Paper from the Young Offenders' Unit that builds on an earlier analysis (Fergusson et al., 1976) of the extent to which juvenile offending could be predicted by age 17 years. That earlier analysis found maximum prediction could be obtained by combining three sources of information: the boy’s 'delinquency proneness', socioeconomic status, and race. This paper developed a model that considered several predictor variables which fell into two clusters of correlated variables: social background variables (SES and race) and school performance and adjustment measures at age 10. Results show that the association between race, SES and juvenile offending cannot be explained by intervening psychological variables relating to the child’s social adjustment. The child’s level of social adjustment is related to his offending behaviour almost independently of his social background, showing that the variation in juvenile offending between groups is explained by social factors, while the within groups variability in offending reflects psychological factors. The results imply that differences in offending rates for various social groups can only be influenced by ameliorating the social conditions that give rise to these differences but offending rates within social groups can be reduced by treating those factors which tend to give rise to maladjusted or disturbed behaviour. It is noted however that the precision of the model is low with the entire model accounting for less than 10% of the total variation in offending behaviour (due to 90% of the sample having committed no offences and the lack of sufficient effective predictor variables).