Empirical Investigation of Ranking vs Best–Worst Scaling Generated Preferences for Attributes of Quality of Life: One and the Same or Differentiable?

Objectives: The objective of this study was to investigate the degree of inconsistency in quality-of-life attribute preference orderings generated via successive best–worst scaling (a form of ranking whereby the respondent chooses the best and worst attributes from a choice set, these attributes are then eliminated and the best and worst attributes from the reduced choice set are then chosen and this process is continued until all presented attributes are eliminated) and conventional ranking methods (whereby the respondent chooses the best, second best and third best from a choice set until all presented attributes are eliminated). Methods: An on-line survey was developed for administration to two general population samples comprising younger people (aged 18–64 years) and older people (aged 65 years and above). Data were analysed in STATA through an empirical examination of the relative level of choice inconsistency (randomness in responses or the variability in choice outcomes not explained by attributes and their associated preference weights) for successive best–worst in comparison with the conventional ranking method for the younger and older person samples. Results: For the younger person sample, both methods were found to be similarly consistent. In contrast, for the older person sample, ranking performed relatively worse than best–worst scaling with more inconsistent responses (tau = 0.515, p < 0.01). Conclusions: These findings lend some support to the hypothesis initially propagated by the developers of best–worst scaling that it is a comparatively easier choice task for respondents to undertake than a traditional ranking task.