An Analysis of Tweets Mentioning Black Lives Matter

Anecdotally, we know that the Black Lives Matter movement in 2020 is by far the largest of its kind.
After analyzing a combination of aerial footage and self-reported poll data, the New York Times estimated that between 1526 million people in the US were in some way form or shape active in the protest.
Even if only half of the self-reported numbers were proven to be true, this would mark Black Lives Matter 2020 as the biggest protest in US history.
While the New York Times covered the magnitude of in-person protests, they neglected to tell the story of what was happening online.
The Black Lives Matter movement has been just as active online, as an analysis of public Twitter data will show.
At the peak on June 2, more than 10.2 million tweets in a given day mentioned Black Lives Matter.
When compared to the second-largest Black Lives Matter movement in 2016, the maximum daily tweet count on June 2, 2020, was 8.5x larger than its predecessor. The peak weekly tweet count was 50x greater than the second-highest week in 2016 while the peak monthly tweet count was 15x higher.
By dividing the number (#) of BLM tweets by the total amount of tweets within the same time frame, we can normalize the data and obtain a more consistent view of BLM intensity, irrespective of whether or not the monthly active number of Twitter users have grown.
At the peak in 2016 , 1.4% of all tweets in the US mentioned BLM, whereas in 2020, the numbers show that on #blackouttuesday, 3.24% of all tweets in the US mentioned Black Lives Matter.
BLM Intensity by State
George Floyd was murdered on May 25, 2020.
By May 28, the Twitter form was in full swing with median tweet intensity across all states at 1.4%. This would peak at 3% on June 2nd and fade by July.
For the purpose of easy reading and length, I have omitted the data from the main body. The data for States ranked by median and peak BLM intensity can be found in the appendix.
All Lives Matter as a Response to Black Lives
In almost any progressive social movement, there is a dissenting voice e.g. anti-LGBTQ protestors, climate change deniers, womens rights subjugators, etc.
To the frustration of BLM protestors and allies worldwide, ALM posts have been jutting their unwanted presence into the social sphere.
The combined voices chanting ALM on Twitter are orders of magnitude lower than BLM. So much smaller, that is impossible to compare them on the same scale. Therefore, the y-axis has been logarithmically transformed to the two parallel time-series data more comparable.
Whenever Black Lives Matter tweets spike, All Lives Matter tweets behave similarly. ALM seem to only surface as a reaction to Black Lives Matter. That is to say, the counter-movement lacks any sort of unique identity.
We can fit a pretty clean linear regression model that predicts the # of ALM tweets on any given day, by simply looking at the corresponding number of BLM tweets.
Again you can find the data for States ranked by median and peak ALM tweet intensity in the appendix.
The Voices are Fading. Have We Done Enough?
Twitter activity for BLM has dropped off significantly since its peak. In fact, one could categorize this as a classic example of exponential decay.
To more easily visualize the speed of decay and compare the drop-off rate to 2016, were going to again apply a logarithmic transformation to the Y-axis. As a result, this converts an exponential relationship into a linear one.
In the graph below, the slope (steepness) of the line represents how fast the drop-off rate is.
BLM 2020 was not only a much larger movement than its predecessors, but a more persisitent one as well.
Besides comparing 2016 to 2020, we can apply our statistical golem to approximate the rate of drop-off for each state. Remember the drop-off rate is expressed in logarithmic terms, as such it is an exponential relationship and not a linear one.
How has the BLM Movement Impacted Legislation?
Did the protests and rallying cries from the people amount to legislative change? If so where?
This infographic put together by Jordan Caspersz , he outlines the bills that have been submitted and approved by government as a direct result of BLM.
Data Source
I pulled this data from the Twitter Premium API using primarily the counts endpoint for tweets mentioning Black Lives Matter and All Lives Matter. I was able to pull the the total # of Tweet mentions over time for these topics, as well as # of Tweets by State over time.
I cleaned the data using Python, and made the graphs using a combination of Seaborn and Plotly. For mapping visualizations, I used geopandas, and ImageMagick to convert a series of individual maps into an animated GIF.
Special Thanks To
Special thanks to Adam Brown for brainstorming math with me. Thanks to George Ceciarelli and Jordan Caspersz for helping research on important BLM-driven legislative changes. Thanks to Christina Xie, my sister for your awesome proofreads.
Appendix
BLM Tweet Volume by State
ALM Tweet Volume by State
BLM Tweet Intensity by State
ALM Tweet Intensity by Stateread more