Contextual Findings
How Much of Overall Crime Was Offender-Identified (Identified to Specific Known Offenders)?
Figure 1 shows the differences between classification concentrations of reported crime volume for offences overall and for what we call “offender-identified crime” or “offender-detected crime”, i.e. offences for which at least one offender was identifiable. With violence against the person offences, for example, there was a greater percentage of the offender identified crime (28%), compared to the percentage of overall reported crime (22.1%). Theft offences made up the highest percentage of overall crimes reported (48.9%) and the highest percentage of offender identified crimes.
A different picture was observed in terms of crime harm, as shown in Fig. 2. Crime classifications with lower rates of offender identification were less represented in the distribution of crime harm by offence categories.
Repeat and One-Time Offenders Over Time
Of the 39,545 unique offenders named in our dataset, 14,548 (36.8%) committed more than one offence; the remainder were one-time offenders (within our time period). Between 2010 and 2016, almost three-quarters of all offender-identified crimes were committed by a repeat offender, a pattern that remained relatively stable year on year (as shown in Fig. 3).
Specialist and Generalist Offenders
As shown in Fig. 4, no offender offended in all nine crime categories, with the most being eight categories. Specialist offenders (one category only) represented 31.8% of all repeat offenders and 19.2% of crimes committed by repeat offenders. Generalists with just two crime categories represented 41.6% of repeat offenders and accounted for a further 29.6% of the total crimes committed by repeat offenders. Smaller percentages of crime and associated offences were found as the number of categories increased.
Repeat Offenders and Harm
Repeat offenders accounted for 68.7% of all offender-identified crime harm in the dataset, at a level which was largely consistent throughout each of the 7 years analysed. Among repeat offenders, “specialists” accounted for 22.5% of harm, slightly more than their proportional contribution to volume (shown in Fig. 4) but considerably less than the proportion of overall repeat offenders that they make up. Figure 5 shows that, as with volume of crime, the more “generalist” an offender is (i.e. the more types of crime they were linked to), the greater the proportional difference of their contribution to harm committed by repeat offenders.
Our findings also show that the more diverse an offender’s behaviour in terms of the range of crimes committed, the more harmful they were. An ANOVA test of the results shown in Fig. 6 indicated a statistically significant variance in the sequence (F (7, 14,540) = 102.63, p < 0.001). Tukey’s honestly significantly difference (HSD) post-hoc test indicated that 26 of the 28 possible combinations inferred by this figure were statistically significant.
Offending Frequency, Escalation and Intermittency
Figure 7 shows that the initial probability of a first time identified offender being linked to a second offence was 37%. The conditional probability of further attributions of crimes to offenders rose with each subsequent offence up until the thirteenth event, when the risk level flattened. After that point, conditional probability ranged between 88 and 92%. On the thirtieth event, it becomes 99% likely to report a thirty-first event. The maximum number of offences identified for any one offender was 138 offences over the 7-year period. After the thirtieth event, the numbers of offenders at those levels of linked crimes become very low. Nonetheless, probability scores continue to range mostly between 90 and 99%. After just twelve events, all subsequent probabilities exceed 88%, which indicates that chronic offending is highly probable to remain chronic.
Intermittency
A related point about future offending is that with each additional offence, the next offence occurs with increasing speed after the last offence. In the criminological concept of “intermittency”, the crime-free time in between crime declines.
Figure 8 shows that on average, there were 333 days between the reported first and second offence. Between events 2 and 3, this time period dropped to 256 days. At the 36th event, intermittency dropped to its lowest average of 27 days. A one-way ANOVA was performed on the first ten pairings for number of days between offences and showed a significant difference between the groups (F [8, 41,491] = 232.7, p < .001). Post-hoc comparisons were undertaken using the Tukey HSD test, which indicated the most pair relationships were significantly different (see Liggins 2017: Table 3).
Escalation
While future offending becomes more likely and sooner as the number of prior offences increases across each offender’s career (on average) within our 7-year window, it does not become more harmful. To the contrary, recidivism becomes generally less harmful with repeated offending, as Fig. 9 shows.
For all first-time offences, Fig. 9 shows an average crime harm score of 74 days of recommended imprisonment (n = 39,545 first-time offences). This was the highest average crime harm score across the chronological sequence of offending. The trend for average crime harm scores appears to mostly show a downward trend with peaks and troughs but no average crime harm score higher than the first reported offence. While the fluctuations become very wide after the 12th offence, the drop from an average of 74 at offence #1 to 36 at offence #12 is a steady drop of over 50% in CCHI scores. While there will be individual exceptions, the best bet across all offenders under management in that time period was that future harm levels would drop. That would include the felonious few, whose participation may persist well into the sequence shown in Fig. 9.
The “Felonious (Power) Few”
Figure 10 shows that for the 7 years combined, 20% of offender identified harm came from 0.6% of unique offenders, 50% of offender identified harm came from 2.7% of unique offenders and 80% came from 7.6% of unique offenders. Of those who were responsible for 80% of the offender-identified harm, this equated to 5.2% (17,033) of all reported crime offences and 31.7% of all crime harm (i.e. including those offences where an offender had not been attributed). Compared to offender identified crime, 20% of crime volume came from 10.3% of known offenders, 50% came from 21.2% of offenders and 80% came from 54.2% of identified offenders.
Figure 11 shows the total percentage of all felonious few offenders (producing 80% of crime harm) broken down between repeat and one-time offenders. This shows that the latter make up a smaller proportion of power few offenders and that typically there were approximately two repeat offenders in the power few for every one-time offender.
The contribution of individual offenders to overall offender-attributed crime harm had broadly the same ratio of 2:1 in favour of repeat offenders, with the notable exception of 2015, as shown in Fig. 12.
Tracking the list of the “felonious power few” cohort responsible for 80% of offender-identified harm in 2010 (n = 610) into subsequent years, we found a very high rate of attrition from the first year onward. As Fig. 13 shows, by 2011 just 3.4% of the 2010 felonious power few offenders remained in the most 80% of CCHI cohort calculated for the next calendar year, and more than three-quarters of that 2010 felonious few had no subsequent arrest in 2011. This pattern continued in every subsequent year up to 2016, by which time less than 1% of 2010’s power few remained among the most harmful offenders and almost nine in 10 had had no further detected offences.
What this analysis cannot tell us, however, is what proportion of the 2010 felonious few was in prison for all of 2011 or all or any part of 2012, 2013 or any year through 2016. Theoretically, the higher the CCHI value attributed to an identified offender in 2010, the more likely they were to be convicted and sentenced to prison. By definition, the higher the CCHI value among convicted offenders, the longer any prison sentences would be. Yet from a policing standpoint, it does not matter whether these offenders have been removed from the population. What matters is whether there is a means by which police can track the whereabouts of any recent list of a felonious few and specifically whether they are in prison or not.