Integrating and Evaluating Multiple Integrating and Evaluating Multiple PSN Strategies in ChicagoPSN Strategies in Chicago

Trace MearesTrace Meares

Andrew PapachristosAndrew Papachristos

University of ChicagoUniversity of Chicago

Jeffrey FaganJeffrey Fagan

Columbia UniversityColumbia University

The Northern District of IllinoisThe Northern District of Illinois Taskforce included:Taskforce included:

Chicago Police DepartmentChicago Police Department ATFATF US Attorney’s Office US Attorney’s Office Cook County State’s Attorney’s OfficeCook County State’s Attorney’s Office Illinois Department of CorrectionsIllinois Department of Corrections Cook County ProbationCook County Probation City OfficialsCity Officials Multiple Community Organizations/ActorsMultiple Community Organizations/Actors Research Partners (University of Chicago Research Partners (University of Chicago

and Columbia University)and Columbia University)

The ProblemThe Problem

Homicides: Guns and Gangs Homicides: Guns and Gangs on the West-Sideon the West-Side

Goal 1aGoal 1a – – Reduce Gun HomicidesReduce Gun Homicides

Annual Homicide Counts in Chicago, 1990 to 2004Annual Homicide Counts in Chicago, 1990 to 200440

050

060

070

080

090

0

Tot

al N

umbe

r of

Hom

icid

es

1990 1995 2000 2005Year

PSN Begins

Gangs, Guns, and DrugsGangs, Guns, and DrugsTotal and Gang Homicides, 1990 to 2004Total and Gang Homicides, 1990 to 20040

200

400

600

800

1000

To

tal N

umbe

r of

Ho

mic

ide

s

1990 1995 2000 2005

Year

All HomicidesAll Homicides

Gang HomicidesGang Homicides

Goal 1bGoal 1b: Going : Going after the guns after the guns

and gun and gun offendersoffenders

Moran’s I = .378 , p =.001

Control Beats

PSN Beats

Guns and Guns and Homicides are Homicides are notnot randomly randomly distributed distributed

Treatment:Treatment: 24 beats on West 24 beats on West

SideSide

Control:Control: 30 beats on 30 beats on

South sideSouth side

ATF Gun Seizures

0 - 15

16 - 31

32 - 48

49 - 67

68 - 117

Crime IndicatorsCrime Indicators

 City (All Beats)

(N = 281)Control Beats

(N = 30) PSN Beats

(N = 24)

2002 Homicide Rate22.3 (648)

49.6 (102)

75.5 (115)

2002 Gang-Related Homicide Rate

4.5 (133)

7.8 (16)

13.8 (21)

2002 Aggravated Assault & Aggravated Battery

862.2 (25005)

1851.9 (3812)

2005.4 (3053)

2002 Average ATF Gun Seizure Rate

215.6 (6252)

438.2(902)

620.8 (945)

PSN StrategiesPSN Strategies

Ex-Offenders, Gun Teams, and Ex-Offenders, Gun Teams, and ProsecutionsProsecutions

PSN Chicago - Program Design PSN Chicago - Program Design

Total PopulationTotal Population

Ex-Offender Population

Crime

Arrest Arrest

State Court

Federal Court

Disposition Disposition

General Model of Offending and Criminal Justice

System

PSN Programs and Initiatives Multi-

Agency Gun Teams and Task Forces

Multi-Agency Gun Teams and Task Forces

Case Review of all gun cases by local

and federal prosecutors.

Decision at which level to prosecute.

Case Review of all gun cases by local

and federal prosecutors.

Decision at which level to prosecute.

Offender Notification Letters

Offender Notification Meetings

Community, Media, & School Programs

PSN InterventionsPSN Interventions

1.1. Increased Federal Gun ProsecutionsIncreased Federal Gun Prosecutions

DeterrenceDeterrence

2.2. Increased Federal Prison SentencesIncreased Federal Prison Sentences

IncapacitationIncapacitation

3.3. Increased Firearms Policing Increased Firearms Policing

Supply-side strategiesSupply-side strategies

4.4. Offender Notification MeetingsOffender Notification Meetings

DeterrenceDeterrence

Normative changeNormative change

EvaluationEvaluation

Quasi-Experimental DesignQuasi-Experimental Design

Research DesignResearch Design Quasi-Experimental Panel DesignQuasi-Experimental Panel Design

Use Near-Equivalent Control GroupUse Near-Equivalent Control Group

Panel data of every “neighborhood” in Chicago from Panel data of every “neighborhood” in Chicago from

1999 to 20041999 to 2004

Multi-level modeling Multi-level modeling strategy to assess within and strategy to assess within and

between group variationbetween group variation

DataData Multiple sources – CPD, ATF, IDOC, and the CensusMultiple sources – CPD, ATF, IDOC, and the Census

Geocoded to Geocoded to police police beat – (~ to “neighborhoods”). beat – (~ to “neighborhoods”).

