Final_Project: Police Shootings in the United States, 2015-2023
Brussiar
03/30/2023
Introduction
This dataset, obtained from the Washington Post, provides a detailed account of police shootings in the U.S. from January 1, 2015. You can access the data source via this link. The dataset reveals that there have been 8,229 fatal police shootings since 2015, and it is updated weekly on the Post’s website.
My analysis aims to explore the patterns and trends embedded within these incidents. Specifically, I am keen on determining the proportion of victims who were armed, thus potentially posing a threat to police officers. This dataset also provides insights into cases that show signs of mental illness. By identifying instances where victims were unarmed yet displayed signs of mental illness, the analysis bolsters the case for deploying mental health first responders who could potentially de-escalate these situations, thereby mitigating the chance of fatal outcomes. This document concludes with a summary of my key findings and their implications for policy and future research.
Exploratory data analysis
The following is a summary of the total data:
## date name age
## 1/28/2019 : 9 : 364 Not reported: 532
## 1/6/2018 : 9 Not reported : 125 32 : 268
## 11/28/2021: 9 Michael Johnson : 4 31 : 264
## 2/1/2018 : 9 Anthony Nunez : 2 34 : 261
## 4/1/2018 : 9 Brandon Jones : 2 27 : 260
## 5/26/2020 : 9 Christopher Johnson: 2 33 : 259
## (Other) :8175 (Other) :7730 (Other) :6385
## gender armed race
## female : 362 gun :4717 White :3643
## male :7829 knife :1393 Black :1914
## non-binary : 1 unarmed : 465 Hispanic :1289
## Not reported: 37 undetermined: 364 Unknown :1110
## vehicle : 288 Asian : 142
## replica : 276 Native American: 110
## (Other) : 726 (Other) : 21
## city state flee body_camera
## Los Angeles: 133 CA :1179 car :1313 Mode :logical
## Phoenix : 106 TX : 752 foot :1056 FALSE:7044
## Houston : 91 FL : 522 not :4511 TRUE :1185
## Las Vegas : 72 AZ : 370 Not reported:1038
## San Antonio: 70 GA : 312 other : 311
## Albuquerque: 59 CO : 300
## (Other) :7698 (Other):4794
## signs_of_mental_illness
## Mode :logical
## FALSE:6521
## TRUE :1708
##
##
##
##
## police_departments_involved
## Los Angeles Police Department, CA : 120
## Phoenix Police Department, AZ : 100
## Los Angeles County Sheriff's Department, CA : 97
## Houston Police Department, TX : 64
## Las Vegas Metropolitan Police Department, NV: 63
## New York Police Department, NY : 62
## (Other) :7723The data set has 12 variables and 8229 observations. Nine of the variables are factors (date, name, gender, armed, race, city, state, flee, and police department involved). There is only one numeral value or int, for the variable age; and two logical vectors for the variables, body_camera, and signs_of_mental_illness.
The variable “armed” identifies if the person was armed or not at the moment at the moment of the shooting and what type of arm they carried if they were armed.
The variable “flee” identifies in the person tried to flee the police or not, and if they did so by car, foot or other.
The variable “body_camera” identifies if the police agent was wearing a body_camera or not at the moment of the shooting.
The variable “signs_of_mental_illness” identifies if the person shot presented signs of mental illness or not at the moment of the shooting.
The data on the variable “State” is shown in its abbreviated form. Please see the following link to understand the meaning of these abbreviations.
Even though the person’s name is important for the general public, for the purposes of this analysis, I will not account for the variable “name”. To better understand the data, I will also change the name of each variable to a more descriptive/cleaner version. Additionally, I will change the logical observations so that they are descriptive of the type of variable represented.
Here are various aspects worth highlighting from these data. Based on the summary table and tables shown below, the following can be identified:
The median age for those fatally shot by the police was 35 years old.
The majority (95%) of people fatally shot were men.
94.3% of people were armed, with guns and knives being the most carried weapons.
When compared to other individual races, White people accounted for the majority of fatal shootings, 44.2% in total. Followed by Black people with 23.2% and Hispanic people with 15.6%. However, something to note about the limitations of these data is that there is an “unknown” category, which accounts for 13.4% of the total population. As a result, these unidentified people could skew the data towards one racial group more than others.
