Berkeley, CA – The tragic deaths of Ahmaud Arbery, Breonna Taylor, George Floyd, and countless other black lives lost to police brutality and senseless violence have spurred calls for policy reform and police accountability among the public. Driven to action, a group of UC Berkeley students led by UC Berkeley senior Tiffany Yu set out to create a scoring system that provides a quantitative measurement of unjust practices for police departments across the country. The project aims to improve accountability and transparency within police departments as well as amplify public awareness by identifying bias and unjust practices over time through detailed analytic reports.
Addressing Police Violence by Tracking Racial Bias
The students investigated face-to-face interactions between officers and civilians by examining traffic stop data, excessive force reports, and population demographics for seven different U.S. cities over a four year period. While some police departments publicly release the racial makeup of traffic stop arrests and excessive force victims, these datasets are large, disparate, and self-inconsistent, making it difficult for civilians to quantify discrepancies in police treatment towards communities of color. Tiffany and her team used a robust data collection and cleaning process, then implemented the information into concise scores and visualizations to highlight any patterns of inequality. Using a carefully-tuned scoring algorithm, the scores measure disparities in the race of arrestees as well as use of force during arrests, ultimately evaluating whether police departments exhibit bias.
Austin, Texas; Bloomington, Indiana; Chicago, Illinois; Cincinnati, Ohio; Lincoln, Nebraska; New York City, New York; and Portland, Oregon were selected as cities due to their geographic variance and the availability of arrest and traffic stop data. A significant component of this project is to illustrate the progression of police department scores over time to determine trends in police departments’ behaviors.
Differences in Police Treatment for African Americans
The results of this project have already revealed disparities in police conduct, particularly found in differences between racial demographics of arrests compared to racial demographics of the city population. The chart above illustrates that black residents were involved in traffic stops disportionately compared to the population of black residents in each city. For example, while black residents make up ~40% of Cincinnati’s population, more than 60% of traffic stops involve black drivers.
Our visualizations reveal that black drivers were stopped disproportionately in almost every city. Furthermore, once African Americans were stopped, they were subjected to force more often than white Americans. For instance, 30% of the drive stops involving African American drivers in Portland involved use of force compared to only 10% of the white drivers.
Perhaps most notably, the team discovered that 6 of the 7 cities had statistically significant differences in excessive force scores between African Americans and white Americans, confirming that race influences the likelihood of an officer using excessive force. The team recognizes that bias may be introduced in the algorithm since this data is self-reported by police departments, so future steps for the project include analyzing more cities to increase the representation among different geographic and population characteristics.
A New Age of Data Science and Social Change
UC Berkeley Assistant Professor of Public Policy Erin M. Kerrison, SCET Chief Scientist & Faculty Director Iklaq Sindhu, and SCET Manager Ed Henrich have all recognized this project as a novel endeavor to apply data science principles to a pressing social issue. Tiffany and her team have made their diagnostic tool and code publicly available, so they encourage others to build off of their work and help lead the charge for change. They hope that with more quantitative methods for assessing racial bias in law enforcement, research into the causes of improvements or deterioration in practices can be used to guide efforts in improving police conduct.
Team members Tiffany Yu, Ethan Shang, Inola Cohen, Lena Bertozzi, Anissa Rashid, and Madeleine Liu are all upperclassmen at UC Berkeley. They all are dedicated to utilizing their backgrounds in data science to tackle the most important social issues of today. This project is supported by UC Berkeley’s SCET Data-X course.
Link to the SCET’s GitHub repository: https://github.com/scetx/datax/tree/master/student-projects/fall-2020.
Those interested in contributing to this effort are encouraged to reach out to Tiffany Yu or Ethan Shang – contact information can be found below.