CVM SPELLS RESPONSES TO HYPOTHETICAL QUESTIONS

Part of this note takes a new look at an earlier study of mine (1972), sometimes used-incorrectly-in support of CVM; it is shown that while Cummings and Harrison are right in criticizing this study for not prov iding such support, they do so for the wrong reasons. Moreover, it is argued here (a) that even if it could be shown that CVM provides accep table accuracy in trial runs, this property cannot be counted on when CVM is used for actual decisionmaking; (b) that the need for CVM has b een exaggerated, since CVM is neither the only approach to estimating non-use values, nor obviously better than using methods that estimate only use-values.