Abstract
In contrast to street offending/delinquency, no study to date has explored how white-collar offending develops across the life course using self-report data. This paper fills this gap in the literature by exploring latent developmental profiles of self-reported white-collar offending across a 16-year period. These trajectories are then compared in terms of concurrently reported street crimes. Data for analysis were culled from Waves 7 through 11 of the National Youth Survey Family Study (NYSFS) when the original respondents were between 22 and 44 years old. Six latent offending profiles were identified and were characterized through the lens of previous research and theory.
Introduction
For many decades, criminologists have sought to understand the relationship between criminal offending and the aging process (Piquero, Farrington, & Blumstein, 2003). One can easily argue that some of the most important advancements in theoretical explanations of crime stem from the criminal career paradigm and its progenies, one of which involves the developmental approach to understanding criminality (Benson, 2012). However, a shortcoming of most studies following this tradition is that the bulk of the attention, both empirical and theoretical, has focused solely on street-level offending (Piquero & Benson, 2004). In reality, there is no arguing against the fact that some people engage in criminal behavior well into adulthood and many of these crimes differ from street crimes; rather, they fall under the umbrella of white-collar crime (Menard, Morris, Gerber, & Covey, 2011).
Though we are only beginning to explore and understand the developmental process for white-collar crimes, life course/developmental criminology has paved the way for this line of research to progress. In fact, there have been a handful of studies that have applied a developmental perspective to white-collar offending (e.g., Piquero & Benson, 2004; Piquero & Moffitt, 2012; Weisburd & Waring, 2001). One reason underlying the small amount of attention to this area is the rarity of longitudinal data collection efforts explicitly accounting for white-collar crimes. Further, much of the data utilized in the study of white-collar crimes are pulled from official court records, rather than customized data collection efforts (e.g., Benson & Kerley, 2000; Morris, Copes, & Perry-Mullis, 2009; Weisburd & Waring, 2001). Only recently have longitudinal data become available that capture information through adulthood, and measure self-reported white-collar crime offenses (e.g., Menard & Morris, 2012; Menard et al., 2011).
The current study extends this line of research by estimating and interpreting the developmental trajectories of self-reported white-collar offenders over a 16-year period in adulthood. Additionally, we explore street crime co-occurring with white-collar crime to preliminarily explore issues surrounding offense specialization.
Review of the Literature
Defining White-Collar Crime
Variation within the definition and extension of Sutherland’s (1940) original version has been the subject of debate for many years (i.e., whether white-collar offending is a matter of status (the offender) or behavior (the offense)). In the end, however, most would agree that white-collar criminals are different from street offenders (Friedrichs, 2002; Menard et al., 2011; Weisburd, Wheeler, Waring, & Bode, 1991), though the fact remains that people who engage in these types of crimes are generally not of high status. Definitional debates aside, we follow an offense-based definition for the purposes of this study as we simply do not have a complete understanding of how these behaviors develop across the life course and whether the individuals who engage in them simply transition from street offending in to more fraud-related crimes, or whether such offenders specialize in this type of offending. Only until we can confirm such patterns of development can we attempt to explain the process through theory.
Life-Course Criminology and White-Collar Crime
There is no question that continued research is needed to understand the criminal careers of white-collar offenders (Piquero & Benson, 2004). Fortunately, the life-course/developmental framework of explaining the age/crime relationship provides for a solid foundation from which to begin. Marked progress has been made among developmental/life-course criminologists in recent years toward our understanding of the relationship between age and crime, but as noted earlier, most of the attention has focused on street-related offending or traditional delinquency (Piquero et al., 2003). This line of research has been successful in reaching its goal for many reasons including the availability of quality longitudinal data, and advancements in computing technology and sophisticated quantitative techniques for identifying latent patterns of development among groups of people. Specifically, the use of techniques such as trajectory analysis, among others, allows for latent behavioral patterns to be identified from within a sample. Trajectory analysis is a quantitative technique that accounts for within- and between-individual change, also known as persistent heterogeneity, which manifests through repeated observations of the same people over time. The application of this tool to life-course studies has provided for an impressive body of literature surrounding the criminal career paradigm.
