Uncertainty in the Economic Appraisal of Water Quality Improve

The value of the services provided by environmental quality improvement projects are subject to large margins of error. These services are not sold in markets and have no established prices, and often have to be valued through stated preference surveys (i.e. contingent valuation). These uncertainties are necessarily transmitted into uncertainties about the net present value (NPV) of environmental investments. The NPV distribution can and, we argue, should, be quantified and evaluated during the project design and approval process. The statistical techniques for imputing benefits using referendum contingent valuation data are sensitive to the analyst’s assumptions. Equally justifiable parametric and nonparametric evaluation routes to average benefits may produce very different mean estimates, estimates that are outside the 99.5% statistical confidence intervals generated by any given route. In short, average (and, by implication, total) benefit estimates based on contingent valuation are fraught with statistical uncertainty. This source of uncertainty can be measured by the standard error of the mean, computed using any particular approach. However, the more important methodological uncertainty about which approach to extracting mean benefits is best has no such easy resolution, and both types of uncertainty about benefits matter. In addition, environmental investments often are accompanied by uncertainties about execution timing provoked by institutional obstacles, divergent interests of stakeholders, and the behavior of the natural world the project operates on and in, as well as the more familiar uncertainties about costs and economic prices. To reflect all of these uncertainties, the economic cost-benefit analysis demonstrated in this paper employs Monte Carlo simulation, which permits their effect on the distribution of project net present value to be quantified. The paper argues that Monte Carlo risk analysis offers a more comprehensive and informative way to look at project risk ex-ante than the traditional (and often arbitrary), one-influence-at-atime sensitivity analysis approach customarily used in IDB analyses of economic feasibility. The case for probabilistic risk analysis is made using data from a project for cleaning up the Tietê river in São Paulo, Brazil. A number of ways to handle uncertainty about benefits are proposed, and their implications for the project acceptance decision and the consequent degree of presumed project risk are explained and illustrated.