Estimating Net Present Value variability for deterministic models

For decision makers the variability in the net present value (NPV) of an investment project is an indication of the project's risk. Risk analysis is one way to estimate this variability. However, risk analysis requires knowledge about the joint-distribution function of the inputs. Modeling this is often very difficult if not impossible for large long-term investment projects, such as energy infrastructures. In practice, the analysis of the variability is then usually restricted to deterministic sensitivity analysis, such as ‘one-factor-at-a-time’ and scenario analysis. These deterministic analyses, however, do not account for the total variability in the NPV. This paper shows that the use of experimental design (taken from statistical theory) in combination with regression (meta) modeling is a better approach.