INTRODUCTION 821 I. LITERATURE REVIEW 824 A. Criminal Networks and Co-Offending 824 B. Messages in Networks 828 C. Specific Deterrence 833 D. Desistance 836 E. Vicarious Deterrence 837 II. THE SACRAMENTO EXPERIMENT 838 III. METHODS 840 A. Research Settings 840 B. Participants. 840 1. Criminal Networks in Sacramento. 840 2. Prolific Offenders 842 3. Co-offenders 843 C. Random Allocation 845 D. Treatments 845 1. Vicarious Deterrence 847 2. Total Network Effects 847 3. Control conditions 847 E. Measures 848 F. Statistical Procedure 850 G. Statistical Power 852 IV. RESULTS 852 A. Baseline Comparability 852 B. Main Effects 854 V. DISCUSSION 857 A. Theoretical Implications 857 B. Policy Implications 862 C. Future Research 863 D. Additional Study Limitations 864 VI. SUPPLEMENTARY MATERIALS 866 A. Front Side Of Leaflet 866 B. Backside Of Leaflet 867 INTRODUCTION
How are messages communicated in networks of people? When does a message become powerful enough to alter behavior? To what extent does a message conveyed by Person A to Person B influence Person C? The assumption across disciplines, including psychology, economics, sociology and public health policy, has been that emotions and behaviors are contagious, cascading between individuals or groups. (1) Concepts like peer pressure, persuasion, or susceptibility have been used since Aristotle, yet identifying the causal influences of contagion in social networks has remained empirically elusive. In part, singling out the effect of a sole message about behavior from the countless bits of information that are transmitted in social group interactions is difficult to describe, let alone to quantify within research settings. Still, the evidence we do have seems to lay out an avenue for testing these concepts in controlled yet real-life settings. Recent studies published in both Science and Nature have shown how online messaging can alter real-life behavior--for instance voting behavior, consumerism, as well as emotional contagion. (2) These type of studies, however, are rare.
Can social institutions, such as the police, cascade messages to individuals that reduce their criminal propensity in such a manner that the crime control effects also reduce offending levels throughout their criminal networks? This question becomes particularly pertinent when the 'target population' is explicitly reluctant to modify its behavior. Some evidence suggests that text message nudges are effective in the cessation of smoking (3) or as reminders to take medications, (4) but these are generally self-initiated behavioral modifications with some degree of self-motivation. What is not known is whether such messages can become an effective instrument to modify behaviors involuntarily, particularly if the intended recipients are members of criminal groups or gangs that they are unwilling or unable to leave. While influencing members of society to vote has been successful, (5) one might expect the social and personal costs of desisting from crime, quitting drugs, or managing anger to be substantially higher. This research therefore utilized strong and robust messages in order to maximize the psychological impact on the individuals and groups involved, but in settings that are potentially challenging to change.
Measurement of the contagion effect with quantifiable data on human behavior remains difficult when the research settings are outside of the laboratory environment. Recent experimental studies have attempted to measure the causal effect of peer influence under field-controlled settings, (6) but there have only been a handful of these (7) and none in criminology. The lack of randomized controlled trials in this field is interesting by itself because, without proof of causal effect, the underlying foundations of living in groups and the fundamentals of communication require us to assume that messages are both conveyed and change behaviors. As well-reasoned as these assumptions may be, a rigorous scientific exploration of these communication patterns in social networks of offenders is essential to provide evidence upon which informed policy decisions can be made.
