Diving into the EPSRC’s “detailed ethnicity data”

[This post was written by Francesca Firth, Michael Sulu, Carla Figueira de Morisson Faria & Rachel Oliver]

EPSRC have recently published the most detailed data they have ever released on the ethnicities of their grant applicants and awardees. They’ve also pledged action to increase racial equity in grant funding including two broad-ranging consultations, one with senior management at universities, and another with the general physical science community. They’ve launched a “‘Have your say’ community survey” and are seeking views from engineering and physical sciences academic researchers, postdoctoral researchers and doctoral students, with the goal of understanding the lived experience of researchers who are marginalised because of their race or ethnicity.

We’re excited that they are looking beyond the bald statistics in the funding data, and hope that the consultation will lead to them taking real leadership to change policies, practices and behaviours in the sector. We want to encourage readers of this blog to fill in the survey and so we’re going to have a dip into some of the data here, to draw out some of the troubling findings and hopefully motivate people to give their views. The greater the diversity of the people responding to the survey, the better EPSRC will understand the range of issues facing minoritised researchers. White researchers have an opportunity to act as allies, by responding to the survey, giving an honest account of their own experiences, highlighting the privilege they experience within the funding system.

So, what is different about the new “ethnicity data”? Well, previous data releases basically only compared “white” and “ethnic minority” researchers. (We’re aware that catchall terms like “ethnic minority” can be unhelpful. Here and elsewhere we’re using the terms used in the data releases for accuracy and clarity). The new data disaggregates “ethnic minority” into, “Black”, “Asian, excluding Chinese”, “Chinese” and “Mixed” and then further disaggregates within those broad groupings. This allows us to see, to some extent, differences in representation and access to funding between different identities. We’re going to look at three aspects of the data:

1. How does academia compare to the general UK labour market in terms of the representation of people from different ethnic groups?

2. How does the rate of grant applications vary with the ethnicity of the applicant?

3. How does the rate of grant awards vary with the ethnicity of the applicant?

How does academia compare to the general UK labour market in terms of the representation of people from different ethnic groups?

Here, we’re reproducing some data straight from the EPSRC’s report.

This aggregated data suggests that Black researchers are hugely under-represented at all levels in the physical sciences. There are so few Black Principal Investigators (PIs - the people who lead and supervise research projects) and Black research fellows, that the data for both is “suppressed”, since there are less than 5 of each in the entire EPSRC cohort. Whilst the “suppression” of the data is intended to protect the privacy of people in groups where there are very small numbers, it leaves us questioning just how few Black PIs there actually are running EPSRC projects currently. If “<5” actually means zero, we think the community should be aware of this. Here, the small numbers are the story, and although they are suppressed they cannot be ignored. The so-called “leaky pipeline” for Black scientists sees the representation of Black people drop sharply through academia’s career stages, which may in some sense “explain” the data, but leaves us asking why UK Universities are hemorrhaging Black talent out of their systems.

Other ethnicities appear to have much better representation, but the data for those who are “Asian (excluding Chinese)” are broken down further and show that there are other areas of under-representation:

Specifically, researchers who identify as Pakistani or Bangladeshi are also notably under-represented, and again the number of PIs in the EPSRC cohort is so low that the data is suppressed.

You’ll have noticed that in graphs above, there is no data for white researchers. This makes us feel uncomfortable. Ignoring whiteness in data sets like this risks ignoring white privilege, which is at the core of much of the under-representation of Black and other ethnic identities which we’re discussing. It also risks ignoring the under-representation of specific white groups such as the traveller community. This tactic strikes us as obfuscatory: it removes a key element of the data set and makes it difficult to make meaningful comparisons between trends in the experience of white researchers and those of other ethnicities. In the next section we’ve analysed the data a bit differently compared to the EPSRC report, to help draw out the over-representation of white people in grant applicants and awardees.

How does the rate of grant applications vary with the ethnicity of the applicant?

Here, we have taken the data provided by the EPSRC on the proportion of grant applicants in different “ethnic minority” groups and the “white group”, and have normalised it by dividing by the proportion of the engineering and physical sciences workforce who belong to the relevant group (using the data EPSRC took from HESA). The resulting numbers compare the representation of different ethnic groups in the grant applicant pool, with values above 1 suggesting that the particular group is over-represented amongst grant applicants, and values below 1 suggesting under-representation. We note that we don’t compare here to the number of people who are eligible to be Principal Investigators (PIs), for two reasons: (a) We don’t have any data on eligibility to be a PI, only on those are eligible and who have actually made an application, and (b) because there are so few Black, Bangladeshi and Pakastani PIs that the relevant data are suppressed (i.e. <5 in each group), we wouldn’t be able to include those folks in an analysis based on the EPSRC PI population.

These data address PIs, Co-Investigators (CIs), and Fellowship applicants. There’s a lot there, but we can pick out some trends. Firstly for PIs, CIs and Fellowships, the white group is consistently over-represented. The normalised application rate is almost always above 1. On the other hand, as PIs, the Black ethnic groups (Black African, Black Carribean) are consistently under-represented among applicants. There’s no data in the CI or Fellowship graphs for Black applicants, because the application rate is so low for Black CIs and Fellows that the data are suppressed. We think it’s safe to assume this is indicative of a severe under-representation of Black researchers amongst the applicants for these types of funding. There are also other missing data points: for example application numbers from Bangladeshi and Pakistani scientists for fellowships are again so low as to have been suppressed in all years.

