On August 9, 2014, Michael Brown, an unarmed man, was shot and killed by a police officer in the St. Louis suburb of Ferguson, Missouri. Police initially called the shooting regrettable, but justified. Other accounts suggested that Michael Brown had his hands raised and was in the process of surrendering when he was shot. Protests were organized on August 10. Most of the protests were peaceful, but some crowd members became unruly, leading to civil disorder in Ferguson which lasted for several days.
In reporting on the event, the news media started to craft a racial narrative to the shooting. Michael Brown was black. The police officer who shot him was white. The FBI opened a parallel civil rights investigation on August 11, which further bolstered the idea that the shooting might have racial overtones. This became the news story: unarmed black man shot by a white police officer. It did not help that Ferguson, the suburb where the shooting took place, is predominately black, while the police officer hailed from the mainly white suburb of Crestwood.
I can understand why the media chose this narrative for the events. When trying to make sense of a tragedy it might be easier to accept racial prejudices as the cause. But as the story progressed, the news narrative took an interesting turn. St. Louis emerged as one of the top 10 most segregated cities in the United States. Perhaps St. Louis was a racial powder keg ready to explode? Was the shooting in Ferguson merely bringing deep racial tensions to the surface?
An alternate narrative (one not as headline-grabbing) was that there were already a distrust of the police within the St. Louis communities they were sworn to serve. The tragedy of Michael Brown was perhaps not racially motivated, but was instead a natural outcome for residents and peace officers who, unfortunately, suspect the worst in each other.
But which narrative are we to believe? If this incident is the result of racial tensions, we should be able to measure these tensions before the event took place. Likewise, if the shooting was due to distrust between the citizenry and police, this too should be evident. By looking at the Twitter record of people in St Louis before and after the August 9 shooting, I hope to understand how race and law enforcement played a role in the subsequent events.
Share of Voice
I collected a sample of 1,123,611 tweets from 15,534 people in St. Louis between July 28, 2014 and August 25, 2014. From these tweets I created three groups: Tweets mentioning Ferguson, tweets mentioning the police, and tweets mentioning the racially polarizing terms black or white.
Figure 1 shows the percentage of people in St. Louis mentioning Ferguson, police, or black/white. All three terms have a dramatic increase in mentions following Michael Brown’s death. At its peak, over one-third of people in St. Louis were discussing the incident. Ferguson sees a 280x increase in mentions, police a 14x increase in mentions, and black/white a 2.5x increase. This is compelling evidence that these terms are related to the shooting events.
Figure 1: Share of Voice for Ferguson, police, and black/white in St. Louis
We can see that the mentions reach a peak following the start of the FBI investigation on Aug. 11. It appears that this is when there was the most intense media attention on the events in St. Louis. Discussions taper off following the formation of a grand jury, signifying the end of the story’s media cycle.
The data shows that St. Louis residents were nearly two times more likely to mention police along with the shooting than black or white. Based on mentions, the data shows that St. Louisians felt the shooting was a police issue – not a racial one. While local leaders and the media may have pushed for the racial narrative, the people rejected it by a margin of 2-to-1.
Next, I looked at the sentiment behind each term. The share of voice shows that the terms Ferguson, police, and black/white are related to the shooting, but gives no indication of how people are using the terms. Are the police viewed more negatively after the shooting? Has the shooting had an impact on racially polarized conversation?
I measured the sentiment for all tweets and calculated an average sentiment for each day. To give a sense of what the sentiment value is and how much variation that value is, I looked at the statistics for average daily sentiment in a one-week window centered on each day between July 28 and August 25. An animated movie of the dynamic sentiment is shown in Figure 2. and Figure 3 shows a filmstrip by day for the data in the movie.
Figure 2: Animated Average Daily Sentiment
Figure 3: Filmstrip of Average Daily Sentiment
The first thing to note is how much variation there is in sentiment for Ferguson before the shooting. This is to be expected as, before the shooting, there is very little reason to cause people to discuss Ferguson either positively or negatively. After the shooting the sentiment drops to very negative with little variation. This too makes sense as most tweets mentioning Ferguson are now correlated with the shooting and as such are expected to have negative sentiment.
The next interesting feature is the change in black/white over the period of the news cycle. Before the shooting, black/white mentions are mostly neutral. After the shooting, and during the period of the most intense media reporting, the is a drop in black/white sentiment. Following the end of the media cycle with the grand jury, the black/white sentiment is recovering towards pre-shooting neutral levels.
Not only did the shooting increase the number of people mentioning black/white, but those mentions are also more negative in their sentiment. Naturally, this raises the question of whether the shooting caused a worsening of race relations? I believe this is not the case. If the shooting prompted more negative racial feelings in St. Louis residents, this negativity should persist after the media cycle. I note that following the media cycle, mentions of race still remain high, suggesting that there is still active conversation on this topic. Instead, the negativity is explicitly tied to the media reporting of the incident. Up to the grand jury, when media attention is at its peak, the racial sentiment is low. Once the media spotlight shifts (as evidenced by the drop in tweets mentioning Ferguson) the racial sentiment improves to pre-shooting levels. Negative racial sentiment appears to be a by-product of the media narrative and not the true feelings of St. Louisians.
Finally, look at the sentiment for the police. There is no change in average sentiment and variation of sentiment before or after the shooting! Let’s pause on that for a moment. People in St. Louis already spoke of the police with such contempt and negativity that the shooting of an unarmed man in broad daylight did not lower their opinion of law enforcement. In light of this, it is quite obvious that there was an issue with the police pre-dating the shooting of Michael Brown. This event is probably a result of ongoing tension rather than racial prejudices.
The data seems pretty clear. The racial motivation behind Michael Brown’s shooting seem very directly tied to the media narrative. While the people of St. Louis could have easily followed the media’s reporting and turned the incident into a racial trigger, it seems that most people did not rise to the bait. This is borne out by the swiftness that the civil unrest was brought back under control and the lack of subsequent racially motivated incidents.
It does seem obvious from the data that there is a problem with police perception. I did not separate tweets that mentioned police based on race, so there may be a racial component to the low sentiment scores. However, given the low scoring of “police” overall, it is likely that St Louis has an atmosphere of distrust between its people and those sworn to protect them. It is this distrust – not underlying racial tensions – that probably led to the tragic shooting of Michael Brown.
Featured Image courtesy of Freebase