To address this challenge, we investigated one particular form of potent messages--proactive engagement by police officers with criminals with an extensive offending history (prolific offenders)--that carried the aim of preventing their delinquent behavior. The police are in a position to convey a critical message to offenders: "stop committing crimes." They can do so by letting the subject know that the police have placed him or her under intensified surveillance. The theory of deterrence predicts that offenders will reduce their criminal activity under these conditions, as it becomes too ''risky" to commit crimes. Some research (reviewed below) on specific deterrence--that is, when the state punishes or threatens to punish a particular offender in order to reduce the likelihood of his or her recidivism--does exist, but has yielded conflicting recommendations. (8) We want to revisit this approach by looking at preventative rather than reactive specific deterrence initiatives. But we explicitly go beyond that and ask: can the deterrence message "travel" to other members of the criminal community? And if so, is it effective enough to deter these co-offenders from committing crimes? More specifically, can this approach be done preventatively as a way of reducing criminality by those who are more likely to commit crimes, because they have committed them in the past but may not have necessarily committed a recent crime?
We conducted a randomized controlled field trial in Sacramento, California, by assigning half of all prolific offenders known to the police into "preventative specific deterrence" and in turn measuring not only the effect of officer-offender interactions on the targeted criminal (Person B) but also the effect of the messages on his co-offenders (Person C) and even other members of the criminal network (Person N). Crucially, the police did not engage with Person C or Person N, but we measured the criminal behavior of Person B, C, and N against the criminal behavior of the control persons.
In order to build our theoretical argument, we combine several intertwined conceptual frameworks. First, we discuss criminal networks more broadly and how this line of inquiry illustrates the relationship between group members, referred to hereafter as "nodes." Directly linked to this is the study of co-offending and criminal groups, which is a rather mature body of literature in the sociology of crime that dates back to the days of the Chicago School; (9) however, we place a greater emphasis on the more recent evidence. Second, we present the concept of dyad and group communications more specifically, and what the literature tells us about the ways in which messages are conveyed in social networks. We then move on to discuss the specific type of messages we are interested in manipulating: "preventative specific deterrence." This type of intervention is meant to increase the perceived likelihood of the arrest of individual offenders as a result of increased sanction threats, but also to incorporate "turning point pathways" as well, as part and parcel of desistance. This then leads us to discuss the idea of vicarious deterrence, as our focus is on the ways in which messages are delivered within social networks.
We then go on to describe our research design; a field-randomized, controlled trial to test these theoretical hypotheses. Directly linked to the theoretical components of our approach, the operationalization of the intervention included both carrots and sticks; a threat of increased surveillance, on the one hand, and a referral to assistance as a way out of crime, on the other. Thus, we tested how a message from a formal social control apparatus (the police) travels informally in criminal networks. We then discuss the results of this experiment and present the implications of the findings for both theory and practice.
CRIMINAL NETWORKS AND CO-OFFENDING
The literature on co-offending--that is, the committing of crime in dyads or larger groups--is one of the most evolved and mature areas of research in criminology. (10) The term "co-offending" was introduced by Professor Reiss (11) not too long ago, yet the scholastic enterprise on committing crimes in groups has existed for some time. (12) Recent empirical evidence suggests co-offending is widespread. (13) Co-offending is a broad term both in terms of group dynamics and the types of criminality that such groups engage in. Previous research includes groups that can vary in size, ranging from two members (14) to many dozens, including gangs, mafias, and organized crime groups (15) often working together in furtherance of their criminal endeavors. Further, while van Mastrigt and Farrington identified co-offending as particularly prevalent in relation to juvenile delinquency, robbery and burglary, most offenders--up to 56% (16) and even 80% (17)--have committed a crime with others at a certain point of their careers. Co-offending can be a life-long partnership or a "one-off' association for a particular job, depending on the social capital involved (18) or the opportunistic nature of the offense. (19) The magnitude and stability of the co-offending network is related to a number of factors, but it has been shown that the amount of crime committed with others is more profound than by solo careerists, with larger groups generating more crimes and more serious harm to society. (20)
It has been assumed across several lines of investigation that what sets co-offending apart from solo offending is a social process of association with other co-offenders or prospective co-offenders. (21) For this (and other) reasons, co-offending can be contextualized as a learning process, coupled with differential association behaviors. (22) This exchange of ideas, or...