Equal versus Differential Weighting for Multiattribute Decisions: There are No Free Lunches

Abstract : Recent work on parameter insensitivity in linear models (equal weighting arguments) is examined for its implications for multiattribute decision making. The key factor for the case equal weighting. It is argued that the non-negative attribute intercorrelations upon which the case for equal weighting of attributes is based does not generally hold for multiattribute decisions because the very tradeoffs which create the decision problem imply negative intercorrelations. After an examination of the likely values of such negative intercorrelations, the effects of using incorrect weights in multiattribute decision making on the correlation between true and estimated evaluations and on the expected utility loss due to use of incorrect weights are evaluated. Implications of the theoretical results are discussed in relation to the precision with which multiattribute methods must assess the weights. Suggestions are included for how to use the theorems in this paper for determining the required accuracy of weight estimation for any gived applied problem