(Bloomberg Opinion) -- I often consider data to have a narrative quality, an emotional valence. Some tell a story of triumph, some of heartbreak. What keeps me up at night, though, are the missing data – and particularly data on crime and police behavior.
Omissions are often neither coincidence nor accident. Rather, blind spots arise where violence or abuse of power occurs. Consider the data on sexual assaults and rapes of Hispanics in Houston. Reported incidents declined 43% in the year after Donald Trump’s election. A great success? Probably not: Houston Police Chief Art Acevedo suggested Hispanics did not feel safe reporting crimes in their community because they didn’t trust the policing system, or were afraid of deportation. Crimes reported by non-Hispanics actually increased over the same period.
Another deeply troubling example: Child abuse in New York City. Reports have dwindled by more than half since the pandemic hit, most likely because kids aren’t going to school, where evidence of abuse typically gets noticed and reported. With no teachers, nurses or other adults to bear witness, abusers can effectively hold their victims prisoner.
Which brings us to the issue of police misconduct. How often do police in the U.S. act badly? How often do they abuse or kill people? Which forces or officers are the worst offenders? Sadly, data are lacking to offer clear and complete answers to any of these crucial questions.
The number of deaths at the hands of police simply wasn’t comprehensively collected until the Guardian started to do so a few years ago. The data cover only 2015 and 2016 and rely on news reports and crowdsourcing (as opposed to the official data, which rely on information submitted voluntarily by police departments). Still, they helped Harvard researchers show that the U.S. had been misreporting about half of the deaths.
Identifying the officers responsible is even more difficult. When, for example, activists won the release of data showing that stop-and-frisk tactics in New York City disproportionately targeted people of color, all information that could identify the officers involved (such as badge numbers) was removed. Most recently, police dealing with protests over the death of George Floyd actually covered or removed their badges, a practice that prevents people from reporting them for bad behavior.
Video footage helps expose what’s going on, but doesn’t offer a complete picture. Consider the now-famous case of the 75-year-old man who fell and was injured after being shoved by police in Buffalo, New York. Officials initially explained that he “tripped and fell,” and it’s not clear how or whether the incident – and dozens of others caught and released on social media – will be reflected in official data. So the efforts of bystanders to document misconduct aren’t necessarily enough. We must ensure that the data are being collected and used.
One bright spot: This week, the New York State Legislature voted to repeal section 50-A of the state’s civil-rights law, which allowed police officers’ disciplinary records to be kept secret. Now, the public will be able to monitor the database to ensure that complaints are recorded. Ideally, the disclosure could lead to rules on how long an officer can keep behaving badly and stay on the force.
I don’t say this lightly. I often criticize the way that allegedly objective – but actually deeply flawed -- data are used to assess job performance. The grading of teachers, for example, went too far. With cops, though, we haven’t gone far enough. The blind spots will remain dangerous until we fill them in.
This column does not necessarily reflect the opinion of the editorial board or Bloomberg LP and its owners.
Cathy O’Neil is a Bloomberg Opinion columnist. She is a mathematician who has worked as a professor, hedge-fund analyst and data scientist. She founded ORCAA, an algorithmic auditing company, and is the author of “Weapons of Math Destruction.”
©2020 Bloomberg L.P.