Measuring Diversity: Notes from the Field

Interest in measuring diversity, particularly workforce diversity, has grown exponentially in recent years. While numerical diversity doesn’t tell the whole story, nor is it a sufficient condition for inclusion and equity, it can still be a meaningful exercise. 

One common reason organizations have for studying numerical diversity is to address this question: Is our workforce representative of the community we serve? To date, few organizations regularly collect employee diversity data and those that do, generally limit it to age and sex. The gap in data is often filled by a workplace survey. In turn, organizational diversity data can then be compared to population level data, such as that available through the census or labour market figures.

However, in doing this work, we have encountered a number of challenges and places where professional judgements needed to be made. Here are a few we think are worth mentioning:

• Sample Size: Getting people to respond to surveys is a challenge, moreso when surveys ask questions about personal diversity. Individuals from diverse backgrounds may not take part because they fear the consequences of disclosure. Others who are resistant to DEI efforts may also be unwillingly to complete a survey with DEI aims. In our experience, expect a 50% response rate. All the incentives and communications in the world will not make up for fear of disclosure or resistance to DEI work, especially if trust is lacking in the workplace. 

• Comparisons: Comparing survey diversity to population diversity may seem like a logical way to answer the question of representativeness. However, without a 100% response rate, you have a non-representative sample making this type of comparison highly problematic. For example:
  • You have 100 employees and 50 respond to a survey. Black employees are very motivated to take part and all 10 of these employees complete the survey. This could make it appear that 20% of employees are Black (10/50) when in fact, the actual proportion is 10% (10/100). In this case, a comparison to the local population using the 20% figure would be inaccurate and would over-represent the proportion of Black employees in the organization.

For this reason, great caution needs to be observed when reporting on numerical diversity and making comparisons to population diversity. If reporting is necessary, it needs to be clear that what is being reported is the diversity of the sample which may not correspond to the actual relative diversity in the organization. Similarly, comparing organizational diversity to population diversity with less than a 100% sample can lend itself to inaccuracy and, worse yet, erroneous claims about representativeness. We have actively discouraged clients from publishing organizational diversity data for those very reasons.

So, does that mean measures of organizational diversity are useless? Not at all! While you may not learn everything about workplace diversity you can definitely learn something about diversity in the organization. This data can also provide a baseline for ongoing measures of workplace diversity with the associated aim of building trust and response rates over time. It is also provides a lens for understanding the organization from the perspective of different groups and a jumping off point for more targeted learning.