Studies of criminal development have found that street offenders/delinquents commonly embark on a criminal trajectory during adolescence or early adulthood (Elliott, 1994). The duration of their criminal career is relatively short, with most offenders desisting from crime in their early 20s (Blumstein, Cohen, Roth, & Visher, 1986). While not the norm, some offenders continue to offend well into adulthood. Moffitt (1993) refers to this group as life-course persistent offenders. However, subsequent research discovered that chronic offenders might actually manifest through two subgroups (high rate and low rate chronic; Nagin & Land, 1993). The difference being that the low-rate chronic group may actually start out committing less crime (i.e., during adolescence), but end up having longer criminal careers comparable to high-rate chronic offenders by the age of 30.
Life course theories attempt to explain why an individual may progress (onset) down one trajectory over another, why persistence or changes in offending patterns may occur along the way, and ultimately what leads to desistance from offending altogether (e.g., Hagan, 1997; Hagan & Palloni, 1998; Moffitt, 1993; Sampson & Laub, 1993; Thornberry, 1987). Due to limited space, we refer those interested in life course theory applications to white-collar crime to Piquero and Benson (2004).
In the end, studies of the criminal career need to extend beyond the generally agreed upon peak years associated with offending (i.e., youth and early adulthood) and strive to capture the entire life course (Hagan & Palloni, 1998). This is particularly relevant to white-collar offending in that offending patterns may not reflect those more common among youthful offenders (i.e., street crimes; Menard et al., 2011; Piquero & Weisburd, 2009; Weisburd & Waring, 2001). In fact, when white-collar offenses are included along with street crimes, the age-crime curve is elongated and peaks much later than generally found in research that focuses exclusively on street crimes—see Menard et al., 2011).
While we do not yet know whether the criminal careers of white-collar offenders operate in a similar manner as street crime offenders, a few studies have provided for some interesting findings on the topic. For example, studies based on official court records have shown that the average age of onset for white-collar offenders is 35 (Weisburd & Waring, 2001). These offenders were found to desist from crime around the age of 43. 1 More recently, white-collar crime scholars have capitalized on the quantitative techniques popularized by criminal career/developmental researchers in an effort to understand the development of white-collar offending. Piquero and Weisburd (2009) estimated trajectory models using the follow-up recidivism data collected by Weisburd & Waring (2001). Their findings provided evidence largely supporting the qualitative categorizations of Weisburd and Waring’s three main patterns of white-collar offending; these include low-frequency offenders with few arrests on record (crisis responders and opportunity takers), a group of intermittent offenders (opportunity takers) and a group of persistent offenders (stereotypical criminals)—for a more detailed review, see Piquero and Weisburd. However, their findings diverged in terms of the prevalence of each classification within the sample. Specifically, Piquero and Weisburd found that “opportunity seekers” were more prevalent than suggested in the original analysis and that the “stereotypical criminal” group was not as common as previously thought. While groundbreaking, Piquero and Weisburd’s assessment was limited by the fact that the data was not age-specific and did not account for periods of imprisonment. Nonetheless, their support for the qualitatively established patterns identified by Weisburd and Waring go a long way in describing the criminal careers of white-collar offenders.
The present study extends the work of Piquero and Weisburd (2009) and contributes to our understanding of white-collar offending by assessing developmental trajectories of adult offending based on self-report data collected over an extended period among a cohort of Americans. It is our hope that the findings presented here provide the next step in understanding the development of white-collar offending and help to establish the next phase of research on this important topic.
Method
Data and Sample
Data for the present study were culled from Waves 7 through 11 of the National Youth Survey Family Study (NYSFS), formerly known as the National Youth Survey (NYS). The NYSFS followed a multiple cohort sequential design covering 12 waves of data across 27 years, beginning in 1976 when the original respondents were between 11 and 17 years old. The NYSFS is considered nationally representative to American youth in 1976 (for more on the study design of the NYS and NYSFS and issues surrounding retention, see Elliott, Huizinga, and Menard (1989) and Menard and Morris (2012)). For present purposes, we analyzed data reported by original respondents who participated in each of Waves 7 through 11 (n = 968) and who reported at least one form of white-collar crime during the observation period (n = 220, or 18.5% of eligible respondents). 2 Excluding those who reported no white-collar offending (i.e., abstainers) was important since the focus here was to assess the development of actual white-collar offending. This approach served to eliminate heterogeneity by excluding those never reporting such acts (see Morris, Carriaga, Diamond, Piquero, & Piquero, 2012).
