Identify and Critique the fatal flaws in

'Toward a Quantitative Typology of Burglars: A Latent Profile Analysis of Career Offenders' by Michael G. Vaughn,

The substantial economic and social costs of burglary warrant increased scientific advancement in the classification and subtyping of burglars. Last year, more than 2.1 million burglaries were reported for law enforcement, accounting for 21.2% of property crimes, resulting in an average financial loss of $1,725 (1). In terms of criminal justice expenditures and collateral victimization costs, the average social cost per burglary has been estimated at nearly $20,000 (2). Most burglars are also involved with other offenses, such as robbery, assault, and drug selling (3,4). Despite being a relatively common felony, few studies have attempted to develop quantitatively based taxonomies of burglar subtypes. As a result, much information on burglars is derived from the offender's perspectives via ethnographic research and case studies. In contrast, the criminal careers paradigm has emerged as one of the foremost paradigms in criminology. Rather than viewing criminal activity such as burglary as a solitary event from the personal perspective of a single offender, the criminal careers perspective studies antisocial behavior longitudinally, usually quantitatively, and explores the ways that criminal offending patterns emerge, continue, escalate, desist, and ultimately end. A major area of study within criminal careers research centers on the utility of categorical classifications of offenders or typologies that organize offenders according to their discrete offending patterns. Although some evidence of offense specialization has been found which would support the use of typologies, most research indicates that criminal offenders are versatile in their offending patterns (5,6). This means that burglary usually occurs along with other types of antisocial behavior; unfortunately, the understanding of how burglary relates to other types of offenses and what its place is in the criminal career is still in progress. Several studies of burglars from the offender's perspective have found that burglars are deliberate in choosing targets and often engage in careful planning (7-13). Findings relevant to the development of typologies of burglars have emerged from these investigations. In a study of 105 residential burglars, Decker et al. (9) found that many burglars engaged in burglaries because it was perceived as a relatively easy way to make money for lifestyle support often involving significant and expensive quantities of drugs and alcohol. Moreover, many perpetrators found the enterprise of burglary to be thrilling, a finding that is concordant with ethnographic research (3,8,14,15). According to a typology developed by Maguire and Bennett (10), three types of burglars can be identified: low-level amateurs, mid-level professionals, and high-level professional burglars. Most of them are low-level or mid-level burglars. In a forensic psychiatric study of sexually oriented burglars, Schlesinger and Revitch (16) found that sexual burglars could be usefully classified into two groups: overt sexual burglars who seek sexual gratification via direct involvement in sexual acts committed during burglaries and covert sexual burglars who are more voyeuristic. Taken together, these studies indicate a broad pattern of themes involving material gain, sexual drive, and substance abuse. Although generating useful insights, these typologies were not developed through rigorous quantitative analysis and were not situated within a criminal careers framework. The aim of the present investigation was to take a rigorous quantitative approach to taxonomic classification of burglars. We identified burglar subtypes using latent profile analysis (LPA) of data on the criminal careers of a sample of 456 burglars. An empirically derived typology such as that developed in this study could serve to synthesize competing conceptualizations of burglar subtypes and provide an initial starting point for future investigations that attempt to subtype burglars using latent variable modeling or qualitative inductive techniques.

