Comparing Hierarchical and Nonhierarchical Weighting Methods for Eliciting Multiattribute Value Models

A value tree relating general values and concerns to specific value relevant attributes was constructed to compare three energy options: nuclear, coal, and a combined geothermal and conservation package. Thirty-seven nonexpert subjects provided judgments of the relative importance of attributes in the tree using both hierarchical and nonhierarchical weighting procedures, and they rated the energy options on all attributes and all levels of the tree. From these importance weights and ratings several additive multiattribute value models were constructed and compared with holistic rankings and ratings of the three options. The experiment had three basic findings: First, hierarchical weights were steeper (higher weight ratios) than nonhierarchical weights. Second, groups that differed in their holistic first choice nevertheless showed substantial agreement in their assessment of attribute weights. Third, attribute level ratings of the relative desirability of energy options also agreed rather well across groups, although there was a tendency for each group to favor their holistic first choice. This convergence of multiattribute value model parameters resulted in a "common model" that was most consistent with holistic evaluations of the pro-conservation group, and generally inconsistent with those of the pro-nuclear group. This third finding of differential consistency between model composites and holistic evaluations is interpreted as a result of weight parameter distortions due to social desirability and/or a neglect to consider attribute value ranges when making weight judgments.