019. This was derived on the grounds of the statistical robustness. The purpose of this was ensure that the calculation of the median gender pay gap statistic would be reasonably reliable when interpreted by a non-statistician who would not be able to appreciate or measure the extent to which the statistic is affected by random chance. In turn, this recommendation was influenced my blog post of 2018 “Life on Mars” which measured the extent to which the pay gap can fluctuate even when an employer is not discriminating in any shape or form.
When I submitted my evidence to CRED last year, I stated I was more relaxed about a minimum category size of 100 since I was recommending that Fingerprints (as per recommendation 1 of my 7+5 recommendations) be the focus of EPGR rather than the mean or median pay gap statistic. This year I introduced the concept of a swap number here and here and I have undertaken further research based on the swap number to see whether I was right to be more relaxed. I will publish my analysis in a separate blog post in due course but I have concluded that the statistical minimum per category should remain at 100 employees which is disappointing but I can’t argue with the laws of mathematics here. Note this minimum continues to be based on the assumption that the people analysing an employer’s data will not be statisticians. A statistician is capable of drawing from categories with a much smaller minimum of 20 (sometimes fewer) employees but not all employers will be able access such expertise.
Ultimately this is a question that parliament needs to decide, namely what the minimum number of employees per ethnic category should be? My professional opinion is that any minima less than 50 will be irresponsible on statistical grounds and will result in unreliable and misleading numbers for most employers. Should parliament decide to go with a number less than 50, I will speak out and oppose any such move unless parliament also requires employers to take statistical advice before publishing their pay gap report which should include a statement from the statistician as to which statistics are reliable and which are not. I recognise that there is a tricky tradeoff here that Parliament will have to consider and so I will return to this point later on.
Why the discrepancy between my two recommendations highlighted in bold here? The former (>100) is when I am focusing on statistical reliability and I concluded the RSS’s recommendation 8 from 2019 continues to be correct. The latter (>50) is when I focus on the public interest and as I will show later on, a minima of 100 all but destroys the case for mandatory EPGR. So the 2nd minima is me making a professional judgement on how much statistical reliability can be sacrificed to allow for a potential public benefit from such data being made public. I am confident that other statisticians would back my >100 recommendation from the point of view of statistical reliability but there would be a wider range of opinions as to how to make the trade off between public interest and statistical reliability.
2 – Should all employers be compelled to use the same ethnic categories?
The UK ethnic diversity varies considerably around the country and it is unreasonable to expect every employer to have the same ethnicities in their workforce. In my article “No employer has a gender or ethnicity pay gap, let’s celebrate!” I pointed out that one reason why the current GPGR has flaws is that Parliament fell into the trap of believing that it was necessary for employers to be compared with each other. In fact comparability is not needed nor is it desirable since it leads to a league table mentality when in fact what is needed is a Continuous Improvement mentality. For this reason, employers should use ethnic categories that are most relevant to them.
3 – How can an employer decide how many categories to report?
Suppose parliament did choose a minimum per ethnic category of 50 employees, how would an employer decide which ethnic categories to report?
One option is for all employers to start by asking each of their employees to tick which of the 18 (now 19) ethnicities used by the ONS in the census (which I will denote as ONS18) is closest to their identity. Note employees must be free to not answer this question and so the employer will in all probability end up with an extra category of N/A for those who did not respond. The employer should then identify which of the ONS18 categories can be reported separately and which need to be aggregated into larger categories to ensure that the minimum category size is met.
I will illustrate my proposed process with 3 fictional employers, Bristol Makers Plc, Bath Legal LLP and Mendip Foods Ltd.
Bristol Makers Plc has 2,336 employees and the breakdown by the ONS18 categories is shown here. Only 3 of the 18 ONS18 categories have 50 or more employers (W-B, W-O and B-A) which means the other 15 categories need to be aggregated somehow to result in a wider category that has at least 50 employees. A suggested aggregation is shown in the graphic and the end result is an employer who could submit data for 7 categories in total namely, W-B, W-ITO, Mixed, Asian, B-A, B-CO and Other.