What are these data telling us? Low application rates may in part reflect the low numbers of Black scientists at senior levels in academia. Only 155 out of more than 23,000 university professors in the UK are Black - a statistic which we should all be ashamed of. However, you don’t need to be a Professor to apply for a Fellowship. The majority of Fellowships are early career awards, some of which can be applied for by PhD students towards the end of their studies. Although Black researchers are under-represented at all levels, this is insufficient to explain the very low applicant numbers. The lack of Black applicants, and also applicants from several other minoritised ethnic groups, suggests that these minoritised researchers experience barriers which prevent them from making applications. Lia Li, Hope Bretscher, Rachel Oliver and Erinma Ochu have written about some of these barriers for Science in Parliament journal. We hope that the EPSRC’s current survey will help the research council to understand further the lived experiences of researchers which prevent innovative minoritised scientists from accessing funding.

How does the rate of grant awards vary with the ethnicity of the applicant?

In the graphs below, we have normalised the data for those who actually get awarded a grant to the representation of the different “ethnic groups” in engineering and the physical sciences, just like we did for the applicant data. So again, values of over 1 represent an ethnic group being over-represented, and a value below 1 represents under-representation. Once again, if data for a particular group is missing from the chart, that implies there are so few grants awarded to people from that group that the data have been suppressed.

The trend we saw for applicants - consistent over-representation of white researchers - continues here. There is no data on any of these graphs for ANY Black ethnic identity, because numbers of grant recipients are so low in ALL categories and in ALL years that the data are suppressed. That’s a shocking result. Given low award rates (the probability that a grant proposal actually leads to a grant being awarded) overall for EPSRC grant applications in recent years, it’s difficult to comment from these data on the extent to which this simply reflects low applicant numbers, or whether it suggests bias in the peer review process. For researchers from the “Chinese” ethnic group, however, we see something quite different in the PI data. Applicants identifying with this group are well-represented in the EPSRC applicant pool as PIs, but under-represented in the grant awardee pool. This means that the award rates experienced by “Chinese” applicants must be much lower than those experienced by white PIs, and that’s borne out if we plot the data simply in terms of award rates:

In fact, except for rare occasions, “white” researchers have a higher award rate than those from any other “ethnic group”. (Remember, award rate here is the probability of a proposal resulting in the award of a grant, so the higher award rate is not related to the higher overall number of white researchers). This must lead us to question the fairness of the (non-anonymised) peer review and panel process, particularly because EPSRC note that in the EPSRC Peer review college, the “proportion of Ethnic Minorities” has been lower than the “HESA EPS Ethnic Minority Population” for all years from 2014-2020. It is particularly notable that the “award rate” for Fellowships for all non-white ethnic groups appears to be zero for all years from 2014-2018. Compared to the grant application process, the Fellowship application process has a greater focus on the applicant rather than the project they aim to pursue, and also includes an interview. Hence, the very low award rates to “ethnic minorities” here are particularly troubling, since they stem from a process which is likely to leave more room for bias against specific protected characteristics.

The EPSRC’s report indicates that racially minoritised scientists are under-represented on the panels which allocate these fellowships, compounding the problem. We encourage scientists from minoritised groups to apply to join the EPSRC Peer Review College to help address this imbalance, but acknowledge the existence of systemic problems with peer review which may not be overcome simply by better representation.

Before we stop talking about the data, we need to acknowledge that although they are much more detailed than anything previously published, they are still incomplete. We know that women are also under-represented in the EPSRC applicant and awardee cohort, particularly at large grant values. We also know that disabled researchers are under-represented. What these data do not allow us to address is how being - for example - both Black and a woman affects access to grant funding. There’s no data at all on the impact of being LGBTQ+ on grant applications or awards (as far as we know EPSRC does not collect data on this question), so analysing the intersection of sexuality and ethnicity is a long way off. We are concerned that the most marginalised people in our community remain invisible in the data set. This is not just a problem with the EPSRC data, but typical of the wider culture in our research community where initiatives such as Athena Swan and the Race Equality Charter address separate aspects of diversity independently, and hence fail to tackle the challenges of intersectionality.

What now?

These data are not a pleasant read, and will be particularly distressing to colleagues from marginalised ethnicities. The data in and of themselves have no intrinsic worth. They only have value as a call to action, a warning klaxon alerting us to systemic racism in the engineering and the physical sciences, and to the privilege experienced by white researchers. The data suggest that minoritised researchers face both barriers to application and bias in the application process. Both problems need to be addressed. Encouraging more applications from minoritised researchers is only a waste of applicants’ time and energy if the panels who judge those applications do so through a biased lens. Working to eradicate panel bias will have limited impact if institutional gatekeeping and other systemic racist barriers mean that applications from minoritised scientists never reach the panel.

In a previous blog, Lia Li, Hope Bretscher, Rachel Oliver and Erinma Ochu made recommendations for achieving racial justice in STEMM funding, which included “recognising the value of qualitative research and lived experience within evidence about racial injustices”. We hope that is what EPSRC will do with their survey on these issues, and we would encourage everyone, whatever their ethnicity, to engage with this. There are different surveys for different career stages: PhD students, post-doctoral researchers and academics. It’s important that white people also engage with this and show empathy towards racially minoritised colleagues in so doing: this will help persuade the EPSRC of broad community support for action towards racial justice, and hopefully allow them to take action rather than just gathering more and more distressing data.

If you are a writer on Medium and you are a member of the Tigers, just let us know and we can add you as a writer for the Medium version of this blog*.

*blog posts submitted by members for this channel must adhere to our code of conduct.

You can find us most easily via @tigerinstemm, or our website, or via email.