May 11, 2024

30 years of Compstat, falling homicide rates, use of AI in criminal investigations, and PERF events and opportunities


PERF members,

This week I’d like to cover a few different topics, including the 30th anniversary of Compstat, the recent decrease in homicides, scrutiny of AI in criminal investigations, and a few opportunities from PERF.

30 years of Compstat

In the early 1990s, my friend Frank Hartmann, who chaired the Harvard Kennedy School’s Executive Session on Policing and Public Safety, recommended I go up to New York and sit in on one of the NYPD’s “crime meetings.” He said he hadn’t seen anything quite like them and thought I should see one for myself.

I took Frank’s advice and traveled up to New York. At 7:00 a.m. I walked into a large room full of about 300 members of the NYPD. The new NYPD commissioner, Bill Bratton (who I shared a desk with when we both worked at the Boston Police Department), sat at the center of the table below a screen with the outline of a precinct and its crime numbers — homicides, shootings, robberies, etc. The precinct commander would then explain to the entire leadership team of the largest police department in the country what he or she was doing about crime in their area. This was my first experience with Compstat.

At the time, Joe Dunne was the commander of the 75th Precinct in East New York, an area sometimes referred to as the “killing fields.” Dunne told me that before Bratton’s arrival, a precinct commander would rarely have a face-to-face interaction with the bureau chief, much less the commissioner. Now every senior official in the department was both evaluating and supporting every precinct commander.

And in the past, impoverished communities that saw greater violence, like East New York, may not have received as many resources or top commanders. As Bratton implemented Compstat, he placed promising commanders in more difficult assignments, provided them with the resources needed to do the job, and tracked their progress. When it came time to make promotions, he showed the department that the way to advance was to succeed in difficult assignments.

Through Compstat, Bill Bratton and senior NYPD officials like Jack Maple, Lou Anemone, and John Timoney sent officers the message that they had the ability to reduce crime. This went against the prevailing thinking at the time, which was that no matter what the police did, they couldn’t do anything about crime. This an excellent example of what business management author and consultant Jim Collins calls a “Big Hairy Audacious Goal.”

Last week I traveled back up to New York for an event celebrating the 30th anniversary of Compstat. As Bratton and others spoke about its origins, I was struck by the way this concept has permeated policing over the past 30 years and become part of many agencies’ DNA. It is how police departments expect to be measured, and how they hold themselves accountable to the community. At last week’s event, former NYPD Chief of Department Lou Anemone said, “Compstat gave everyone a sense of urgency.”

That sense of urgency is as relevant today as it was 30 years ago. Lives have been saved by the effective management and good policing brought about by Compstat.

For more on how Compstat shaped the profession, read the 2013 report PERF and the Bureau of Justice Assistance wrote about Compstat’s origins and evolution.

Falling homicide rates

Some recent good news is that all signs indicate we’re in the midst of a substantial drop in the country’s homicide rate. FBI Quarterly Uniform Crime Report data shows homicides fell 13 percent from 2022 to 2023. This figure is in line with other sources; the Major Cities Chiefs Association (MCCA) reported a 10.4 percent drop in homicides in 2023 across 69 U.S. cities, and AH Datalytics, which tracks murder data from more than 200 cities, found a roughly 12 percent decline.

The news has been even better this year! AH Datalytics is reporting an 18.8 percent drop in murders so far this year across 247 cities, and MCCA data from 68 U.S. cities shows homicides decreased 17.3 percent in the first three months of the year.

Data from some individual cities is even more remarkable. Philadelphia, Baltimore, New Orleans, and Austin all report homicides have dropped at least 35 percent in their cities so far this year. And Boston, a city of roughly 650,000, just recorded its third homicide of the year, an 80 percent decrease from last year.

Last month Jeff Asher, a co-founder of AH Datalytics, wrote a good piece on Substack about the drop in murders. His bottom line was that “a murder decline of even half the magnitude suggested by the early 2024 data would place the U.S. murder rate this year largely on par with or below where it was from 2015 to 2019 prior to the surge in murder in 2020.”

We may never be able to parse out exactly what has led to this decrease, just as it may not be possible to pinpoint the specific cause of the sharp rise in homicides during the turmoil of 2020. But it’s welcome news, and I know you all are doing everything you can to further reduce the homicide rate below where it was in the 2010s.

Understanding AI Tools

Monday’s Daily Clips included an NBC News article about growing scrutiny on the Cybercheck AI tool, which analyzes open-source data to assist with criminal investigations. As NBC reports, “defense lawyers have questioned its accuracy and reliability. Its methodology is opaque, they’ve said, and it hasn’t been independently verified.”

I will leave the legal analyses around due process and discovery concerns to the lawyers, but what struck me most from this story was that it sounded so familiar. I believe this is another case of well-intentioned officers trying to solve a crime with a piece of technology that is not fully understood, hasn’t been subjected to independent verification, and whose use isn’t guided by policy or best practices. As I’ve recently written in this column, there are many ways AI can help police more effectively and efficiently fight crime, but departments need to understand the strengths and limitations of any tool they consider using. Working with stakeholders, agencies need to develop policies detailing when the technology can and cannot be used, and how it will be audited and evaluated. Most importantly, any AI tool must always be used with a “human in the loop.” No AI tool should make enforcement decisions or be the sole basis of an arrest or identification of a suspect. Law enforcement officials must be able to explain how each technology works and verify any output with other information as they build a case.

News from PERF

I’ll close with a reminder about a few opportunities from PERF. First, we’re offering free training and technical assistance (TTA) on using DOJ’s Violent Crime Reduction Roadmap strategies to reduce violent crime while building community trust. PERF has been funded by the DOJ Bureau of Justice Assistance (BJA) to coordinate this TTA, which will be tailored to each community’s unique needs and violent crime issues. Click here for more information, to view a recent webinar on this opportunity, and to request TTA.

And the PERF Annual Meeting is happening on May 29-31 in Orlando! We’ll begin with a welcome reception Wednesday evening. On Thursday, we’ll hold a full-day Town Hall Meeting, including: a discussion on artificial intelligence in policing led by Harvard Business School Professor Mitch Weiss; free-flowing conversations about falling homicide rates, campus protests, officer review of body-worn camera footage after critical incidents, challenges facing new police chiefs, civilianization and alternative response, PERF’s recent police staffing survey and how agencies are addressing staffing challenges, and more; and the presentation of PERF’s Gary Hayes and Leadership awards. We’ll have another reception on Thursday evening where you can continue the day’s discussions with your colleagues. On Friday morning, we’ll hold two panel sessions: one on officer wellness, and the other on police use of drones. I hope you can join the hundreds who have already registered to take part in the Annual Meeting. Click here for more information and to register.

Finally, we still have some seats available in our June 25-26 ICAT train-the-trainer session in Decatur, Illinois, and have just opened registration for our August 13-14 and September 10-11 sessions. The training is free; click here for more information about ICAT and to register for the sessions.

Have a nice weekend!