The Individual-Level Deterrent Effect of “Call-In” Meetings on Time to Re-Arrest

Abstract

Focused on deterrence popular model to address community-level violence, however little research has examined the individual-level effect of deterrent messaging on subsequent offending. To answer this question, we utilize data on 254 gang- and group-involved probationers and parolees who attended offender “call-in” meetings as part of the Detroit Ceasefire. We employ inverse-probability weighting to construct a counterfactual comparison group from a sample of gang-involved young adults who were not subject to the Ceasefire call-in. We then use a Cox regression to estimate time to re-arrest. We find that individuals who were delivered a deterrent message at a call-in meeting had a longer time to re-arrest compared to a weighted comparison group for up to 3 years following the meeting.

Keywords

focused deterrence gangs gun violence survival analysis

Introduction

A continuing and pressing concern in many communities is how best to address violent crime. Police strategies of the 1980s and 1990s often focused on aggressive order maintenance (so-called “broken windows” policing), increased stop-and-frisk interactions, and preventive patrols (Sousa & Kelling, 2006). An unintended consequence of these strategies was declining citizen trust in the police—especially within minority communities (Taylor, 2006). At the same time, a growing recognition that a small number of individuals and groups are responsible for a disproportionate share of crime indicated the need for new strategies (Papachristos, Wildeman, & Roberto, 2015). Given the limited nature of police resources, focused deterrence emerged as a promising, data-driven strategy to address violent crime by concentrating on the most at-risk, chronic offenders (Sherman, 2007). As a “data-driven” strategy, focused deterrence uses information about known offenders to selectively drive police and community responses. This is intended to communicate the program as a fair and transparent response to crime. In light of promising results in a number of other locales, the city of Detroit began its own focused deterrence program in 2013 modeled after the “Boston Ceasefire” initiative (Braga, Kennedy, Waring, & Piehl, 2001). Detroit Ceasefire was intended to address very high rates of violent crime in two of its most violent precincts on the city’s East and West side. Consistent with the focused deterrence model, Detroit Ceasefire focused police and community resources on a small number of gangs and street groups involved in a disproportionate share of violent crime, with the intended outcome of reduced community-level violent crime.

While most evaluations of focused deterrence strategies like Ceasefire have focused on the community level, relatively few have examined the individual-level effect of focused deterrence. This is an important and overlooked area in the focused deterrence literature. To address this question, we utilized data from the Detroit Ceasefire from 2013 to 2016, examining the outcomes of 254 gang- and group-involved individuals who were subject to a Ceasefire call-in meeting. In the following, we discuss prior research on the focused deterrence model and how it applies to the implementation of Ceasefire in Detroit. We then describe our evaluation design, results, and discuss how these results apply more widely to focused deterrence as both an individual-level and community-level strategy for reducing violence.

Focused Deterrence

Focused deterrence, or “pulling levers,” is an interagency approach popularized by the success of Boston’s Operation Ceasefire in the mid-1990s. This strategy targets gangs and violent street groups in an effort to reduce and prevent violent crime (Kennedy, 1997). The name “pulling levers” derives from the strategy of utilizing any available source of leverage against known offenders, often blending both formal and informal sanctions. The goal of focused deterrence strategies is to alter offender behavior by addressing individual- and community-level factors that facilitate crime and using a “blended strategy” which relies upon the resources and actions of law enforcement, the community, and social services (Braga, Weisburd, & Turchan, 2018). Critical to this goal is the direct and repeated communication of a deterrence-based message to high-risk individuals involved in gangs and violent groups by law enforcement and community officials in the form of a “call-in” meeting or forum (Crandall & Wong, 2012; McDevitt et al., 2006). This message is intended to deter future criminal behavior by altering offenders’ perceptions associated with the cost-benefits of offending. In particular, the deterrence-based message is specifically constructed to enhance offenders’ perceptions of the certainty, severity, and swiftness of sanctions—particularly for violent crime. A specific goal of focused deterrence is to maintain a sense of procedural fairness among offenders, meaning that severity—while one part of the message—is not the only message taken. This is achieved by notifying offenders that law enforcement is closely monitoring them, are aware of their illegal activities, and that continued violence will result in aggressive enforcement and the “pulling” of all available “levers” to hold the group accountable (Kennedy, 1997). Importantly, the manner and context in which the message is delivered couples the focus on deterrence with efforts to enhance offenders’ perceptions of procedural justice and police legitimacy (Braga et al., 2018). This is accomplished through an emphasis on respectful delivery of the message and the notion that the message is fair in the sense of providing clear notice of what will occur if violence continues. An offer of community support and social services, such as employment assistance, housing, and substance abuse treatment, often accompanies this message and is intended as a disincentive to criminal behavior (Braga et al., 2018). Finally, the communication typically includes a community voice describing the impact of serious gun-related violence on individuals, families, and communities.

Since Boston’s Operation Ceasefire, a series of evaluations of focused deterrence strategies implemented across the United States have shown decreases in youth homicide (Braga et al., 2001; Corsaro & McGarrell, 2010; Kennedy, 1997), gun homicide (Corsaro & McGarrell, 2010; Kennedy, 1997; McGarrell, Chermak, Wilson, & Corsaro, 2006; Papachristos, Meares, & Fagan, 2007), and violent crime (Corsaro, Hunt, Kroovand Hipple, & McGarrell, 2012; Papachristos et al., 2007; Saunders, Kilmer, & Ober, 2015). More limited research has examined the impact on violent gang offending, with promising results (Braga, Apel, & Welsh, 2013; Braga, Hureau, & Papachristos, 2013; Corsaro & McGarrell, 2009; Papachristos & Kirk, 2015). Beyond the more common application to group violence, focused deterrence strategies have been creatively applied to eliminate drug markets (Corsaro, Brunson, & McGarrell, 2010, 2013; Corsaro et al., 2012), prevent repeat domestic violence incidents (Sechrist & Weil, 2018), and reduce violence among parolees (Clark-Moorman, Rydberg, & McGarrell, 2019; Wallace, Papachristos, Meares, & Fagan, 2016).

