The Drift From Convention to Crime

Abstract

This article conceptualizes intermittency in the form of Matza’s drift and assesses the relationship between co-offending and intermittency to determine whether the gap between offenses is influenced by a situation of company. Using the 1958 Philadelphia Birth Cohort, we explore the age/intermittency curve for the entire sample and lifetime co-, solo-, and mixed offenders to determine whether co-offending during the life-course influences intermittency. We devote particular attention to lifetime mixed offenders, who exhibit variation between co-offending and solo-offending, by using survival analysis to predict the risk of re-offending (i.e., time to re-offense) when the immediately prior offense was a co-offense. Findings suggest that lifetime mixed offenders have the shortest average gaps between offenses. Among mixed offenders, an immediately prior co-offense is related to a significantly lower risk of re-offending (longer time between offenses). The results do not support a relationship between a situation of company and persistent offending behavior.

Keywords

criminal careers co-offending intermittency drift theory life-course criminology

Introduction

Developmental/life-course criminology focuses on crime as a career, which implies a beginning, variable duration or length, and an end (Blumstein & Cohen, 1987; Gottfredson & Hirschi, 1990). Within this basic career foundation, criminal careers can be further characterized along other dimensions, such as specialization, amount of time and effort devoted to crime, level of accomplishment, productivity, current direction, overall shape, and time out for other activities (Gottfredson & Hirschi, 1990). While much attention has been devoted in the literature to multiple aspects of the criminal career, including onset, persistence, and desistance, less attention has been focused on the time out for other activities, or the gaps that occur between offenses, where an individual drifts from conventional behavior to unconventional behavior and vice versa (Baker, Falco Metcalfe, & Piquero, 2013; Farrington, 2003; Matza, 1964; Piquero, 2004).

These brief lapses and sporadic episodes of crime that occur at sometimes unpredictable intervals, or intermittency, are characteristic of all criminal careers. According to Farrington (1986), “even a five-year or ten-year crime-free period is no guarantee that offending has terminated” (p. 201). As a result, scholars have noted the importance of intermittency and the need to explore it in more detail (Bushway, Piquero, Broidy, Cauffman, & Mazerolle, 2001; Piquero, Farrington, & Blumstein, 2007). While there generally is continuity in offending, Laub and Sampson (2001) recognize that there does exist a great deal of heterogeneity in criminal behavior. Offenders vary in the time between each of their offenses—some drift between conventional and unconventional behavior rather quickly, while others experience prolonged periods of conventional behavior before resuming unconventional behavior or desisting (Baker, Falco Metcalfe, & Piquero, 2013).

The question remains as to what can explain these gaps that exist between offenses, or what Laub and Sampson (2001) refer to as “off-time” onset, whereby an individual resumes offending after a period of abstinence from offending. Shorter time periods between offenses imply persistent offending, whereas longer time periods between offenses suggest a pattern of desistance, which may eventually lead to termination of offending behavior. Factors that lead to greater gaps between offenses would then presumably increase the likelihood of desistance. Developmental criminology stresses the importance of temporal, within-individual changes in offending, and the recognition of causal factors that predict development and impact the life course of individuals (Loeber & Stouthamer-Loeber, 1996). Considering the importance of intermittency in the criminal career framework and its close ties to persistence and desistance, the causal factors that influence the time between offenses should be explored further (Baker, Falco Metcalfe, & Piquero, 2013; Piquero, 2004).

To this end, it is the purpose of this article to integrate the concept of intermittency into a theoretical framework that can potentially explain these time gaps that occur between offenses. Seeing that Matza (1964) was one of the first to recognize that criminal behavior is sporadic, and that there exists this drift between conventional and unconventional behavior (Luckenbill & Best, 1981), we conceptualize intermittency in terms of Matza’s (1964) drift, since delinquents, according to him, are “casually, intermittently, and transiently immersed in a pattern of illegal action” (p. 28). Matza (1990) also recognizes that while offenders have an affinity to offend, they can be provoked into continued offending through affiliation, or what he calls a situation of company (Matza, 1964), and the learning of delinquent behavior in the context of a delinquent subculture. In this sense, the drift and the affiliation arguments call attention to the effect of peer groups and the implication of group offending, or co-offending, on the criminal careers of offenders, including their abandonment of conventional behavior for unconventional behavior (Matza, 1964, 1990). Following Matza’s (1964, 1990) propositions, we use the 1958 Philadelphia Birth Cohort Study to assess the link between co-offending and intermittency, and the significance of this relationship in understanding criminal careers.

Theoretical Background

Several criminological theories add insight into the predictors of intermittency (Piquero, 2004). The elements of Hirschi’s (1969) social bond—attachment, commitment, involvement, and belief—can be used to explain conforming behavior, and when this bond to society is weakened or broken, criminal behavior should be expected. Sampson and Laub (1993) and Laub and Sampson (2003) extend these elements of the social bond to also explain adult criminal behavior, and how bonds in adulthood, such as marriage, employment, and military service, can serve as turning points in the life course, leading to a gap in offending behavior or even desistance. According to deterrence theory, the gaps in offending can be explained by specific and general deterrence, or objective and subjective deterrence that dissuades future offending behavior, influencing the time between offenses (Paternoster, 2010). Agnew (1992) would suggest that the presence and absence of particular strains explains the time between offenses, and “off-time” onset into subsequent offending. While each of these frameworks has implications for intermittency, we focus on Matza’s (1964) drift theory, in that the concept of drift most closely approximates the idea of intermittent offending behavior.

Matza’s Drift Theory

According to Matza (1964), deviance is something that individuals drift in and out of during periods of the life course (Piquero, 2004). The drift into criminal or unconventional behavior is characteristic of those individuals with an affinity toward delinquent behavior and is worsened by affiliation with offending groups (Matza, 1990), or what Matza (1964) calls a situation of company. The situation of company fosters a mutual misconception that all members of the group are committed to their misdeeds and the tenets of the delinquent subculture (Matza, 1964). As a result, delinquents become adherents to the delinquent subculture, which entails customs that are borderline conventional and criminal at the same time, leading to a state of limbo between convention and crime. These drifters are just as likely to resort to criminal or conventional behavior at any point in time (affinity), but based on the process outlined above, this decision seems to be partially dependent on the influence of co-offenders (a situation of company or affiliation).