Methods Sampling and Procedures The study of active, career offenders who are burglars poses a multitude of methodological difficulties for researchers. Such persons are not easily identifiable in general population samples. For example, DeLisi (17,18) found that little was known about the offending careers of extreme habitual criminals because most criminal career datasets lacked sufficient numbers of the most serious offenders. Consequently, an offender sample must be large enough to yield enough offenders with burglary offenses to support multivariate statistical analyses. From 1995 to 2000, the second author of this report was employed as a bond commissioner at a large urban jail located in Colorado. In this particular jurisdiction, bond commissioners served as judicial officers and worked in conjunction with sheriff deputies within the county jail. Their function was to interview all criminal defendants brought to the jail and to obtain employment, residency, and criminal history data for setting bond. This work experience permitted continuous access (the bond commissioner unit was staffed 24 h a day) to all arrestees who were brought to the jail during this time period. Data pertaining to the present study sampling frame were derived from an effort to identify the most recidivistic offenders to determine their eligibility for various social services and to facilitate prosecutorial efforts. A pilot study of c. 50 offenders comprised the original frequent offender sample; their criminal histories contained an average of 30 arrest charges. Using this selection criterion, any offender whose record contained 30 or more arrest charges was classified as a frequent offender contingent upon the approval of the chief district judge and district attorney's office. From 1995-2000, the bond commissioner unit processed 25,640 defendants, 500 of whom (less than 2%) qualified for frequent offender status. These 500 offenders were, in effect, the frequent offender population of a 6-year census of official criminal offenders processed in this jurisdiction. Importantly, although offenders were processed at one facility, their criminal activity could and did occur in multiple jurisdictions. Of the 500 frequent offenders identified, the final study sample comprised of 456 adults who possessed a history of burglary offenses. Sample characteristics are presented in Table 1. Data and Measures During bond interviews (which were legal proceedings conducted under oath), defendants self-reported their criminal history, including all police contacts, arrests, court actions, and sentences. Self-reports can yield arrests and other criminal activities that do not appear on official records, arguably rendering them a more accurate reflection of an individual's true criminal past (19,20). However, the self-reporting method is not without its problems. For example, one problem involves the fact that most serious career criminals have offending careers that include potentially hundreds of arrests, convictions, and various punishments. Their criminal careers often span decades and involve many events precipitated or accompanied by drug and alcohol use. Thus, not only are there many criminal events, but career offenders may suffer from memory and other cognitive impairments stemming from an antisocial lifestyle. Therefore, self-reported criminal histories were supplemented with official records from the Interstate Identification Index (III) system. Under the III system, the FBI maintains an automated criminal record containing an FBI number and state identification number for each state holding criminal history information on an individual. The III records are accessed using the National Crime Information Center (NCIC) telecommunications lines that retrieve criminal records from repositories. Increasingly, researchers studying criminal careers are using both official and self-reported measures of criminal offending to assess and enhance validity of offender self-report data (21-23). In addition, this technique reduces missing data, increases reliability, and minimizes deficiencies of self-reported and official data collection strategies. Farrington et al. (24, p. 953) concluded, ''there was a significant overlap between chronic offenders identified in court referrals and chronic offenders identified in self-reports. Therefore, to a considerable extent, selfreports and court referrals identified the same people as the worst offenders.'' Data Analysis Variants of finite mixture modeling such as latent class analysis (LCA) and LPA have become increasingly popular methods because of their ability to identify underlying patterns in data based on unobserved quantities (25). LCA uses categorical variables while LPA is employed with continuous variable measures. Latent variable modeling possesses numerous advantages not found in the related techniques of k-means or hierarchical cluster analysis. In LCA, results are model-based rather than based on ad hoc distance measures and as such employ maximum likelihood estimation procedures. In addition, covariates can be entered simultaneously with indicator variables without the additional computational steps necessary with traditional cluster analytic procedures. Similar to cluster analysis, however, final class solutions depend entirely on variables that are originally entered into the analysis. As continuous variables were used to form latent subgroups, LPA was used to quantify the underlying patterns of homogeneity among burglars. The underlying assumption of LPA is that the relationship among continuous indicators can be explained by a categorical latent variable. The continuous indicators are considered to be locally independent, meaning that the observed items are statistically independent within each latent subgroup or class (26,27). Models were run using