Following the promising research findings in the original Boston Ceasefire initiative (Braga et al., 2001), a series of studies were conducted in other cities utilizing the focused deterrence strategy. The impact of these strategies was first reviewed in a meta-analysis conducted in 2012 (Braga & Weisburd, 2012) and later in 2018 (Braga et al., 2018). These reviews found focused deterrence strategies to be associated with statistically significant crime reductions, ranging from small to moderate, with the most rigorous study designs associated with smaller effect sizes. Despite offering strong support of the effectiveness of focused deterrence strategies, these reviews focused on evaluations of focused deterrence strategies at the community level. However, the primary mechanism through which focused deterrence strategies operate to reduce and prevent violent crime is by altering offenders’ perceptions associated with the cost-benefits of offending. Thus, chronic offenders who are the subjects of focused deterrence strategies should experience observable decreases in the likelihood and severity of their criminal behavior (Uchida et al., 2019).

One of the areas where focused deterrence research has examined individual-level effects is in relation to call-in meetings conducted with individuals, typically with histories of violence and gun offenses, returning to the community from prison. Typically, the data used to select invitees to call-in meetings are based on individuals’ prior criminal history (Wallace et al., 2016), group affiliation (Braga et al., 2001), and centrality in relation to other group/gang members (Wheeler, McLean, Becker, & Worden, 2019). McGarrell, Banks, and Hipple (2003) utilized a matched control-group design to examine the effectiveness of Indianapolis’ re-entry project—part of a broader lever-pulling intervention strategy of the Indianapolis Violence Reduction Partnership (IVRP)—on reducing individual recidivism. Although this was a fairly small-scale study, they found that probationers and parolees who received a deterrence-based message upon re-entry were less likely to reoffend and had greater lengths of time to re-arrest than probationers/parolees of the control group. In a similar study, Wallace and colleagues (2016) evaluated Chicago’s Project Safe Neighborhoods’ (PSN) use of parolee forums (call-ins) to communicate a focused deterrence message to high-risk individuals returning to the community from prison. Utilizing a matched control-group design, they found that parolees who attended a call-in meeting that communicated a deterrence-based message strongly rooted in the principles of procedural justice and legitimacy stayed out of prison for significantly greater lengths of time and were significantly less likely to commit a technical violation, new criminal offense, or serious crime than similar returning offenders who did not attend a call-in meeting (Wallace et al., 2016).¹ Thus, there are at least promising findings of the individual-level effects of the focused deterrence model in the context of re-entry. These studies do not, however, address the issue of individual-level effects in the more common context of gang- or group-involved individuals participating in call-in meetings.

The first examination of individual-level effects was Chermak and McGarrell’s (2004) study of probationers and parolees attending a call-in meeting as part of the Indianapolis IVRP-focused deterrence initiative. Although they did find evidence of significant declines in violence at the community level, they found that attendees of the call-in meetings were no less likely to recidivate compared to a control group of probationers/parolees. Given that call-in attendees were purposely selected based on their high-risk status, however, Chermak and McGarrell (2004) argued that a more ideal evaluation of the effectiveness of the IVRP approach would involve utilizing an experimental design to randomly assign high-risk offenders to receive the IVRP deterrence-based message or standard probation/parole.

In a later evaluation of the IVRP, Chermak (2008) was able to implement an experimental design where probationers were randomly assigned to either attend standard probation or one of two meeting types. One meeting type communicated a deterrence-based message, while the other focused on the negative effects of violence in the community and available services. Chermak (2008) found that the treatment groups did not commit significantly fewer crimes, were not significantly less likely to be convicted of a new offense, and did not experience a significantly greater length of time to conviction failure. However, he identified several problems in the delivery of the program that likely compromised its effectiveness, such as an inability to carry out the deterrent message as delivered, or monitor probationers’ post-call-in. In part, these issues were exacerbated by a large number of probationers managed by a limited number of program staff. In light of these problems, Chermak (2008) argued that future research should consider the importance of program integrity on the effectiveness of focused deterrence strategies and provide a detailed description of what the treatment resembles beyond attendance at call-in meetings.

A subsequent randomized experimental design by Uchida et al. (2019) examined the impact of the Regional Analytics for the Safety of our Residents (RASOR) project on probationer’s time to new arraignment. They identified high-risk offenders from police reports to receive either the treatment or control (Uchida et al., 2019). The treatment group was invited to attend a call-in meeting in which they received a deterrence-based message. They were also assigned a Focused Deterrence Team which consisted of a police detective and social worker. This team helped connect participants with services, developed individual treatment plans, conducted follow-up meetings, and monitored participants’ progress (Uchida et al., 2019). Furthermore, a measurement which captured treatment dosage was calculated based upon the time the team members spent on each individual offender. In comparison, the control group only received standard law enforcement activities. Overall, Uchida et al. (2019) failed to identify a significant difference in time to a new arraignment between groups. However, they found that as the client load increased, case managers were increasingly unable to manage probationers. Over time, later cohorts in the project had worse outcomes compared to earlier cohorts that had substantially more contact with program staff (Uchida et al., 2019). They argued that the success of focused deterrence strategies may require a balance “between the amount of time spent by case managers in monitoring offenders, supporting police activity, or providing incentives and services for avoiding criminal activity” (p. 19).

