Reporting bias and heterogeneity in self-assessed health. Evidence from the British Household Panel Survey