You should be able to see that ONS18 has a higher level of 5 categories (White, Mixed, Asian, Black, Other) which is more commonly known as Big5, and in general, aggregations should take place within the Big5. However, I believe it is up to the employer to decide what works best for them (perhaps after consultation with their employees) and in this example, I defined Asian to consist of A-I, A-P and A-B and separately combined A-C & A-O with O-A & O-O to produce Other. Note in the 2001 census, Chinese was actually in the Other category and was only moved to Asian in 2011. Notice also that I chose to combine W-I & W-T with W-O rather than put them in Other. Another option might have been to combine them with W-B instead to get W-BIT.
Let’s now look at Bath Legal LLP who have 749 employees. Unlike Bristol Makers, they are only able to produce 2 categories, W-BI combining W-B & W-I and then an all other category which I call Not-W-BI. It might have made more sense to have White and Not-White but then the Not-White category would have less than 50 employees.
When I look at Mendip Foods, I end up with only 1 category W-B which has more than 50 employees. It is clear for this employer that having only 407 employees makes it harder to end up with 2 or more categories with at least 50 employees.
This is why I say Parliament has a trade off to make here. If the minimum had been 20 employees instead then both Bath Legal and Mendip Foods would end up with 2 or more categories. But such categories would result in pay gap statistics that would be hugely subject to the vagaries of chance and only a professional statistician would be qualified to interpret such data.
4 – CRED Recommendations 23 & 24 – My Opinions
At this point, I want to review CRED’s recommendations 23 & 24 and give my opinions since this has a bearing on the next section. The background to these can be found in the Data section of the CRED report on pages 49 to 51 and I would summarise the two recommendations as follows :-
- 23 – Use Data in a responsible & informed way – this is a call for ONS and the Race Disparity Unit (RDU) to work together to create common standards for reporting on ethnicity, especially across the public sector and those in receipt of public funds. In addition, a plea is made to the press to exercise more responsibility when reporting ethnicity data.
- 24 – Disaggregate the term BAME – throughout the report, it is clear CRED prefers that analysis is done at a granular level as much as possible e.g. the ONS18. If this is not possible then the Big5 is an acceptable alternative but a White v BAME analysis should be the absolute last resort and avoided as much as possible. They note that a White v not-White analysis will “usually lack analytical value“.
I completely concur with CRED’s recommendation 24 to avoid a binary approach as much as possible. I can relate to this through being registered blind and partially deaf and thus someone who is disabled. When it comes to disability pay gap reporting (which I will be write about in a future blog post), I am sceptical of an abled v disabled comparison since I know my own experience as someone with sensory impairments is very different from a wheelchair user, say. Our issues almost do not overlap though we may have the same issues in terms of unhelpful employers. Therefore it could be the case that an employer is excellent on wheelchair accessibility but rubbish for deaf and blind people. Their lack of a pay gap could simply be the result of being selective as to which disabled people they employ rather than being truly inclusive. A binary approach hides this for disability and I am sure it is the same for ethnicity.
Having said that, I was struck by the extent to which CRED wanted analysis to be conducted at the ONS18 level as much as possible. Indeed, they noted that the ONS added a 19th category Roma for the 2021 census just completed and are recommending 3 new categories (White – West European, White – East European, Black – Somali) going forward. I am sure that as the demographics of the UK change, more categories could be created. For this reason, I endorse recommendation 23 for the creation of common standards for ethnicity categories. A recommendation I would make regarding 23 is that thought should be given to the current intersection of race, heritage, nationality & religion which seems to be how the current ONS18 has come about. I am particularly mindful of those who see themselves as Black British or Asian British only i.e. not Black African say, and the current categories do seem to be forcing such people to put themselves in a box that doesn’t quite describe them. I would also call upon ONS & RDU to take advantage of any EPGR system that is introduced as that will give them access to a large pool of employees and employers who will be telling you how they identify their ethnicity.
5 – How many employers will be able to report more than 2 categories?
