Back on the Beat

By Douglas A. Samuelson

A two-day symposium recently held in Arlington, Va., helped reinvigorate an important but neglected field of operations research: criminal justice applications. The symposium brought together about 40 police officers and about 35 operations researchers to explore their common interests in analyzing operations to increase effectiveness. The National Institute of Justice (NIJ), a component of the U. S. Department of Justice, sponsored the event to attract and encourage more research and to publicize NIJ's upcoming solicitation of grants on Criminal Justice Operations Research next year, the first in several years. WINFORMS, the Washington, D.C., chapter of INFORMS, co-organized the symposium.
The symposium featured a mix of "classic" examples of work in the field and work by relative newcomers. A particular focus was two challenge problems posed by criminal justice practitioners, responses and lively breakout-group discussions of the proposed approaches.

Each day started with welcoming remarks and a keynote address of general interest, followed by a couple of presentations of successful applications of O.R. to law enforcement problems. Lunch featured another keynote address of general interest, followed by a challenge competition and response, a panel discussion of the responses and small-group discussions. Each concluded with previously designated reporters' summaries of the insights from the discussion groups.

Glenn Schmitt, acting director of NIJ, led off Day 1 with welcoming remarks and thanks. He was joined by Stanley Erickson and Iara Infosino, who coordinated NIJ's efforts in putting the conference together. (The idea for the symposium began with former INFORMS President Richard Larson of MIT of INFORMS at a meeting at NIJ in the summer of 2005 with Assistant Director of NIJ John Morgan, who agreed to support it at that time. The details of the symposium plan matured at a meeting among Infosino, Larson and this reporter at the 2005 INFORMS national meeting.)

Al Blumstein, University Professor and dean emeritus of the Heinz School of Public Policy and Management at Carnegie Mellon University and a former president of ORSA, TIMS and INFORMS, then recounted his 40 years of analysis of crime and the operation of the criminal justice system. He noted, with amusement, a famous but unfortunate analysis identifying an apparent "steady state" in the number of people the United States was willing to imprison. The analysis was published just before changes in policy, especially with regard to drug offenses, triggered a dramatic rise in the number. Other predictions, however, highlighting the importance of the size of the population of males in their late teens and early twenties, have proven more accurate, yielding useful insights into criminal careers and how to alter them.

Blumstein also told how his use of a simple model including the traditional rate-parameter using the Greek letter, lambda, to denote the rate at which individuals commit crimes, had been "abstract" enough to generate protests among his audiences. One wit, who will remain nameless to protect the guilty, inquired whether this rate was lower for married individuals. Upon learning that it was, he observed, "So, married had a little lambda?"

Next were presentations by Linda Green (Columbia University) and Ken Chelst (Wayne State University in Detroit) about patrol car allocation and management in New York City and Buffalo, N.Y., respectively. In both cases, understanding union rules and officers' preferred responses to certain situations proved essential, and "optimization" was not the answer. Adjusting duty shifts and shift lengths to correspond to peaks and troughs in calls resulted in substantial cost savings and improved effectiveness in Buffalo.

Larson delivered the lunchtime keynote address. He described work he and several colleagues did in streamlining the arrest-to-arraignment process in New York City, among other projects. He emphasized the need for analysts to immerse themselves in practical problems rather than relying on technical prowess. He quoted Philip Morse, one of the founders of O.R., as insisting, early in World War II, "the most important thing we can do is get good analytically trained people out on station, where they can observe for themselves what the problems are and what solutions are likely to work."

Challenge No. 1:
Balancing COMPSTAT and Community Policing.

The first challenge problem was how to balance the recommendations of COMPSTAT, a widely used quantitatively based, computer-implemented method for prioritizing calls and assigning responses, against the community policing activities that seem effective in preventing crime over time. Hubert Williams, president of the Police Foundation, offered this challenge.

Professor Jitamitra Desai, currently a visiting faculty member at the University of Arizona, and Laura McClay, Virginia Polytechnic Institute and State University, presented their proposed approaches. Debra Piehl, COMPSTAT director for the Massachusetts State Police, then presented suggestions for how to integrate the technical approaches with better subject-matter understanding. She argued forcefully that COMPSTAT and community policing complement each other more than they compete for resources. The selection panel liked her submission so much that they awarded her an "honorable mention" and one of the challenge prizes, then asked her to collaborate with the two honorees prior to the symposium. What resulted was inclusion of more problem detail in the two technical approaches and more ideas about analysis in Piehl's presentation.