California and the city of Los Angeles accounted for the majority of shootings.
More than half of people did not intend to flee the police encounter. However, 12.6% of observations did not report this variable.
An overwhelming majority of police officers, 85.6% in total, did not wear a body camera.
20.7% of people shot presented a sign of mental illness.
Missing data appears as “not reported”.
##The following tables support these conclusions.
For a better grasp of the proportion of the data, the tables below are shown in percentage numbers:
| Gender | Percentage |
|---|---|
| female | 4.40 |
| male | 95.14 |
| non-binary | 0.01 |
| Not reported | 0.45 |
| Armed Status | Percentage |
|---|---|
| Armed | 94.35 |
| Unarmed | 5.65 |
| Race | Percentage |
|---|---|
| Asian | 1.73 |
| Black | 23.26 |
| Black,Hispanic | 0.01 |
| Hispanic | 15.66 |
| Native American | 1.34 |
| Other | 0.24 |
| Unknown | 13.49 |
| White | 44.27 |
| Flee Status | Percentage |
|---|---|
| car | 15.96 |
| foot | 12.83 |
| No | 54.82 |
| Not reported | 12.61 |
| other | 3.78 |
| Body Camera Status | Percentage |
|---|---|
| Police officer did not wear a body Camera | 85.6 |
| Police officer did wear a body Camera | 14.4 |
| Mental Illness Status | Percentage |
|---|---|
| Did not have a sign of mental Illness | 79.24 |
| Sign of mental Illness | 20.76 |
To explore the relationships between variables, let’s examine the presence of mental illness among individuals and whether they were armed or not. Specifically, we can identify how many people with signs of mental illness were unarmed:
## Person presented signs of mental Illness
## Armed No Yes
## blunt_object 1.68 0.81
## blunt_object,blunt_object 0.01 0.01
## blunt_object,knife 0.01 0.01
## gun 46.98 10.34
## gun,knife 0.30 0.11
## gun,vehicle 0.41 0.04
## knife 11.24 5.69
## knife,blunt_object 0.02 0.00
## knife,vehicle 0.02 0.01
## Not reported 2.35 0.21
## other 0.75 0.26
## other,blunt_object,knife 0.00 0.01
## other,gun 0.01 0.02
## replica 1.99 1.36
## replica,vehicle 0.01 0.00
## unarmed 4.59 1.06
## undetermined 3.97 0.45
## unknown 1.37 0.17
## vehicle 3.32 0.18
## vehicle,gun 0.17 0.01
## vehicle,knife,other 0.01 0.00Based on the above data, it can be seen that only 1.06% of the people fatally shot were identified as both mentally ill and unarmed . While this represents a relatively small percentage, it highlights the significance of police officers being able to identify cases where victims show signs of mental illness and are not armed. In such situations, providing alternative forms of support with a focus on mental health becomes essential. By recognizing and appropriately addressing these specific scenarios, law enforcement can contribute to offering more targeted and specialized assistance to individuals in need.
Graphical EDA
The visual distribution of the data is presented in the following graphs, providing a clear representation of the patterns and trends within the dataset.
## Warning: NAs introduced by coercion
The graph above indicates that the majority of individuals fatally shot by the police were aged between 20 and 40 years. Further analysis reveals additional patterns. Within this age range, a higher likelihood of fleeing by car or on foot is observed compared to individuals aged 20 or younger, or 40 and older. Guns were the most common weapons used across all age groups, followed by knives. Notably, individuals aged 20 or younger carried the fewest weapons . These findings underscore the importance of considering age-related dynamics and their impact on behaviors and outcomes in fatal police shootings.