Given CRED recommendation 24 to avoid a binary White v BAME approach as much as possible, the natural question to ask is that if EPGR was introduced for all employers with 250 or more employees, how many of the 11,000 employers currently required to report GPGR could submit data for at least 3 or more ethnic categories using the process I laid out above?
Unfortunately, there is little data on ethnicity breakdowns by employer currently available so until EPGR is mandated, we won’t know the answer. However we do have the 2011 Census broken down by ONS18. Is it possible to use this data to estimate how many employers’ will be able to report 3 or more categories?
I believe it is possible and I have created 676 fictional employers using the 2011 Census for England only. You can take a look at these and draw your own conclusions by downloading this spreadsheet Simulated Employers Ethnicity Profiles v1.1. I will be writing a separate article to explain what I have done in more detail but here is a bullet point summary.
Please note there was an error in the formulas in an earlier version v1.0. Please download v1.1 instead if you already have v1.0.
- All fictional employers have between 250 & 4999 employees.
- I have assumed that all of the ~600 employers with 5000 or more employees currently reporting gender pay gaps will be able to report 3 or more ethnic categories.
- In my article “How could ethnicity pay gap reporting be introduced?” written in February 2020, I gave 5 options, the 3rd of which was to restrict EPGR to employers with 5,000 or more employees on this basis.
- I’ve used the 2011 census for England only rather the UK for two reasons.
- Scotland did not collect data for all ONS18 categories in 2011 so their data is incomplete.
- I am assuming that the 2021 census will show a lower proportion of white people in the UK and given England was less white than the other 3 nations in 2011, I thought it might be a predictor of 2021.
- 326 fictional employers were generated using the ethnic breakdown of the Lower Tier Local Authority (LTLA)
- The 3 fictional employers I showed earlier, Bristol Makers Plc, Bath Legal LLP and Mendip Foods were in fact Bristol City Unitary Authority, Bath & North East Somerset Unitary Authority and Mendip District Council.
- For each LTLA, I took the 2011 population, divided it by 1,000 and raised the answer to the power of 1.28 to get the number of employees which was set to be a minimum of 250.
- The distribution of employer size generated using this formula matches very well with the known distribution of employers who come under GPGR regulations.
- The fictional employees were then given ethnicities based on the ONS18 census results for that LTLA.
- I am not claiming that any employer from Bristol say will be the same as Bristol Makers Plc. I am claiming that as a collection, the 326 LTLAs can be a good estimate of the kind of variation we will see among employers with 250-4999 employees both in terms of number of employees and their ethnicity profiles.
- These fictional employers are likely to be representative of highly local employers e.g. a local authority workforce, a university, a hospital, a manufacturer with a single large factory e.g. Nissan, etc where the majority of employees can be expected to come from the LTLA in which the employer is based.
- 350 fictional employers were generated using the ethnic breakdown of the 533 English parliamentary constituencies
- Unlike LTLAs which vary considerably in size, parliamentary constituencies nominally have the same population of 100,000 people.
- 175 employers with 250-499 employees were created by merging 3 neighbouring constituencies together and using the 2011 census to work out the ethnicity profile. For example, Cornwall has 6 seats and a natural division into east Cornwall and west Cornwall could be made and used to create two fictional employers.
- 108 employers with 500-999 employees were created on a regional basis by recruiting each employee in turn by selecting a constituency at random and then selecting an ethnicity at random within the seat based on the 2011 census. For example, 12 employers were generated on this basis using the 55 constituencies in the South West region.
- 67 employers with 1000-4999 employees were created on a super-regional basis by selecting a constituency at random and then selecting an ethnicity at random within the seat based on the 2011 census. For example, 18 employers were generated on this basis using the 158 seats in the North super-region consisting of North West, North East & Yorkshire & Humber. The other 2 super regions were London + South East + East and East Midlands + West Midlands + South West.
- The number of employers based on London seats was boosted to reflect the fact that most headquarters are based in London though of course, the workers in these HQs will often be commuters who live in the East and South East regions.
