Aggregation Issues in Conjoint Analysis

In applications of conjoint analysis, results obtained at the individual level have to be aggregated for managerial purposes. For a main-effects part-worth model such aggregation can take place by averaging estimated part worths, averaging derived attribute importance weights, and by aggregating choice predictions to obtain market share estimates. In this paper, we discuss aggregation issues, and we show that substantive conclusions may be affected by the number of intermediate levels used for an attribute, if rank order preference judgments are collected.