Thus, in contrast to fairly consistent violence reduction findings of the focused deterrence model at the community level, there is much less research and the findings are much more mixed at the individual level. Consequently, this research presents the findings of a quasi-experimental evaluation of focused deterrence with the unit of analysis being high-risk individuals attending a call-in meeting.

Detroit Ceasefire

At the outset, Detroit Ceasefire was explicitly modeled on the focused deterrence initiative originally developed in Boston Ceasefire (Braga et al., 2001). The Detroit implementation team received training and technical assistance from personnel at the National Network for Safe Communities led by David Kennedy, who was heavily involved in Boston Ceasefire (Kennedy, 2009). The call-in meetings focused on individuals identified as being actively involved in gangs and street groups involved in serious violence. Call-in invitees were primarily identified using existing Detroit Police Department (DPD) data on known group- and gang-involved individuals. Prior to the first call-in, DPD had already conducted a group audit where line officers and gang intelligence officers identified offenders who were known to belong to one or more groups or gangs. These data were shared with the Ceasefire working group and were used to select call-in candidates along with an existing internal gang database managed by DPD. At planning meetings, candidates were selected based on their prior criminal history, their recent behavior in the community, and their interactions with law enforcement. Almost all attendees were under probation or parole supervision and were ordered to attend the call-in. Call-in meetings typically involved 10 to 35 attendees of almost exclusively young men. The structure of the meetings followed that of meetings conducted in Chicago, Cincinnati, and New Orleans, and described in detail in studies of those initiatives (Corsaro & Engel, 2015; Engel, Tillyer, & Corsaro, 2013; Papachristos & Kirk, 2015). The theoretical foundations behind the call-in message draw upon focused deterrence (Kennedy, 2009). That is, the message combines the principle of specific deterrence with a direct delivery to people believed to be at high risk for violence. The goal is to shift the perception of the certainty, severity, and celerity of sanction, particularly related to continued violence. This deterrent message also includes a group accountability theme that attempts to increase the perception of the likelihood of future sanction for violence involving the social network and thereby attempts to utilize the group’s informal social control. The deterrent message is coupled with offers of services that can be considered as efforts to increase social support (Cullen, 1994) and decrease negative strain and susceptibility (Agnew, 2005, 2016). The meetings included the voice of a community member affected by violence. Most commonly this involved the mother of a victim of gun violence and can be considered as an attempt to address techniques of neutralization (Matza, 1964). Additional emphasis is placed on delivering the message in a respectful and fair tone consistent with tenets of procedural justice (Tyler, 2006). The research team’s observations of Detroit Ceasefire call-in meetings suggested that the themes consistent with these theoretical frameworks were consistently included in the call-in meetings. In the following, we discuss the methodology and design of our evaluation of the Detroit Ceasefire’s impact on individual-level outcomes for those who attended a call-in meeting.

Method

For analysis, we relied on data from individuals who were invited and attended a Ceasefire call-in meeting. Between August 29, 2013 and December 1 2016, the Detroit Ceasefire hosted 14 individual call-in meetings of which 254 gang- and group-involved individuals attended. As stated above, individuals invited to these meetings were chosen during meetings of the Ceasefire working group. The intention was to identify high-risk, chronic offenders associated with known gangs and groups involved in gun violence. In nearly all cases, these individuals were already under correctional supervision and, therefore, could be compelled to attend the call-in meeting as a condition of their probation or parole. Table 1 shows demographics for all call-in invitees by call-in meeting. Attendance at call-in meetings varied from 9 to 26—with later call-ins having larger attendance, as the scope of the Ceasefire initiative expanded city-wide. The average proportion of individuals invited to call-ins who actually attended was 84.7%. On average, call-in meetings included about 18 individuals, which is consistent with the program’s focus on individuals associated with gangs or groups. The average age of invitees was 24.6 years with an average of about 5.6 prior arrests (including traffic-related arrests). Of these, invitees had an average of .85 violent arrests, 1.1 property arrests, 1.4 disorder arrest, .93 drug arrests, and .46 weapon arrests.

Table 1.

Ceasefire Call-In Demographics.

Call-in meeting	Attended (n)	Age (years)	Client prior arrest history (M)
Call-in meeting	Attended (n)	Age (years)	All arrests	Violent	Property	Disorder	Drugs	Weapon
August 29, 2013	16	23.88	5.25	1.31	1.19	1.31	0.69	0.38
December 12, 2013	13	23.54	5.69	0.92	0.77	1.69	0.69	0.54
May 20, 2014	9	24.67	4.44	0.44	1.67	1.00	1.11	0.11
August 28, 2014	15	24.53	5.60	0.80	1.00	1.80	0.67	0.53
December 4, 2014	13	26.07	4.71	0.64	1.14	1.00	0.43	0.71
March 25, 2015	16	24.18	5.94	1.12	1.06	1.53	0.94	0.41
August 27, 2015	15	25.31	5.63	0.75	0.88	1.50	0.88	0.44
November 14, 2015	24	23.54	6.33	1.00	1.38	1.46	1.21	0.42
January 26, 2016	19	25.26	6.84	0.74	1.26	2.32	1.21	0.42
May 18, 2016	24	25.00	5.04	0.75	1.13	1.04	0.88	0.58
June 16, 2016	22	24.64	6.50	0.59	1.27	1.32	1.23	0.41
August 25, 2016	20	24.35	5.45	0.85	1.00	1.25	0.90	0.50
September 29, 2016	22	24.50	5.77	1.45	0.82	1.45	0.77	0.45
December 1, 2016	26	25.42	5.54	0.46	1.15	0.96	1.42	0.50

Our primary outcome variable, time to arrest post-call-in, was measured as the number of days to any (non-traffic) arrest, any violent crime arrest, and any weapon crime arrest. For each crime type, we considered a separate survival model. We observed invitees for up to 3 years until their arrest or censoring time. The minimum and maximum number of days individuals were observed until censoring was 395 and 1,095, respectively. Data on invitees’ criminal histories were obtained from the Michigan State Police, which maintains statewide data on criminal arrests and adjudications. We matched individual-level records to this dataset, which contained arrest data as far back as 1947. This allowed us to compute invitees’ criminal history prior to attending a call-in meeting, as well as their involvement with the criminal justice system post-call-in. A noted limitation of these data is that it was constrained to offenses or adjudications which occurred in Michigan.

