Fatal Police Shootings: A Red-Blue State Divide?
Over the last year, I’ve been trying to learn more about issues I care about by looking at relevant data. I’ve decided to use this Medium blog to share some of what I have learned, about police shootings, homelessness and mass incarceration.
I first looked for data about fatal police shootings in 2019. Following the tragic death of George Floyd in May this year, as the topic of police shootings became a huge national issue, I was motivated to go back and take a closer look. (We were also doing school remotely in NYC at the time because of the lockdown, so I had more spare time than I usually do.) I had found two sources of data about shootings perpetrated by police in the U.S., one from the Washington Post, and the other from Fatal Encounters. All of what follows uses the Washington Post data, which I downloaded on June 12, 2020, and then again on October 7, 2020. I did most of the analysis on the first data set over the summer, and updated it in October.
When I began to compare the shootings data with factors like race, political leaning, state budgets and so on, I had a couple of notions that I expected to be confirmed and some that I wasn’t completely sure about. I had expected that there would be specific kinds of differences between the racial breakdown of police shooting victims across red and blue states. I also thought that factors like gun ownership rates would have predictable effects on the rate of police shootings.
What I found was surprisingly inconclusive to me, but still a great learning experience. I used data from a number of different sources of data as I explored the effects of different factors on shootings, and I will mention specific sources every time I discuss results from new data.
Fatal police shootings seem to have gone down since the protests in May 2020 started — from 252 in June, July and August 2019 to 234 in June, July and August 2020, and from 244 in Q3 2019 to 216 in Q3 2020. At first glance, it is hard to tell if this was a normal fluctuation or a longer term trend. However, when one looks further back, this change in 2020 could just be normal up and down fluctuation in the numbers.
In May 2020 however, just as the nationwide protests were starting, monthly fatal police shootings were the highest they’ve ever been in any single month (pictured below).
These 110 fatal police shootings in May 2020 represents the highest monthly total in the database. This was only the fifth time the monthly total has exceeded 100 deaths since January of 2015.
How shooting rates vary by race
The primary focus of my analysis was on differences in fatal police shooting rates by race of the shooting victim, which is identified in about 90% of the cases in the Washington Post data set (5083 out of 5673 entries as of October 7th, 2020). Of these 5083 entries, the race identified for over 95% (4862 entries) victims was one of three values — Black non-Hispanic, White non-Hispanic, or Hispanic. Since the number of victims identified as Asian or Native American was quite low, I did not include that data in my analysis.
After taking into account the racial breakdown of the population in each state, it is very clear that across the board, Black non-Hispanic people in the U.S. are getting fatally shot by the police at higher rates than White non-Hispanic or Hispanic people.
A common counterargument when racially motivated murders by the police occur is pointing to the fact that more white people in total are shot by the police than black people. This may be true, but this is also why it is important to factor in the white and black population in each state. I used data from the US Census Bureau on the breakdown in population by race and Hispanic origin across the country to calculate these annual fatal shooting rates per capita.
The Census data also includes the racial breakdown of population by state, which I used to analyze per capita fatal police shooting rates across states and by race of the victim. I only considered states which reported a total of at least 60 fatal police shootings (32 out of the 50 states, which collectively account for over 90% of the fatal police shootings).
While there is in fact a lot of variation in fatal shooting rates across states, the per capita rate of fatal police shootings with a Black victim is consistently higher than the corresponding rate for White victims across 30 of the 32 states, with Mississippi and New Mexico being the exceptions. The fact that these per capita rates are much higher across the states for Black victims is illustrated in the histograms below.
The red-blue state divide?
Next, I explored whether the political leanings of a state might have a relationship to either the per capita overall fatal police shooting rate, or its variation by race of the victim. I categorized the states based on the outcome of the 2016 Presidential election, while labeling six states — Arizona, Florida, Michigan, North Carolina, Pennsylvania and Michigan — as swing states. (I realize there are lots of different ways to label political leaning, but slight variations don’t make a significant difference to my analysis.)
