Johnnatan Messias at the Federal University of Minas Gerais in Brazil and a few pals began the study began by filtering the Twitter stream during the three months to September 2016. In total, they collected 341,457,982 tweets posted by 50,270,310 users.
They filtered this by time zone, geolocation, and those with a profile picture to leave them with 1.6 million users based in the U.S. They then fed the pictures through a state-of-the-art face recognition application called Face++, which reports the sex and race (black, white, or Asian) of each user. The overall makeup of the group revealed that 53 percent were women and 47 percent men. It also revealed the race breakdown, which was 18 percent Asian, 14 percent black, and 68 percent white. Messias and co began by looking at the ratios of men and women who had the most followers on Twitter. Of the top 1 percent of Twitter users with the most followers, 57 percent were male and 43 percent female. Researchers say inequalities surfaced when the group studied the distribution of races among the most popular Twitter users. “At the highest levels of [Twitter] visibility, users perceived to be White come out on top position.”
The most privileged group turns out to be white males, who are overrepresented by 20 percent among popular Twitter users. White females are also more privileged albeit to much lesser extent, just 3 percent. The most underprivileged groups are Asian females and black females, who are underrepresented by 31 percent.
Moreover, The study reveals that there are significant biases in the way genders and races link to each other. This is an effect known as “homophily”—the tendency of people to seek out others like themselves. White people tend to follow more white people than expected by a margin of 16 percent. Black people tend to follow more black people than expected by a very significant margin of over 200 percent. However, Asian people tend to follow fewer Asian people than expected by a margin of 10 percent.