As previously discussed, an offer of community support and social services (such as employment assistance, housing, and substance abuse treatment) often accompanies the focused deterrence message. The Detroit call-in meetings provided access to services with service providers in attendance at the meeting. Service providers performed follow-up calls with invitees within 24 to 48 hr following the call-in. If invitees were not contacted, follow-up calls were made within 5 to 7 days afterwards. Among call-in participants, 48 (19%) individuals had an intake screening or discussion of needs with service providers or outreach workers and 25 (10%) received some type of service (e.g., counseling, education assistance, legal assistance).²

Quasi-Experimental Design

A noted weakness of observational studies is that measured or unmeasured covariates may be partially or wholly responsible for the difference in outcomes between treated and comparison groups. While random assignment of a treatment substantially reduces these threats to a study’s internal validity, in practice they are often impractical or impossible (Shadish, Cook, & Campbell, 2002). Indeed, as part of the Ceasefire initiative, invitees were specifically chosen to attend meetings based on their gang or group involvement, prior arrest history, and calculated risk to the community. Therefore, simply comparing these call-in invitees to other individuals on probation or parole would likely ignore profound pre-treatment differences. First, we developed a comparison pool of individuals who were not subject to a Ceasefire call-in meeting, but who could have plausibly been targeted. This group consisted of other known gang- and group-involved individuals drawn from DPD’s gang intelligence unit. We selected individuals who were also under correctional supervision at the same time and obtained their arrest records—yielding a comparison group of 335 individuals. Thus, the comparison group was limited to other identified gang- and group-involved individuals who were also under correctional supervision at the same time. We computed prior arrest histories for the comparison group based on the date of the first Ceasefire call-in meeting on August 29, 2013. The demographics of all Ceasefire call-in participants and comparison individuals were Black males.³

The second issue we addressed was the common support assumption. This assumes that for each individual in the treatment group, there exists a similar individual in the comparison group (Imbens & Rubin, 2015). To account for this, we employed a quasi-experimental study design using inverse-probability weighting (IPW) on the propensity score (Cunningham, 2018; Imbens & Rubin, 2015). In contrast to matching methods, IPW utilizes all cases in the comparison and treatment groups and re-weights them such that the pre-treatment differences between the two are conditionally ignorable (Hernán & Robins, 2010). In addition, IPW minimizes some of the limitations noted in the use of propensity score matching, as well as avoiding the discarding of any data (see King & Nielsen, 2019 for more). The process of IPW involves a two-step procedure. First, a logistic regression model is used to estimate the probability of receiving treatment, conditional on observed pre-treatment covariates (Imbens & Rubin, 2015). In our case, we considered the invitees’ age at the call-in and their prior arrest history (for violent, property, disorder, drug, and weapon crimes). After fitting the logistic regression model, we estimated stabilized inverse-probability weights where the propensity score is given as follows:

p (x) = P (T = 1 | X = x)

and the weights for individuals in the treatment group are

w (x) = \frac{P (T = 1)}{p (x)}

and the weights for individuals in the comparison group are

w (x) = \frac{P (T = 1)}{1 - p (x)}

where $P (T = 1)$ is equal to the marginal probability of treatment assignment. Including this in the numerator of the weights produces stabilized weights which have the useful property of preserving the original sample size and reducing the impact of cases with very large or very small propensity scores (Hernán & Robins, 2010).

This method produces weights which assign lower values to individuals in the treatment group with low propensity scores and higher weights to individuals in the comparison group with high propensity scores. If successful, this weighting allows an estimate of a causal effect under conditional ignorability (Hernán & Robins, 2010). To verify whether the weighting procedure effectively reduced covariate imbalance between the treated and comparison individuals, we computed the absolute standardized difference for the weighted and unweighted datasets. Table 2 shows the descriptive statistics for the unweighted comparison group, the comparison group (post-weighting), and the treatment group (post-weighting). Figure 1 shows that all covariates utilized in the estimation of propensity scores were weighted below the 10% threshold—suggesting that the weights successfully reduced covariate imbalance (Austin, 2009).

Table 2.

Treatment and Comparison Group, Post-Weighting.

Variable	Comparison groupN = 335				Treated groupN = 254		SD
	Unweighted		Weighted		Weighted
	M	SD	M	SD	M	SD
Age	25.50	6.16	24.94	4.72	25.38	5.56	0.01
Prior arrest history
Violent	0.42	0.60	0.50	0.53	0.60	0.87	0.05
Property	0.56	0.85	0.80	0.99	0.78	1.11	0.04
Disorder	0.47	0.70	0.62	0.70	0.83	1.43	0.08
Drugs	0.10	0.37	0.20	0.49	0.53	1.13	0.07
Weapon	0.22	0.46	0.26	0.42	0.37	0.59	0.05

Figure 1.

Covariate balance and weighted and unweighted samples.

