Stanford big data study finds racial disparities in Oakland, Calif., police behavior, offers solutions

Stanford researchers analyzing thousands of data points found racial disparities in how Oakland Police Department officers treated African Americans on routine traffic and pedestrian stops. The researchers suggest 50 measures to improve police-community relations, such as better data collection, bias training and changes in cultures and systems.

New Stanford research on thousands of police interactions found significant racial differences in Oakland, California, police conduct toward African Americans in traffic and pedestrian stops, while offering a big data approach to improving police-community relationships there and elsewhere.

Go to the web site to view the video.

Video by Kurt Hickman

Stanford researchers found racial disparities in how Oakland Police Department officers treated African Americans on routine traffic or pedestrian stops. The researchers suggest 50 measures to improve police-community relations, such as better data collection, bias training, and changing cultures and systems.

The report makes 50 specific recommendations for police agencies to consider, such as more expansive data collection and more focused efforts to change the nature of mindsets, policies and systems in law enforcement that contribute to racial disparities.

Among the findings, African American men were four times more likely to be searched than whites during a traffic stop. African Americans were also more likely to be handcuffed, even if they ultimately were not arrested.

Across the United States, the report noted, police agencies are guided by the commitment to serve communities with fairness, respect and honor. Yet tensions between police and communities of color are documented to be at an all-time high.

Oakland, located in the San Francisco Bay Area, has a population of 390,724; 34.5 percent is white, 28 percent is African American, and 25.4 percent is Latino, according to 2010 census data.

The Oakland Police Department has been under federal monitoring for more than a decade since the so-called Riders case involving police misconduct. The Stanford researchers, led by psychologist Jennifer Eberhardt, were engaged to assist Oakland in complying with the federal order to collect and analyze stop data by race. Oakland police started wearing body cameras in 2010 – one of the first departments in the U.S. to do so.

The two-year Stanford study, conducted in close cooperation and collaboration with the Oakland Police Department, examined data from body camera footage, police stops and reports, and community and resident surveys.

Jennifer Eberhardt speaks to Oakland PD

Jennifer Eberhardt, an associate professor of psychology at Stanford who led the research study, offered training to the Oakland Police Department on how implicit bias can influence decision making and behavior. (Image credit: Kurt Hickman)

Eberhardt applauded the willingness of the Oakland Police Department to share its data and the department’s interest in identifying new ways to build better ties between law enforcement and the local residents. “Transparency and data will set you free,” she said.

Oakland Police Assistant Chief Paul Figueroa said, “This report provides a road map forward for the Oakland Police Department and police agencies across the country. This critical work moves from data collection to action. Oakland has already implemented many of the recommendations in the report and will move quickly to implement the remaining items.”

He added, “I would like to thank Stanford University and SPARQ for their commitment to this multi-year project. Their strong research methodologies are evident throughout the report. Throughout this project they involved numerous stakeholders, held listening sessions, and included voices at all levels. This has not been an easy task when fairness in policing and all criminal justice systems are in question.”

Eberhardt is a faculty director of Stanford SPARQ, which stands for “Social Psychological Answers to Real-world Questions.” The Oakland police project is an example of how SPARQ partners with government, businesses and nonprofits to craft solutions to pressing issues in communities.

Research findings

The study analyzed traffic stop data from police body cameras that occurred between April 1, 2013, and April 30, 2014. During this period, 28,119 traffic and pedestrian stops were recorded by 510 police officers. Police can legally stop people on the basis of traffic violations, probable cause, reasonable suspicion, or for being on probation or parole, among other reasons.

With audio recordings from a one-month period, researchers scrutinized more than 157,000 words, including the specific language and tone officers used with residents during traffic stops. They tracked words related to respect or anxiety, and words that reveal how the interaction went and how the resident experienced the interaction with the officer.

The researchers also examined more than 1,000 police reports or “narratives” on traffic stops and surveyed more than 400 Oakland residents about their views on police-community issues.

They found that 60 percent of police stops in Oakland, or nearly 17,000 stops, were made of African Americans. This rate is more than three times that of the next most common group, Hispanics (whites accounted for 13 percent). The research also showed that:

  • When officers report being able to identify the race of the person before stopping them, the person stopped is much more likely to be African American (62 percent) than when officers couldn’t tell the race (48 percent).
  • African American men were more likely to be handcuffed during a stop (1 out of 4 times) than whites (1 out of 15 times), excluding arrests.
  • African American men were also more likely to be searched (1 in 5 times vs. 1 in 20 times for whites), though officers were no more likely to make a recovery from those searches.
  • African American men were more likely to be arrested after a stop by police –1 in every 6 vs. 1 in 14 for white men.

