Inaugural Ingram Olkin S3 Forum: Gun Violence - The Statistical Issues

June 26 and 27, 2019




Gun Violence — The Statistical Issues

In honor of Ingram Olkin (1924-2016), a principal founder of NISS and an internationally prominent statistician whose work stressed the importance of statistical thinking in studying major societal problems, NISS is proud to be the sponsor of the Ingram Olkin SForums. These Forums are aimed at engaging scientists and stakeholders in addressing the compelling issues facing contemporary society and exploring the roles that statisticians and data scientists can bring to address these issues.

With the cooperation and support from the American Statistical Association (ASA) and the Statistical and Applied Mathematical Sciences Institute (SAMSI), NISS initiates the S3 Forums with a workshop examining gun violence and how statisticians can contribute to its understanding. That gun violence is a vexing problem in the U.S. needs little explanation. Numerous foundations and governmental bodies have recognized that adequate information and insight is lacking about ownership and use of firearms, the causes and consequences of their use, and the effects of interventions and technological innovations.  The sociological context and the implications on health and on law enforcement involve a variety of disciplines; the spectrum of questions that arise affect policymakers at all levels.  Implicit is that obtaining adequate information and understanding requires data of high quality and analyses that can withstand scrutiny by all concerned.  

To address the twin concerns, adequate data and informative analysis, this S3 Forum brings a working group together consisting first, of criminologists and others with experience in a multiplicity of gun violence issues and second, statisticians with interest and expertise to tackle the driving concerns. The workshop program suggests that there are many problems but not sufficient numbers of involved statisticians. A principal purpose of the Forum is to expose the opportunities and needs for statistical efforts and to engage a substantial number of statisticians to work across disciplinary lines to help in understanding and mitigating the effects of gun violence.


Each of the five topic sessions will be chaired and have statisticians as discussants who will set the stage for an ensuing open moderated roundtable discussion involving panelists and audience members. This discussion will be guided by a moderator with a focus on “next steps” for the discussion’s topic. The moderator with other participants at the roundtable will prepare a written report concerning the session topic to serve as the material for a final report by this S3 Forum, which will be distributed widely.


Copies (papers, preprints) of the materials leading to the individual talks will be made available by June 1 to the discussants and the chair. The chair of each session may not necessarily be a statistician and could be drawn from the other speakers. Time allotments: session speakers have about 25 min; discussant about 10 min followed by open discussion.  As indicated, roundtable discussions will be included to follow up on next steps and to form collaborations.  

Participants - How to Register  (REGISTRATION IS CLOSED as 6/2/2019)

The session speakers and statistician discussants will be invited participants to this S3 Forum.

Other participants can apply to attend but will be limited in number due to site capacity. For the latter, first please complete the Pre-Event Registration web form below:


After submitting Pre-Event registration information, NISS will contact you regarding instructions for using the event registration links which includes payment for this event.

Registration Fee:  

NISS Affiliate members - $150
Non-NISS Affiliates - $250

Those individuals who have been notified by NISS may use the Registration options at the top right of this event page to register for this event..


If you are calling to make a reservation (703-836-4700), please say:

"I would like to make a Reservation for the NISS Forum 2019."

If you use the website link it is automatically filled with the block details so you only need to select your date range that you plan to be there.

Please Note: All reservations must be completed by May 28, 2019, to receive the discounted rate of $159 plus taxes. 


This event has taken place.  Review the news story "Statisticians Focus on Data Related to Gun Violence at Inaugural Ingram Olkin Forum" that includes links to slides and recordings of presentations on the NISS website.

Wednesday, June 26, 2019

8:30am – 9:00am


9:00am – 10:30am

Topic 1

Emerging Data Sources: Statistical analysis depends on good data sources.  This session aims to familiarize attendees with those data sources, some classic and some new.  Attendees will hear about federal data collection efforts and their potential for analysis of gun violence.  Attendees will also hear about modern police data collections including gun violence networks and acoustic gunshot detection.

  • Erica L Smith, Bureau of Justice Statistics, U.S. Department of Justice.