During the observation period, the age range of respondents varied depending on the NYSFS cohort (birth year). The age range across the observation period varied from 22 to 38 years (for the 11-year-old cohort at Wave 1) and 28 to 44 years (for the 17-year-old cohort at Wave 1). During the observation period, the sample representation was distributed across the cohorts as follows: age 11 (15.9%); 12 (18.2%); 13 (14.6%); 14 (15.0%); 15 (16.4%); 16 (9.1%); and 17 (10.9%).
Measurement
Our primary interest resided in exploring how repeated observations of self-reported white-collar offending developed over the 16-year period. That said we calculated one composite measure of white-collar offending, which was operationalized identically across the five waves of data. Here, white-collar crime is represented by the summed frequency (count) of self-reported acts of check fraud, credit card fraud, sales fraud (conning, or cheating someone for monetary gain), employee theft, and embezzlement reported at each wave, respectively. Specifically, respondents were asked to report the number of times they had participated in the behavior during the 12-month period preceding the interview. All respondents were asked to report “How many times in the past year have you . . . ”: (a) “used checks illegally or used phony money”, (b) “used or tried to use credit cards without the owner’s permission?,” (c) “tried to cheat someone by selling them something that was worthless or not what you said it was?”, (d) “stolen money, goods, or property from work?”, and (e) “embezzled money, that is, used money or funds entrusted to your care for some purpose other than intended?” Prior to summing the scale, each white-collar crime indicator was truncated to a value of 10 (generally the 99th percentile) to reduce the influence of outliers. 3
Between-group (i.e., trajectory classification) differences were also explored on concurrent participation in street crimes. As with the white-collar offenses, respondents were asked to report the number of street crimes they had engaged in during the 12 months prior to the interview. These included aggravated assault, sexual assault, gang fights, auto theft, burglary, theft of goods greater than US$50, and buying stolen goods. Reports from each indicator of street crime were summed and truncated (99th percentile) resulting in a count of self-reported street crimes reported at each of the five waves. The array of street crimes comprising the measure follows past research using the NYSFS (e.g., Menard et al., 2011). Table 1 presents the descriptive statistics for the above noted variables.
Means and Standard Deviations for White-collar Crime Indicators (offenders only) and Street Crimes, W7-W11.
Analytical Procedure
This study was based upon two analytical components. First, we examined whether different developmental profiles exist for adult white-collar offending across five waves of the NYSFS (Waves 7 through 11). Second, we compared differences in street offending levels that co-occurred during the observation period for each of the developmental profiles. We leave the exploration of factors predicting assignment to one group over another to future studies.
The first component was based on a trajectory analysis (i.e., group-based trajectory modeling, or GBTM—see Nagin, 2010; Nagin & Land, 1993). This procedure has been used extensively by life-course criminologists as a means to better understand transitions in offending behavior, the development of such behavior over time, and to account for unobserved group heterogeneity in offending (i.e., the extraction of similar developmental profiles for subsequent analyses)—(see Bushway, Piquero, Broidy, Cauffman, & Mazerolle, 2001; Cohen, Piquero, & Jennings, 2010; Hay & Forrest, 2006; Laub, Nagin, & Sampson, 1998; Morris & Piquero, 2011; Nagin, Farrington & Moffitt, 1995; Sampson & Laub, 2003). The first step in this procedure was to estimate the trajectories. Here, we utilized longitudinal latent class analysis (LLCA), which is one form of group-based trajectory modeling (GBTM) that has recently been used in longitudinal studies of criminality (Morris et al., 2012; for more detail on this technique see Feldman, Masyn, & Conger, 2009). What is unique and advantageous about this form of GBTM is that it assumes no functional form of development. Rather, it classifies cases based on patterns, or states, of development, regardless of functional form. This relaxed form of trajectory analysis may be quite fitting to studies of white-collar crime development as past work has consistently identified a large amount of intermittency in offending that may hamper models requiring specification of a particular functional form (Piquero & Weisburd, 2009; Weisburd & Waring, 2001). Subsequent steps in this procedure are presented along with the findings below.