Latent GOLD 4.0 software (Statistical Innovation, Belmont, MA) (28) with the goal of analyzing one to five classes. We could have included more classes, but maintaining parsimony is critical. Final optimal class solutions were based on several fit indices including maximum likelihood estimation using the Bayesian Information Criterion (BIC) where lower BIC values indicate model improvement. Additional fit indices, such as class error, number of parameters, and entropy were also examined. Class assignment probabilities were also evaluated to assess class homogeneity. Conceptual fit of models is critical and was examined by using visual representations of the offense indicators to assess their interpretability and practical implications. Final class solutions should be theoretically interpretable and not merely reflect statistical fit optimization. Bivariate residuals were examined to ensure that the assumption of local independence was not violated; no violations were detected. ANOVA and chi-squared tests were TABLE 1Characteristics of burglary offenders (n = 456). n (%) Mean ( SD) Age 38.50 (10.66) Gender Male 424 (93.0) Female 32 (7.0) Ethnicity White 256 (56.1) Hispanic 125 (27.4) African-American 49 (10.8) Native American 23 (5.0) Asian-American 3 (0.7) Offense onset 18.41 (4.99) Burglaries 4.77 (4.89) Criminal career 20.09 (10.01) Total offenses 55.52 (32.61) 1388 JOURNAL OF FORENSIC SCIENCES

potentially a precursor for even more serious homicide offending and an important avenue for future research. In addition to these empirical linkages to criminal careers research, the current study also bears on important theoretical issues. Early offender typologies soon fell out of favor because they were empirically weak and often not mutually exclusive (5,6,18,29). The typology issue relates to a large theoretical question centering on the validity of general theories that explain criminal careers from an underlying propensity perspective (32) and developmental theories that explain criminal careers as they relate to life circumstances and involvement in social institutions (33,34). Unfortunately, because the current study lacked data on employment, education, and marital status, we were unable to assess how and whether the offending behavior of the four classes changed according to life circumstances. For instance, McGloin et al. (35) found that criminal careers are best understood when short-term institutional relationships are known. In their study, offenders who were unmarried tended to be more versatile in their offending patterns whereas married offenders had constrained opportunities to offend and were more specialized. Developmental theories and research thus provide context to criminal careers and illuminate ways that getting married, having a job, and going to school decrease antisocial behaviors. The lack of data on life circumstances and the informal social controls rendered by marriage, job, and school is an important limitation of the current study. It is also important to note that the LPA of the four burglary groups is specific to the current data and that replications with other datasets could probably produce different estimates and groupings. In no way should current burglary classes be reified. Interestingly, quantitatively sophisticated analytical techniques are producing typologies of offenders that in some ways are reminiscent of early, qualitative efforts. For instance, semi-parametric group-based and LCAs allow for the study of criminal careers data to compare discrete groups of offenders who are homogeneous within their grouping or trajectory but different from other groupings or trajectories. Based on the shape of the offending trajectories over time, scholars are identifying ''groups'' of offenders with an assortment of labels, such as nonoffenders, adolescence-peaked, low-rate-, high-rate-, and late-onset chronics; classic desisters, moderate-rate desisters, low-, moderate-, and high-rate chronics. Yet even today, there appears to be dissatisfaction with the substantive worth of these approaches. For example, Eggleston et al. (36) concluded that ''there is always a danger when a particular methodology approaches hegemonic status within a field of inquiry. At such a point questions are no longer asked and researchers unthinkingly apply the method of the day.'' Despite advances in data analysis techniques, the easy compartmentalization of offender groups that typologies promise has yet to be realized. Although young versatile, vagrant, drugoriented, and sexual predators were the four groups to emerge from the current analyses, it should be noted that these groups generally had extensive criminal careers characterized by involvement in diverse forms of offending. Even significant evidence for categorization must be tempered with additional evidence of versatility that would seem to challenge categorization. Although this investigation is one of the largest studies of burglary typologies to date, a larger sample would have provided greater statistical power to detect small to moderate effect size differences between subgroups. Future studies should attempt to include some of the aforementioned variables and perhaps, personality inventories as well. Combined quantitative and qualitative approaches in a single study would be able to further illuminate and refine future typologies. Further, little is known about the interactional dynamics of burglars in terms of their decision-making and opportunities (3) and therefore information on these processes would yield practical benefit. Given that burglary is a fairly prevalent crime causing extensive economic and mental harm to victims, larger and more intensive studies of the types of burglars that exist would provide much needed information for crime prevention services and aid forensic criminal investigation.