While most offending at all stages of the life course can be considered intermittent, intermittent spells, or the drift between unconventional behaviors, can vary in length over the life course (Baker, Falco Metcalfe, & Piquero, 2013). Particularly during adolescence, when offending is situationally motivated by co-offenders, the time to re-offense may be partially explained by the presence of co-offenders. Matza (1964) says that “persistent deviance typically is not a solitary enterprise; rather it best flourishes when it receives group support” (p. 63). If shorter gaps between offenses suggest persistent offending behavior, Matza’s (1964) statement and theoretical propositions would imply that (a) individuals that co-offend have relatively shorter gaps between offenses and (b) co-offending in the immediately prior offense should increase the risk of re-offending, or shorten the time between offenses. Before addressing these implications, we begin by reviewing what is known about the drift between offenses, that is, intermittency, and co-offending across the life course.

Intermittency

While much attention has been devoted in the literature to desistance, Laub and Sampson (2001) actually identify studies of desistance as studies of both persistent and intermittent offending behavior. According to Bushway et al. (2001), desistance can be viewed as a process in which the propensity to offend changes with age. In this process, it is illogical to assume that criminality remains constant and then suddenly drops to zero. Rather, the process toward desistance would appear to be gradual, characterized by increasing time between offenses until the point of termination. Several studies have even noted offenders that follow a particular pattern of intermittent behavior that eventually ends in desistance (Barnett, Blumstein, & Farrington, 1989; Bushway, Thornberry, & Krohn, 2003).

Intermittency, or the time between offenses, is recognized as a temporary abstinence from criminal activity followed by a resumption of criminal activity (Piquero, 2004). While few studies of intermittency exist, the literature has drawn several conclusions about intermittency within offending careers, usually based on analyses of cohort data. According to these studies, frequent or chronic offenders, typically identified as offenders with five or more offenses, experience gaps in offending that are characterized by relatively short time intervals (Baker, Falco Metcalfe, & Piquero, 2013; Barnett et al., 1989; Piquero et al., 2007). The average time between offenses continues to decrease with an increasing number of criminal events, meaning the first transition has the longest time between offenses, whereas the fifth or sixth transition has a shorter time between events (Piquero et al., 2007; Raskin, 1987). Offenders who are chronic with an early age of onset also have significantly shorter gaps between offenses (Baker, Falco Metcalfe, & Piquero, 2013).

In addition to frequency of offending, there exists a relationship between age and intermittency, in that the gaps between offenses grow as individuals age (Baker, Falco Metcalfe, & Piquero, 2013). Individuals that offend less frequently (have four or fewer offenses) tend to age into longer gaps between offenses earlier in the life course than chronic offenders (Baker, Falco Metcalfe, & Piquero, 2013). While age and offense frequency serve as significant predictors of the time between offenses, intermittency has its own effect on offense seriousness as well. Among chronic offenders, shorter time periods between offenses result in more serious offending behavior, although the same is not true for less frequent offenders (Baker, Falco Metcalfe, & Piquero, 2013).

In addition to time between offenses, scholars have analyzed intermittency in the form of change in offending and as a parameter within models to control for periods of activity or inactivity. When examining change in offending behavior, Kempf (1989) found that delinquency dropouts were less likely to be adult offenders than individuals with more stable delinquent careers. Horney, Osgood, and Marshall (1995) computed the odds of starting and stopping criminal behavior and found that changes in offending depended on changes in local life circumstances. Alternatively, Nagin and Land (1993) and D’Unger, Land, McCall, and Nagin (1998) incorporated an intermittency parameter to replace the typical controls for onset and desistance in order to capture “discontinuous jumps from a state of zero criminal potential to positive criminal potential” Nagin and Land (1993) recognized that the models incorporating the intermittency parameter fit the data better, and the intermittency parameter added greater explanatory power to their models. Whether operationalizing intermittency as the time between offenses, change in offending, or a parameter in larger models, these findings demonstrate the importance of accounting for intermittency when studying criminal careers.

Co-Offending

Like intermittency, co-offending has been identified as an important aspect of the criminal career, but studies of co-offending are not as common as studies of other facets of the criminal career (Farrington, 2003; Hochstetler, 2001). In 1931, Shaw and McKay discovered that more than 80% of juveniles appearing before the Chicago Juvenile Court had accomplices. While early studies following this finding focused primarily on peers and their effect on criminal involvement (see Matsueda, 1988; Matsueda & Anderson, 1998; Warr, 1996), Reiss (1988) brought co-offending into the forefront as a potentially critical factor related to onset, persistence, and desistance from offending.

Citing Van Mastrigt and Farrington (2009) and McGloin, Sullivan, Piquero, and Bacon (2008), McGloin and Povitsky Stickle (2011) note that most analyses of co-offending “have been cross-sectional and aggregate; focused discussions about the meaning of co-offending for the crime career, as opposed to viewing it simply as an attribute of the crime event, are rare” (p. 422). In other words, researchers and theorists rarely integrate group offending into discussions about continuity and change over the life course (McGloin & Povitsky Stickle, 2011), since most analyses are cross-sectional and aggregate (McGloin et al., 2008; Van Mastrigt & Farrington, 2009). Of the studies that do exist, there has been a general consensus among several facets of co-offending within the criminal career.

Co-offending has been recognized as an adolescent phenomenon, occurring less often in adulthood. Generally, the majority of offenders follow a co-offending trajectory similar to the age/crime curve, where they have accomplices until the early 20s or the mid to late teens, after which the majority of offenses are committed alone (Hood & Sparks, 1970; Reiss, 1988; Van Mastrigt & Farrington, 2009; Warr, 1993). In this sense, offenders tend to age out of co-offending. This change within individual criminal careers seems to reflect a switch from co-offending to solo-offending, and not necessarily the persistence of solo-offending and/or the dropping out of co-offenders (McGloin et al., 2008; Reiss & Farrington, 1991). It is for this reason that adolescence-limited and life-course persistent offenders may be difficult to distinguish in adolescence, considering that chronic offenders are just as likely, or maybe even more likely, to engage in group offending during this time (McGloin & Povitsky Stickle, 2011; Piquero, 2004).

