Personal Democracy Plus Our premium content network. LEARN MORE You are not logged in. LOG IN NOW >

Bill Bratton and the Ideology of Data: NYC and the Future of Policing

BY Ingrid Burrington | Friday, December 6 2013

New York City Compstat weekly crime data, 11/18-24/13

The disappointment and outrage at Bill de Blasio's appointment of Bill Bratton as NYPD Police Commissioner isn't just about stop-and-frisk. It isn't just about Bratton's profiteering in the private security industry. It's also about data--how governments think about data, how they use (and misuse) data, and what happens when that data means more to governments than the human beings and lives it's supposed to represent.

The great promise of data-driven policy (in cities, in advertising, in online dating) is that it's neutral. Numbers and algorithms don't lie, so decisions made through collecting data are going to be the best ones. While the idea of data as an instrument of state legibility deserves its own essay, the contemporary perfection and normalization of it really comes back to Bratton and to CompStat, the program for decreasing crime through data-driven analysis that Bratton implemented in 1994 as Police Commissioner during the Guiliani administration.

CompStat was, for a long time, credited with dramatically decreasing crime in New York City in the 1990s. Its model has been appropriated by police departments, city agencies, and even entire state governments throughout the country, including Los Angeles, Austin, and San Francisco (Bratton often played a direct role in implementing the program in other cities, either as an employee of the department as in LA or as a consultant). CitiStat, the version of CompStat implemented in Baltimore (depicted as "ComStat" on The Wire), later became the model for StateStat, the data-driven program implemented for the entire state of Maryland. It's even implicated in the burgeoning open data industry, as seen in the software company Socrata's GovStat tool.

Tellingly, CompStat is often mentioned but rarely unpacked in the breathless evangelizing over data-driven policy. At a conference on The Measured City at NYU in October, keynote speakers from city agencies described their "own versions" of CompStat, from NYCHA to the DEP. The person who might be most qualified to speak about CompStat, commissioner Ray Kelly, was notably absent from the proceedings. While Kelly's name appeared among speaker bios, he was a no-show. This was shortly after the incident in which students at Brown University protested Kelly giving a lecture there, a fact that, like CompStat's negative implications, went undiscussed at the keynotes.

CompStat is in principle data-driven, and like data, its principles are generally perceived as apolitical. Here's a depiction of CompStat's key principles from Jeff Godowon's article in The Police Chief magazine:

Accurate and timely intelligence: Know what is happening.

Effective tactics: Have a plan.

Rapid deployment: Do it quickly.

Relentless follow-up and assessment: If it works, do more. If not, do something else.

Those all sound like pretty good ideas, right? CompStat is all about having more regular meetings, lots of communication, focusing on results--honestly, it kind of sounds like a startup. And of course, the intended application of the "Stat" model is generally toward things people like. We like having less crime, we like governments that are more efficient and communicative, we like Google aesthetics and drag-and-drop charts. But case studies of CompStat have suggested those principles are often at odds with other entrenched beliefs inherent to policing--beliefs in necessary hierarchies, in necessary authoritarianism, in accountability for some but not all, in citizens as adversaries.

Numbers don't lie, but humans do--and methodologies for collecting and analyzing data can occlude realities. In the case of CompStat, what's occluded in the public narrative are all the other contributing factors that led to a decline in crime during the Guiliani administration. Furthermore, whistleblowers like Adrian Schoolcraft demonstrated that in the 81st precinct, CompStat became an end in itself seeking better numbers, not better neighborhoods--from completely fabricated stop-and-frisks in order to push numbers up to arbitrary arrests for "blocking the sidewalk" (i.e. congregating on it).

It is not enough, however, to say that CompStat "went wrong" in the hands of misguided precincts. The fact that Bratton strongly supports broken windows theory and connects CompStat to that theory indicates that his model of data-driven policing has at its foundation an expectation of the worst in citizens and a refusal of systemic, prefigurative approaches to addressing the needs of a community. It means the data methodology and the data analysis are designed to work against the people it claims to protect. The broken windows theory is a blunt instrument for understanding nuanced problems, and it's utterly baffling that self-proclaimed progressives like de Blasio would support it.