"Department of Justice Statistical Data Collections Related to Gun Violence” - The Department of Justice, through the Bureau of Justice Statistics and the Federal Bureau of Investigation, has for decades collected data on crime and the administration of justice. This presentation will discuss four key sources of federal data that collect information gun violence—the National Crime Victimization Survey, the Supplementary Homicide Reports, the Survey of Prison Inmates, and the National Incident-Based Reporting System. The presentation will showcase what we can learn about gun violence from each collection and how the data can be used by researchers.

  • Jonathan Lewin, Chief of the Bureau of Technical Services, Chicago Police Department.

"Chicago Crime Fighting Initiative" - In response to an increase in crime in 2016, the Chicago Police Department (CPD) launched the Strategic Decision Support Centers (SDSCs) in 2017. Developed in partnership with the Los Angeles Police Department and University of Chicago Crime Lab, the program began as a pilot in two of the city’s busiest districts.  Analytics Solution:   SDSCs combine technology and process, at the district level, and utilize predictive modeling to inform daily deployment missions. This provides officers with focused priorities and allows the district to monitor and modify strategies based on data. SDSCs integrate mapping, place-based crime forecasting, gunshot detection, crime cameras, and newly-hired criminal intelligence analysts to develop precision policing strategies. Officers are equipped with smartphones, which provide a real-time two-way connection between field and SDSC personnel. SDSCs provide a collaboration hub where prosecutors, officers, detectives, and law enforcement partners can work together to solve problems.

Discussion Leader:  David Banks -  Statistical & Applied Mathematical Sciences Institute (SAMSI), Director;Duke University, Profesor, Department of Statistical Science

10:30am – 11:00am


11:00am – 1:00pm

Topic 2

Gun Violence Trends: Crime has been on a decline for 20+ years, but what are the trends in gun violence? The session aims to reveal the ups, downs, and cycles of gun violence. Some places have experienced steep declines while others sharp increases. Attendees will hear from statisticians and criminologists what different data sources say about gun violence trends.

  • Charles Loeffler, University of Pennsylvania

"​Is Gun Violence Contagious?" -  Existing theories of gun violence predict stable spatial concentrations and contagious diffusion of gun violence into surrounding areas. Recent empirical studies have reported confirmatory evidence of such spatiotemporal diffusion of gun violence. However, existing space/time interaction tests cannot readily distinguish spatiotemporal clustering from spatiotemporal diffusion. Using point process data from an acoustical gunshot locator system and a combination of Bayesian spatiotemporal point process modeling and classical space/time interaction tests, this paper distinguishes between clustered but non-diffusing gun violence and clustered gun violence resulting from diffusion.

  • Rick Rosenfeld, Founders Professor, Department of Criminology and Criminal Justice, University of Missouri-St. Louis

​"Gun Homicides and the Impact of the Opioid Epidemic" -  An increase in homicide is an overlooked collateral consequence of the contemporary opioid epidemic in the United States. The current study fills that research gap in a state-level panel analysis spanning the years 1999 to 2016 of the impact of the opioid epidemic on gun and non-gun homicide among Non-Hispanic whites, the population in which the epidemic has been disproportionately concentrated. We find that elevated homicide rates among whites are associated with increased demand for opioids, measured by opioid-related overdose death rates. We estimate that, had their opioid-related overdose death rate remained constant over time and across states, an average of nearly 1,000 white homicides would have been averted each year between 1999 and 2016. We interpret the results as reflecting the violent dynamics of street drug markets, although more research of the kind conducted on the crack-cocaine markets is needed to draw definitive conclusions about the mechanisms linking opioid demand and homicides.

  • Janet Lauritsen, Curators’ Distinguished Professor, Department of Criminology & Criminal Justice, University of Missouri-St. Louis

"​National and Local Trends in Serious Violence, Firearm Victimization, and Homicide" -  The recent upturn in U.S. homicide rates may have resulted from increases in the number of serious violent incidents, growth in the percentage of those incidents involving firearms, or increases in lethality. To assess these possibilities, we use data from the NCVS and SHR to examine trends in nonlethal forms of serious violence, injury, gun use, and homicide. We also use data from St. Louis to examine localized patterns in serious violence, gun crimes, and homicide trends.  Our national-level findings suggest that recent increases in homicide were not driven by increases in serious violence or the proportion of these crimes involving firearms, but may be related to an increase in lethality. The trends in St. Louis differ and reveal an increase in the rates of homicide and nonlethal violence committed with firearms, but no increase in the non-firearm rates for these crimes. Homicide increases in St. Louis were concentrated in communities that previously had the highest rates of violence, but there is little evidence that changes in the structural conditions or population composition of these areas were associated with the homicide increase.  Overall, the national and St. Louis trends suggest that increases in lethality may have played a role in the recent homicide rise, but that this increase began before the upturn in homicides.