- These fictional employers are intended to be representative of employers whose employees are more widely dispersed across the country, perhaps because they have multiple premises, workers can work remotely, etc.
- In generating these 676 fictional employers, I have made the following assumptions:-
- The employment rate is the same for all ONS18 ethnic categories. This is a big assumption as it is known that this varies due to age profile, cultural reasons, differences in health/education/etc.
- There are no differences in preferences for industry sectors by ethnicity. By this I mean that Asians do not prefer to work in medicine, Eastern Europeans do not prefer to work as farm workers, Chinese people do not prefer to setup their own restaurant business, etc. Such stereotypes may or may not be true but for my fictional employers I have assumed no stereotypes whatsoever.
- There are no differences in preferences for working for large employers. You might think to yourself there are no shortage of businesses with significant ethnic minority representation e.g. your local Chinese supermarket, Indian restaurant, taxi company. But these will often be small businesses with less than 250 employees whereas EPGR will be with larger employers.
- All employees at these fictional employers will declare their ethnicity i.e. a 100% response rate. This is not going to happen and my experience in other surveys is that between 5% & 40% of employees will fail to declare their ethnicity.
I would be very happy for others to critique this spreadsheet using the links I give at the end of the article. I have made a number of assumptions that do need challenging but overall I think I have created something that can answer the question I’ve posed in this section i.e. how many employers will be able to report 3 or more categories? If anything, I think my collection of fictional employers is likely to overestimate the number of employers who will be able to do this rather than underestimate it.
The graphic below shows what a nationally representative (England only) employer of 1,000 employees would look like if it perfectly matched the ONS 2011 Census. I’ve presented breakdowns by ONS18, Big5 and Binary.
Such a nationally representative employer with 1000 employees would just about be able to report 4 categories if the minimum was 50 for each category. If the minimum was 100 instead, the only option would be a White v BAME which CRED recommend against. If the employer had only 500 employees instead, then even a White v BAME would not be possible with a minimum of 100 and if there were only 250 employees, White v BAME would not be possible with even a minimum of 50.
It is beginning to look like the answer to the question as how many employers with 250-4999 employees will be able to report 3+ categories if the minimum category size is 50 is not going to be that high. The graphic below summarises the outcome for my 676 fictional employers and the traffic light colouring is intentional here. This table can be found in the spreadsheet you can download but I will provide more explanation when I publish the separate blog post explaining my analysis in more detail.
If an employer uses ONS18 and treats White British as the main category, then I estimate 7 out of 10 employers with 250-499 employees will be unable to create a 2nd category with at least 50 employees. In other words they will be unable to report any ethnicity pay gap data that could be considered semi-reliable. If the employer wanted to use White as the main category then this figure rises to 8 out of 10 employers with 250-499 employees. The bottom line is that EPGR is a complete non-starter for employers with less than 500 employers and this backs up the 1st recommendation I made in my submission to the CRED report in November last year.
When we look at employers with 500-999 employees, the proportion who will be unable to report any category than White-British falls to 2 in 10. A further 2 in 10 will only be able to report White British v non-White British. Note this is not White v BAME because non-White-British also includes White-Other. What is especially notable here is that although 6 in 10 employers in this size class will be able to report 3 or more categories, over half of such employers are likely to find themselves reporting White-British, White-Other and Not-White only.
I must confess until I did this analysis I had not realised that White-Other is the 2nd largest ethnic category of ONS18 after White-British. When people talk about Ethnicity Pay Gap reporting, I am sure people’s mental image is that of comparing White v Black and I suspect some people will struggle with the concept of EPGR being about a comparison between two White categories. This could become an even larger effect with the news published last month that over 5 million EU citizens applied for settled status in the UK following the UK’s departure from the EU. If all of these are applicants are white (unlikely) and none hold British nationality that basically doubles the White-Other category compared to 2011.
