There has been increasingly widespread coverage in recent years about the deaths of people at the hands of law enforcement. Movements such as Black Lives Matter have arisen based upon the deaths of black citizens by police. The Counted is a website run by The Guardian which during 2015 and 2016 has worked to track every such death within the United States and compile them into an online data set that is available to everyone. This data is available for download, allowing for others to perform their own set of analyses on it.
In order to visualize the data I used Cytoscape, a network analysis tool, to create a network of the data for each year based on which races or ethnicities had died at the hands of which police departments. By doing so, this allows us to examine and analyze the dataset as a whole in a way that is difficult to do when in tabular format.
Above, one can see the two visualizations formatted in an easy to understand way. Each blue dot of the network is a police department responsible for the deaths of one or more people, and each of the red circles represents the race of one of the deceased. The visualizations of the two years are quite similar as well.
For both, the node for “White” is the most densely connected by edges, followed by “Black”, then “Hispanic/Latino”, with “Asian/Pacific Islander”, “Native American” and “Arab-American” being the least densely connected. This gives use a general idea at a glance of how many people of each race were killed. However, this is not perfectly accurate either. By mapping directly from police department to race/ethnicity, these visualizations exclude any indication of whether a single police department is connected to one multiple deaths of people of the same race, or only a single one. So, if a single police department is connected to the deaths of 10 black people and 1 white person, it would only appear as if they were connected to 1 of each.
However, if we still assume that our network approximates the actual totals, then we can compare the visualizations of the two years to see changes between the two. For example, there is a clear increase in the number of deaths of Native Americans. This brings about several questions. For example, how many of the police departments involved overlap between the two years? What external causes might have led to this increase?
The visualization also allows us to see which nodes are connected to multiple races or ethnicities, and which are not. Most of the centrally located nodes are connected to multiple races, whereas the one on the outside are only connected to a single one. Those that are connected to multiple are more likely to be in areas of higher population which have higher crime rates and greater diversity. This would lead to there being multiple deaths connected to the same police department. In contrast, those nodes with a connection to only one race or ethnicity are areas with either lower crime which would lead to fewer deaths by police, or are more likely to be in areas of less diversity.
While this visualization of the data gives an interesting view, I think it fails to highlight an important distinction. There are some deaths by police that are justified and happen as the officers defends themselves or others, but other deaths are unjustified and were preventable. I believe this network would benefit from displaying which of the deceased were armed vs unarmed, and by doing so would bring to light possible discrepancies related to race.