Twitter on Friday published data that appeared to show its workforce was more diverse in 2017, but the way Twitter currently measures diversity makes it hard to tell how much progress the company is actually making.
According to Twitter’s diversity data, “underrepresented minorities” — which Twitter defined as non-white and non-Asian — now make up 12.5 percent of the company’s total workforce, up from 11 percent in 2016.
But that 12.5 percent includes people who specifically declined to identify their ethnicity on Twitter’s internal survey, according to a company spokesperson. That unidentified group makes up 2.9 percent of Twitter’s workforce and is being classified as “underrepresented minorities” even though this contingent specifically declined to be classified. There is no way to tell what ethnicity they actually are.
If you subtract that group from Twitter’s list of “underrepresented minorities,” it shows non-white and non-Asian employees at Twitter made up just 9.6 percent of the company workforce, a decline from 11 percent in 2016.
So while Twitter said its ranks were more diverse last year, that may not actually be true.
After Recode approached Twitter with our findings, the company took down the chart that showed “underrepresented minorities” made up 12.5 percent of its workforce and issued the following statement.
As has been done in years past, the chart with the URM data included all employees who did not self identify as white or asian. This includes employees who declined to self-identify in our systems. The race/ethnicity breakdown is a more granular and transparent one and more reflective of the data we’ll report moving forward. To eliminate any confusion and in the interest of transparency, we’ve removed the URM chart from our diversity report.
Update: A Twitter spokesperson sent Recode a longer statement Friday afternoon reiterating the company’s approach to measuring diversity. Here’s part of that updated statement.
“We released our diversity report this morning which included a graph that caused some confusion. We want to apologize and clarify.
“The race/ethnicity breakdown in the report is more accurate and granular than URM which we had already decided to move away from. Our new goals put specific focus on increasing overall women, black, and latinx representation at Twitter, and we’re holding ourselves accountable.”
Twitter also countered that the data on its 2016 report was rounded, so it’s not as precise. Twitter adds that its 2016 survey also allowed employees to decline to identify their ethnicity and were consequently lumped into its “underrepresented minorities” group, which Twitter tabulated as accounting for 11 percent of staff in 2016.
But if you look at Twitter’s breakdown of individual ethnic groups in that year, it’s clear the group of employees who declined to identify their race or ethnicity was less than 1 percent in 2016, a much smaller group than 2017.
Even if we assume that group was a full 1 percent of Twitter employees, which it was not, underrepresented minorities made up a minimum of 10 percent of Twitter’s workforce in 2016. That’s still higher than the 9.6 percent of employees who fell into that category in 2017.
It’s a small decline, but no matter how you slice it, Twitter is not as diverse as it was claiming to be this morning.
The company reported 3,372 full-time employees as of December, according to regulatory filings, but that figure includes its global workforce. The internal diversity survey only looked at underrepresented minorities that are part of Twitter’s U.S. workforce.
Silicon Valley companies have struggled with diversity despite vowing to do better. The rate of improvement has been slow, and tech education among minority students lags better-funded schools in affluent districts, which are still mostly white. As a driving force of the economy, tech could end up creating a deeper racial and class divide.
In the case of Twitter, its latest improvements are obscured. For example, unlike in 2016, the portions of the company’s circle charts do not add up to 100 percent. A spokesperson says those charts do not include the percentage of employees who either specifically checked the box declining to share their ethnicity or didn’t take the survey at all.
That means roughly 18 percent of Twitter’s employees didn’t take the survey or self-report their race or ethnicity. We have no way of knowing what percentage of Twitter employees fell into this same group in 2016, because Twitter’s circle chart did not include them or leave a space for them.
Overall, it’s a bad look. Twitter is clearly trying to diversify its workforce and touted internal groups and programs for minority employees, as well as efforts to reach underrepresented minorities still in college as ways it’s trying to diversify. But having consistent, transparent numbers should be part of that process.
Twitter says its process for sharing this data is supposed to change. The company’s new set of diversity goals focused only on hiring three specific groups: Women, black and Latinx employees. That should, theoretically, eliminate the odd “underrepresented minorities” measurement that Twitter is now reporting.
Here’s a look at where Twitter wants to be in two years:
This article originally appeared on Recode.net.