The public has a right to question whether police are biased when they shoot and kill unarmed people of color.

To do this, the public needs data about when police shootings occur, but that kind of data isn’t collected nationally. That’s why several news outlets, such as The Washington Post and The Guardian, have begun collecting news and crowd-sourced data on police shootings and killings.

Analyses of these data by researchers have suggested that African-Americans are more frequently shot by police than whites. Researchers have inferred that the difference is likely due to what’s called implicit racial bias. This kind of bias exists outside of our conscious awareness, and so may impact our decisions without our realizing it.

However, such databases have a fundamental limitation. They only examine negative outcomes – when an officer involved shooting occurs – and ignore cases when officers choose to not shoot at a suspect.

Based on a recent study of data from the Dallas Police Department on not only when police shoot but also when they don’t shoot, my colleagues and I have suggestions for police departments, policy-makers and the public on how to analyze if there is racial bias in officer-involved shootings, as well as how to identify factors to reduce excessive shootings.

What influences the decision to shoot?

My colleagues and I are criminologists, and we wanted to understand whether a suspect’s race influenced an officer’s decision to shoot. So we examined data from the Dallas Police Department from 2013 through 2016. Officers in Dallas are required to submit a report when they draw their firearms and point it at a suspect, but decide not to fire. We compared these cases with the 56 times officers decided to shoot.

This comparison helped us identify aspects of the situation and suspects that might explain why officers decide not to pull the trigger, including the race of the suspect.

We found officers were less likely to shoot at African-Americans compared to white suspects. In our sample, white suspects were shot 4 percent of the time (11 out of 250 white suspects), Latinos 5 percent of the time (22 out of 476 Latino suspects), and African-Americans were shot 2 percent of the time (23 out of 1,005 African-American suspects). The differences in the suspect’s race were not significant. This suggests race is not a factor in an officer’s decision to shoot, and officers are not racially biased in their decision to shoot African-Americans.

It also suggests that factors other than race have a much greater influence on when officers decide to shoot at a suspect.

When suspects are armed with a firearm, officers fire at the suspect in 23 percent of the cases across all races, compared to only 1 percent when they are unarmed. When an officer is injured, they fire at a suspect 11 percent of the time, compared to only 3 percent of the time when they are not injured. Using regression analysis that controls for all of these factors, we find a suspect being armed and a officer being injured have the strongest relationship to whether an officer decides to shoot.

Our findings that African-Americans are less likely to be shot are possibly the result of officers being more likely to draw their firearm when the suspect is a minority, even if the situation does not call for it. That would also be consistent with the data we show that African-Americans are much more likely to have a firearm pointed at them in the Dallas data.

Another potential explanation though is simply geographic. That is, minorities are more likely to come into contact with police because of the way crime rates and police resources are currently distributed. Crimes often repeatedly take place in the same areas within neighborhoods or cities, and police tend to target extra resources at those particular locations. Minorities tend to disproportionately make up the resident population of those locations. This approach to policing could explain the fact that African-Americans have an officer point a gun at them more frequently.

Policy implications

To be clear, existing databases of police shootings can demonstrate the magnitude of racial disparity in police shootings. African-Americans represent a higher proportion of officer-involved shootings, compared to the proportion of African-Americans living in the U.S.

However, our findings shed light on why this alone is not enough to show that police shootings are a result of racial bias.

To assess whether police officers are racially biased in their use of deadly force, researchers need to collect information on when officers choose not to use deadly force, too.

Instead of constructing a database of only officer-involved shootings, all levels of use of force – including shootings – could be regularly collected and made public by police agencies. That would allow assessments of racial bias at all levels of decision-making, like using a TASER or simply an officer’s hands.

In additional research, we found that officers with a greater number of complaints of mistreatment made by the public were more likely to shoot at suspects. Our findings suggest that identifying officers who behave badly could help reduce the number of police shootings, at least in Dallas.

Police departments may be more likely to prevent excessive use of force against the public either through better training of such officers, or in more serious cases, firing an officer.

Limitations

Our findings have important limitations. Our study only examined officer-involved shootings in one city: Dallas. We cannot make the generalization that all police in the United States are unbiased in their decision-making based on the findings in one police department.

Also, our study only looked at only one decision – when officers decide whether to shoot after they’ve already drawn their weapons. Our research cannot show whether officers are racially biased in other instances. For example, we did not look at bias when officers use other types of force, or when they decide to draw their gun in the first place.

Identifying racial bias in the decision to shoot is important, but lower levels of force are much more regular. Monitoring their use will have more power to identify problem behavior by individual officers or disparity in use of force toward any particular racial group.

Other researchers have similarly used Dallas Police Department’s use of force database to examine lesser levels of force and have come to similar conclusions. This suggestion, of course, relies on officers reliably reporting such information and for departments to disseminate that information to the public.

Andrew P. Wheeler does not work for, consult, own shares in or receive funding from any company or organisation that would benefit from this article, and has disclosed no relevant affiliations beyond their academic appointment.