It would be pleasant but inaccurate to report that any breakthrough insights ensued. The criminal justice practitioners and O.R. analysts generally reaffirmed that this is a hard problem, and that specific insights by analysts with more opportunity to familiarize themselves with actual operations were likely to be the most useful contributions. Several participants reiterated Larson's recommendation to "put analysts on station" and cited this challenge as a good example of why that is necessary. A particularly promising topic for further research, it appears, would be how to incorporate long-term, non-specific benefits, such as those from community policing programs, into resource-directing methods such as those used in COMPSTAT.

Broadening the Scope

If the first day seemed focused on applications of O.R. to "classical" police applications, the second day provided a counter-balance. This reporter, leading off with welcoming remarks on behalf of WINFORMS, took the opportunity to urge more attention both to new methods, such as adaptive, learning solutions and agent-based simulation, and new topics, such as mass egress and evacuation, crowd dynamics and control, threat investigations, court and prison management concerns and mental health issues. Developing appropriate metrics continues to be an underrated part of the work, and developing insight, in conjunction with domain experts, is often more useful than developing models.

Randy Robinson, former executive director of INFORMS, followed with a general "Science of Better" overview of what O.R. is and what benefits it can provide. James "Chips" Stewart, former director of NIJ and now a senior fellow at the Center for Naval Analyses, then reviewed the benefits O.R. — especially under NIJ sponsorship — has already provided to the criminal justice system, adding still more encouragement to expand the use of O.R. methods in this subject area.

Colleen "Kelly" McCue presented recent work on using predictive analysis and data mining of crime statistics and patterns to direct police activities in Richmond, Va. Her title asked, "Why Count Crime When You Can Prevent It?" She provided some details about one initiative that reduced complaints about random gunfire by nearly half and more than tripled the recovery of firearms during an eight-hour initiative to reduce gun crime on New Year's Eve. Stephen Huxley, of the University of San Francisco, followed with an interesting extension of scheduling to include patrol offices, not just patrols — a complex but rewarding effort.

At lunch, INFORMS Executive Director Mark Doherty told a number of stories about his experiences as a Cincinnati, Ohio, police officer early in his career. Once again, the value of having a trained analyst immersed in the real problems was readily apparent. His analysis of when and how to patrol drastically reduced a spree of armed robberies, apparently an early and less complex version of what McCue called "Just In Time Policing."

Doherty reported having counseled his department, despite the attractions of the offer, to decline the gift of a new helicopter, when he learned that annual maintenance costs would approximately equal the purchase price. He told, as well, of an occasion when he and his partner wanted to run a license plate search and were rebuffed by the computer department because the lengthy print run for the city's water bills could not be interrupted. Suffice it to say that his solution of this problem was creative, unorthodox and remarkably efficient, if perhaps a bit unpleasant, and owed absolutely nothing to traditional O.R. methods. In less than a minute, the computer operators were persuaded to find a way to run the license plate search immediately. (Reporter's note: Details are available on request. Members of INFORMS who contemplate tangling with the executive director would do well to learn the details of this story and keep it in mind.)

Challenge No. 2:
Call Service Optimization.

The second challenge problem was how to analyze call center records to prioritize responses and possibly eliminate responses to some calls. Ed Flynn, chief of police in Springfield, Mass., offered this challenge.

M. Kevin Geraghty, president of Revenue Research, an O.R. consulting company based in Atlanta, showed his winning analysis of some of Chief Flynn's data. Sorting calls into categories of urgency, he was able to display breakouts of types of calls versus locations versus true urgency, and then to propose a "triage" scheme that appeared promising. As he conceded, however, no system can completely eliminate the remote but frightening possibility of neglecting a true emergency call. (Chief Flynn revealed that in some cases, when a particular telephone has been used for a number of bogus 911 calls, the Police Department simply negotiates the removal of that telephone. This policy, he conceded, has its potential drawbacks, but "we're drowning in 911 calls, many of them non-emergency or downright illegitimate, and the tyranny of the telephone is interfering with getting police work done.")