Most criminal careers are characterized by a mix of offending alone and with accomplices, but there are some individuals who always offend alone or always offend with others (Reiss, 1988; Reiss & Farrington, 1991; Warr, 2009). Researchers have noted that participation and interaction with co-offenders may moderate fear of punishment and increase chances of offending (Cloward & Ohlin, 1960; Eckland-Olson, Lieb, & Zurcher, 1984; Erikson & Jensen, 1977; Hochstetler, 2001; Short & Strodtbeck, 1965; Shover & Henderson, 1995). Individuals commit acts when they are with others that they would not have committed had they been alone (Warr, 2009). For this reason, studies have found that co-offenders are more likely to recidivate, or persist in offending (Piquero et al., 2007; Reiss and Farrington, 1991), and those who engage in illegal behavior in groups are likely to engage in this behavior more frequently than those offending alone (Hindelang, 1976; Reiss, 1988).

Within the criminal career literature, co-offending has also been used to explain initiation and changes in offending behavior. In general, offenders with an early age of onset have more co-offenders (Piquero et al., 2007), suggesting again that co-offending is just as likely to occur among chronic or persistent offenders, who are generally characterized by careers with early ages of onset. Among those individuals who do commit their first offense alone, they are likely to commit fewer offenses on average than individuals who commit their first offense with others (Reiss & Farrington, 1991). However, co-offending is most common at onset, with higher average numbers of co-offenders at early conviction numbers, but as the conviction number increases, the average number of co-offenders decreases (Carrington, 2009; Piquero et al., 2007). Together, these studies suggest that the group nature of offending may increase the ability of research to test the degree to which prior behavior affects future behavior, as well as explain continuity and change in offending over the life course (McGloin & Povitsky Stickle, 2011; Zimring, 1981).

Current Study

Using Matza’s (1964) drift theory, the current study connects intermittency and co-offending as concepts that vary with each other over the life course. Specifically, we focus on the meaning of co-offending for the criminal career in the context of Matza’s (1964) propositions relating a situation of company to the drift from periods of conventional behavior to unconventional behavior. Since Matza (1964) suggests persistent offending is characteristic of those who offend in groups, we assess (a) whether individuals that co-offend have relatively shorter gaps between their offenses than those who do not co-offend and (b) whether co-offending in the immediately prior offense affects the time to re-offense, that is, the drift between one unconventional behavior to the next, over the life course. Essentially, we address three main research questions:

Research Question 1: Do offenders that co-offend have shorter gaps between their offenses throughout the life course?

Research Question 2: For lifetime mixed offenders (offenders that co-offend and solo-offend throughout the life course), what effect does co-offending in the immediately prior offense have on the risk of re-offending?

Research Question 3: Does the risk of re-offending change depending on whether the re-offense is a solo-offense versus a co-offense?

Data

To explore the relationship between co-offending and intermittency, we use the 1958 Philadelphia Birth Cohort Study (Tracy & Kempf-Leonard, 1996; Tracy, Wolfgang, & Figlio, 1990). This data set contains various measures for individuals born in Philadelphia in 1958 to the age of 26. Because 7 years old is generally considered to be the minimum age at which individuals may be held accountable for their actions, we limit the study to offenses committed after the age of 7 (Baker, Falco Metcalfe, & Piquero, 2013; Bernard, 1992, 2006; Farrington & Welsh, 2007; Tanenhaus, 2004; Zimring, 2005). Police contact data before the age of 18 was taken from “rap sheets” developed by the Philadelphia police that include records of “remedial, or informal, handling of the youth by an officer” (Tracy & Kempf-Leonard, 1996, p. 65). Once the cohort reached age 18, “court files included police reports, so data on adult crime are comparable to that for delinquency,” but there were no remedial reports for adults (Tracy & Kempf-Leonard, 1996, p. 65). Therefore, the data is comprised of police reports of contacts with youth, as well as actual arrest data, constituting a “much fuller record of delinquency than data based solely on arrest information” (Kempf-Leonard, Tracy, & Howell, 2006, p. 458). The primary data are police record data, and the largest number of offenses is at the police contact level (Zimring, Jennings, Piquero, & Hays, 2009). Although not all contacts with the police resulted in arrest, contact by an officer is presumed to be made due to offending behavior by an individual, and therefore this contact is relevant in the context of both co-offending and intermittency, since neither of these concepts assumes arrest and adjudication. The individual measures and police contact/arrest information combine to form a nested data set of offenses within individuals.

We recognize that the 1958 Philadelphia Birth Cohort has several limitations. Some of these limitations include the use of official statistics, age of the data set, and inability to perfectly account for street time. We also know that we are unable to measure certain concepts possibly relevant to Matza’s (1964) drift theory, including time spent with peers and peer commitment, which are concepts typically operationalized within self-report data. Despite these limitations, we feel that the use of this data set for the research questions presented has more benefits than drawbacks, and is important for the progress of research in the area of developmental/life-course criminology (Farrington, 2005). This particular data set provides the month and year of each offense reported to the police for 26 years of a person’s life span, meaning that intermittency can actually be operationalized over 26 years using this data set, which, to our knowledge, is not available with any other data set using either self-report or official statistics. In addition, while we cannot capture some of the peer concepts delineated by Matza (1964), we can measure co-offending and the extent to which offending in groups influences the time that occurs between offenses, which is Matza’s (1964) situation of company and drift. Within these limitations, we shed some light on the relevance of intermittency within criminal careers, and in doing so, hope that future data will allow us to explore this issue further using self-reported crime and delinquency that includes better measures of peer involvement and other life events.

Unit of Analysis

The data is multilevel in that individual offenses are clustered within individuals’ criminal careers. Therefore, offenses are the unit of analysis. Outlined below are the offense-specific and offender-specific variables used in the models presented.

