Background: Carcinomas in children are rare and have not been well studied. Methods: We conducted a population-based case–control study and examined associations between birth characteristics and childhood carcinomas diagnosed from 28 days to 14 years during 1980–2004 using pooled data from five states (NY, WA, MN, TX, and CA) that linked their birth and cancer registries. The pooled data set contained 57 966 controls and 475 carcinoma cases, including 159 thyroid and 126 malignant melanoma cases. We used unconditional logistic regression to calculate odds ratios (ORs) and 95% confidence intervals (CIs). Results: White compared with ‘other' race was positively associated with melanoma (OR=3.22, 95% CI 1.33–8.33). Older maternal age increased the risk for melanoma (ORper 5-year age increase=1.20, 95% CI 1.00–1.44), whereas paternal age increased the risk for any carcinoma (OR=1.10per 5-year age increase, 95% CI 1.01–1.20) and thyroid carcinoma (ORper 5-year age increase=1.16, 95% CI 1.01–1.33). Gestational age <37 vs 37–42 weeks increased the risk for thyroid carcinoma (OR=1.87, 95% CI 1.07–3.27). Plurality, birth weight, and birth order were not significantly associated with childhood carcinomas. Conclusion: This exploratory study indicates that some birth characteristics including older parental age and low gestational age may be related to childhood carcinoma aetiology.