I found no definitive separation in fatal police shooting rates per capita by a state’s political leaning. Red states seem to have slightly higher fatal police shooting rates overall, but blue states seem to have more variation in this rate. (This comes across more clearly in the scatter plots that follow.)
To examine this more closely, I looked at the contrast across states between the per capita fatal shooting rates for Black victims and for White victims:
The ratio of fatal shooting rates by race of victim is highest in Utah, although this is partly driven by the overall relatively small numbers and the very low fraction of the population that is Black — 7 of its total 69 shootings had Black victims, while just 1.1% of its population identifies as Black. To focus on a majority of the data points more clearly, I dropped the “outlier” Utah, and rescaled the axes so that the diagonal line has a slope of 3, since nationwide, the per capita rate of fatal police shootings with Black victims is about 3x that of fatal police shootings with White victims.
Although there is a lot of variation across states, the figure illustrates that the political leaning of a state has mixed correlation with the rates of police shootings. Among this group, the chances of being shot fatally by the police seem highest in Oklahoma. New York State is shown to have the lowest per capita rates of fatal police shooting overall, and with White victims as well as Black victims. I’ve labeled a few other states that had noticeably higher rates of fatal police shootings with Black victims.
Overall, the biggest differences between blue and red states seems to be that red states appear to have slightly higher fatal police shooting rates, and higher rates with White victims than blue states do, while blue states seem to have more variation in these rates. More red states are below the diagonal (representing a smaller difference in the per capita rate of shootings with Black victims versus with White victims), while more blue states are above the diagonal (representing a larger difference).
This difference across red and blue states is puzzling, and I am not certain about its cause, but I’ve explored a number of possible explanations. I obtained data about firearm ownership rates by state from a RAND Corporation study. Gun ownership could have both a positive or a negative correlation with fatal police shootings. Some believe that gun ownership makes a state safer and therefore there would be less violent encounters with police since citizens have access to “protection.” However, more people owning guns is a safety threat in itself and could be an indicator of a state generally being less safe both of which would lead to more situations in which the police are called.
Looking at the data for the 32 states mentioned earlier, I found no correlation I could see between gun ownership and per capita shooting rates where the victim was Black. However, as illustrated above, there does seem to be a positive correlation between gun ownership and the overall police shooting rate per capita, though the conclusion from the chart most visible is probably that people in red states own more guns per capita.
Next, I turned to examining how spending on state and local law enforcement might be related to fatal police shooting rates. I found data for 2017 state and local spending at the Tax Policy Center (Urban Institute and Brookings Institute). More police spending can lead to better training, but on the flip side, more money going into the police force could mean greater militarization or a tendency to deploy more officers in neighborhoods that are seen as “unsafe,” which then results in more shootings.
Perhaps all these effects are cancelling each other out, because the data indicates that the amount a state spends on its state and local police forces has little to no correlation on the per capita shooting rates where the victim was Black, or on overall per capita fatal police shooting rates (illustrated above). Again, the main finding seems to be a stark contrast between red and blue states — blue states seem to spend more per capita on their police forces.
Finally, I looked into whether a higher fraction of urban residents in some blue states might explain why there is more variation in their fatal police shooting rates. Since metropolitan areas have higher violent crime rates than suburban or rural areas and perhaps this causes a greater militarization of police or a different attitude. I got data about the fraction of a state’s population that lives in urban areas from the US Census (downloaded from Iowa State). Yet again, there was no clear correlation, and the main finding seems to be to confirm that blue states are much more likely to have a higher fraction of their population living in urban areas, which makes sense.
I’ve concluded that there’s a lot more data analysis to be done before I can identify any serious correlation or infer anything for sure about what causes higher rates of fatal police shootings, specifically where the victims are Black. It’s important that we are able to identify causes outside of racial bias harbored in police officers which is hard to truly get rid of and find what factors result in more shootings in order to attempt to implement effective solutions.
I did all of my number crunching in Google Sheets (the pivot tables feature came in especially handy). I created most of my charts using Seaborn in Python. I got help from a friend to create the map.