Dependent Variable

The dependent variable for this study is intermittency, operationalized as the months between offenses, or the months to re-offense, given that the data provide the month and year of each offense. For intermittency to be measured, there must be more than one offense; thus, individuals that committed only a single offense were excluded from the study. Similarly, the first contact with the police (t₁) is not included in measures of intermittency, as there is no gap between offense zero and offense one that is meaningful within the conceptualization of intermittency.¹

Independent Variables

Co-offending Variables

Two separate co-offending measures are created for each offense to account for time-variant and time-invariant aspects of co-offending. The first measure, designated Immediately Prior Offense Co-Offense, is a dichotomous variable indicating whether the immediately prior offense was a co-offense (1) or a solo-offense (0). This measure is designed to determine whether immediately prior co-offending affects the time to re-offense (Research Question 2). The second measure of co-offending, identified as First Offense Co-Offense, is dichotomized based on whether the first offense of an individual’s criminal career was a co-offense (1) or a solo-offense (0). Reiss and Farrington (1991) found that individuals who commit their first offense alone are likely to commit fewer offenses on average than those who commit their first offense with others. In other words, when an individual’s first offense is a co-offense, he or she should commit more offenses on average. Therefore, whether the first offense was a co-offense or solo-offense may be associated with the time between an individual’s offenses.

Control Variables

In an effort to control for other aspects of the criminal career that are related to co-offending and may affect intermittency, we include various criminal career and demographic measures as potential confounders. Following prior research (Baker, Falco Metcalfe, & Piquero, 2013; Patterson, Crosby, & Vuchinich, 1992; Piquero et al., 2007; Tibbetts & Piquero, 1999), Age of Onset is a dichotomous variable coded early onset (1) for offenders with an onset age before 14 years, and late onset (0) for offenders with an onset age of 14 years or older. A dichotomous measure of onset age is chosen, as opposed to a continuous measure, to address the potential confounding effects of age and age of onset suggested by prior research (Piquero, Paternoster, Mazerolle, Brame, & Dean, 1999). Chronicity of Offending represents a count of the number of offenses committed by an individual over the life course as a proxy for an individual’s propensity to offend. Finally, Violent Offense is a dichotomous measure indicating whether the offense was violent (1) or nonviolent (0), since the co-offending literature suggests that nonviolent offenses may be more conducive to co-offending behavior (Carrington, 2009; Stolzenberg & D’Alessio, 2008; Warr, 2009). Offenses were operationalized as violent if they were classified by the police as homicide, rape, robbery, assault, weapons possession, or domestic violence, and all other offenses were operationalized as nonviolent.

As for demographic information of the individuals in the study, Age is a continuous variable measured in years, and reflects the age at which a particular individual committed his or her offenses. Socioeconomic status (SES) is a dichotomous measure based on a 10-item scale obtained from census tract data, where individuals are subsequently designated as high SES (1) or low SES (0; see Tracy, 1981 and Kempf, 1983, for more details). Race and Sex are dichotomized as White (1) and non-White (0), and male (1), and female (0).

Analysis

To assess the effect of co-offending in the immediately prior offense on the time to re-offense, we use survival analysis to predict the risk or hazard of re-offending. The variance-correction model for repeated events, known as the conditional risk-set model (Prentice, Williams, & Peterson, 1981) or the “gap-time” model (Box-Steffensmeier & Zorn, 2002), examines the risk of covariates for the occurrence of repeated events. This method can be “most easily thought of as generalizations of survival data techniques in which the hazard function modeling is continued beyond a subject’s first failure to second and subsequent failures” (Prentice et al., 1981, p. 373). This modeling strategy takes into account that individuals cannot commit a third offense without having committed at least a second offense. Because of this, the model is stratified by Event Rank, or in this context, the ranked order of an offense in an individuals’ criminal career (second offense, third offense, etc.). Essentially, this model examines the intermittency of an event by determining the risk or hazard of identified covariates on the next occurrence of an event. For purposes of this study, the gap-time model allows us to predict if an immediately prior co-offense, when controlling for the other covariates, increases or decreases the risk for re-offending (i.e., shortens or lengthens the time between offenses, respectively).

Reiss (1988) noted that the majority of offenders are mixed offenders, offending alone and with others, while there are some individuals who only co-offend and individuals who only solo-offend. Because of this, we parcel the sample into three types of offenders: lifetime solo-offenders (about 20% of the sample), lifetime co-offenders (about 13% of the sample), and lifetime mixed offenders (about 67% of the sample; Reiss, 1988; Reiss & Farrington, 1991). To address the first research question, we describe each offender group and explore the age/intermittency curve for the entire sample and for each group of offenders.

As there is no variation in co-offending among lifetime solo- and lifetime co-offenders, we then conduct all multivariate analyses using the group of mixed offenders, who have variation in solo- and co-offending. To address our second research question, we determine whether co-offending in the immediately prior offense shortens or lengthens the time to re-offense. Finally, to answer our third research question, we further disaggregate lifetime mixed offenders based on whether the offense committed is a solo- or co-offense. The solo-/co-offending disaggregated models are designed to determine whether an immediately prior co-offense results in a higher or lower hazard (shorter or longer gap between offenses) when the re-offense is a co-offense versus a solo-offense.

Results

The descriptive statistics broken down by offender group are presented in Table 1. Generally, the average offense in all three groups was committed by a non-White male between the ages of 17 and 18 from a low SES background. Lifetime solo-offenders have the highest proportion of offenses committed by females (43%) and lifetime mixed offenders have the lowest (9%). Lifetime mixed offenders are also more chronic offenders than lifetime solo- and lifetime co-offenders, averaging approximately 11 offenses compared to 4 and 3 offenses, respectively. In addition, and perhaps most importantly, lifetime mixed offenders have a shorter average intermittency (13 months) than lifetime solo- (19 months) and lifetime co-offenders (19 months).