Since Geraghty's was the only entry received for the challenge competition, this reporter offered another analytical view, based on experience with call centers. The idea here is to identify caller characteristics early in the encounter and develop a workable classification scheme similar to the market targeting many retail call centers have. Characteristics of interest might include location of origin, recent events in that area, the nature of the problem, how quickly and cogently the caller describes the problem, and how panicky the caller sounds. Probably the most useful recommendation was to get an analyst into a call center for a few weeks to understand, first-hand, what incoming calls sound like, and what experienced dispatchers have figured out about what to do in response.

The discussion and the small-group sessions brought out the interesting observation that the two challenges can be viewed as different aspects of the same general problem: How best to assign resources to perceived needs. This, too, is a hard problem, most likely best addressed via incremental improvements rather than by an attempt to "solve" everything with one big model. The political implications — police chiefs don't get fired for slow decline of everything nearly as quickly as for one mishandled life-threatening emergency — make any meaningful "triage" risky. This problem, too, looks like a fine candidate for further research, especially research that includes extensive collaboration between the analysts and the subject matter experts.


The comments the organizers received back after the conference were overwhelmingly enthusiastic. NIJ expects to announce a new grant solicitation for CJOR in early 2007 and encourages those who might be interested to keep checking the NIJ Web site,, or to request e-mail announcements of NIJ solicitations at NIJ and WINFORMS also hope to have another symposium like this one, tentatively planned for the spring of 2008. O.R. and criminal justice are back in business together, and NIJ is poised to play a vital role in expanding the collaboration.

Blumstein to Receive Stockholm Prize
in Criminology

Al Blumstein, who transformed the field of criminology by pioneering the use of systems and quantitative analysis as well as the modeling of criminal careers, has been named one of two recipients of the 2007 Stockholm Prize in Criminology. The prize, considered the equivalent of the Nobel Prize for criminologists, will be awarded during the International Stockholm Criminology Symposium in June 2007. Blumstein and co-winner Terrie E. Moffitt, a psychologist honored for her studies of population cohorts, will share the 1 million SEK ($140,000) award.
The Stockholm Prize in Criminology was established in 2005 in order to recognize achievements in the field of criminological research or in the practical implementation of researcher findings in order to combat crime and promote human rights. An independent, international jury of 11 criminologists selected Blumstein and Moffitt.

The jury cited Blumstein for making "a major contribution to the research into criminal careers" and for his "analyses of variations in the frequency of offending in the careers of active criminals in U.S. jurisdictions. His research has had a global impact on justice policies and practices, as well as on the rapid growth in the influence of development and life-course criminology."

Asked his reaction to being named a recipient of the prestigious prize, Blumstein said, "It was a surprise, a delightful surprise. It's probably the crowning achievement of my career. It's certainly the largest from a fiscal standpoint."

Blumstein, a major player in criminology for several decades, hasn't decided exactly what to do with the approximately $70,000 he'll receive as his share of the prize, but he's certain it "won't be used for pure Blumstein entertainment." Instead, he says, he'll share the money with former students and colleagues as a means of "paying back the larger community of researchers."

Blumstein, a past president of INFORMS, ORSA and TIMS, is the J. Erick Jonsson University Professor of Urban Systems and Operations Research and former dean at the Heinz School of Public Policy at Carnegie-Mellon University. His contributions include much of the language used to describe the key features of criminal careers, including the concept of "lambda" as the underlying true frequency of criminal offending that can only be estimated on the basis of such indirect measures as arrests, convictions or self-reported offending.

As the chair of the National Academy of Sciences Panel on Research on Criminal Careers in 1986, he was senior editor of a landmark report on the state of knowledge about "criminal careers" and patterns of offending, which stressed methodological uncertainties about predictions of future offending in any individual case and the ethical limitations of basing sentencing practices upon prior criminal records. In co-authored work with many of the leading contributors to knowledge on criminal careers, he provided insightful and inspiring leadership for understanding the development of criminal careers — with dramatic implications for public safety and crime prevention.

Editor's Note:
As part of the continuing story of "O.R. & Criminal Justice," Al Blumstein will be profiled in the February 2007 issue of OR/MS Today, which will also feature an interview with John Morgan, the incoming director of the National Institute of Justice.

Douglas A. Samuelson is a senior analyst for the Homeland Security Institute, Arlington, Va.; president of WINFORMS, the Washington, D.C., chapter of INFORMS; author of the "ORacle" column; and a frequent feature writer for OR/MS Today.