Random Error and Simulation Models With an Unobserved Dependent Variable as applied to the Benefits and Costs of the Clean Air Act

Most empirical simulation models used in benefit-cost or risk analysis investigate uncertainty based on variability in parameters and conditioning factors. A pure random error term is frequently omitted. Ex-ante benefit-cost analyses create a particular problem because there are not historically observed values of the dependent variable, such as net present value. An estimator for the error variance is developed based on analysis of variance measures and R-squared. When applied to a model of the net benefits of the Clean Air Act, the probability that the net present value is negative increases from .2 to 4.5 percent.