In the stated preference method called graded pairs comparisons respondents are asked to rate the intensity of their preference for their preferred alternative in a pairwise comparison of alternatives. Econometricians anticipate that the additional information will improve statistical efficiency compared to the standard DCE format. However, this paper reveals that added information inherent in graded pairs scale does not provide smaller standard deviations for the WTP estimated. Secondly, the ordered-response regression models employing the full range of the graded pairs data tend to overestimate WTP, which presumably is caused by the inherent tendency of the ordered-response models to 'predict to the extremes'.
Copyright (c) 2006 John Wiley & Sons, Ltd.