Commuting time to work and behaviour-related health: a fixed-effect analysis