How (And When) Should We Ask About Ethnicity? Use These Principles for More Inclusive Demographic-Gathering

How (and when) should researchers ask about ethnicity? As importantly: How do people want their ethnicity asked about?

These are questions that deserve more attention, more care, and more consideration than they’re often given.

At dscout, we wanted to design more inclusive participant profiles—rethinking the section of our platform where “scouts” detail mandatory demographic information.

In 2019, we worked to restructure how our users reported their gender-identity (and came away with key design principles for gender-identity inclusion). Recently, we realized there were critical flaws in the way we (and many products, tools, and surveys) asked users to indicate their race and ethnicity.

When a form asks participants to identify their race and ethnicity, the options rarely allow a user to select multiple options. More commonly, they’re provided a list of potential selections, and the isolating catch-all, category of “other” (at dscout, we instead asked participants to self-identify). This forces users to “choose” which identity best represents them in that moment, excluding individuals who identify with multiple races.

We wanted to do better. And in the process of learning how, we learned a lot more broadly about how products, services, tools, and surveys can more ethically and accurately ask about race and ethnicity.

Here’s a summary of our findings—the whats, hows, whys, and “so what’s” that inform the way we ask about identity moving forward.

Jump to:

• Study design: How do participants want to report their racial and ethnic identity?
• What we built with this information
• What we learned (and why this work is important)

Study design: How do participants want to report their racial and ethnic identity?

We had a hunch that we weren’t the only folks questioning how we were asking about race and representing race in our data. We first wanted to do secondary research to learn from other research practitioners in the space. And we wanted to talk directly to our participants and ensure that we weren’t making assumptions on their behalf.

Secondary research:

We first turned to research communities for pre-existing perspectives—noting how other research tools handle demographic questions.

While there are many different versions of the race/ethnicity question inside and outside the research industry, there is no norm. Even the US Census is influx and has different proposals for how to ask about race/ethnicity.

Overall, a big finding from our secondary research is that there wasn’t much out there. We were surprised and a little disappointed to not see more in the space.

Still a few learnings did emerge, at least in considering how race and ethnicity data should be collected and used:

• Use a race and ethnicity question only if you'll be doing something with that data.
• If you are using a race and ethnicity question, inform your participants why you're collecting this data, or how it will be used.
• Having an "other" category, might be, well, othering. Allowing participants who don’t fit into any of the existing categories a chance to answer in a way that is still accurate to them is critical.

This informed the primary research undergone with both researchers and participants.

Primary research:

Interviews with researchers

First, we ran a series of in-depth interviews with researchers—learning about how they talk, ask, and think about race as part of their research practice. We tried to better understand the context both in the US and internationally, and to get a better understanding of why and how racial data was used and solicited.

Some key findings here:

• It was important to researchers that participants can accurately self-identify with however many races they identify as
• Generally, researchers collect participant demographics to ensure they have representative samples (rather than analyzing along demographic lines)
• At the same time, researchers want to have a clear way of representing the demographic data from their sample (ie. the numbers in their sample need to add up in a way that makes sense)

Feedback from participants

We then ran a study on Express (our media-rich survey tool) with scouts who identify with more than one race. We asked a series of open and close ended questions to learn what language felt most appropriate to them (i.e. is Multi-racial the right identity label?), what options they felt were limiting and more representative when it came to identifying themselves, and generally how they hoped researchers were understanding and utilizing their identity data.

Some key findings here:

1. For participants who identify with more than one race, "multi-racial" is the right term to use. A slight majority of scouts would be prefer to identify with their individual races as opposed to "multi-racial."

2. Being able to select as many race categories as you identify with (as opposed to a single select of “multi-racial”) is critically important. We had no business doing any work on our demographic questions if we made a different choice. Here, flexibility was critically important for many participants. Specifically, they didn’t want to feel pigeon-holed into an option that doesn’t apply to them.

3. Participants want to be represented in any of their race categories (i.e. if Jane Doe identifies as “asian” and “white” she wants a researcher to see her amongst “white” scouts and amongst “asian” scouts). That may sound like a no-brainer, but we wanted to understand if perhaps Jane Doe would want to be viewed in a unique category of “White and asian” with only other participants who identify as “white and asian” specifically.

Generally though, representation emerged as important to participants—they wanted their identity to be reported in a way that felt comfortable to them, but still made sense for researchers seeking to understand specific identity subsections.

4. Participants are comfortable with researchers, and research tools, collecting and sharing their demographic data.

Evaluative research

We used that initial data to create new designs for our participant profiles. From there, we conducted evaluative research with actual designs to see how certain theoretical situations would play out when talking about a participant pool (i.e. How summary charts of participant demographics would look, how filters based on demographic data within the platform would work, etc…)

So what? Here’s what we built.

Before the new release, between 2-5% of our participant pool self-identified their race and identity. Many of these self-identifications are to show the multiple ethnicities that are a part of their identity.

We are now moving to a multi-select model, and providing more context as to why the data is collected, and how it will be used. Then, participant profiles will feature all selected identities.

This will have a positive impact on participant’s ability to be accurately recruited for studies where researchers dictate a particular sample demographic balance.

We also noted that it was important to note this change to researchers, to better explain any perceived discrepancies in their data (ie. percentages that add up to over 100%).

What we learned (and why this work is important)

At dscout, we had a company-wide OKR (Objective Key Result) to focus on internal improvements to diversity, equity, and inclusion. Creating more accurate and representative scout profiles was critical to this work.

Some thoughts from our research team...

Taylor Klassman, UX Research Lead: "We were determined to build the best thing we could for our participants, and we knew that meant building something really hard. We didn’t cut corners from a technical perspective, even though that meant this project took us much longer than anticipated and scoped for.

Something we were asking ourselves was: "Why are we even asking scouts to identify along racial/ethnic terms?" We made a pretty intentional decision that by having scouts self-identify (instead of having researchers make assumptions based on photos) we would actually promote more equitable processes. I feel like representation is about not being colorblind.

We also learned a lot as a team about our edges and comfort about the very human-based work we do everyday—despite it being housed in the 1s and 0s of technology. I felt ill-equipped to be making certain judgement calls, but our working team was so supportive and thoughtful that we figured out this complex feature together. We needed to figure out how to best represent the complexity of human identity in a pretty rigid system—and I think, I hope, we did our best.

We’re excited to get feedback from our scouts and researchers on how we did and if there is anything else we can do to make the people behind our tool feel heard and loved and seen."

Tymmarah Anderson, Research Advisor: "Additionally, I think the lack of research and examples out there felt a bit daunting—which contributed to the feeling of inadequacy when making these decisions. Ultimately though, we had to get comfortable with the idea that we may be looking at something relatively new to this space. And just because it hasn’t happened yet doesn’t mean it shouldn’t be done.

It’s also possible we may need to iterate on this feature in the future, especially when we begin to think about our international scout pool. What’s right for now could change and we need to be able to adapt."