Table 1:

Descriptive Statistics by Criminal Career Offending Pattern

	Lifetime Solo Offenders (1,588 Offenses)		Lifetime Co-Offenders (879 Offenses)		Lifetime Mixed Offenders (13,302 Offenses)
	M	SD	M	SD	M	SD
Sex (Male = 1)	0.57	0.49	0.87	0.34	0.91	0.29
Race (White = 1)	0.26	0.44	0.34	0.47	0.25	0.43
Age	18.17	3.84	16.90	2.77	17.51	3.50
SES (High = 1)	0.32	0.47	0.36	0.48	0.28	0.45
Early onset	0.24	0.43	0.24	0.43	0.47	0.50
Chronicity of offending	4.09	2.47	3.35	1.72	10.67	7.72
Violent offense (Violent = 1)	0.22	0.42	0.25	0.43	0.30	0.46
Immediately prior offense co-offense					0.47	0.50
First offense co-offense					0.56	0.50
Intermittency	19.22	27.47	18.73	24.05	12.98	19.83

Note. SES = socioeconomic status.

Figure 1 shows the age/intermittency curve for the entire sample, lifetime mixed offenders, lifetime solo-offenders, and lifetime co-offenders. Lifetime mixed offenders have the shortest average gaps in offending across the life course, meaning they have a higher risk of re-offending than the other groups of offenders. Lifetime solo- and co-offenders have sporadic offending patterns until about age 12, at which time both types of offenders have longer average gaps between offenses than lifetime mixed offenders throughout the remainder of the life course. Generally, lifetime co-offenders have the longest average gaps between offenses, followed by lifetime solo-offenders, while lifetime mixed offenders have the shortest average gaps between offenses. However, lifetime co-offenders do experience slightly smaller average gaps in offending than lifetime solo-offenders at ages 16 to 17. It appears that offenders who both co-offend and solo-offend have a higher risk of re-offending, or shorter time between their offenses, and this risk is slightly greater in adolescence, considering that all offender groups appear to age into longer gaps between offenses (see also Baker, Falco Metcalfe, & Piquero, 2013).

Figure 1:

Average Intermittency by Age

The results of the conditional risk-set model for lifetime mixed offenders are presented in Table 2. The findings indicate that Age (HR = .809) and Early Onset (HR = .499) significantly reduce the hazard of re-offending.² If the offense is Violent (HR = .932) or the immediately prior offense was a co-offense (HR = .793) there is also a significantly lower hazard of re-offending (increased time between offenses).³ Essentially, for those that co-offend and solo-offend, co-offending in the immediately prior offense does not increase the risk of re-offending, that is, shorten the time to re-offense. Alternatively, when the first offense is a co-offense (HR = 1.056), the risk of re-offending is significantly increased (there is a shorter time between offenses). Chronicity of Offending is also positively and significantly associated with re-offending.⁴

Table 2:

Hazard of Re-Offending Among Lifetime Mixed Offenders

	HR	SE
Sex (Male = 1)	1.006	0.027
Race (White = 1)	0.966	0.021
Age	0.809***	0.005
SES (High = 1)	1.068**	0.022
Early onset	0.499***	0.010
Chronicity of offending	1.029***	0.003
Violent offense (Violent = 1)	0.932***	0.019
Prior offense co-offense	0.793***	0.016
First offense co-offense	1.056**	0.021
Log-pseudolikelihood	−72,872.092
Wald χ²	2,163.850***
Stratified by event rank

Note. SES = socioeconomic status. Stratified by event rank.

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

Addressing our final research question, we disaggregate by solo- and co-offenses. Results of this disaggregated model are presented in Table 3. The findings are generally consistent with the results in Table 2. Co-offending in the immediately prior offense increases the time to re-offense (lowers the hazard of re-offending) for solo- and co-offenses (HR = 0.755 and HR = 0.821, respectively). Unlike the previous model, if the individual’s first offense is a co-offense, there is a significantly lower hazard of a re-offense that is a solo-offense (HR = 0.930), but a significantly higher hazard of a re-offense that is a co-offense (HR = 1.223).⁵

Table 3:

Predicting the Hazard of Re-Offending by Offense Type for Lifetime Mixed Offenders

	Solo Offense (7,191 Offenses)		Co-Offense (6,111 Offenses)
	HR	SE	HR	SE
Sex (Male = 1)	0.978	0.038	1.082	0.047
Race (White = 1)	0.987	0.030	0.962	0.031
Age	0.811***	0.006	0.805***	0.006
SES (High = 1)	1.070*	0.031	1.067*	0.032
Early onset	0.529***	0.015	0.459***	0.014
Chronicity of offending	1.025***	0.004	1.030***	0.003
Violent offense (Violent = 1)	0.923**	0.025	0.941*	0.027
Immediately prior offense co-offense	0.755***	0.020	0.821***	0.024
First offense co-offense	0.930**	0.025	1.223***	0.035
Log-pseudolikelihood	−34,505.868		−29,410.336
Wald χ²	1,260.920***		1,320.110***
Stratified by event rank

Note. SES = socioeconomic status. Stratified by event rank.

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

Discussion and Conclusion

This study uses Matza’s drift theory to link co-offending with intermittency. According to Matza (1964), offenders drift between conventional and unconventional behavior. This drift is influenced by a situation of company, suggesting that offending within groups, or co-offending, impacts drift and persistent offending behavior. We characterize drift in terms of intermittency, since it represents a downtime period following criminal behavior before an individual drifts back into unconventional behavior or desists. Since Matza (1964) emphasizes the importance of a situation of company in affecting the tendency to drift, we determine whether co-offending has an impact on the time to re-offense, with shorter gaps suggesting persistent offending behavior and longer gaps suggesting the possibility of eventual desistance.

In response to our research questions, we can draw three important conclusions. First, we found that lifetime mixed offenders, who both co- and solo-offend, have shorter gaps between offenses. These gaps were shorter at younger ages for all offenders, not just among those who co-offend, although the average time between offenses was shorter for lifetime co-offenders and lifetime mixed offenders at ages 16 and 17. Second, among the lifetime mixed offenders, co-offending in the immediately prior offense decreased the risk of re-offending, or lengthened the time to re-offense. Third, the risk of re-offending does not change depending on whether the re-offense is a solo-offense versus a co-offense. Co-offending in the immediately prior offense lengthens the time to re-offense for all offenses. Together these findings do not lend support for Matza’s claim that persistent offending is linked to group offending.