##
## (0,20] (20,30] (30,40] (40,50] (50,Inf]
## car 0.08 0.31 0.32 0.18 0.10
## foot 0.11 0.38 0.32 0.13 0.06
## No 0.05 0.25 0.29 0.20 0.21
## Not reported 0.06 0.25 0.30 0.19 0.20
## other 0.09 0.34 0.36 0.15 0.06##
## (0,20] (20,30] (30,40] (40,50] (50,Inf]
## blunt_object 7 44 67 51 23
## blunt_object,blunt_object 0 1 0 1 0
## blunt_object,knife 0 0 0 0 2
## gun 269 1185 1323 793 846
## gun,knife 2 6 11 8 4
## gun,vehicle 3 10 14 2 5
## knife 62 392 403 271 190
## knife,blunt_object 0 1 0 1 0
## knife,vehicle 0 2 0 0 0
## Not reported 22 74 61 28 20
## other 3 14 30 18 13
## other,blunt_object,knife 0 1 0 0 0
## other,gun 0 0 3 0 0
## replica 28 74 68 46 48
## replica,vehicle 0 0 1 0 0
## unarmed 56 171 134 61 37
## undetermined 23 88 93 60 38
## unknown 4 32 30 20 25
## vehicle 25 75 92 46 22
## vehicle,gun 1 6 5 0 1
## vehicle,knife,other 1 0 0 0 0When examining individuals who attempted to flee, the highest proportions were observed in the age groups of 20-50, while the lowest proportions were seen in the age group of 50 and above. This pattern remained relatively consistent across most categories of people who tried to flee. Similarly, when considering armed individuals, the age group of 20-30 had the highest proportions, followed by the age groups of 30-40 and 40-50, whereas the age group of 0-20 had the lowest proportion. These findings suggest that age may play a role in both fleeing behavior and being armed during police incidents, highlighting potential associations worth exploring further.
##
## Attaching package: 'lubridate'## The following objects are masked from 'package:base':
##
## date, intersect, setdiff, union## 2015 2016 2017 2018 2019 2020 2021 2022 2023
## 995 958 983 992 997 1019 1048 1096 141## 2016 2017 2018 2019 2020 2021 2022 2023
## -3.72 2.61 0.92 0.50 2.21 2.85 4.58 -87.14When analyzing the number of fatal police shooting incidents per year, the data reveals some interesting trends. The highest number of incidents occurred in 2022, with 1,096 cases, closely followed by 2021 with 1,048 cases. On the other hand, 2016 had the lowest number of reported shootings at 958. Examining the percentage changes in shooting incidents from year to year, there was a modest decrease of 3.72% from 2016 to 2017. However, subsequent years experienced positive increases ranging from 0.92% to 4.58%. It is important to note that the data for 2023 only includes information up until March, resulting in a lower count for that year. Overall, since the tracking of data began, there has been a cumulative increase of 10.1% in fatal police shootings until 2022. These findings underscore the importance of continued analysis and monitoring to fully comprehend the underlying factors contributing to these trends in fatal police shootings.
## [1] 3643## [1] 3476## [1] 1110
As mentioned before, even though white people represent 44.27% of the fatally shot cases, according to the U.S. Census Bureau, white people represent 61.6% of the total population in the country (U.S. Census Bureau, 2021). Thus, from the data, we can observe that non-white people are being killed at a higher rate than white people. It is important to note that the presence of individuals categorized as “Unknown” in the data may affect the percentages for these two population groups.
Lastly, as we can see from the “Fatal Police Shootings by State” graph, the states of California, Texas, Florida, Arizona, and Georgia had the highest number of fatal police shootings in the country. These data make sense for the first three states, as they are the most populated in the country. However, Georgia ranks 8th and Arizona ranks 14th in terms of population size. Further research would be required to analyze why these two states have such disproportionate numbers based on the size of their population.
Conclusion:
The analysis of the data has offered some surprising insights. Initially, I had assumed that most fatalities in police shootings involved individuals from Black and Hispanic communities. While this is not the case in absolute numbers, these communities are indeed disproportionately affected when considering their relative share of the total U.S. population.
The proportion of victims who were both mentally ill and unarmed was lower than I initially anticipated. Regardless of this group’s size, I believe that more effort is necessary to prevent lethal force being used against individuals who are unarmed or in situations where it’s unclear whether they’re armed.
The age distribution of victims was as expected, with the majority of fatalities occurring among individuals aged 20-40. This fact aligns with the notion that this age group typically has greater access to firearms.
A notable finding was the prevalence of fatal police shootings in the states of Georgia and Arizona. Their inclusion in the top five states for such incidents was unexpected. This discovery underscores the importance of considering regional variations when analyzing these incidents.
Through this analysis, I have gleaned valuable insights about police shootings in the U.S. These findings can inform policy discussions and direct future research towards areas requiring further investigation.