In population-based and general offender-based samples, criminal careers that are one standard deviation more than the average criminal career length are roughly the same duration as the mean criminal career length in samples of serious, at risk, and repeat offenders. For example, when this formula is applied to males in population-based and general offender-based samples, a criminal career can be considered persistent at 24 years in the CSDD and 15 years in the study of Ontario male offenders (Day et al., 2012). In comparison, the mean criminal career duration in samples of male serious, at risk, and repeat offenders can range from 16 years in the CYA (Warren & Rosenbaum, 1987), 15 years in the Montreal Longitudinal Study (Kazemian, LeBlanc, Farrington, & Pease, 2007), 21 years in data from the National Crime Information Centre (DeLisi, 2001), and 25 years in the Glueck
study on delinquent men (Sampson & Laub, 2003). Using this method, 9.2 percent of the subjects in the population-based CSDD have a criminal career equal to or more than one standard deviation longer than the average criminal career length in the sample, or 24 years. Using the same cut-off for samples of serious and violent offenders, around 40 percent of the offenders in the CCLS and CYA study have or exceed criminal career durations of 24 years (Piquero, Brane, & Lynam, 2004; Blokland, 2005).
As a whole, this evidence supports our argument that persistent offending can be usefully defined as a criminal career that exceeds the average duration for a criminal career in population-based and general offender-based samples. We expect there to be a much smaller proportion of persistent offenders in population-based and general offender-based samples than in samples of serious and violent offenders. We have identified across multiple studies that when using population-based or general offender-based samples, a criminal career one standard deviation (SD) longer than the mean (x̅ ) criminal career length (ccl), or x̅ccl + 1 SD, approximates the mean criminal career duration for serious, at risk, and repeat offenders. To move towards a consistent method of identifying persistent offenders, we posit that in population-based and general offender-based samples:
Persistent offending = ≥ (x̅ccl + 1 SD)
But in serious and violent offender samples it should be: Persistent offending = ≥ x̅ccl
The empirical utility of operationalizing persistent offending as x̅ccl + 1 SD has been demonstrated by Whitten and colleagues (2017), who examined the criminal careers of persistent offenders identified by 35 distinct operationalizations used in the research
literature. Their findings were based on the annually reported official convictions of 411 men from the CSDD. The x̅ccl + 1 SD operationalization identified all offenders with exceedingly long criminal careers as persistent, and excluded all offenders with a criminal career less than
24 years. Persistent offenders identified by x̅ccl + 1 SD were the only group to have lengthy criminal careers, comprise between 5 to 10 per cent of the sample (9.2 per cent), and were responsible for around half of all convictions (45.9 per cent).
We have produced a snapshot of Whitten and colleagues (2017) findings to
demonstrate the advantage of the x̅ccl + 1 SD method over four competing operationalizations. These operationalizations identify offenders as persistent if they: (1) offend before and after the age of 21 (e.g., Bergman and Andershed, 2009; Pulkkinen et al, 2009); (2) offend before the age of 14 (e.g., Hagell and Newburn, 1994; Hay and Forrest, 2009); (3) commit five or more offences (e.g., Kempf-Leonard et al., 2001; McGloin and Stickle, 2011), and; (4) belong to the highest and longest group-based trajectory (e.g., Blokland et al., 2005; Day et al., 2012). Table 16 details the criminal career dimensions for the groups of offenders identified by each operationalization. To put these results into perspective, almost half of the men in the CSDD (43.27 per cent) had been convicted at least once. These offenders, on average, were first convicted at 19.3 years, had a criminal career of 11.15 years, and were convicted 5.14 times.
Table 16. Criminal career statistics of persistent offenders identified by five operationalisations.
Criminal career dimension
Criminal career at least 1SD longer than mean
criminal career
Convicted before and after the age
of 21
Convicted at or before the age of
14 years
Five or more convictions
High rate group- based trajectory Prevalence (%) 38 (9.2) 72 (17.5) 8 (1.9) 62 (15.1) 21 (5.1) Duration Mean (SD) 33.13 (6.5) 23.66 (11.8) 22.12 (13) 22 (12.7) 33.29 (9.24) Minimum - Maximum 24-46 2-46 2-45 1-46 6-46 % duration 1-10 years 0 12.5 12.5 19.4 4.8 % duration 31+ years 60.5 31.9 25 29 61.9
Age of first conviction
Mean (SD) 15.03 (2.8) 14.73 (2.7) 11.37 (.7) 15.4 (5.8) 13.76 (2.1) Minimum - Maximum 10-24 10-20 10-12 10-45 10-18 % onset ≤14 years 44.7 45.8 100 48.4 57.1 % onset ≥18 years 15.8 17.7 0 12.9 4.8 Number of convictions Mean (SD) 10.42 (7.2) 9 (6.7) 13.5 (5.9) 10.9 (5.9) 14.91 (6.5) Minimum - Maximum 2-31 2-31 4-22 5-31 5-31 Lambda .31 .46 .93 .82 .47 Total convictions (%) 396 (45.9) 644 (72.5) 108 (11.9) 674 (74.2) 313 (34.4) % convicted 1-4 21.1 31.9 12.5 0 0 % convicted 15+ 26.3 19.4 37.6 22.6 47.6
The operationalizations ‘convicted before and after the age of 21’ and ‘five or more convictions’ identified far more than 10 per cent of the CSDD men as persistent (17.5 and 15.1 per cent, respectively). Some of these persistent offenders also had criminal careers that were shorter than the CSDD mean (12.5 and 19.4 per cent, respectively, had criminal careers 10 years or less in duration). The operationalizations ‘convicted at or before the age of 14 years’ and ‘high rate group-based trajectory’ also identified persistent offenders with short criminal careers. These offenders also made up a much smaller portion of the CSDD (1.9 and 5.1 per cent, respectively), and were responsible for a relatively smaller proportion of the sample’s total convictions (11.9 and 34.4 per cent, respectively).
A small number of persistent offenders identified by x̅ccl + 1 SD had a late age of onset and infrequently offended. Just over 15 per cent of the persistent offenders identified by x̅ccl + 1 SD were first convicted at or after the age of 18 years. Likewise, slightly more than 20 per cent were convicted four times or less. This is consistent with Moffitt’s (2006) assertions that some persistent offenders would offend at a low rate, and Thornberry’s (2005) presupposition that relatively few persistent offenders may commit their first offence at a late age. The reasons why a persistent offender may have a late onset or offend infrequently are discussed later in this paper.