Notes1. We thank colleagues in statistics and econometrics for discussion and suggesting this correction.2. We ran a series of diagnostic checks for normality, linearity and constant variance. In addition, we found no influential observation and no undue effects from multicollinearity.