Quarterly from January 1999 to December 2004Quarterly from January 1999 to December 2004

VariablesVariables Dependent Variables / Outcome Measures:Dependent Variables / Outcome Measures:

Homicide Rate (logged)Homicide Rate (logged) Gun Homicide Rate (logged)Gun Homicide Rate (logged) Gang-Related Homicide Rate (logged)Gang-Related Homicide Rate (logged) Aggravated Assault and Battery Arrest Rate Aggravated Assault and Battery Arrest Rate

(logged)(logged)

Neighborhood Social Indicators:Neighborhood Social Indicators: DeprivationDeprivation Concentrated ImmigrationConcentrated Immigration Residential StabilityResidential Stability Spatial Lag of beat-level Homicides (Moran’s I)Spatial Lag of beat-level Homicides (Moran’s I)

PSN Treatment VariablesPSN Treatment Variables

1.1. Offender Notification Meetings Offender Notification Meetings – percent of offenders – percent of offenders in beat who have attended a meetingin beat who have attended a meeting

2.2. Federal Prosecutions Federal Prosecutions – number of prosecution in beat – number of prosecution in beat (logged)(logged)

3.3. Federal Prison Sentences Federal Prison Sentences – person-month sentences – person-month sentences (logged)(logged)

4.4. ATF Gun SeizuresATF Gun Seizures – number of guns seized by beat – number of guns seized by beat (logged)(logged)

5.5. Index of Components Index of Components – additive score of where each – additive score of where each beat falls on quintiles of other predictors (ranges from beat falls on quintiles of other predictors (ranges from zero to 20). zero to 20).

Annual Homicide Rates, 1991 to 2004Annual Homicide Rates, 1991 to 2004

PSN Group

Control Group

Other areas of City

PSN BeginsPSN Begins

2040

6080

100

Ho

mic

ide

Rat

e p

er 1

00,0

00

1990 1995 2000 2005year

Change in Neighborhood Homicide Rates, Before Change in Neighborhood Homicide Rates, Before and After Start of PSNand After Start of PSN

PSNPSNNeighborhoodsNeighborhoods

Control Control NeighborhoodsNeighborhoods

-10

-50

510

mea

n o

f ch

ang

e

Change in Monthly Homicide Rates in PSN AreaChange in Monthly Homicide Rates in PSN Area

1st Parolee Forum

1st PSN Prosecution

02

46

810

Hom

icid

e R

ate

per

100,

000

01 Jan 00 01 Jul 01 01 Jan 03 01 Jul 04 01 Jan 06

Figure 3. Change in Quarterly Homicide Figure 3. Change in Quarterly Homicide Rates by Assignment GroupsRates by Assignment Groups

PSN GroupPSN Group

Control GroupControl Group

Rest of CityRest of City

12

34

56

Ho

mic

ide

Rat

e p

er 1

00,0

00

01 Jan 00 01 Jul 01 01 Jan 03 01 Jul 04 01 Jan 06

Summary of FindingsSummary of Findings Find Find negative negative effect on Homicides for:effect on Homicides for:

Forums – largest effect.Forums – largest effect. ProsecutionsProsecutions ATF SeizuresATF Seizures Additive IndexAdditive Index

No significant effect of:No significant effect of: Sentence Length on HomicidesSentence Length on Homicides Any predictor on Agg. Battery and Agg Any predictor on Agg. Battery and Agg

Asslt. Asslt.

Summary of PSN “Treatment” Effects on Quarterly Summary of PSN “Treatment” Effects on Quarterly Neighborhood Homicide RatesNeighborhood Homicide Rates

-1

-40.1

-2.2 -2.7-0.8

-11

-45

-40

-35

-30

-25

-20

-15

-10

-5

0

5

10Control

Variables ForumsGun

Recoveries

Prosecutions

Sentence Length

Combined Effect

**N.B.: Random Effects Poisson Regression

Percentage decrease in Log(Homicide Rate) associated with a one-unit increase in PSN “treatment”

ConclusionsConclusions The observed drop in homicide is larger The observed drop in homicide is larger

in the treatment vs. control area (35 in the treatment vs. control area (35

percent). percent). Find a significant negative relationship Find a significant negative relationship

between various PSN interventions and between various PSN interventions and

homicidehomicide Forums appear to be having the largest Forums appear to be having the largest

effect, effect, butbut there is a clear additive there is a clear additive

effect.effect.

This paper can be found at. . .This paper can be found at. . .

Social Science Research Network:Social Science Research Network:

www.ssrn.comwww.ssrn.com

DirectionDirection Individual Level data:Individual Level data:

Recidivism of those who attend the forum as Recidivism of those who attend the forum as

compared to non-attendees?compared to non-attendees?

The Chicago Gun ProjectThe Chicago Gun Project Offender perceptions of legitimacy/legal Offender perceptions of legitimacy/legal

authority?authority? The role of social networks in:The role of social networks in:

Gun marketsGun markets Patterns of OffendingPatterns of Offending Nexus between community and offender capitalNexus between community and offender capital