Findings
Trajectory Analysis
The goal of LLCA, as with all forms of finite mixture modeling, is to assess and classify the development of some observable or latent phenomenon as it changes over a specified period of time. The end result is an approximation of a finite number of developmental profiles, or trajectories, from within the population. The first step in this procedure was to determine an appropriate number of latent classes, or groups, based on unconditional trajectory models (i.e., models without predictor variables). These models were developed by utilizing the negative binomial link-function as the outcome variables were over-dispersed counts of white-collar crime. Following convention, Bayesian information criterion (BIC), posterior probabilities, group proportions, entropy, visualization plots, in addition to consideration for model parsimony, were used in tandem to settle on a final model solution (i.e., the number of classes to retain). We also utilized visualizations of the estimated trajectory profiles against the observed data of 20 randomly selected individuals within each class to assess model fit and to help guide interpretation of the findings. Fit statistics and group proportions are presented in Table 2. A preponderance of evidence from these criteria suggested that the 6-class solution was the best fit to the data, which is visualized in Figure 1. Analyses were carried out using MPlus version 6.
Longitudinal Latent Class Analysis Results for NYSFS Original Respondents Reporting White-collar Offending, W7-W11.
Notes: BIC= Bayesian information criterion.
Estimated white-collar offending trajectories, 6-group solution.
In our final model solution, we can see that among the six groups of offenders, intermittency in offending appears to be a common group characteristic. Individuals assigned to Group 1 (21.4% of the sample), for example, are expected to have a low level of participation in white-collar offending in most waves except for Wave 9, and to a lesser extent in Waves 7 and 10. Reviewing the observed individual trajectories for individuals within this class supports the notion that this is a highly intermittent group (see Figure 2). Group 2 (31.3% of the sample), the most prevalent group, follows a similar pattern, with the only marked difference being the timing at which expected offending his highest (i.e., at Wave 7), and again these individuals are generally intermittent in their offending.
Estimated group trajectories versus 20 random individual observations* (n = 20) for white-Collar offenses.
Group 3 (13.2% of the sample), represented by a small number of intermittent respondents, does not generally engage in white-collar crime until Wave 10 and continues through Wave 11. Group 4, the smallest group (6.3% of the sample), is characterized by the highest amount of activity at some Waves (7 and 8), but dormancy (or perhaps desistance) beginning at Wave 9, carrying through Wave 11. Similarly Group 6 (13.1% of the sample) follows a pattern of one period of activity with remainder being dormant, very similar to Group 4 in terms of dormancy and the timing of a spike in offending.
Group 5 (14.5% of the sample) is the most criminally active group represented by relative continuity in white-collar offending at a generally higher rate. Comparing the observed trajectories against the estimated class trajectory, we see that this characterization holds for most individuals in the class, though some individuals are intermittent between periods of high-rate activity.
Overall, these results partially corroborate those of Piquero and Weisburd (2009) and partially fall in line with the typologies identified by Weisburd and Waring (2001). Unlike both of those studies which found that the vast majority of white-collar offenders fell into a “low-rate” group, represented by long periods of dormancy and brief reactions to “crisis” or “opportunity” at one period, we found that most white-collar offenders are in fact intermittent (52.7%, Groups 1 and 2), reporting offenses at multiple waves well into mid-adulthood, but at a low rate. As discussed by Piquero and Weisburd, these individuals are perhaps the “opportunity seekers” outlined by Weisburd and Waring, but at this point, we are more confident in identifying this subset as an intermittent group of white-collar offenders.
