Background Child mental health problems have been linked to a variety of negative outcomes over the lifecourse. In recent years there has been a focus on exploring the development of mental health problems during childhood and adolescence. However, these analyses often use data from cohort studies, which suffer from differential attrition, and parent-reported data, which is liable to be affected by the parent’s own mental state. The current study used routine data to explore what happened to children’s mental health problems during this time; whether this differed for children from different backgrounds; and whether schools made a difference to the development of these problems over the first three years. Methods This study used data collected through nurseries and primary schools for all children in Local Authority funded nurseries or schools in Glasgow City. The study used Goodman’s Strengths and Difficulties Questionnaire to assess internalising and externalising symptoms. Data were collected for children in their preschool year (age 4–5) in 2010 and then in Primary 3 (P3, age 7–8) in 2013. Data were available for 2131 children. Multilevel binomial models were fitted in MLwiN in order to explore the impact of child and school level characteristics on mental health problems at P3. Results Data were assessed for representativeness. Children from an ethnic minority, as well as affluent children with mental health problems, suffered from greater attrition. Children’s mental health in Glasgow was similar to the UK, though with heightened levels of hyperactivity at P3. Results showed a widening of inequalities in mental health problems by deprivation: levels of difficulties in children from the most affluent quintile of area deprivation stayed level between the two time points, whilst levels for children from the most deprived quintiles started higher and increased over time. Multilevel models showed evidence of school effects. Significant predictors of mental health problems at P3 included being male, having been under supervision of the state in early childhood, having an abnormal score at Preschool and being in a school with higher levels of Free School Meals (p < 0.01). Conclusion In conclusion, children in Glasgow show a similar pattern of difficulties compared with UK norms, though with increased levels of Hyperacitivity/inattention by P3. Patterns differ for children from different backgrounds, with children who have experienced adversity having increased levels of difficulties by P3. There does seem to be an effect of schools in the first few years, though a large proportion of this variance is accounted for by the demographics of the pupils within them.