Broken windows and CompStat's ideological roots in it set the stage for the next phase of data-driven policy: the predictive model. The techno-utopic dream of using data to identify crimes before they happen requires a tremendously low opinion of human beings and an unrelenting faith in algorithms. Predictive policing justifies tactics like mass surveillance of people based solely on things like where they live, the color of their skin, or what they believe. It's also worth considering who stands to turn a profit with the rise of predictive policing--including Bratton himself, who makes considerable profit from his consulting work and is on the board of both police department contractor Motorola and gunshot-detection system company ShotSpotter (which received a $12 million investment from Motorola in 2012). Bratton has also praised predictive policing startups like PredPol, a company that began through a strategic partnership between UCLA researchers and the LAPD while Bratton was LAPD commissioner.

Putting faith in mathematics over faith in human beings is a problem that is bigger than stop-and-frisk. The Bloomberg legacy is not merely stop-and-frisk, high-stakes testing, or public health, it's data evangelism--which makes sense, given that Bloomberg's entire media empire is built on gatekeeping data. Selecting Bratton as NYPD commissioner suggests that de Blasio is just as willing to bow before the altar of data as panacea, and that's incredibly troubling.

To be clear, all this is not to say that data is inherently designed to oppress citizens. After all, data was crucial to the stop-and-frisk debate itself--analysis of years of stop-and-frisk data was a huge part of shifting public perception on the policy. That shift happened when the data was forced into the public by a court order.

Perhaps if the NYPD is asked to make all of its data public--not just curated CompStat numbers, not just court-ordered stop-and-frisk data, but literally everything in granular detail--the metrics can be used to maintain and not obfuscate accountability. Perhaps if the deBlasio administration and the progressive cheerleaders of open data are willing to think of civic data not merely as apps, but as an instrument for holding those in power accountable, especially the police, the narrative of data-driven policy can become genuinely people-powered policy.

"Better data theory" is a demand that admittedly stirs far less people than "stop stop-and-frisk." But it's a demand for the kind of radical shift in thought that de Blasio campaigned on, that his administration needs to be held accountable to, and that, it seems, New Yorkers are ready to fight for.

Ingrid Burrington lives and works on a small island off the coast of America. More at

News Briefs

RSS Feed thursday >

NY Study Shows How Freedom of Information Can Inform Open Data

On New York State's open data portal, the New York Department of Environmental Conservation has around 40 data resources of varying sizes, such as maps of lakes and ponds and rivers, bird conservation areas and hiking trails. But those datasets do not include several data resources that are most sought after by many New York businesses, a new study from advocacy group Reinvent Albany has found. Welcome to a little-discussed corner of so-called "open government"--while agencies often pay lip service to the cause, the data they actually release is sometimes nowhere close to what is most wanted. GO

Responding to Ferguson, Activists Organize #NMOS14 Vigils Across America In Just 4 Days

This evening peaceful crowds will gather at more than 90 locations around the country to honor the victims of police brutality, most recently the unarmed black teenager, Michael Brown, who was shot and killed by a police officer in Ferguson, Missouri, on Saturday. A moment of silence will begin at 20 minutes past 7 p.m. (EST). The vigils are being organized almost entirely online by the writer and activist Feminista Jones (@FeministaJones), with help from others from around the country who have volunteered to coordinate a vigil in their communities. Organizing such a large event in only a few days is a challenge, but in addition to ironing out basic logistics, the National Moment of Silence (#NMOS14) organizers have had to deal with co-optation, misrepresentation, and Google Docs and Facebook pages that are, apparently, buckling under traffic.


wednesday >

NDI Launches Open Source DemTools for International Development

Yesterday the National Democratic Institute launched a suite of web-based applications created for their partner organizations, mostly pro-democracy groups and political parties around the world. These “DemTools,” which are ready-to-use but can also be customized, will give organizations in developing countries some of the capabilities that political activists and parties in the United States have had for years. Moreover, since the National Democratic Institute (NDI) is making the promise to host partner organization's applications in the cloud essentially forever, they hope these applications will help usher in a period of more sustainable tech.