Measuring preferences with hybrid conjoint analysis: The impact of a different number of attributes in the master design

Abstract

Experiments, using approximately 300 college seniors, were run to compare the cross-validations of a traditional conjoint model, two self-explicated models, and several hybrid conjoint models. This experiment varied both the number of respondents across which the hybrid task was aggregated and the number of attributes used to describe each concept. The hybrid conjoint model displayed strong cross-validation evidence. Furthermore, the sample sizes, across which the hybrid models were aggregated, had little impact on the predictive power of the hybrid models. Contrary to expectations, increasing the complexity of the design, i.e., the number of attributes in the profiles, did not lead to higher cross-validations for the hybrid models relative to the traditional conjoint model. Consistent with previous studies, segment-level hybrid models did not lead to higher cross-validations than corresponding aggregate models.