Survival Model

To estimate the time to re-arrest following a call-in meeting, we used a Cox proportional hazards regression to model 3-year survival curves. A Cox regression is suited to questions of this type as it represents a flexible method to estimate “time-to-event” or “time-to-failure” occurrences (Fox, 2002). In contrast to Kaplan–Meir survival curves, which estimate a non-parametric survival curve across one or more groups, Cox regressions allow for the inclusion of other covariates which are likely to influence the time to failure (Hernán, 2010). In our case, we wished to estimate the time to re-arrest, adjusting for differences in age and prior arrest history. While the use of IPW substantially reduced the impact of pre-treatment covariates, including these variables in the regression model allows for additional adjustment—providing a “doubly-robust” estimate of causal effects (Bang & Robins, 2005). Prior to analysis, we evaluated the Kaplan–Meir curves for the data to verify that the proportional hazards assumption was met (Hernán, 2010). We found that the proportionality assumption held for all variables utilized in the models.

We used the “survival” package in R to implement a Cox proportional hazards regression where the risk (hazard) of arrest $h$ at time $t$ is estimated as a function of an underlying risk $h_{0} (t)$ , an indicator variable $β_{c a l l_{i n}}$ which takes on the value “1” if an individual was subject to a Ceasefire call-in meeting, and “0” if an individual was not. If an individual was not arrested during their observation period, they were considered right-censored. A matrix of pre-treatment covariates $β_{1} x_{1} + \dots + β_{k} x_{k}$ contains data on individuals’ age and the number of prior arrests. Our equation is written as follows:

h (t) = h_{0} (t) \times \exp (β_{c a l l_{i n}} + β_{1} x_{1} + \dots + β_{k} x_{k})

Results

The interpretation of the weighted Cox regression can be viewed as the difference between the observed and counterfactual states: $T = 1$ where all individuals in the sample were exposed to the call-in versus $T = 0$ where all individuals in the sample were not exposed to the call-in (Pezzi, Cavo, Biggeri, Zamagni, & Nanni, 2016), providing estimates of the average treatment effect (ATE). The treatment indicator reflects the estimated difference between survival times in the treatment group and comparison group. Table 3 displays the results of the Cox regression for each of the three outcome cases (all arrests, violent arrests, and weapon arrests).

Table 3.

Cox Proportional Hazards Regression on Time to Re-Arrest, 3-Year Estimates.

Variables	Model outcome
Variables	All arrests	Violent arrest	Weapon arrest
Ceasefire call-in	−0.331* (0.129)	−0.636* (0.248)	−0.182 (0.318)
(log) Age	−1.558*** (0.366)	−1.88** (0.659)	−3.477*** (0.971)
Prior arrest history
Violent	−0.034 (0.085)	0.067 (0.146)	0.044 (0.197)
Property	0.172** (0.05)	0.36*** (0.079)	−0.272 (0.164)
Disorder	0.143* (0.062)	0.088 (0.103)	0.182 (0.156)
Drugs	0.131 (0.079)	0.16 (0.136)	−0.085 (0.229)
Weapon	0.167 (0.109)	0.044 (0.21)	0.284 (0.255)

p < .05. **p < .01. ***p < .001.

The estimates shown in Table 3 are on the log-odds scale, where negative values represent a decreased hazard of arrest and positive values represent an increased hazard of arrest. For dichotomous, non-time-varying treatments, the hazard ratio (HR) reflects the hazard in the treatment group divided by the hazard in the comparison group (Hernán, 2010). This can be compared analogous to incident rates and incident rate ratios (IRR). To remain consistent with prior research, we set our minimum α level for a statistically significant effect to .05 (Lakens et al., 2018).

After adjusting for age and prior arrest histories, we found individuals who attended Ceasefire call-in meetings were less likely to be re-arrested up to 3 years following the meeting. For any (non-traffic) arrest, individuals who attended a Ceasefire call-in meeting had a reduced .331 log-odds risk of re-arrest or a roughly 29% lower risk of arrest (HR: .71, 95% confidence interval [CI] = [.558, .925]). Of particular interest to this evaluation was whether invitees who attended call-in meetings were subsequently less likely to be arrested for violent crimes—given this was the primary outcome of the Ceasefire intervention. Here, we also found that call-in invitees had a decreased .636 log-odds risk of arrest—equating to a 47% reduced risk of arrest (HR: .529, 95% CI: [.325, .861]). For weapons arrests, we found a small, but not statistically significant effect on time to re-arrest (HR: .833; 95% CI: [.447, 1.55]). In all three models, the strongly negative coefficient on age reflects the fact that older individuals were less likely to be re-arrested during the 3-year follow-up period than younger individuals. This likely reflects the age-based differences in offending prevalence and changes across the life course (Warr, 1998).

Figure 2 shows the covariate adjusted estimated survival times for the call-in invitees and the comparison group, with the 95% CIs shaded. These survival curves display the estimated time to re-arrest between the Ceasefire call-in group and the comparison group, holding all other covariates fixed to their average value (Hernán, 2010). The tick marks indicate censoring (that is, when cohorts were no longer observed). These estimates provide a visual indicator of the differences between the call-in invitees and comparison group over time.

Figure 2.

Covariate-adjusted survival curves: time (in days) to re-arrest.

Discussion

As noted at the outset, the focused deterrence model first developed in Boston Ceasefire has developed a strong empirical foundation for violence prevention effects at the community level. This has been witnessed in a large number of cities including Chicago, Cincinnati, Indianapolis, Los Angeles, Lowell, New Orleans, Rochester, Stockton, and others. A recent systematic review of the evaluations of these initiatives finds consistent effects on community levels of violence (Braga et al., 2018). These prior studies have generally relied on aggregate numbers of reported violent crimes, homicides, and non-fatal shootings (Braga et al., 2001; Corsaro et al., 2012). Implicitly, these evaluations have assumed that the dispersal of a deterrence-based message to high-impact offenders would have a subsequent impact on overall crime rates. Given research finding tight networks of individuals disproportionately involved in both offending and victimization, this is a theoretically valid strategy (Papachristos et al., 2015). The related question of the impact of the call-in component of the focused deterrence model on the individuals participating in the meetings has received much less attention with more mixed results (Chermak, 2008; Chermak & McGarrell, 2004; McGarrell et al., 2003; Uchida et al., 2019; Wallace et al., 2016).

