A formal and empirical comparison of two score measures for best-worst scaling

Best-worst scaling (BWS) is a method that asks individuals to choose the most and the least preferred option from a set of available options. There has been extensive discussion and evaluation of the use of scores (data summaries) in the analysis of such data. Here we motivate, summarize, and compare the usefulness of two such score measures: the analytical closed form solution (Lipovetsky and Conklin, 2014, Journal of Choice Modelling) and normalized best-worst scores (Louviere et al., 2015, Cambridge University Press). We conclude that both have underlying motivations in the maxdiff model of best-worst choice and that the analytical closed form solution provides better fits to the aggregate choices in several best-worst choice data sets.