We also found that about one third (32.6%; groups 3, 4, and 6) of individuals might be similar to the “low-rate” offenders of previous studies who generally, but not always, report one or two concurrent periods of activity, albeit at varying rates and at different points during the observation period. Perhaps these groups are the “opportunity takers” or “crisis responders” or perhaps these findings highlight the reality that these responses to situations vary in magnitude (i.e., number of reports) and perhaps duration (e.g., the crisis may last longer than 3 years or the opportunity may remain for an extended period). Also interesting about these groups is that they report extended periods of no white-collar crime. Group 3 stands out in this regard as very few of these individuals report any activity prior to Wave 10. Groups 4 and 6 appear to either go dormant from Wave 9 through Wave 11, or they have desisted altogether. Additional waves of data would be required to confirm the latter possibility. Within this particular subset of groups, Wave 8 (1990) stands out. During this period, the respondents were aged between 25 and 31. It is possible that the observed spike in offending was in response to a perceived crisis that is correlated with this age range. This may have been a period in which some of these individuals were attempting to become established in life, experienced a crisis, and reacted with crime as a result of their vulnerability to losing pace on the path to becoming established. For some, small crises could have a lasting impact in terms of reaching aspirations. Whether these abrupt changes in criminal behavior result from crisis or opportunity, or whether some positive turning point resulted in desistence, are questions for future research. Future analyses will be carried out to assess such issues.
Our high-rate/persistent group (14.6%, Group 5) is more predominant than the high-rate group reported by Piquero and Weisburd (2009; 5%) but perhaps similar to the high-rate offenders reported by Weisburd and Waring (2001, p. 175; between 6 and 16%). On average, this group is stable in terms of prevalence (i.e., at least one report), but tends to vary in magnitude and perhaps by brief periods of intermittency for some, but not others. Weisburd and Waring identify this group as the “stereotypical offenders” who are actively criminal throughout their lives. In an effort to determine whether these offenders are actually “stereotypical,” we extend our analysis to the assessment of concurrently occurring street offending among each latent class of offenders, respectively.
Concurrent Street- and White-Collar Crime
Figure 3 presents the observed average number of street offenses reported at each wave for each white-collar latent class. With very few exceptions, white-collar crimes are more common at any given wave, however some interesting patterns warrant further attention. First, two of the classes (Groups 1 and 2) report very low rates of street crimes in spite of clear intermittency with white-collar offending. Second, three white-collar profiles (Groups 3, 4, and 6) exhibit increased levels of street crime during periods of increased white-collar crime. However, during these periods, white-collar crimes are far higher in magnitude. Group 4 (the smallest group) reports the highest involvement in both street and white-collar crime at any single wave. Third, our high-rate/persistent group (Group 5) tends to remain modestly and stably involved in street crimes through the observation period.
Observed street crime trajectories by latent class, W7-W11.
In sum, these findings suggest that most white-collar offenders tend to specialize in white-collar offenses during this period of adulthood (Groups 1 and 2), but they do so intermittently, perhaps seeking opportunities, or simply responding spontaneously to other situational circumstances. Others (Groups 3, 4, and 6) seem to have periods of generally heightened criminality (both street and white-collar), the bulk of which is white collar. These periods tend to fall between periods of criminal inactivity, or dormancy. Such individuals may be the opportunity takers or crisis responders outlined by Weisburd and Waring (2001), but it is likely that some people respond to such opportunities with crime in general, rather than white-collar crime exclusively. For others, opportunities for crime during this period of life may be more tied with legitimate employment or status, than with the streets, which would perhaps partially explain the diverging trajectories between the two forms (i.e., street and white-collar). Finally, there is one group of white-collar offenders who are consistently active in crime of all types (Group 5). White-collar forms clearly take precedent for these individuals, but street crimes are consistently reported throughout Waves 7 through 11.
Conclusions
The goal of this study was to determine whether distinct developmental patterns of white-collar offending could (a) be identified and (b) whether such development coalesced with participation in street crimes. These findings were based on white-collar offending trajectories of NYSFS original respondents across five waves of data covering a 16-year period during adulthood and contribute to the white-collar crime literature in several ways. First, this is the only longitudinal assessment of white-collar crime developmental profiles generated from self-report data of like-aged individuals to date. In doing so, we identified six classes of development for white-collar crimes each of which varied in magnitude, timing, and/or shape, but generally reflected three patterns, or subsets; intermittency, isolated periods of amplified criminality, and for some, persistent offending. Some of the findings parallel the small number of past studies that have sought to explain the criminal careers of white-collar offenders while other findings suggest otherwise or somewhere in between (Piquero & Weisburd, 2009; Weisburd & Waring, 2001).