The results here suggest that Ceasefire call-in meetings significantly decreased the probability of re-arrest for all crimes, but especially for violent crimes for up to 3 years. Among all arrests and violent crime arrests, individuals who were invited and attended a Ceasefire call-in meeting were about 29% (HR: .71) and 47% (HR: .529) less likely to be re-arrested for those offenses, respectively. However, the call-in meetings did not appear to have any effect on arrests for weapons offenses. The lack of an effect on weapons offense arrests may be partially explained by the fact that call-in attendees were also subject to increased surveillance by police, probation, and parole agents. One of the Ceasefire program components was that when a violent incident occurred following a call-in, particularly if it involved a group that was subject to a call-in meeting, there would be an enforcement response. These responses involved not only the offender himself but also the offender’s wider group of associates and group/gang members. Responses manifested as directed police patrol in the area of the incident and/or the group territory, home checks by joint police–probation–parole teams, and the delivery of custom notifications (individual deterrent messages). A consequence is that call-in participants may have been more likely to experience a proactive enforcement arrest for charges such as carrying a concealed weapon or felon in possession of a firearm.

Putting this research into context, estimates from our study largely agree with a prior evaluation of recidivism among individuals who participated in an offender notification parolee forum as part of Chicago’s PSN parolee forums (Wallace et al., 2016). While our estimates of overall re-arrest were lower in magnitude (HR: .71 vs. .431), our estimates of the effect on violent rearrests were nearly identical (HR: .529 versus .513). Given that so little research has examined the effect of deterrent messaging on offenders, this provides more consistent evidence that these programs do indeed have an individual-level effect. This study’s ability to utilize a comparison group of gang-involved probationers and parolees is a strength that may account for the findings in contrast to the above-mentioned Indianapolis studies (Chermak, 2008; Chermak & McGarrell, 2004). The results are, however, tempered by the findings from the evaluation of RASOR program that utilized a randomized experiment (Uchida et al., 2019). Clearly more research on individual-level effects is warranted.

One potential explanation for these findings is that our observations suggest that Detroit Ceasefire followed the principles of the focused deterrence model, at least after several years of learning from its implementation. The call-in meetings were conducted using a script with what members of the research team perceived as a serious but very respectful delivery. Outreach workers and community members met privately with call-in participants just prior to the actual call-in to establish a one-on-one relationship and let the participants know what to expect. The call-ins included a focused deterrence message delivered by law enforcement and prosecutors, a community voice that often began with the Mayor and included local residents typically including someone harmed by violence (“voice of pain”), and a services message offering social support. The meetings concluded with an invitation to break bread. Outreach workers, who were individuals who had previously been involved in gang life, encouraged participants to meet with them the following day to develop a plan for services and continued to work with those individuals who expressed an interest in services. The focused deterrence model is complex with multiple components. Failure to include these elements may mitigate the individual-level effects observed here, in the Chicago Violence Reduction Strategy (Papachristos & Kirk, 2015) and in the Chicago parolee forums (Wallace et al., 2016). As mentioned previously, one limitation of the current study is that we could not include information on service utilization given the small number of call-in participants who received services and the variability in services received.

This noted complexity of the focused deterrence model also points to one of the limitations of this research. We do not have data from the perspective of the call-in participants—such as their perceptions of the call-in meeting or its deterrent messaging, the perceived fairness of the meeting, or a change in their perceived risk of future victimization or incarceration. We also have little information on the role that social services and community support play. These questions, of course, relate to the theoretical foundations of the focused deterrence model and provide a clear direction for future research in this area. Our findings, combined with consistent research on community-level effects, suggest something of meaning is occurring in the call-in meetings but much more research is needed to understand the theoretically based components of the focused deterrence model.

Another remaining question relates to the gang and group component of the focused deterrence, call-in model. The current study focused on individuals embedded in gangs and violent street groups. Such involvement in gangs and groups increases the risk level for future violence (Matsuda, Melde, Taylor, Freng, & Esbensen, 2013; Melde & Esbensen, 2013). The call-in meetings and the focused deterrence model seek to take advantage of this gang and group structure by using the group as a mechanism to spread the deterrence message and by the threat of holding the entire group accountable. The current research does not address the question of the individual-level impact on high-risk individuals who are not part of a gang or group. An example of such an approach is observed in Madison, Wisconsin where call-in meetings are used to communicate the focused deterrence message to chronic violent offenders who are not necessarily involved in a gang or violent street group (Scott, 2017). Future research should examine the efficacy of the call-in approach upon such chronic violent offenders.

Focused deterrence strategies are a comparatively recent area of scholarly research within the crime control and prevention literature. The current study expands upon this area of evaluative research by examining the individual-level effect of deterrent messaging on recidivism, finding that participation in call-in meetings significantly decreased the probability of re-arrest for all crimes, but especially for violent crimes. While determining whether focused deterrent strategies have their desired individual-level impact on criminal behavior is essential, future research is needed to illuminate why these strategies seem to work (Braga, Hureau, & Papachristos, 2013). Future research should consider additional factors associated with offender decision making including perceptions of legitimacy (e.g., of the deterrent message, the system, services offer), involvement in services, and techniques of neutralization, among others.

