Do experimental protocols in Conjoint Analysis matter in non Hypothetical settings?

This paper aims at comparing the performance of three conjoint analyses (CA) in terms of estimated partworths, predictive power and estimated WTP: choice experiment (CE); ranking conjoint analysis (RCA) and best-worst scaling (BWS). Comparisons are made in a non-hypothetical setting. For comparison purposes in the last two formats only the information on the most preferred option is considered. The hypothetical CE is used as the benchmark. Olive oil is the food product used in our experiment. Results reveal preferences regularity between samples’ responses across the formats implying not statistically differences in the marginal participants’ WTP. Moreover, in an incentive compatible context, RCA and BWS compared with CE provide similar results regarding to the in-sample and out-of-sample predictive power and also in terms of decision consistency when just only the first rank data is analyzed.