How Black Lives Matter and Common Core Evolved on Social Media
We analyzed how both discussions became popular on Twitter over time.
This article is part of CSMaP’s ongoing Research Summary series. We’ll provide quick, digestible versions of our research articles for peer researchers, journalists, policymakers, and all those interested in the relationship between social media and politics. This week, we wrote about our recent data report, which analyzes how issues become popular online.
The Black Lives Matter movement was borne out of a hashtag that surfaced online after the 2013 acquittal of George Zimmerman — a man with both white and Peruvian ancestry who shot and killed Trayvon Martin, a Black teenager who had walked to a convenience store in Sanford, Florida.
The movement rose to prominence after Michael Brown, another Black teen, was similarly killed by white police officer Darren Wilson in Ferguson, Missouri — sparking protests that roiled the city for weeks and captured the national news cycle. With no centralized hierarchy, its organizers have called for community oversight of policing, less incarceration, and more funding for public education among other policy goals.
The movement continued to snowball after more high-profile killings of Black people by law enforcement. It faced many detractors, but slowly won public opinion to its side, ultimately achieving widespread support after George Floyd was killed by a white police officer in 2020.
Four years before the first Black Lives Matter hashtag surfaced on Twitter, the federal government tied Great Recession-era stimulus funding to the adoption of a standardized set of education policies. This led to 45 states and the District of Columbia adopting the Common Core State Standards in 2010. Touted as a collaboration among state government, educators, researchers, and other stakeholders, Common Core established new guidelines for what K-12 students should be able to achieve in reading, writing, and math.
Unlike increasing support for the Black Lives Matter movement, public approval of Common Core weakened over time among all voters. Democrats opposed the standards because they believed it constrained teachers who sought to tailor curriculums to students’ needs, while Republicans didn’t like the federal government’s interference with local K-12 education. By the 2018–2019 academic year, some states had dropped the standards from their education programs.
Our researchers studied how both issues were discussed — and politicized — over time on Twitter. We chose them for their similarities (each topic lacked an initial partisan or policy framing and attracted different types of social media users) and their differences (Black Lives Matter is about state-sanctioned law enforcement killing members of a minority group and is far more existential in nature than Common Core). We published our findings in a data report earlier this month.
“Why do some issues get giant pick-up and become dominant on social media?” said Sean Kates, a former postdoctoral researcher at NYU’s Center for Social Media and Politics and lead author of the study. “It ties into a lot of what we talk about in politics more generally — like how issues become politicized and how they become polarized. I think the report approaches this question in a way that hasn’t been done before.”
In other words, we examined how influential particular tweets were in each discussion and measured that influence along multiple dimensions. These include: Which users and which messages are most likely to be retweeted? What kind of tweets tend to attract and retain new participants? What role does the tone or sentiment of a tweet play in politicizing an issue? And do some topics tend to amplify the discussion compared to others?
“In some ways, it matters more who sent the tweet than what they sent.”
Here’s what we found: The popularity of a user — marked by attributes such as follower count and verification status — is a more powerful predictor of influence than the substance of what they’re tweeting. But substance still matters: Angry tweets can lure new participants to a discussion and polarizing topics such as individual freedom can drive further engagement.
Here’s how we arrived at our findings: We used keywords and hashtags to gather as close to the full population of tweets about each issue as possible, spanning the period from January 1, 2010 to December 31, 2018. We trained a machine learning classifier to weed out irrelevant tweets and culled about 6.3 million tweets about Common Core and nearly 150 million tweets about Black Lives Matter. We then performed a descriptive analysis of these two datasets, which included examining the distributions of variables that determine influence — such as the ones outlined above — and any potential predictors of those outcomes.
The timelines for the development of each discussion tell two different stories. Discussion of Common Core begins in the early 2010s, picks up speed in 2013, and grows steadily throughout the sample period. Discussion of Black Lives Matter on the other hand starts later, around mid-2014, and is much more responsive to events. There are at least four dramatic spikes in the discussion that coincide with the Ferguson protests, the deadly clash between white nationalists and counter-protesters in Charlottesville, Virgina, and other events that fueled the movement (see figures 1 and 2).
Discussion of both issues followed roughly the same pattern for retweets, which is one of our measures of influence: The majority of original tweets — about 79 percent in both cases — were not retweeted, though the Black Lives Matter data contains a higher percentage of tweets that were widely shared. Retweets by new participants in the discussions — another measure of influence — were an even rarer occurrence at just 4 percent for both issues.
We further probed the impact of retweets by calculating what we call downstream engagement. This measure captures how many tweets in our sample come from users who entered the discussion for the first time by retweeting the analyzed tweet — and distinguishes between casual, short-term retweeters and those who are more invested in discussing each issue. Most tweets that attracted new participants could not retain them, but we observed the following traits among the few that did: Common Core tweets with high downstream engagement were negative and rallied support against the standards, while Black Lives Matter tweets with high downstream engagement were general — allowing them to be retweeted at any point during the movement’s timeline — and focused on the failings of institutions such as the press or the police.
Hashtags fueled engagement: They were associated with a 25 percent increase of retweets and downstream tweets in the Common Core sample. URLs on the other hand nearly halved engagement among downstream tweets, and reduced it by one-third in the overall sample. Common Core tweeters were also more likely to use URLs and multiple hashtags per tweet, while Black Lives Matter tweeters focused on amplifying single hashtags to drive discussion — such as the names of individuals killed by police (e.g., #PhilandoCastile).
“If hashtags help people locate the conversation, then you want to use them, but use them sparingly,” said Kates. “This is how Black Lives Matter moves,” he added, noting the movement’s audience is likely “younger, progressive, and more accustomed to using Twitter language.”
We also expected the tone or sentiment of a tweet to affect its influence and capacity to engage new participants: Using a classifier for words associated with anger, we found that at least one of these words was present in nearly a third of tweets in the Black Lives Matter sample, compared with just 17 percent of tweets in the Common Core sample (see Figure 11). It’s not surprising that anger plays an outsized role in a discussion that centers victims of police violence, but we found it’s also a powerful driver of engagement. Each additional anger word, for example, corresponds with about 25 percent more downstream tweets in our sample.
The two most popular subtopics within Common Core — education policy and individual values — were associated with the largest increases in downstream tweets. But we found that even a rarely mentioned subtopic could be a major politicizing force: Critics of Common Core often cast the standards as the product of philanthropic organizations who were disconnected from everyday people. These elites were mentioned in fewer than 1 percent of tweets in the Common Core sample, but were associated with an outsized increase in new entrants to the discussion.
Perhaps least surprisingly, elites have the upperhand in shaping and influencing a Twitter discussion: We hypothesized that tweets by verified users would be more influential than tweets by non-verified users, and indeed a tweet by a verified user would be expected to receive five times as many retweets and downstream tweets as a tweet by an otherwise identical non-verified user. A user’s follower count is an even stronger predictor of influence, with the median user having up to 12 times the influence of a user with no or few followers.
“In some ways, it matters more who sent the tweet than what they sent,” Kates said, though he caveated that a tweet’s substance can still drive influence in all the ways outlined above.
He added that future research might take on what we left out of this data report: “Which issues didn’t grow in the same way?” he asked. “Why did they effectively wither on the vine or stay in their niche communities? That would be a natural next step for this kind of research.”
