The relationship between physical activity and child health and development is well-documented, yet the extant literature provides limited causal insight into the amount of physical activity considered optimal for improving any given health or developmental outcome. This paper exploits exogenous variations in local weather conditions observed across random time use diary dates for the same individuals over time to investigate the causal impact of physical activity on a comprehensive set of health, non-cognitive development, and academic outcomes of children and adolescents. Applying an individual fixed-effects instrumental variables model to a nationally representative panel dataset from Australia, we find that physical activity leads to widespread benefits in child development. These include improved health, social and emotional development, and lower health expenditure. The results further indicate that physical activity offers greater developmental benefits for females. However, we find no evidence that physical activity improves academic performance. Our study highlights that the 'optimal' amount of time that children and adolescents should spend physically active each day varies by the health or non-cognitive development outcome of interest. The results are robust to a series of specification and sensitivity tests, including an over-identification test and controlling for weather conditions recorded on the day when development outcomes were assessed.