Education and social care predictors of offending trajectories: An administrative data linkage study

Criminal behaviour is a global public health problem associated with a wide range of poor health and social outcomes for victims and perpetrators. Such behaviour typically follows distinct pathways or trajectories, with some individuals behaving antisocially throughout their life, and others for only short periods of time such as during their adolescence. However, accurately predicting which pathway an individual child or adolescent may follow remains difficult. Accurate prediction of those who are most at risk of behaving antisocially throughout their life would help to inform targeted interventions in educational, social care or criminal justice settings. Routinely collected educational and social care data may be very important in informing such predictions. Important information on learning development, school attainment, school exclusion, child protection involvement and special educational need is available for every child in state education in the UK. Such information can now be linked with the child's later offending records. This type of data has enormous potential public benefits by being truly representative of the whole population and highly cost effective because the data has already been collected. This study will establish whether it is possible to use routinely collected education and social care data to identify those children and adolescents who are more likely to become persistent offenders before involvement with the criminal justice system begins. This will help influence decisions on how best to support them, potentially reducing criminal offending and its associated social and economic costs.

In this 12-month project I will use routinely collected information on crime records for individuals aged 10-32 years born on, or after, 31st August 1985 p to and including 31st August 2007 which are linked to the same individuals' prior educational and social care records when aged between 4 and 18 years. First, I will use a statistical analysis approach called latent class analysis to identify different trajectories of offending behaviours following a first recorded conviction or caution. Second, I will adopt a statistical learning approach, which is a form of machine learning, to see if it is possible to predict the offending trajectories identified in step 1, using prior education and social care information. With machine learning, computers can learn to make decisions and predictions without being directly programmed to do so and can potentially identify important factors that we as humans may miss.

The identification of children and adolescents at higher risk for persistent offending will be used to inform early intervention approaches and criminal justice responses to reduce offending and by extension contribute to evidence-based policy making. The project findings will also highlight how routinely collected data can be used to improve public services for children and adolescents. In order for these benefits to fully realised, I will document my work on the project for publication on ADR UK, Ministry of Justice and Department of Education websites. I will produce two academic publications and give a series of presentations over the course of the project to other academics and project beneficiaries. I also plan to engage with youth charity organisations and wider project stakeholders to discuss project results and discuss future directions of this work. For example, findings from this project, in consultation with stakeholders, could be used to inform the development of a tool that could discriminate between the likelihood of future offending trajectories in an educational setting, that will help children get the help they need early on.