Discussion Leader:  Tom Belin - Professor, Department of Biostatistics, UCLA Fielding School of Public Health

1:00m – 2:00pm


2:00pm - 3:30pm

Topic 3

Policing Gun Violence: Communities often look to their police to address gun violence. One avenue for reducing gun violence is better investigations of gun assaults. In this session, attendees will hear about a statistical analysis assessing the value of police investigations on solving gun crimes as well as the latest research on statistical analysis of ballistic evidence.

  • Philip J. Cook, ITT/Terry Sanford Professor Emeritus of Public Policy Studies, Duke University

"​Policing Gun Violence: The case for investing in investigations" -  The prevailing view is that detective work has limited value in clearing cases or preventing crime.  We test this claim by using a quasi-experimental design to compare investigative resources invested in clearing gun homicide cases relative to non-fatal gun assaults in Boston. We find the large gap in clearances (43% for gun murders vs. 19% for nonfatal gun assaults) is primarily due to greater and more sustained investigative effort in homicide cases, especially after the first two days.  This evidence suggests that allocation of additional resources to investigating shootings would increase clearance rates, despite the well-known challenges of gaining witness cooperation.   A higher arrest rate would arguably prevent gun violence through four mechanisms:  deterrence, incapacitation of high-risk individuals, interruption of tit-for-tat dynamics, and improved police-community relations. Based on a paper forthcoming in Criminology & Public Policy, by PJ Cook, AA Braga, BS Turchan, and LM Barao.

  • Heike Hofmann, CSAFE/Iowa, Iowa State University.

"Using Machine Learning to Match Striae Pattern on Land Engraved Areas of Bullets" -  Recent advances in microscopy have made it possible to collect 3D topographic scans of bullets, enabling virtual comparisons based on the collected 3D data next to traditional comparison microscopy. Automatic matching algorithms have been introduced for various scenarios, such as matching cartridge cases (Tai and Eddy 2018) or matching bullet striae (Hare et al. 2017b, Chu et al 2013, De Kinder and Bonfanti 1999). One key aspect of validating automatic matching algorithms is to evaluate the performance of the algorithm on external tests. Here, we are presenting a discussion of the performance of the matching algorithm (Hare et al. 2017b) in three studies. We are considering matching performance based on the Random forest score, cross correlation, and consecutive matching striae (CMS) at the land-to-land level and, using Sequential Average Maxima scores, also at the bullet-to bullet level. Cross correlation and Random Forest scores both result in perfect discrimination of same-source and different-source bullets. At the land-to-land level, discrimination (based on area under the curve, AUC) is excellent (≥ 0.90).

Discussion Leader:  Kimberly Sellers - Associate Professor, Statistics, Department of Mathematics & Statistics, Georgetown University

3:30pm - 4:00pm


4:00pm - 5:00pm

Working Roundtables

  • Emerging Data Sources Table 1:  Moderator David Banks; Recorder: Lingzhou Xue
  • Emerging Data Sources Table 2:  Moderator: Erica Smith; Recorder TBA
  • Gun Violence Trends Table 3:  Moderator: Tom Belin; Recorder TBA
  • Gun Violence Trends Table 4:  Moderator: TBA; Recorder TBA
  • Policing Gun Violence Table 5:  Moderator: Kimberly Sellers; Recorder TBA
  • Policing Gun Violence Table 6:  Moderator: Heike Hofmann; Recorder TBA

A reception is included in the registration following the working roundtable on Wednesday.


Thursday, June 27, 2019

8:30am - 11:00am

Topic 4

Assessing Gun Violence Risks and Evaluating Initiatives: Several communities have proposed interventions aimed at reducing gun violence. Communities have changed laws, altered policing strategies, and redeveloped community infrastructure in hopes of minimizing or eliminating gun violence. Statistical analysis is essential in the determination of whether these initiatives work and critical in building a base of evidence to support future gun violence reduction plans. Attendees will learn about specific interventions and methods for evaluating those interventions.

