Since January 1, 2016, there have been 714 fatal police shootings in the United States. That comes to 79 deaths a month, 18 a week, and three a day. For context, the US recorded 43 cases of locally transmitted Zika virus and 25 deaths from West Nile in the same time.

For most, if not all, public health issues, some segments of the population are more vulnerable than others. Americans who engage in unprotected sex are more likely to contract Zika virus, while older people are more likely to die from West Nile.

Maimuna Majumder


Maimuna Majumder is an Engineering Systems PhD candidate at MIT and computational epidemiology research fellow at HealthMap. Her research interests involve probabilistic modeling, Bayesian statistics, and “systems epidemiology” in the context of emerging infectious diseases.

And when it comes to fatal police shootings, black Americans are more likely to be shot and killed.

Earlier this month, police shot and killed three black Americans in six days. Tyre King, Terence Crutcher, and Keith Scott are among 174 black Americans to die in such shootings so far this year.

Computational epidemiologists like myself specialize in transforming numbers like these into public health insight, usually by way of mathematical modeling. Our main goals are identifying at-risk groups and proposing mitigation strategies. Though we typically focus on disease, our methods—and interests—apply to anything that injures or kills. So we decided to apply them to the United States’ fatal police shootings.

Though black Americans comprise 12 percent of the US population, they account for 27 percent of the victims in fatal police shootings reported between January 1, 2015 and July 7, according to a database compiled by the The Washington Post. Moreover, police officers were twice as likely to shoot and kill unarmed black individuals than unarmed white individuals.

tkRace representation in the US population and in fatal police shootingsWIRED/MAIMUNA SHAHNAZ MAJUMDER

This broader phenomenon—the over-representation of black Americans within the context of fatal police shootings—is heterogeneous across the United States. To understand how this trend plays out in different parts of the country, we’ve defined an “over-representation ratio” for each state: the percentage of fatal police shootings in which a black American was killed divided by percentage of black Americans. Though the national average over-representation ratio is 2.3 (27 percent divided by 12 percent), some states clock in at several times that. In Ohio, where 13-year-old Tyre King was killed, it’s 4.0.

In a truly race-neutral universe, you would expect an over-representation ratio of 1. However, some argue that over-representation of black Americans—both in armed and unarmed fatal police shootings—is justified due to a “concentration of criminal violence” within black communities. According to 2009 data from the Bureau of Justice Statistics, black Americans were charged with 62 percent of robberies, 57 percent of murders, and 45 percent of assaults in the nation’s 75 biggest counties (despite comprising 15 percent of the total population in these counties).

This doesn’t necessarily mean black Americans are more likely to commit these crimes. In fact, black individuals are more likely than white individuals to be charged, convicted, and incarcerated for the same crime. And perhaps unsurprisingly, states with higher black-to-white incarceration ratios have higher over-representation ratios.

So, why are some states so much worse off than others?

To address this question, we collected data on a number of important socioeconomic indicators for each state that saw at least one fatal police shooting involving a black victim between January 1, 2015 and July 7 (there were 39). We then used multivariate linear regression to determine which, if any, indicators were statistically significant determinants of state-by-state over-representation ratio—the outcome of interest.

Along with the percentage of eligible black Americans registered to vote, we examined indicators like black-to-white incarceration ratio, median household income, the Gini index (as a measure of income inequality), and percentage of population with college education. By including these additional indicators in the model, we could glean how much influence each had on the over-representation ratio while controlling for the others—critical in situations where the indicators themselves are correlated, like household income and college education.

tkMultivariate linear regression results. KFF, Mauer & King, US CensusWIRED/MAIMUNA SHAHNAZ MAJUMDER

What we found was staggering. The number one determinant of over-representation in fatal police shootings—after controlling for all other aforementioned indicators—was the percentage of eligible black Americans registered to vote within the state in question. In other words, the higher the percentage of eligible black Americans registered to vote, the lower the over-representation ratio in a given state. Furthermore, states suffering from increased rates of income inequality (i.e. Gini index, median household income) demonstrated higher over-representation ratios, while states with increased diversity (i.e. percentage of noncitizen residents) demonstrated lower over-representation ratios.

What does this all mean? In addition to promoting diversity and reducing income inequality, these preliminary results suggest that increasing voter registration among black Americans could potentially reduce the risk of fatal police shootings of black victims.

Of course, this is easier said than done. Minority groups are routinely excluded from voter registration drives due to higher rates of unlisted individuals. Furthermore, most states require voters to register several weeks before Election Day—a practice that disproportionately suppresses minority registrants. (Voter registration remains open in most states ahead of Election Day 2016; check whether you’re registered to vote here.)

Despite controlling for several socioeconomic variables, these results don’t conclusively imply causation. But they’re a reasonable starting point. Expanding on that knowledge and finding other potentially actionable strategies will require reframing the issue as a public health crisis, with a focus on data-driven research and policy recommendations.

Yet the Centers for Disease Control and Prevention hasn’t touched gun violence research—let alone fatal police shootings research—since 1996, when the National Rifle Association accused the agency of promoting gun control. Though the government lifted the ban two years ago, federal funding remains hard to come by, and government-sponsored gun violence research is at a standstill.

In the interim, non-governmental research groups are working hard to fill the void. Among them is HealthMap, a computational epidemiology lab based out of Boston Children’s Hospital and Harvard Medical School, which will soon launch an interactive gun violence platform. By harnessing digital data sources—like online news, Internet search query trends, blogging and micro-blogging sites—this platform will provide real-time, open access gun violence information to researchers, policy makers, and the general public.

If society wishes to reduce the burden among those most at-risk, public health organizations must be given the opportunity to carefully examine gun violence—and fatal police shootings—just as they do disease.


An Intriguing Link Between Police Shootings and Black Voter Registration