Fair Policing Analysis of the Chicago Police Department

Aryaman Sinha
Published in
9 min readMar 25, 2021


Image Source: https://www.thetruthaboutcars.com/2016/02/eighty-percent-chicago-police-dashcam-videos-suffering-accidental-sabotage/

Police Departments across the US have been in the news for all the wrong reasons, especially in recent times due to instances of excessive force being used. The deaths of George Floyd and Breonna Taylor among others sparked a debate around the role race/ethnicity plays in the treatment an individual gets from the police. As a part of the University of California, San Diego’s Data Science Student Society’s Projects Committee, we sought to gather, analyze, and visualize data from the Chicago Police Department (CPD) to do a Fair Policing Analysis. The purpose of this analysis is not to make any conclusions on policing in the city of Chicago, but simply to analyze a large dataset with the incentive to promote further data-driven research of policing.

Our team worked with a dataset from the Invisible Institute which is a journalism company from Chicago, which can be found here. This is one of the largest publicly available datasets on a police department in the US, with data spanning from 1940 to 2017. The most relevant features used in our analysis were demographics of the city of Chicago, officer ranks/promotions, officer years served, officer salaries, and complaints against officers. Working with this dataset proved challenging as some of the files were poorly named; the naming conventions and the methods in which data were recorded changed over time.

Demographics Analysis

The first step in our analysis was to analyze the racial demographics of the CPD. In comparing the racial breakdown of the City of Chicago and the CPD, we found alarming disparities.

As seen in Figure 1A, White officers made up about 32% of the CPD, while about 62% of the people living in Chicago were White as of 2017 as displayed in Figure 1B. The demographic groups of Black, Hispanic, and people of other races/ethnicities were significantly under-represented in their city’s police force. To test the validity of this claim, we conducted a chi-square goodness-of-fit test which yielded a significant p-value of < 0.0001 suggesting that these 2 sets of racial demographic counts come from differing distributions. However, the sample sizes here are essentially the entire population so the assumption of independence and randomization needed to conduct a chi-squared test is not met.

Following this analysis of demographics, our team compared the ranks of officers currently serving in Chicago by race and once again found notable differences.

As shown by the purple bars in Figure 2A, the proportion of White officers that served as low-rank officers is lower than all other races/ethnicities, which means that White officers were more likely to get promoted. Also, the data suggest that White officers hold higher ranks (Medium and High-Rank Positions) as the orange and teal bars appear higher for White officers in comparison to Black, Hispanic, and officers of other races. Intrigued by this result, we checked if a similar disparity exists concerning gender.

Here we can see that the purple bar is lower for male officers, once again indicating that they tend to get promoted more often. In addition, the data suggests that male officers have a higher likelihood of holding Medium and High-Rank Positions compared to female officers as the orange and teal bars are higher for male officers.

Years Served Analysis

As a part of our project, our team was interested in seeing if there were significant differences in the years served by officers of different races/ethnicities and genders. Using our dataset of officer profiles, we calculated an officer’s years of service by subtracting their resignation date (or the date the dataset was last updated, meaning the officer had no resignation date but was still serving), by their first appointed date, and created a column titled “years_served” in our dataset. To analyze race/ethnicity first, we divided the data into the three main racial/ethnic groups: White, Black, and Hispanic/Other, and using this, we created an overlaid histogram of the years served of 33,324 officers using the Seaborn package in Python:

The mean of our overall dataset as seen in Figure 3A was 20.94 years, and the median was 23.07 years. Through our interpretation of the data, we found three interesting points. First, many officers, specifically Hispanic/Other officers, were recorded as working for one year or less. Intrigued by this, we looked at our data and found that a large proportion of officers that were labeled as Hispanic/Other races/ethnicities had an appointed date and a reasonable/accurate resignation date. We speculate that our data for the Hispanic/Other racial/ethnic groups could be skewed and limited. Next, we saw a sudden increase among officers of all races/ethnicities that retired after 20 years in the Chicago PD. Doing some further research (publicly available on the city of Chicago’s website), we found that the minimum threshold for a police officer in the Chicago PD to receive a pension plan of 50% of their latest yearly salary for the rest of their lives was after 20 years of service. We also noticed a large portion of officers retiring after 30 years of service; our team hypothesizes that this trend is due to the pension plan that maxes out after 29 years of service (which is up to 75% of an officer’s latest yearly salary for the rest of their lives). We did a similar analysis on the years served by gender and found relevant results seen in Figure 3B:

In this histogram, the mean for male officers was 21.9 years, while the mean for female officers was 16.2 years. Just like the previous analysis with race/ethnicity, we noticed a large number of officers, particularly female officers, who only served for one year or less. This was especially interesting to us because we couldn’t find any causality for this spike with the data we were provided. Also, we saw a sudden increase in the number of female and male officers resigning after 20 years, and the largest portion of officers, especially male officers, resigning after about 30 years. This trend also aligns with our previously stated hypothesis that officers retire between the 20–30 year range to maximize their pension benefits.