As Reiss (1988) notes in his work on co-offending, the majority of offenders in our sample are mixed offenders, offending alone and with others. At face value, these offenders exhibit what Matza (1990) would call affinity and affiliation. They have an affinity to offend in that they are committed to offending, whether it is on their own or with others. At the same time, it could be argued that affiliation with others influences their offending behavior, since just under half of their offenses are co-offenses. However, while it is obvious that affiliation with others is common, it does not influence persistent offending behavior. The situation of company actually delays the time to re-offense, which is contrary to Matza’s (1964) prediction, especially considering that longer gaps between offenses can provide evidence toward desistance. In essence, there seems to be more evidence to demonstrate a lack of commitment to misdeeds among those who offend in groups than a negative impact of offending with others (Matza, 1964).

Lifetime mixed offenders seem to most closely resemble chronic offenders. The descriptive statistics demonstrate that mixed offenders commit more offenses on average than lifetime solo- and co-offenders combined. In comparison to the other offenders, they also have a shorter average intermittency, a greater number of them onset early, and more of their offenses are violent in comparison to the other groups. From these statistics, it is apparent that the most chronic offenders, and maybe what Moffitt (1993) would call life-course persistent offenders, are included within this group of mixed offenders. These type of offenders are typically characterized by (a) a willingness to solo-offend or a tendency to transition into solo-offending (McGloin et al., 2008; Moffitt, 1993; Reiss & Farrington, 1991) and (b) shorter gaps between offenses (Baker, Falco Metcalfe, & Piquero, 2013). For these reasons, the results are not very surprising. The link between solo-offending and chronicity may explain why solo-offending in the prior offense is related to a higher risk of re-offending, that is, shorter gaps between offenses.

While we do not find strong support for an increased risk of re-offending when the immediately prior offense is a co-offense, we do find some support that affiliation in the first offense can have an impact on persistent offending behavior. According to the results, the drift between unconventional behaviors is shorter among co-offenses when the first offense is a co-offense, and longer among solo-offenses when the first offense is a co-offense. It appears that affiliation with others in the first offense can lead to shorter gaps between offenses, suggesting persistent offending behavior. It can also be argued that individuals who co-offend in their first offense have a greater affinity to offend. For instance, Reiss and Farrington (1991) found that individuals who commit their first offense alone are likely to commit fewer offenses on average than individuals who commit their first offense with others.

While these findings have interesting implications for the impact of co-offending on criminal careers, there are some limitations to the study and directions to consider for future research. As noted above, we acknowledge that the most important limitation of our study is the data set used. Yet, at the same time, it is one of the most valuable aspects of our study, considering that it allows us to measure intermittency. The 1958 Philadelphia Birth Cohort Study (Tracy & Kempf-Leonard, 1996; Tracy et al., 1990) is composed of official statistics on offenses committed over a 26-year time-span in Philadelphia, and the presence of co-offenders is only known for a subset of these offenses. Because of this, the data only covers those offenses known to the police and known to have co-offenders. Self-report data may cover more offending instances and prevent some of these inaccuracies, but unfortunately, most self-report data do not contain dates that offenses occur for each individual, which is needed to calculate intermittency, along with information about co-offending. To our knowledge, the 1958 Philadelphia Birth Cohort Study (Tracy et al., 1990; Tracy & Kempf-Leonard, 1996) is the only longitudinal data set that provides this information for at least a subset of offenses.

Due to this limitation, it is important to note that the use of official statistics can confound the relationship between co-offending and intermittency. As an anonymous reviewer noted, time between police-recorded events may have more to do with the contingencies of police detection and recording than the offender’s criminal behavior. These detection and recording issues may then be correlated with group offending. For this reason, there is a possibility that co-offending could lead to shorter gaps between offenses. Replication of this study with appropriate self-report data could add further insight into this possibility.

Other lines of research should also be considered in the context of co-offending and intermittency. As suggested by McGloin and Povitsky Stickle (2011), there should be research differentiating the impact of co-offending versus the acquisition of delinquent peers. Co-offenses are often committed by groups of delinquent peers, so knowledge of these peer networks would contribute to a better understanding of criminal careers. Particularly, not every offender within a delinquent peer group offends, and one offender is typically the recruiter, or the most active offender (Reiss, 1988; Reiss & Farrington, 1991; Velarde, 1978). In addition, peer networks can be evaluated based on the frequency, duration, priority, and intensity of involvement with delinquent peers (Warr, 1993, 2009). These aspects of delinquent peer groups and the change of peer contexts across the life course may further contribute to the drift between criminal and conventional behavior and the resulting time between offenses. Criminal career research based on the duration argument, and some focusing specifically on stability of co-offending networks, has found an impact on versatility of offending (Baker, Falco Metcalfe, & Jennings, 2013; McGloin & Piquero, 2010). Unfortunately, data capturing these intricate aspects of delinquent peer networks is rare, making it difficult to operationalize each of these concepts.

An additional line of research, aside from peer networks, would be to explore whether the relationship between co-offending and intermittency varies depending on whether the offender is adolescence-limited or life-course persistent. While controlling for age of onset, chronicity, and event rank account for the differences between these two types of offenders, disaggregating by these groups may produce some interesting findings. Moffitt’s (1993) work does seem to align with Matza (1964) in that adolescence-limited offenders are expected to engage in delinquent behavior based on a social mimicry of others, while life-course persistent offenders should be more willing to solo-offend and continue their offending into adulthood. Because of this, co-offending may have more of an impact on the offending behavior of adolescence-limited offenders than the offending behavior of life-course persistent offenders. Future research may consider trajectory modeling to create these offender groups (Nagin, 2005), and then assess the impact of co-offending on intermittency for each group.

Focusing on the importance of understanding intermittency within criminal careers, we used Matza’s drift theory to explain the gaps between offenses. Applying Matza’s (1964) situation of company and drift, we linked co-offending to intermittency. We failed to find evidence demonstrating an impact of affiliation on persistent offending behavior. Our findings lend credence to the importance of understanding mixed offenders, those who both co- and solo-offend, and what factors influence their time to re-offense.