Footnotes

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This project was supported by Grant No. 2012-PB-FX-K002 awarded by the Office of Juvenile Justice and Delinquency Prevention. Additional support for the criminal history research of this project was provided by Grant No. 2017-BJ-CX-K020 awarded by the Bureau of Justice Statistics. The Office of Juvenile Justice and Delinquency Prevention and the Bureau of Justice Statistics are components of the Department of Justice’s Office of Justice Programs, which also includes the Bureau of Justice Assistance, the National Institute of Justice, the Office for Victims of Crime, and the Office of Sex Offender Sentencing, Monitoring, Apprehending, Registering, and Tracking. Points of view or opinions in this document are those of the authors and do not necessarily represent the official position or policies of the U.S. Department of Justice.

ORCID iDs

Giovanni Circo

Julie M. Krupa

Notes

Author Biographies

Giovanni Circo is an assistant professor in the School of Criminal Justice at the University of New Haven in West Haven, CT. His research focuses on program evaluation, quantitative methods, and gun violence. His research has appeared in Injury Prevention, Policing: An International Journal, Criminal Justice Policy Review, and the Journal of Experimental Criminology.

Julie M. Krupa is an assistant professor in the School of Criminal Justice at Michigan State University. Her research focuses on communities and crime, juvenile corrections, and the intersection between public and juvenile justice.

Edmund McGarrell is a professor in the School of Criminal Justice at Michigan State University. His research focuses on communities and crime with a specific interest in gun, gang, and drug-market-related crime and the efficacy of prevention and control strategies. His recent publications have appeared in Applied Geography, Criminal Justice & Behavior, Criminal Justice Policy Review, Journal of Crime and Justice, Crime and Delinquency, and European Journal of Criminology.

Alaina DeBiasi is a PhD candidate in Criminal Justice at Michigan State University. Her research interests include crime and place, communities, policing, and violence prevention.

References

Agnew

(2005). Why do criminals offend? A general theory of crime and delinquency. New York, NY: Oxford University Press.

Agnew

(2016). A theory of crime resistance and susceptibility. Criminology, 54, 181-211.

Austin

P. C.

(2009). Balance diagnostics for comparing the distribution of baseline covariates between treatment groups in propensity—Score matched samples. Statistics in Medicine, 28, 3083-3107.

Bang

Robins

J. M.

(2005). Doubly robust estimation in missing data and causal inference models. Biometrics, 61, 962-973.

Braga

A. A.

Apel

Welsh

B. C.

(2013). The spillover effects of focused deterrence on gang violence. Evaluation Review, 37, 314-342.

Braga

A. A.

Hureau

D. M.

Papachristos

A. V.

(2013). Deterring gang-involved gun violence: Measuring the impact of Boston’s Operation Ceasefire on street gang behavior. Journal of Quantitative Criminology, 30, 113-139.

Braga

A. A.

Kennedy

D. M.

Waring

E. J.

Piehl

A. M.

(2001). Problem-oriented policing, deterrence, and youth violence: An evaluation of Boston’s operation ceasefire. Journal of Research in Crime & Delinquency, 38, 195-225.

Braga

A. A.

Weisburd

D. L.

(2012). The effects of focused deterrence strategies on crime: A systematic review and meta-analysis of the empirical evidence. Journal of Research in Crime & Delinquency, 49, 323-358.

Braga

A. A.

Weisburd

D. L.

Turchan

(2018). Focused deterrence strategies and crime control. Criminology & Public Policy, 17, 205-250.

10.

Chermak

(2008). Reducing violent crime and firearms violence: The Indianapolis lever-pulling experiment (United States Department of Justice Final Report). Washington, DC: National Institute of Justice.

11.

Chermak

McGarrell

(2004). Problem-solving approaches to homicide: An evaluation of the Indianapolis violence reduction partnership. Criminal Justice Policy Review, 15, 161-192.

12.

Clark-Moorman

Rydberg

McGarrell

E. F.

(2019). Impact evaluation of a parolee-based focused deterrence program on community-level violence. Criminal Justice Policy Review, 30, 1408-1430.

13.

Corsaro

Brunson

R. K.

McGarrell

E. F.

(2010). Evaluating a policing strategy intended to disrupt an illicit street-level drug market. Evaluation Review, 34, 513-548.

14.

Corsaro

Brunson

R. K.

McGarrell

E. F.

(2013). Problem-oriented policing and open-air drug markets: Examining the Rockford pulling levers deterrence strategy. Crime & Delinquency, 59, 1085-1107.

15.

Corsaro

Engel

R. S.

(2015). Most challenging of contexts assessing the impact of focused deterrence on serious violence in New Orleans. Criminology & Public Policy, 14(3), 1-35.

16.

Corsaro

Hunt

E. D.

Kroovand Hipple

McGarrell

E. F.

(2012). The impact of drug market pulling levers policing on neighborhood violence: An evaluation of the High Point drug market intervention. Criminology & Public Policy, 11, 167-199.

17.

Corsaro

McGarrell

E. F.

(2009). Testing a promising homicide reduction strategy: Re-assessing the impact of the Indianapolis “pulling levers” intervention. Journal of Experimental Criminology, 5, 63-82.

18.

Corsaro

McGarrell

E. F.

(2010). Reducing homicide risk in Indianapolis between 1997 and 2000. Journal of Urban Health: Bulletin of the New York Academy of Medicine, 87, 851-864.

19.

Crandall

Wong

S. L.

(2012). Group violence reduction strategy: Call-in preparation and execution. Washington, DC: U.S. Department of Justice, Office of Community Oriented Policing Strategies.

20.

Cullen

F. T.

(1994). Social support as an organizing concept for criminology: Presidential address to the academy of criminal justice sciences. Justice Quarterly, 11, 527-559.

21.

Cunningham

(2018). Causal Inference: The Mixtape.

22.

Engel

R. S.

Tillyer

M. S.

Corsaro

(2013). Reducing gang violence using focused deterrence: Evaluating the Cincinnati Initiative to Reduce Violence (CIRV). Justice Quarterly, 30, 403-439.