If one instead uses Big5 and make White the default category (thus combining W-B & W-O), then of employers with 500-999 employees, then roughly, 1 in 3 will only be able to report data for White employees, another third will only be able to do White v BAME leaving only a final third able to report 3 or more categories of which two will be non-white. Whilst my process for generating fictional employers needs to be challenged I doubt my estimates are that wrong and the end result is that the majority of employers in this size class will be unable to report useful data under EPGR.
I am therefore left with the conclusion that only employers with 1,000 or more employees can be included in a mandatory EPGR. That means only about a quarter of employers that currently report gender pay gaps would have to report ethnicity pay gaps.
6 – CRED Recommendations 9 & 10 – My Opinions
Pages 115 and 116 of the CRED report is where you can find the background to recommendation 9. I have reproduced the full recommendation here.
“RECOMMENDATION 9: Promote fairness – Investigate what causes existing ethnic pay disparities
The Commission recommends that all employers that choose to publish their ethnicity pay figures should also publish a diagnosis and action plan to lay out the reasons for and the strategy to improve any disparities. Reported ethnicity pay data should also be disaggregated by different ethnicities to provide the best information possible to facilitate change. Account should also be taken of small sample sizes in particular regions and smaller organisations.
To support employers undertaking this exercise, the Commission recommends that the Department for Business, Energy and Industrial Strategy (BEIS) is tasked with producing guidance for employers to draw on.”
My opinions can be listed as bullet points.
- On page 115, CRED state they “… are aware of the pitfalls around the execution of ethnicity pay reporting, but feel that this work needs to start somewhere”. What CRED say here completely echoes my feelings i.e. let’s make a start somewhere.
- The extensive comments about sample size look like they’ve been copied and pasted from my 1st two blogs on EPGR here and here. These two blogs were included in the evidence I submitted to CRED so I have no objection to this but I confess I would have liked CRED to have acknowledged me as the source.
- CRED recommend “… all employers that choose to publish their ethnicity pay figures should also publish a diagnosis and action plan to lay out the reasons for and the strategy to improve any disparities”. I commend CRED for explicitly linking Measurement of pay gaps with Analysis of why they exist and identifying actions to Improve things. The capitalisation of M A I I’ve done there is because of DMAIC, an acronym used by many who work in fields of Continuous Improvement (Six Sigma being the most notable). I explained in this blog why DMAIC should underpin efforts to close pay gaps & the part that statisticians can play in this. Just measuring pay gaps for the sake of measurement is a waste of time.
- I completely agree with the recommendation that the government should publish guidance. However, the Government Equalities Office (GEO) is the body tasked with GPGR guidance today and I think they are the better option to produce guidance for EPGR rather than BEIS. I think this needs to be produced as soon as possible and obviously I would be happy to contribute to such an endeavour.
So in short I completely agree with everything in recommendation 9. For some reason some commentators thought CRED recommended against ethnicity pay gap reporting but assuming they have actually read the report then I struggle to find anything in this recommendation that could be objectionable.
It is of course true that CRED did not recommend that EPGR be made mandatory and that voluntary reporting should continue for now. As I demonstrated in the previous section, the minimum sample size per category issue is something that is incredibly difficult to get around. I’ve also demonstrated something that has not been talked at all in that a lot of employers may end doing a white-British v white-Other comparison which is not what people probably had in mind when they consider EPGR. So whilst I did originally call for mandatory EPGR for employers with more than 500 employees, I now have to modify that recommendation to only apply to employers with more than 1,000 employees.
This then raises the question as to whether EPGR should be voluntary rather than mandatory. CRED are entirely within their rights to say EPGR should remain voluntary for now given what I’ve talked about this in this article. Increasingly, it feels like the choice is between a voluntary system with a minority of employers reporting and doing a proper analysis because they want to and a mandatory system where the majority of employers will be unable to report meaningful numbers and end up confused. My opinion is that it is still worth trying out a mandatory system if nothing else to see what we learn but I have no problem with the alternative viewpoint like CRED’s that it should continue to be voluntary until we have a better idea of how to deal with the issues. I should declare at this point that I stand to benefit commercially if a mandatory EPGR system is introduced.
