Elliott, Amy and Francis, Brian (2018) Changing crime mix patterns of criminal careers : longitudinal latent variable approaches for modelling conviction data in England & Wales and the Netherlands. PhD thesis, Lancaster University.
2018AmyElliottAppliedSocialStatistics.pdf - Published Version
Available under License Creative Commons Attribution Non-commercial No Derivatives.
Download (2MB)
Abstract
In criminal career research, there has been a great deal of attention paid to the frequency of offending over the life course. This neglects any changes in the patterns and types of offences being committed. However, it is crucial to explore these patterns of offending in detail and various types of crimes being committed, as this will enhance the understanding of criminal activity and the causes of offending behaviour. This is especially true for policy makers, so they can make better informed decisions when deciding how best to target their resources when it comes to tackling crime. This thesis aims to identify crime mix patterns (different offenders will commit different selection of offences) and how they develop over the life course from two official conviction datasets. The first is the England and Wales Offenders Index (OI). The cohort data of the OI contains the court convictions of offenders from 1963 to the end of 2008 in eight birth cohorts. The other dataset is from the Netherlands Criminal Career and Life-course study (CCLS) which contains data covering the criminal careers of those offenders who were convicted of a crime in the Netherlands in 1977, starting at age 12 and followed up till 2005. The study will provide a contrasting analysis of the two datasets using a Latent Markov Model approach similar to that published in Francis et al. (2010) where the idea of lifestyle specialisation and short-term crime typologies (crime mixes) over five-year age-periods was introduced for female offenders. This approach will jointly estimate the crime mix patterns and the transition probabilities (offenders move from one pattern to another). The study adds methodological innovation in criminology by the use of B-splines in group based trajectory models and in the modelling of Poisson counts in latent Markov models. The thesis also contributes to cross-national research. Not only is it important to be able to identify crime mix patterns in both datasets separately but being able to compare and contrast the results from each country will allow for the examination to check if offender’s crime mix patterns are the same across jurisdictions. These findings will be of great interest to both criminologists and policy makers. The analysis provides a cross- national understanding of progression in the types of crime mixes offenders are involved in, whether some crime mix patterns are more specialised than others in terms of their long term patterns and whether some crime patterns desist earlier than others. The results show that each dataset both have versatile and specialist crime mix offending groups but there are also important differences in the makeup of these groups, with regard to the type of offences. These results are discussed in further detail, along with the issues of how best to carry out analyses upon the two datasets. The additional problems encountered when comparing the two datasets and the strategies used to overcome them are explained. Finally, suggestions for future research are given along with encouragement of replicating the methodologies used in this study upon more recent datasets in other jurisdictions.