Objectives: It is well-known that in cross-sectional analyses, agreement between informants is modest as best when rating ADHD and other disruptive behaviour disorder symptoms. We here aimed to develop recommendations for the use of multi-informant data in the context of longitudinal developmental analyses that examine symptom trajectories over time. Method: Using parallel process modelling, we estimated parent-teacher agreement in inattention and hyperactivity/impulsivity symptom initial levels and slopes across the elementary school years (ages 7, 9 and 11) for a community sample of n=1388 youth. We also used these models to examine whether initial levels and slopes differed significantly across informants. Results: Informant agreement was low to moderate and higher for inattention slopes (r=.47) than for hyperactivity/impulsivity slopes (r=.23). Parents and teachers reported opposite developmental trends for inattention with teachers reporting declines and parents reporting increases over time. Parents reported overall higher levels of hyperactivity/impulsivity but there were no average informant differences in slopes. Conclusion: Of the options available, we recommend specifying separate but correlated factors for different informants in developmental analyses of ADHD. This can be achieved within latent growth curve and growth mixture models.