Also, 77 percent of Oakland police officers who made stops during the 13-month period never discretionarily searched a white person, but 65 percent did so with an African American person.

Likewise, 74 percent of these officers did not handcuff a white person who was not ultimately arrested, yet 72 percent did so with an African American person. Also, the degree of racial disparities in handcuffing and arrests was lower for more experienced officers than less experienced ones.

“Racial disparities are real, as this research shows,” Eberhardt said. “Differences exist in how police officers treated African Americans compared to those of other ethnic groups.”

The researchers point out that racial disparities are not defined as overt racism – in fact, they found no such acts by Oakland police officers while conducting the study. It is not so much an individual as an institutional problem or pattern, they note.

Co-investigators involved in the study were Rebecca C. Hetey, a postdoctoral psychology research associate; Benoît Monin, a psychology professor; and Amrita Maitreyi, a psychology researcher, all of Stanford.

Hetey said, “We found a consistent and persistent pattern of racial disparity, even when we controlled for variables such as crime rate.”

She said that drilling deep into the data allowed the researchers to identify problem areas and evidence-based recommendations.

“We’re using science to diagnose the problem and design solutions that can realistically work,” Hetey said.

Recommendations

The researchers suggest that police departments in Oakland and elsewhere can overcome a subtle bias problem. Using better data, providing education and becoming informed are the first steps.

In fact, the Stanford researchers have already conducted training workshops on the subject of bias for about 700 – or 90 percent – of the sworn officers in Oakland. The researchers suggest brief, frequent training sessions with feedback on effectiveness for all police forces.

In the report, Eberhardt wrote, “Our recommendations are broad but are anchored in our primary mission of pushing agencies to collect more data and to do more with the data they collect. For many agencies, this will require a change in mindset: it requires seeing themselves not only as crime-fighting institutions, but also as institutions of learning.”

In broad terms, the researchers’ recommendations suggest:

  • Use data to measure what matters: Continue collecting traffic-stop data, expand these efforts and update the forms; and standardize, track and analyze crime-related communications provided to officers.
  • Leverage police body-worn camera footage: Use the footage to train officers and evaluate policies and require officers to self-audit racially charged footage.
  • Make data accessible: Build a stop data dashboard; automate stop data and narrative analyses; use automatic speech recognition systems; and improve the back-up systems for footage.
  • Collaborate with data partners: Hire a data manager and partner with experts to analyze traffic stop data.
  • Improve feedback channels: Give officers feedback on their stop performance and more efficient ways for them to communicate with command staff; conduct customer-service audits after routine stops and community surveys.
  • Train officers in social tactics: Expand officer training topics; hold more frequent but shorter workshops; hire a training coordinator; and measure the effects of all trainings.
  • Increase positive community contact: Hold monthly relationship-building meetings; require squad-based community projects; train officers and community members together; show more care in high-crime areas; and hold “critical incident” discussions and trainings and annual conferences on police-community relations.
  • Enhance risk management: Identify officers who may be problems; monitor and reduce time pressures, stress and fatigue on officers; review policies on handcuffing people in searches, searching people who are on probation or parole, and asking people whether they are on probation or parole.

On the latter point, the Stanford researchers suggest that police agencies publish an annual Racial Impact Statement on stop data and analyze the data for trends over time, as well as develop “early warning” systems to head off future problems of bias.

Eberhardt and her colleagues said that by becoming “learning institutions,” police agencies in Oakland and elsewhere can understand how to best change their cultures.

Increasingly, it is a matter of law. In California, law enforcement agencies around the state will soon be required to collect stop data and to track that data by race.

“This new law presents an opportunity for change. The Oakland Police Department leadership has begun to answer the call for change,” the researchers wrote.

What do the findings mean?

Prior research, the Stanford scholars point out, suggests that stereotypes of African Americans lead people to believe they are “dangerous, violent, aggressive – criminal.”

But the problem is even deeper, Eberhardt said. Independent of one’s own values, biased attitudes can arise from observing how other people behave toward African Americans in a given situation or environment.

For example, if officers frequently witness the handcuffing and searching of African Americans, this behavior becomes normal and expected – “a script for what is supposed to happen,” according to the report.

Hetey said racial disparities are not overt, intentional biases, but are best described as something people are generally unaware of or almost knee-jerk reactions to which people have been conditioned as a result of social norms, group pressure, culture, and systemic and policy influences.

Eberhardt pointed out that decades of research in social psychology and sociology suggest that norms and culture are significant drivers of behavior.

The reach of such “implicit bias,” Eberhardt wrote, is enduring and often unrecognized, especially in the context of criminal justice. Implicit bias refers to the attitudes or stereotypes that affect our understanding, actions and decisions in an unconscious manner.