  • Terry Schell / Andrew Morral, Senior Scientists, RAND Corporation

"Justify your model: Identifying the best model assumptions​" - In the current literature estimating the effects of state gun laws, it is rare to see discussions of how various modeling choices were made (such as the selection of link functions, outcome transformations, independence assumptions, or presumed error distributions), and rarer to find empirical evidence justifying those choices. Nevertheless, the substantive findings are frequently sensitive to these modeling choices. Given this sensitivity, it is difficult to evaluate the existing literature without knowing more about the appropriateness of the model assumptions implicit in the selected methods. We have used simulations to determine the statistical properties of causal effect estimates on total firearms death rates across a wide range of commonly used models. This research evaluates these effect estimates in terms of accuracy of variance estimation, statistical power, and two types of bias. We found substantial problems with all the methods currently used in this field when analyzing firearms death rates. We also identified a novel model that performs well on all metrics. We conclude by estimating the effects of three common gun laws on total firearms deaths using a Bayesian implementation of this preferred model.

  • John MacDonald, Professor of Criminology and Sociology, University of Pennsylvania.

​"Place-Based Experiments to Remediate Blighted Vacant Land and Reduce Shootings"  -  Remediating blighted vacant urban land can help reduce firearms violence in US cities.  Place-based interventions to remediate vacant land and abandoned housing can tested in the field through randomized controlled trials.  When a place-based remediation treatment is shown to be effective at reducing firearm violence in a field experiment in a city it can be scaled to the entire population of that setting.  A randomized controlled trial and a quasi-experimental evaluation of place-based interventions to remediate vacant land and abandoned housing in Philadelphia, Pennsylvania are shown to reduce firearm violence.  Cities should experiment with place-based interventions that address blighted spaces to develop effective firearm violence–reduction strategies.

  • Yifan Zhang, Biostatistician, Health Policy, Stanford University

"The Longitudinal Study of Handgun Ownership and Transfer (LongSHOT)" -  More than a dozen previous studies have sought to understand the health risks firearm ownership poses to individuals and households. Nearly all of them are case-control studies.  Collectively, this body of research provides fairly strong evidence that those with firearm access have elevated risks of homicide and suicide.  However, the estimates vary widely, and the study designs have methodological weaknesses that limit our ability to draw causal inferences. We initiated the Longitudinal Study of Handgun Ownership and Transfer (LongSHOT) in 2016. Our aim is to produce the most complete and robust estimates to date of the causal effects of firearm ownership on the health of owners and owners’ family members.  To determine who owned handguns, and when, we are taking advantage of an unusual data resource: California’s comprehensive statewide archive of handgun transfers. The backbone of the LongSHOT cohort is a series of historical extracts of California’s official voter registration database, to which we linked the handgun transfer data and all-cause mortality data for California residents at the individual level.  The cohort consists of 28.7 million adults, who were followed for up to 12.2 years. A total of 1.02 million cohort members purchased handguns during the study period and 1.05 million cohort members died—15,616 from firearm-related injuries. This presentation will describe how the cohort was assembled and report results from preliminary analyses. 

  • Jens Ludwig, Edwin A. & Betty L. Bergman Distinguished Service Professor, Director of the Chicago Crime Lab, Harris School of Public Policy, University of Chicago.

"Blocking and Tackling"  -  This paper examines the potential for reducing crime by improving a police department’s ability to manage and deploy its resources – the basic “blocking and tackling” that is at the heart of any government agency. In response to the surge in gun violence that happened in Chicago in 2016, the Chicago Police Department implemented a number of management and related changes in some of the city’s highest-violence districts designed to strengthen implementation of the department’s pre-existing commitment to focus on high-risk individuals, high-risk places and community policing. To measure the impact of these management changes, a randomized controlled trial was obviously not an option given the pressing public safety crisis facing Chicago and the need to prioritize these changes for the most violent districts. And given the limited number of police districts, a regression discontinuity design is also not feasible. We instead employ a variant of the synthetic controls approach from Abadie, Diamond and Hainmueller (2010), which solves the problem that the “treatment” police districts were outliers with respect to violence rates (and also of the limited number of candidate donor districts) by using data at the beat level to construct simulated donor districts through bootstrapping (random selection with replacement). We find that there was a very large reduction in shootings (34%) in one of the first places to receive a SDSC, district 7 (Englewood on the south side), but not in other districts. This does not seem to have been due to simply ‘flooding the zone’ with extra officers, nor to any increase in total arrests. Instead the Englewood district seems to have increased gun arrests specifically (crowding out arrests for other offenses), arrests of individuals with active warrants, and increases in recorded positive interactions of officers with the community.

