A Practical Measure of Offence Seriousness: Sentence Severity , continued
Study 1. Ideally, the choice of values for the seriousness of each offence should be guided by the goal of reducing measurement error. Measurement error was reduced in Study 1: (a) by eliminating the identified jurisdictional variants and (b) by averaging the NSCS offence seriousness scores for the range of crime descriptions within any single CCC category. The value used as the measure of offence seriousness for each offence category, then, was the mean NSCS offence seriousness value for crime descriptions in that category. This approach assumes that the range of crime descriptions used in the NSCS was representative of the range of criminal acts which typically occur in any given offence category and that the mean sentence length calculated for the sentence severity measure was based on a similar range of criminal acts. Table 1 contains the 19 offenses analyzed in Study 1.
Insert Table 1 about here
Study 2 In Study 2 the robustness of the relationship between sentence severity and offence seriousness was examined under conditions presumed to increase measurement error. In this study, the range of offenses was extended, the jurisdictional variants were included rather than excluded, and the value of offence seriousness was determined by choosing the single most representative crime description associated with each offence category. The latter task was accomplished by submitting the list of crime descriptions to the previously-mentioned expert in inmate classification for a decision on the most representative or modal example of each offence category. The 33 offenses analyzed in Study 2 are presented in Table 2.
Insert Table 2 about here
Two adjustments of the data were considered. Firstly, it had been determined by the psychologist expert in inmate classifications that the NSCS offence categories for murder one and murder two were not differentiated in the NSCS crime descriptions in ways meaningful to the CCC murder categories. For present purposes, therefore, murder one and murder two were combined, the sentence lengths for murder were averaged, and NSCS crime descriptions were found to represent the combined murder category. Secondly, given the range of social harm consequences deliberately built into the NSCS crime descriptions, it might have been justifiable to eliminate crime descriptions with extreme values on the grounds that they might have distorted the representativeness of the derived (averaged) seriousness scores. However, the mean and median values for the seriousness of each offence category were approximately equal, suggesting that no significant distortion due to outliers had occurred. Therefore, no crime descriptions were eliminated from the present dataset on this basis.