With this question in mind, I must confess I found recommendation 10 of the CRED report on pages 116 to 119 to be somewhat unhelpful. The recommendation is headed “Improve understanding of the ethnicity pay gap in NHS England” and basically calls for NHS England to undertake a detailed ethnicity pay gap analysis. I don’t have anything against this recommendation in itself but I find myself questioning why single out the NHS? It is the largest employer in the UK and will obviously be able to undertake a detailed analysis but they are on a completely different scale to the majority of employers who have less than 500 employees.
The reason why I say this recommendation is unhelpful is that when I point out the difficulties with EPGR to others, I often get an answer along the lines of “well, so-and-so has done an ethnicity pay gap report so why not all employers?” and the danger with the NHS being highlighted specifically is that it encourages more comments on those lines. What such people fail to realise until I point it out is that the employer they’ve chosen usually has more than 5,000 employees which as I have highlighted already is a threshold where any employer with that number of employees should have no problem doing EPGR. But such employers are a tiny fraction of the total number of employers and cannot act as an example to much smaller employers.
7 – My Conclusions
Thank you for taking the time to read this longer than normal blog post! I will again use a bullet point summary to pull together the 12 key points I would like you to take away.
- From the point of view of employee anonymity, the minimum number of employees per any category used in Ethnicity pay gap reporting should be at least 20 but I would like the Information Commissioner to comment on this.
- From the point of view of statistical reliability, the minimum category size needs to be 100 employees which confirms the recommendation made by the Royal Statistical Society in 2019.
- From the point of view of the public interest, I will accept Parliament setting the minimum at 50 employees but no lower unless parliament requires employers to take advice from a professional statistician before publishing their data.
- I endorse CRED’s recommendation 23 asking the ONS and RDU to define common standards for reporting on ethnicity across the public domain.
- Employers should ask their employees to say what kind of categorisation they would like to use, select from a commonly agreed list of ethnic categories the category they most identify with but employees must be free not to answer this question. (Note this bullet was slightly different in the draft version of this article)
- Employers will then be free to aggregate the categories in whatever way makes most sense to them until they arrive at a set of ethnic categories each with at least 50 employees or some other minimum as determined by Parliament.
- I endorse CRED’s recommendation 24 which calls for data analysis to avoid as much as possible a binary White v BAME comparison and to use as many categories as possible.
- Using the 2011 Census for England to create 676 fictional (but hopefully realistic) employers, my analysis shows that employers with less than 1000 employees are unlikely to be able to meet the requirements of points 6 & 7 above.
- Employers with more than 1,000 employees are more likely to be able to undertake a meaningful ethnicity pay gap analysis and therefore if Parliament wishes to introduce a mandatory EPGR system, then this should only apply to employers with more than 1,000 employees. Currently only a quarter of employers reporting gender pay gaps would have to report ethnicity pay gaps as well.
- An issue that I think most people are not aware of is that a large number of employers who are able to report 3 ethnic categories will find themselves reporting data for just White-British, White-Other and not-White. The fact that White-Other is the 2nd largest ethnic category of the 2011 census and could conceivably double in size for the 2021 census is something I think people are not aware of and I think it would be ironic if ethnicity pay gap reporting ended up being mostly a White v White comparison at a lot of employers.
- I endorse CRED’s recommendation 9 with respect to EPGR. In particular, I commend their call for what amounts to a continuous improvement approach using DMAIC used in many industries and their call for government guidance to be published.
- Given the difficulties with sample sizes that I have highlighted in previous blogs (and were copied and pasted by CRED in their report!) have turned out to be more problematic as described in this article, I do understand why a voluntary approach may be preferable so as to learn how to overcome the issues that are specific to EPGR. I would still like to see a mandatory system at some point for employers with at least 1,000 employees though.
— Would you like to comment on this article? —-
— 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 & diversity? —
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 gaps & disparities? —
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 disparities in pay, promotion, recruitment, etc. 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 representation & pay gaps.
- Expert Witness – Has your pay gap & diversity 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.