Discussion Leader.  Elizabeth Stuart - Associate Dean for Education, Professor, Johns Hopkins Bloomberg School of Public Health

11:00am - 11:25am


11:25am - 11:40am  

Juan Meza - Division Director, Division of Mathematical Sciences, National Science Foundation

 "Data Science at NSF: Challenges and Opportunities"

11:40am - 1:10pm

Topic 5

Police Shootings: High-profile police use-of-force incidents and shootings have prompted three-quarters of the major urban protests, civil unrest, and riots in the United States over the past century, with recent police shootings continuing this pattern. In this session, attendees will hear about statistical analyses of police shootings aimed at gaining a better understanding of the factors that elevate the risk of police shootings.

  • David Hemenway, Department of Health Policy and Management, Chan School of Public Health, Harvard University.

"Law Enforcement Homicides: across states and by rurality"  -  The goal of this presentation will be to explain the variation in fatal police shootings of civilians across US states, and to describe the variation in fatal police shootings by urbanicity/rurality.

  • Greg Ridgeway, Department of Criminology, Department of Statistics, University of Pennsylvania

"​The Role of Individual Officer Characteristics in Police Shootings" -  Assessing whether individual characteristics of police officers like age, race, and prior performance influence the risk of shootings has been a long-standing topic of social science research. The effect of officer characteristics on the risk of police shootings is confounded by police assignments and the environmental factors associated with those assignments. This paper demonstrates a method to separate out the influence of individual officer characteristics from environmental factors. Shooting incidents involving multiple officers hold constant the environmental factors present at the scene of a shooting, which permits an assessment of the relative importance of individual officer characteristics in predicting who is at the greatest risk of shooting. Although multi-officer shootings are not common, they are a uniquely rich source of information about the role of officer characteristics and can produce estimates that require no measuring or modeling of a shooting’s environment, greatly reducing the risk of confounding. This paper demonstrates the approach using police shooting data from New York City and from member agencies of the Major Cities Chiefs Association.

Discussion Leader:  Lynne Stokes - Professor, Department of Statistical Science, Southern Methodist University

1:00pm - 3:00pm

Lunch & Working Roundtables

Catered on the 2nd floor (break out discussions during lunch with specific topics per table).  Form 4-5 groups around topics to form collaboration groups. Each table assigned a facilitator & recorder.

  • Assessing Gun Violence Risks and Evaluating Initiatives Table 7:  Moderator: Elizabeth Stuart; Recorder: TBA
  • Assessing Gun Violence Risks and Evaluating Initiatives Table 8;  Moderator: Terry Schell; Recorder: TBA
  • Police Shootings: Table 9;  Moderator: Lynne Stokes; Recorder: TBA
  • Police Shootings:  Table 10 Moderator: Lucas Mentch; Recorder TBA

3:00pm - 3:15pm


3:15pm - 4:20pm


Reporting and Writing:  Group synthesis and call to action writing

4:20pm - 4:30pm



Organizing Committee:

  • Greg Ridgeway, Chair
  • Allan Sampson, Vice Chair
  • David Banks
  • Nancy Flournoy
  • Amanda Golbeck
  • Sally Morton
  • Jim Rosenberger
  • Jerry Sacks
  • Lingzhou Xue

Event Type


National Institute of Statistical Sciences


Statistical and Applied Mathematical Sciences Institute
American Statistical Association
The David Bohnett Foundation
Stanford University, Department of Statistics
University of Pittsburgh, Department of Statistics
Penn State University, Department of Statistics


Registration by invitation - please fill out the pre-event registration form
United States
NISS and SAMSI and ASA logos
David Bohnett Foundation and Stanford Department of Statistics
University of Pittsburgh, Department of Statistics, and Penn State Department of Statistics