To many people, it may seem “obvious” that conducting a risk analysis will produce a better analysis and decision-making under uncertainty than just static forecasting. However, management are justified in asking whether the extra effort and potential that is required to do so is really worthwhile. Why and when does risk analysis produce better results?

In each a series of blogs, starting here, I will examine some of the benefits of risk analysis (which are discussed in more detail and with examples both in my training courses at Palisade Corporation, as well as within my recent book Financial Modelling in Practice).

Here I wish to simply point out that one role of risk analysis and modelling is to establish the average of the outcomes. The average is an especially important quantity in financial decision-making, where it represents the “cut-off” point for “go/no-go” decisions, at least in theory. For example, the net present value of a project is the average of its discounted cash flows. Generically, financial theory values an asset or project with uncertain returns based of its average value (discounted in some way). The average is important as a decision criterion where it is believer that the project is of a representative or repetitive nature.

A crucial, but generally overlooked point, is that when building static Excel models, such models (with the inputs set at their base values) do not show the average of the output, and are therefore inherently biased. There are several reasons for this: First, when estimating an input value, one will generally estimate a most likely value (any other estimate is less likely and therefore poorer in that sense). Where a variable is not symmetrically distributed, then the base case model will not show average values (as the most likely is not the same as the average). Second, a model may contain a non-linearity (e.g. division of a quantity by an uncertain input, use of IF, MAX functions); in such models a static base value will not generally be the same as the average outcome that would result from a simulation.

Examples of the biases in base case static linear models are those which involve simple summation of the entries. These include cost budgeting (where static entries may have been estimated at their most likely values), and oil reserves estimation (where the total reserves is the sum of those of individual fields, but such individual base values may have been populated at say P50 or P90 values, so that the sum of these is not the true P50 or P90 of the total).

In some sense, risk analysis (or risk assessment) can be thought of as a tool simply to correct for biases in the base static case model. In subsequent blogs in this series we discuss some other reasons for performing risk analysis.

Dr. Michael Rees

Director of Training and Consulting