Spurious regression problems in the determinants of health care expenditure: a comment on Hitiris (1997)