Footnotes

Acknowledgements

We would like to thank all those who reviewed prior versions of this manuscript, and the anonymous reviewers for their helpful comments.

Notes

Christi Falco Metcalfe is a PhD candidate in the College of Criminology and Criminal Justice at The Florida State University. Her current research interests include courts, the organizational behavior of criminal justice systems, life-course criminology, and punitive attitudes.

Thomas Baker is an assistant professor in the Criminal Justice Program in the L. Douglas Wilder School of Government and Public Affairs at Virginia Commonwealth University. His research interests include perceptions of crime and the criminal justice system, criminal careers, and quantitative research methods. His work has appeared in various outlets including the Journal of Criminal Justice, Crime & Delinquency, the Journal of Quantitative Criminology, and Criminal Justice and Behavior.

References

Agnew

(1992). Foundation for a general strain theory of crime and delinquency. Criminology, 30, 47-87.

Baker

Falco Metcalfe

Jennings

W. G.

(2013). What are the odds? Predicting specialization in offending over the life-course. Criminal Justice and Behavior, 40, 911-934.

Baker

Falco Metcalfe

Piquero

(2013). Measuring the intermittency of criminal careers. Crime & Delinquency. Advance online publication. doi:10.1177/0011128712466382

Barnett

Blumstein

Farrington

(1989). A prospective test of a criminal career model. Criminology, 27, 373-388.

Bernard

T. J.

(1992). The cycle of juvenile justice. New York, NY: Oxford University Press.

Bernard

T. J.

(2006). Serious delinquency: An anthology. Los Angeles, CA: Roxbury Publishing Company.

Blumstein

Cohen

(1987). Characterizing criminal careers. Science, 237, 985-991.

Box-Steffensmeier

J. M.

Zorn

(2002). Duration models for repeated events. The Journal of Politics, 64, 1069-1094.

Bushway

S. D.

Piquero

A. R.

Broidy

L. M.

Cauffman

Mazerolle

(2001). An empirical framework for studying desistance as a process. Criminology, 39, 491-515.

10.

Bushway

S. D.

Thornberry

T. P.

Krohn

M. D.

(2003). Desistance as a developmental process: A comparison of static and dynamic approaches. Journal of Quantitative Criminology, 19, 129-153.

11.

Carrington

(2009). Co-offending and the development of the delinquent career. Criminology, 47, 1295-1329.

12.

Cloward

R. A.

Ohlin

L. E.

(1960). Delinquency and opportunity: A theory of delinquent gangs. Glencoe, IL: Free Press.

13.

D’Unger

A. V.

Land

K. C.

McCall

P. L.

Nagin

D. S.

(1998). How many latent classes of delinquent criminal careers results from mixed Poisson regression analysis. The American Journal of Sociology, 103, 1593-1630.

14.

Eckland-Olson

Lieb

Zurcher

(1984). The paradoxical impact of criminal sanctions: Some microstructural findings. Law & Society Review, 18, 159-178.

15.

Erikson

M. L.

Jensen

G. F.

(1977). “Delinquency is still a group behavior!” Toward revitalizing the group premise in sociology of deviance. Journal of Criminal Law and Criminology, 68, 262-273.

16.

Farrington

D. P.

(1986). Age and crime. In Tonry

(Ed.), Crime and justice: An annual review of research (pp. 189-250). Chicago, IL: University of Chicago Press.

17.

Farrington

D. P.

(2003). Key results from the first forty years of the Cambridge study in delinquent development. In Thornberry

T. P.

Krohn

M. D.

(Eds.), Taking stock in delinquency: An overview of finding from contemporary longitudinal studies (pp. 137-183). Boston, MA: Kluwer.

18.

Farrington

D. P.

(2005). Integrated developmental & life-course theories of offending, advances in criminology, Volume 14. New Brunswick, NJ: Transaction Publishers.

19.

Farrington

D. P.

Welsh

B. C.

(2007). Saving children from a life of crime. Oxford, UK: Oxford University Press.

20.

Gottfredson

Hirschi

(1990). A general theory of crime. Stanford, CA: Stanford University Press.

21.

Hindelang

M. J.

(1976). With a little help from their friends: Group participation in reported delinquency. British Journal of Criminology, 16, 109-125.

22.

Hirschi

(1969). Causes of delinquency. Berkeley: University of California Press.

23.

Hochstetler

(2001). Opportunities and decisions: Interactional dynamics in robbery and burglary groups. Criminology, 39, 737-763.

24.

Hood

R. G.

Sparks

R. F.

(1970). Key issues in criminology. London, England: World University Library.

25.

Horney

Osgood

D. W.

Marshall

I. H.

(1995). Criminal careers in the short-term: Intra-individual variability in crime and its relation to local life circumstances. American Sociological Review, 60, 655-673.

26.

Kempf

(1983, November). Assessment of the relationship between socioeconomic status and delinquency using the 1958 Philadelphia birth cohort. Paper presented at the annual meeting of the American Society of Criminology, Denver, CO.

27.

Kempf

(1989). Delinquency: Do the dropouts drop back in? Youth & Society, 20, 269-289.

28.

Kempf-Leonard

Tracy

P. E.

Howell

J. C.

(2006). Serious, violent, and chronic juvenile offenders: The relationship of delinquency career types to adult criminality. Justice Quarterly, 18, 449-478.

29.

Laub

J. H.

Sampson

(2001). Understanding desistance from crime. Crime and Justice, 28, 1-69.

30.

Laub

J. H.

Sampson

(2003). Shared beginnings, divergent lives: Delinquent boys to age 70. Cambridge, MA: Harvard University Press.

31.

Loeber

Stouthamer-Loeber

(1996). The development of offending. Criminal Justice and Behavior, 23, 12-24.

32.

Luckenbill

D. F.

Best

(1981). Careers in deviance and respectability: The analogy’s limitations. Social Problems, 29, 197-206.

33.

Matsueda

R. L.

(1988). The current state of differential association theory. Crime & Delinquency, 34, 277-306.

34.