Salary Analysis

After performing the Kaplan-Meier analysis of years served, our team investigated whether the salary is impacted by the race/ethnicity of the officer. Using data across all police ranks, we found that the median salary of officers for every race/ethnicity is approximately the same. However, when officers were stratified by rank, we found potential discrepancies in salary. From the data, sergeants of Hispanic or Other races/ethnicities had a lower median salary compared to White and Black sergeants by 3.8% on average. This translates to a difference in income of approximately $2500 as seen in Figure 5A:

Additionally, White commanders had a greater annual salary than Black, Hispanic, and commanders of other races/ethnicities, by 35.2% on average: a difference in income of around

$39,000, which is depicted in Figure 5B:

To test the statistical significance of these findings (Figure 5A and 5B), we conducted four separate ANOVA hypothesis tests, stratified by rank in the police force (technician, officer, sergeant, and commander), of officer salaries by race/ethnicity. Each ANOVA test concerning different officer races/ethnicities. In performing this ANOVA test, we verified that the following assumptions were met: (1) the data in each race/ethnic group is nearly normal, (2) the spreads in each race/ethnic group are roughly equal, and (3) the data are independent within each race/ethnic group and across each rank. The salaries for sergeants and commanders, specifically, are not equal between racial/ethnic groups as both of these tests yielded a significant p-value less than 0.001. Our findings indicate that there is income inequality for race/ethnicity present in the higher ranks of the police force, but for lower-rank positions, such as technicians and officers, there is no statistically significant difference in income. There are limitations in the dataset that have an impact on the conclusions we have drawn; for example, there is significantly less data for higher-ranked officers such as commanders, and as such, the salaries may not accurately reflect the salaries of all commanders in the Chicago Police Department.

Intrigued by these findings, we did further analysis to observe if gender may also have an impact on the salary of a police officer. We conducted hypothesis testing to compare the mean salaries between female and male officers and found no statistically significant difference in salary between genders for any police rank.

Excessive Force Analysis

Out of the 244,000 records of complaints against Chicago Police Department officers between 2000 and 2016, 32,500, or 17.2% of those complaints were for use of excessive force. In addition, 98.2% of excessive force complaints resulted in no consequences of reprimands for the officers. To further our analysis, we wanted to analyze officers accused of excessive force by officer race/ethnicity and gender. Using our complaints dataset, we filtered for complaints of excessive force and removed officers that had more than one complaint against them of excessive force (meaning our analysis analyzes the proportions of officers of different races/ethnicities and genders who have had a complaint of excessive force against them). To start, we analyzed the proportion of officers with excessive force complaints by race/ethnicity, as depicted in Figure 6A:

Our results displayed how Black officers had the largest complaints of excessive force against them at about 34%. Not too far off, the proportion of White officers was 30.2%, with Hispanic and officers of other races/ethnicities following with 27.7% and 22.1% respectively. We also performed a similar analysis with respect to gender shown in Figure 6B:

Through this horizontal bar plot, we found that 31.9% of male officers have had a complaint of excessive force against them, with females following with 24.3%. To put this into perspective, almost 1 in 3 male officers had a complaint of excessive force against them.


Throughout our project, our team highlighted the many disparities with regards to race/ethnicity and gender in the Chicago Police Department listed below:

  • There are significant disparities in terms of racial/ethnic demographics when comparing the Chicago Police Department and the city of Chicago.
  • There is a racial and gender bias in the likelihood of promotion of an officer: white officers and male officers are more likely to get a promotion.
  • There is both a racial/ethnic and gender disparity when it comes to the number of years an officer served in the CPD.
  • There may be income inequality in higher ranks of the police force between racial/ethnic groups. We found no significant disparity in salary between officers of different genders.
  • About 1 in 3 male officers and 24.3% of female officers have had a complaint of excessive force against them, with 98.2% of excessive force complaints resulting in no consequences of reprimands for the officers.

As a team, we brainstormed some possible solutions for these major disparities and the minimal disciplinary actions taken against officers accused of excessive force. First, to promote diversity within the Chicago Police Department, we believe implementing mandatory diversity training, and hiring officers representative of the Chicago population would be beneficial. Also, increased accountability for the use of excessive force through the use of body cameras and citizen review committees is crucial. Incorporating these measures will promote equality and fairness in policing in the city of Chicago, set a precedent for fair policing in the United States.

About the Project/Authors

This project was created as a part of UC San Diego’s Data Science Student Society’s Project Committee. You can check out our full analysis, video presentation, and code on our GitHub here. Feel free to leave your thoughts below, and reach out to us via LinkedIn if you have any questions! Team members include:

Aryaman Sinha (Team Lead, 2nd-year Data Science major and Cognitive Science minor at UC San Diego)

Koosha Jadbabaei (Data Science Intern at The Data Standard, Undergraduate Research Scholar, 2nd-year Data Science major at UC San Diego)

Anika Garg (1st-year Data Science major at UC San Diego)