Handling Uncertainty in Cost Benefit Analysis



Handling Uncertainty in Cost Benefit Analysis

Authors

Marie Aarestrup Aasness, NTNU/NPRA, James Odeck, NTNU/NPRA

Description

A simple statistical approach to assessing uncertainties in CBA’s is developed and tested on project rankings in Norway. Accounting for uncertainties changes project rankings. Hence, uncertainties are necessary as aid to decision making.

Abstract

Cost Benefit Analyses (CBA) are used for evaluating transport investment projects in many countries worldwide. In particular, CBA often plays an important role when ranking individual investment projects against each other, e.g., as in national transport plans. Irrespective of whether CBA's are structured explicitly in terms of contingencies and their probabilities, there is the problem of uncertainties with regards to the magnitude of the impacts being predicted and the monetary unit values that are assigned to them. Taking account of uncertainties in CBA is informative for decision makers â€' some projects may have a high probability of not delivering net benefits even if the basic net benefits are high. Consequently, the additional information regarding uncertainties in CBA's may change the rank order of projects in a portfolio of projects. Unfortunately, the practices of CBA calculus in Europe and most part of the world do not handle uncertainties. A possible reason for the lack of handling uncertainties in CBA is that the CBA calculus is itself complex and hence, handling uncertainties is even more complex.

In this paper we claim that the uncertainties inherent in CBA’s of transportation projects should be addressed and presented to decision makers. This is because handling uncertainties provides relevant information that may be useful for decision making such as ranking of projects. We use a portfolio of 44 projects from the Norwegian road sector to demonstrate how accounting for uncertainties may impact project rankings. The methodological approach we use is the exploitation of simple statistical relationships. Specifically, for each individual project we calculate the standard deviations of impacts across the life time of projects. The standard deviations are then used to construct a confidence interval of which the net benefits are expected to fall. As is well known in statistics, if the confidence intervals of two observations (two projects CBA results) are distinctly different, there is no doubt as to the ranking. However, if they overlap there is no statistical reason to assume or take for granted that one is preferred to the other; a useful information that cannot be derived without a consideration of uncertainty and which, naturally may impact ranking of projects as compared to status quo practices of CBA. Additionally, we conduct a simple sensitivity analyses by varying some elements of BCA and then compare the results to the ones derived through the statistical approach mention above.

Our main results demonstrates that there are overlaps in confidence intervals of BCA results between several projects attesting that decision makers need knowledge of uncertainties and should not base their decisions solely on basic CBA without some form uncertainty assessment; basic CBA results do not discriminate between projects adequately. Furthermore, the simple sensitivity analyses confirms that our approach to handling uncertainties with regards to project ranking is robust. We urge practitioners of CBA to consider addressing the inherent uncertainties in their calculus to make CBA a more complete basis for informed decision making. We are however, the first to point out that there still some future research opportunities in this topic area. This concerns e.g., the use of monte Carlo simulation to ascertain uncertainties in CBA at the project selection level and to compare it with other methods such as ours.

Publisher

Association for European Transport