Guide to COST-BENEFIT ANALYSIS of investment projects - Ramiri

Reference ForecastingThe question of where to look for relevant distributions arises. One possible approach is ‘Reference Forecasting’. i.e.taking an ‘outside view’ of the project by placing it in a statistical distribution of outcomes from a class of similarprojects. It requires the following three steps:- the identification of a relevant reference class of past projects, sufficiently broad to be statistically meaningfulwithout becoming too generic;- the determination of a probability distribution of the outcomes for the selected reference class of project;- a comparison of the specific project with the reference class distribution and a derivation of the ‘most likely’outcome.According to Flyvberg (2005) ‘The comparative advantage of the outside view is most pronounced for non routineprojects. It is in planning such new efforts that the biases toward optimism and strategic misrepresentation are likelyto be largest.’Systematic RiskIn financial and economic literature there is a distinction between variability that is random and, at least in principle,diversifiable, and variability that is correlated with overall market trends and economic growth. Non-diversifiablevariability is usually described as systematic or market risk.Risk that is diversifiable, or non-systematic, is regarded for most practical purposes as costless in the public andprivate sectors. Public sector risks are generally spread across taxpayers, again reducing the variability faced by anyindividual to a small fraction of individual income.In welfare economics the cost (or benefit) of systematic variability is conventionally estimated from first principles,using a utility function in which the marginal utility of extra income declines as the individual’s income increases.This can materially affect the estimated value of the benefits of schemes that produce the highest benefits in yearswhen incomes would otherwise have been very low. Such a utility function usually assumes a constant but plausiblevalue for the income elasticity of marginal utility with respect to income (normally abbreviated to the ‘elasticity ofmarginal utility’).Risk analysisHaving established the probability distributions for the critical variables, it is possible to proceed with the calculationof the probability distribution of the project’s NPV (or the IRR or the BCR). The following table shows a simplecalculation procedure that uses a tree development of the independent variables. In the sample reported in the table,given the underlying assumptions, there is 95% probability that the NPV is positive. The more general approach tothe calculation of the conditional probability of project performance by the Monte Carlo method was presented inChapter 2. See also references in the bibliography.Table H.1 Probability calculation for NPV conditional to the distribution of critical variables (€million)Critical variablesResultInvestment Other costs Benefit NPVValue Value Probability Value Probability Value Probability74.0 0.15 5.0 0.03-13.0 0.2077.7 0.30 8.7 0.0681.6 0.40 12.6 0.0885.7 0.15 16.7 0.0374.0 0.15 2.4 0.08-56.0-15.6 0.5077.7 0.30 6.1 0.1581.6 0.40 10.0 0.2085.7 0.15 14.1 0.0874.0 0.15 -0.7 0.05-18.7 0.3077.7 0.30 3.0 0.0981.6 0.40 6.9 0.1285.7 0.15 10.9 0.05236