Developing a Complete Sentence Severity Scale using Extended Goodman RC models

Abstract Purpose The aim of this paper is to construct a single sentence severity scale incorporating the full range of custodial and non-custodial sentences meted out by the courts. Such a scale would allow us to measure and rank the severity of sentences, relative to other sentences. Methods We use disaggregated individual level sentencing data to model the association between offenses and their associated sentences using the Goodman Row Column (RC) Association Model. We then extend this model to control for three legal factors; conviction history, offense plea, and number of offenses, to produce a series of standardised scores. We use linear interpolation and extrapolation to convert the scores to equivalent days in custody. Results The scores from the model enable the sentences to be ranked in order of severity; longer custodial sentences dominate at the severe end whilst non-custodial sentences congregate towards the lower end. In the middle of the scale, non-custodial and shorter custodial sentences interweave. We then demonstrate one use of the scale by applying it to Crown Court data, illustrating change in sentencing severity over time. Conclusions The Goodman RC Association Model provides a suitable methodology for scoring sentence severity on a single scale. The study found that by extending the model, we were also able to control for three legal factors. The sentence severity scale, as a research tool is specific to England and Wales but the method is universal and can be applied in any jurisdiction where the relevant data is available.