For example, our findings on the most consistently active white-collar criminals seem to parallel those of Weisburd and Waring (2001); about 14% of our sample fell into the persistent, but most of their reported offending was for white-collar crime. Similar to Piquero and Wiesburd (2009), our findings suggest that Weisburd and Waring underestimated the prevalence of those deemed “opportunity seekers.” In fact, most white-collar offenders in the data relied upon here arguably fit into this category by way of intermittency in white-collar offending and relatively low rates of street offending. These individuals reported white-collar crimes at multiple waves and none at others. Other white-collar criminals tended to limit offending one consecutive period of time (one or two waves), but varied in rates of offending.
Our findings also suggest that a majority of those involved with white-collar offending, at least during mid-adulthood, tend to report attenuated levels of street offenses, if at all. Whether these individuals specialize in white-collar crime over the life course exclusively requires analyses beyond the scope of the present study. Still, others tended to desist, or enter a period of dormancy for an extended period of time. What is interesting is that we found this pattern for both low-rate and higher rate offenders. For example Group 4 reported the highest rates of both white-collar and street crime at Waves 7 and 8, but then virtually no crime thereafter.
In the end, we have shown that there is still much to be learned regarding onset, continuity, intermittency, and desistence in white-collar offending over the life course and what we have presented here certainly raises a number of important questions for scholars to consider. For example, what are the processes driving the relationship between adult white-collar offending and offending occurring at earlier points in the life course (i.e., population heterogeneity theories vs. state dependent theories; see Nagin & Paternoster, 1991; Paternoster, Dean, Piquero, Mazerolle, & Brame, 1997; Sampson & Laub, 1993)? Much is to be learned in terms of the causal processes that may explain how one person ends up going down one white-collar pathway versus another, as well as how we might explain varying patterns of development for this particular genre of offending—for example see Paternoster et al. (1997). More broadly, do differences in patterns of offending for individuals who engage in white-collar crimes (definitional dilemmas notwithstanding) parallel those found underlying more general definitions of crime (i.e., street crimes)? Are such differences best explained by static, dynamic, or some combination of developmental theoretical perspectives? For example, do all diverging patterns of white-collar offending stem from one distinct offending profile established earlier in life (e.g., Moffitt’s (1993) developmental taxonomy), from more than one, or is the process driving white-collar crime something altogether different? Do life circumstances vary in effect between developmental trajectories of white-collar offending? Or, are these patterns explained by individual personality traits?
Like the few studies of developmental white-collar crime that have preceded this one, our study demonstrates the complex nature of criminal offending and clearly reveals that much is to be learned about adult offending behaviors. In any case “ . . . it is important when charting the course of criminal careers to recognize that different models of explanation may be needed to provide explanations for different types of offenders” and the questions that our findings raise warrant considerable attention, far more than can be presented in a single article (Piquero & Weisburd, 2009, p. 167).
In closing, several study limitations must be acknowledged. First, our white-collar crime indicators, while unique in terms of content and their being based on longitudinal data, were limited to only five types of offenses. Other offenses, such as tax evasion, are much more prevalent in these data than are those reported here (see Menard et al., 2011), but were excluded for methodological reasons, noted previously. Also, it can be argued that some of the indicators may be qualitatively similar, further limiting the breadth of coverage on white-collar crime (e.g., embezzlement and stealing money from work). It is also important to note that this study was not able to address other important forms of white-collar offending, such as those involving corporations and organizations, such as antitrust violations, securities offenses, health care fraud, etc. Individuals involved in corporate types of white-collar offending may display very different patterns of offending involvement that are represented by the patterns of development illustrated in this study; we must leave that opportunity for research to others. Additionally, the NYSFS data was not collected at equidistant waves. Most striking is the 9-year gap between Wave 9 and Wave 10. We also did not attempt to unpack what predicts an individual from being classified to one latent profile over another or what factors influence changes in the developmental profile from one wave to the next. Despite these limitations, this study provides the first longitudinal assessment of self-reported white-collar offending to date and paves the way for a considerable amount of research to be conducted in the future.
Footnotes
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The authors received no financial support for the research, authorship, and/or publication of this article.