Matsueda

R. L.

Anderson

(1998). The dynamics of delinquent peers and delinquent behavior. Criminology, 36, 269-308.

35.

Matza

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

36.

Matza

(1990). Becoming deviant. Englewood Cliffs, NJ: Prentice Hall.

37.

McGloin

J. M.

Piquero

A. R.

(2010). On the relationship between co-offending network redundancy and offending Versatility. Journal of Research in Crime & Delinquency, 47, 63-90.

38.

McGloin

J. M.

Povitsky Stickle

(2011). Influence or convenience? Disentangling peer influence and co-offending for chronic offenders. Journal of Research in Crime & Delinquency, 48, 419-447.

39.

McGloin

J. M.

Sullivan

C. J.

Piquero

A. R.

Bacon

(2008). Investigating the stability of co-offending and co-offenders among a sample of youthful offenders. Criminology, 46, 155-188.

40.

Moffitt

T. E.

(1993). Adolescence-limited and life-course-persistent antisocial behavior: A developmental taxonomy. Psychological Review, 100, 674-701.

41.

Nagin

D. S.

(2005). Group-based modeling of development. Cambridge, MA: Harvard University Press.

42.

Nagin

D. S.

Land

K. C.

(1993). Age criminal careers, and population heterogeneity: Specification and estimation of a nonparametric, mixed Poisson model. Criminology, 31, 327-362.

43.

Paternoster

(2010). How much do we really know about criminal deterrence? The Journal of Criminal Law and Criminology, 100, 765-823.

44.

Patterson

G. R.

Crosby

Vuchinich

(1992). Predicting risk for early police arrest. Journal of Quantitative Criminology, 8, 335-355.

45.

Piquero

A. R.

(2004). Somewhere between persistence and desistance: The intermittency of criminal careers. In Maruna

Immarigeon

(Eds.), After crime and punishment: Pathways to offender reintegration (pp. 101-120). Cullompton, UK: Willan.

46.

Piquero

A. R.

Blumstein

Brame

Haapanen

Mulvey

E. P.

Nagin

D. S.

(2001). Assessing the impact of exposure time and incapacitation on longitudinal trajectories of criminal offending. Journal of Adolescent Research, 16, 54-74.

47.

Piquero

A. R.

Farrington

Blumstein

(2007). Key issues in criminal career research: New analyses of the Cambridge study in delinquent development. Cambridge, UK: Cambridge University Press.

48.

Piquero

A. R.

Paternoster

Mazerolle

Brame

Dean

C. W.

(1999). Onset age and offense specialization. Journal of Research in Crime & Delinquency, 36, 275-299.

49.

Prentice

R. L.

Williams

B. J.

Peterson

A. V.

(1981). On the regression analysis of multivariate failure time data. Biometrika, 68, 373-379.

50.

Raskin

S. B.

(1987). The measurement of time intervals between arrests. In Wolfgang

M. E.

Thornberry

T. P.

Figlio

R. M.

(Eds.), From boy to man, from delinquency to crime (pp. 59-67). Chicago, IL: University of Chicago Press.

51.

Reiss

A. J.

Farrington

(1991). Advancing knowledge about co-offending: Results from a prospective longitudinal survey of London males. The Journal of Criminal Law and Criminology, 82, 360-395.

52.

Reiss

A. J.

(1988). Co-offending and criminal careers. Crime and Justice, 10, 117-170.

53.

Sampson

Laub

J. H.

(1993). Crime in the making: Pathways and turning points toward desistance. Cambridge, MA: Harvard University Press.

54.

Shaw

C. R.

McKay

H. D.

(1931). Report on the cause of crime, Volume II. Washington, DC: U.S. Government Printing Office.

55.

Short

Jr. Strodtbeck

F. L.

(1965). Group process and gang delinquency. Chicago, IL: University of Chicago Press.

56.

Shover

Henderson

(1995). Repressive crime control and male persistent thieves. In Barlow

H. D.

(Ed.), Crime and public policy: Putting theory to work (pp. 227-246). Boulder, CO: Westview.

57.

Stolzenberg

D’Alessio

(2008). Co-offending and the age-crime curve. Journal of Research in Crime & Delinquency, 45, 65-86.

58.

Tanenhaus

D. S.

(2004). Juvenile justice in the making. Oxford, UK: Oxford University Press.

59.

Tibbetts

S. G.

Piquero

A. R.

(1999). The influence of gender, low birth weight, and disadvantaged environment in predicting early onset of offending: A test of Moffitt’s interactional hypothesis. Criminology, 37, 843-878.

60.

Tracy

P. E.

(1981). Ecology and delinquency: The development of a composite measure of social class. Unpublished manuscript.

61.

Tracy

P. E.

Kempf-Leonard

(1996). Continuity and discontinuity in criminal careers. New York, NY: Plenum Press.

62.

Tracy

P. E.

Wolfgang

M. E.

Figlio

R. M.

(1990). Delinquency in two birth cohorts. New York, NY: Plenum Press.

63.

Van Mastrigt

S. B.

Farrington

D. P.

(2009). Co-offending, age, gender and crime type: Implications for criminal justice policy. British Journal of Criminology, 49, 552-573.

64.

Velarde

(1978). Do delinquents really drift? British Journal of Criminology, 18, 23-39.

65.

Warr

(1993). Age, peers, and delinquency. Criminology, 31, 17-40.

66.

Warr

(1996). Organization and instigation in delinquent groups. Criminology, 34, 11-37.

67.

Warr

(2009). Peers and delinquency. In Krohn

Lizotte

Penly Hall

(Eds.), Handbook on crime and deviance (pp. 383-402). New York, NY: Springer Science + Business Media.

68.

Zimring

F. E.

(1981). Kids, groups and crime: Some implication of a well-known secret. The Journal of Criminal Law and Criminology, 72, 867-885.

69.

Zimring

F. E.

(2005). American juvenile justice. Oxford, UK: Oxford University Press.

70.

Zimring

F. E.

Jennings

W. G.

Piquero

A. R.

Hays

(2009). Investigating the continuity of sex offending: Evidence from the second Philadelphia birth cohort. Justice Quarterly, 26, 58-76.