While many risk factors for child physical abuse are known, little research exists examining these in multilevel contexts including both individual and environmental influences. The authors examined the roles of individual-, family- and community-level factors such as socioeconomic status (SES) in determining the likelihood of child physical abuse in Guangzhou, China. Twenty-four schools were recruited by stratified random sampling, with 6628 junior high-school students aged 13-16 years participating. Parental child physical abuse experience, together with family and community levels of SES among students were measured and their relationships were investigated by applying univariable, multivariable and multilevel logistic regression models. Univariable, multivariable and multilevel logistic regression models were applied. Six-month prevalence of minor, severe and very severe assaults were 23.2%, 15.1% and 2.8%, respectively. A U-shaped association between family SES and likelihood of severe assaults was identified. In the multilevel model, indicators of low family SES, mother's higher occupational and educational status remained significantly independent predictors of physical abuse. Internal migration status was associated with higher risk as was younger age. The authors suggest that previous categories of risk factors for physical abuse may be too simplistic, and that further research on social and environmental influences may usefully inform intervention programs.