Early socio-emotional skills and adolescent offending: Evidence from administrative data

This paper examines which early childhood skill deficits predict adolescent offending in England, through which school pathways, and how robust these associations are to comparisons within schools and within families. Using linked MoJ–DfE administrative records, I estimate an exploratory structural equation model on teacher assessments of child development at age five. This identifies two latent domains: cognitive difficulties (literacy and numeracy) and socio-emotional difficulties (emotional regulation, social relationships and communication). A one-standard-deviation increase in either domain is associated with an approximately 15% higher probability of a caution or conviction between ages 11 and 17.

A mediation decomposition suggests that cognitive difficulties are associated with offending primarily through lower academic attainment. The socio-emotional association runs mainly through school exclusion, the strongest single predictor of offending, yet around half remains unexplained by school pathways. The association between socio-emotional difficulties and offending persists even after accounting for differences in school and family environments. Socio-emotional difficulties also predict the frequency and type of offending, including violent and property-related offences. Finally, skill-offending gradients are considerably flatter among children receiving SEND support, suggesting that risk may be concentrated among children whose needs remain unidentified.

The findings point to differentiated policy responses: academic support for children with cognitive difficulties, and individual-level interventions for those with socio-emotional difficulties, alongside alternatives to exclusionary discipline. In both cases, early identification of needs may be particularly important.

Teams link: https://teams.microsoft.com/meet/391603224522308?p=sx1QMVQPndBvL18MTZ