Skip to main content

Diversity Explorer 2.0

Opinions on diversity are diverse. At ODILeeds we’ve been working with the public and with our sponsors to listen to as many as possible. Our sponsor Yorkshire Water, headquartered in one of the UK’s most diverse cities and serving one of its less ethnically diverse regions, has been particularly important in leading this discussion.

Recently, like many others, we’ve taken more time to listen to debates about diversity beyond the UK. Engaging with discussions in countries such as France and the USA with very different laws and cultural norms has reminded us that data cannot help with many problems in diversity.

In some cases this is because the data that we need cannot exist. The US constitution forbids its census from asking the religion of respondents. The French constitution goes further and forbids all distinction of people by religion or ethnicity by the state.

Examples of the differences in culture can be seen in publications such as France’s annual report on attitudes towards racism, anti-semitism, and xenophobia where public and learned attitudes towards race and religion are substantially different to in the UK. And in the USA we see common societal groupings such as "hispanic" and "indigenous" that are hard to understand in the UK or have no obvious local equivalent.

Beyond these and other challenges we think that there are areas where data can help people.

Our latest update to our Diversity Explorer responds to a request that we have heard repeatedly. People and organisations want us to help them understand the diversity of the society that they operate in and serve so that they can judge whether they reflect and represent it.

What does society look like?

Today’s update to our Diversity Explorer does this. A user selects a region, selects an age range, and selects the number of employees in their organisation.

The tool then does a thousand simulations to calculate the expected number of employees in each ethnic group. The number is returned as a range representing the number of employees we would expect in each ethnic group if hiring was random.

89% of Yorkshire aged 36 to 66 is in one of the White ethnic groups in the UK.
A company in Yorkshire with employees aged 20 to 70 would be expected to have between five and seventeen BAME employees if staff were hired completely without bias.

The tool raises many questions and hopefully shows how data can be used to inform discussion about them.

One such question is whether an organisation should aim to represent the society it serves or the society that it operates within?

This is a common question when analysing the UK government’s excellent civil service statistics.

The senior civil service increasingly reflects the UK’s ethnic diversity at the age range of its employees. The UK civil service reflects the society it serves well. But given that the majority of senior civil servants live and work in London it is a poor reflection of the much more diverse place where it operates.

We hope that our tool can also inform ongoing discussions about additional monitoring of inclusion in the UK economy. Particularly we hope to explain how statistics will help with two live discussions in the UK. Specifically,

  1. Whether the UK government’s gender pay gap dataset should be extended to cover the pay gaps by ethnicity.
  2. To what extent the split between White and other ethnicities (variously described as ethnic minorities, BME, and BAME in the UK) undervalues diversity within those groupings and what advantages and disadvantages there are to more detailed groupings.

Statistics and diversity

The key consideration and the most important feature of our tool is that it reports a range of expected employees in each group rather than a precise estimate.

Reporting a range rather than a number is important. We cannot assume that a company in Yorkshire with ten employees, all of whom are white, is unrepresentative of Yorkshire’s ethnic diversity. We would probably expect only nine employees to be white, but all would be white more than 5% of the time even if hiring had no bias at all.

When we start dealing with small numbers, whether numbers of employees in an organisation, or percentages of the population of a given ethnicity we have to start dealing with problems like this.

My instinct is to support the publication of ethnicity pay gap data in the UK. The gender pay gap data has generated significant unexpected value alongside creating some unexpected problems. But it is important that discussions about how detailed that data should be, both at the level of different ethnic groups and at the level of company size and location, are mindful of data’s limits.

Try Diversity Explorer 2.0 now. Let us know what you think. If enough people like it, we'll transfer it to ODILeeds' site.