How well do socio-demographic characteristics explain variation in childhood safety practices?