23.

Fox

(2002). Cox proportional-hazards regression for survival data (An R and S-PLUS Companion to Applied Regression). Retrieved from https://socialsciences.mcmaster.ca/jfox/Books/Companion-1E/appendix-cox-regression.pdf

24.

Hernán

M. A.

(2010). The hazards of hazard ratios. Epidemiology, 21, 13-15.

25.

Hernán

M. A.

Robins

J. M.

(2010). Causal inference. Boca Raton, FL: CRC Press.

26.

Imbens

G. W.

Rubin

D. B.

(2015). Causal inference in statistics, social, and biomedical sciences. New York, NY: Cambridge University Press.

27.

Kennedy

D. M.

(1997). Pulling levers: Chronic offenders, high-crime settings, and a theory of prevention. Valparaiso University Law Review, 31, 449-484.

28.

Kennedy

D. M.

(2009). Deterrence and crime prevention: Reconsidering the prospect of sanction. New York, NY: Routledge.

29.

King

Nielsen

(2019). Why propensity scores should not be used for matching. Political Analysis, 27, 435-454.

30.

Lakens

Adolfi

F. G.

Albers

C. J.

Anvari

Apps

M. A.

Argamon

S. E.

. . . Buchanan

E. M.

(2018). Justify your alpha. Nature Human Behaviour, 2, 168-171.

31.

Matsuda

K. N.

Melde

Taylor

T. J.

Freng

Esbensen

F. A.

(2013). Gang membership and adherence to the “code of the street.” Justice Quarterly, 30, 440-468.

32.

Matza

(1964). Delinquency and drift. New York, NY: John Wiley.

33.

McDevitt

Decker

S. H.

Kroovand

N. K.

McGarrell

E. F.

Klofas

Bynum

(2006). Offender notification meetings: A Project Safe Neighborhoods strategic intervention. Washington, DC: U.S. Department of Justice, Office of Justice Programs.

34.

McGarrell

E. F.

Banks

Hipple

(2003). Community meetings as a tool in inmate reentry. Justice Research and Policy, 5(2), 5-32.

35.

McGarrell

E. F.

Chermak

Wilson

Corsaro

(2006). Reducing homicide through a “lever-pulling” strategy. Justice Quarterly, 23, 214-231.

36.

Melde

Esbensen

F. A.

(2013). Gangs and violence: Disentangling the impact of gang membership on the level and nature of offending. Journal of Quantitative Criminology, 29, 143-166.

37.

Papachristos

A. V.

Kirk

D. S.

(2015). Changing the street dynamic: Evaluating Chicago’s group violence reduction strategy. Criminology & Public Policy, 14, 525-558.

38.

Papachristos

A. V.

Meares

T. L.

Fagan

(2007). Attention felons: Evaluating project safe neighborhoods in Chicago. Journal of Empirical Legal Studies, 4, 223-272.

39.

Papachristos

A. V.

Wildeman

Roberto

(2015). Tragic, but not random: The social contagion of nonfatal gunshot injuries. Social Science & Medicine, 125, 139-150.

40.

Pezzi

Cavo

Biggeri

Zamagni

Nanni

(2016). Inverse probability weighting to estimate causal effect of a singular phase in a multiphase randomized clinical trial for multiple myeloma. BMC Medical Research Methodology, 16, Article 150.

41.

Saunders

Kilmer

Ober

A. J.

(2015). A comprehensive evaluation of a drug market intervention training cohort. Santa Monica, CA: RAND Corporation.

42.

Scott

M. S.

(2017). Focused deterrence of high-risk individuals (Strategies for Policing Innovation: Problem-Oriented Guides for Police Response Guide Series No. 13). Retrieved from https://popcenter.asu.edu/sites/default/files/spi_focused_deterrence_pop_guide_final.pdf

43.

Sechrist

S. M.

Weil

J. D.

(2018). Assessing the impact of a focused deterrence strategy to combat intimate partner domestic violence. Violence Against Women, 24, 243-265.

44.

Shadish

Cook

T. D.

Campbell

D. T.

(2002). Experimental and quasi-experimental designs for generalized causal inference (2nd ed.). Boston, MA: Houghton Mifflin.

45.

Sherman

L. W.

(2007). The power few: Experimental criminology and the reduction of harm. Journal of Experimental Criminology, 3, 299-321.

46.

Sousa

Kelling

(2006). Of “broken windows,” criminology, and criminal justice. In Weisburd

Braga

(Eds.), Police Innovation (pp. 77-97). Cambridge, UK: Cambridge University Press.

47.

Taylor

(2006). Incivilities reduction policing, zero tolerance, and the retreat from coproduction: Weak foundations and strong pressures. In Weisburd

Braga

(Eds.), Police Innovation (pp. 98-114). Cambridge, UK: Cambridge University Press.

48.

Tyler

T. R.

(2006). Why people obey the law (2nd ed.). New Haven, CT: Yale University Press.

49.

Uchida

C. D.

Swatt

Schnobrich-Davis

Connor

Shutinya

Wagner

(2019). A randomized control trial of a targeted high-risk offender program across three jurisdictions. Police Quarterly, 22, 192-216.

50.

Wallace

Papachristos

A. V.

Meares

Fagan

(2016). Desistance and legitimacy: The impact of offender notification meetings on recidivism among high risk offenders. Justice Quarterly, 33, 1237-1264.

51.

Warr

(1998). Life course transitions and desistance from crime. Criminology, 36, 183-216.

52.

Wheeler

A. P.

McLean

S. J.

Becker

K. J.

Worden

R. E.

(2019). Choosing representatives to deliver the message in a group violence intervention. Justice Evaluation Journal, 2, 90-117.