The British Child and Adolescent Mental Health Surveys (B-CAMHS) of 1999 and 2004 found a substantially lower prevalence of any child mental disorder in Indians compared to the general population (3.4% vs. 9.4%, p<0.001). This PhD sought to understand this apparent Indian mental health advantage through secondary analyses comparing the 16 449 White and 419 Indian children aged 5-16 in B-CAMHS. There was strong evidence (p<0.002) of an Indian advantage for externalising problems/disorders and little or no difference for internalising problems. This was consistently observed for clinical diagnosis and for the Strengths and Difficulties Questionnaire (SDQ) administered separately to parents, teachers and children. Detailed psychometric analyses provided no evidence that measurement bias could account for this observed Indian advantage. There was likewise no evidence that the advantage could be explained by participation bias. In multivariable analyses the unexplained difference between Indians and Whites for externalising problems decreased somewhat after adjusting for the fact that Indian children were more likely to live in two-parent families (92.2% vs. 65.4%) and less likely to have academic difficulties (e.g. 2.9% vs. 8.6% for parent-reported learning difficulties). In models adjusting for a larger number of child, family, school and area variables the difference reduced only by about a quarter (e.g. from 1.08 to 0.75 SDQ points on the parent SDQ) and remained highly significant (p<0.001). In both unadjusted and adjusted models, the unexplained Indian advantage for externalising problems was consistently larger in families of low SEP. There was little or no evidence of an ethnic difference for internalising problems/disorders in unadjusted or adjusted models. In conclusion, the Indian mental health advantage is genuine and is specific to externalising problems/disorders. Family type and academic abilities mediate part of this advantage, but most of the advantage is not explained by major child mental health risk factors.