Cost-Benefit Analysis Based on Referendum CV: Dealing with Uncertainty

Abstract

An interesting and potentially important, but little appreciated, result of the use of dichotomous-choice contingent valuation (CV) techniques is that there are about a dozen ways to deal econometrically with the data on bid acceptance to construct a mean or median estimate of willingness-to-pay benefits, none of which is clearly the correct choice. They produce measures of central tendency that can differ, from lowest to highest, by factors ranging from 5 to 15. This is a source of uncertainty in benefit or damage estimation that can easily dominate the more traditionally recognized statistical uncertainty based on noise in the response (and respondent) data. It potentially can introduce ambiguity into a cost-benefit test of economic feasibility if a prospective investment either passes the test based on some willingness-to-pay measures or fails it based on others. These points are illustrated using actual data from a well-designed dichotomous-choice CV survey administered to residents of São Paulo, Brazil, as part of the application process for a loan to cover a multiphase project to improve the quality of the Tietê River.