You Can’t Forecast Accurately …. But

I lifted the following story from Grant Williams’s article from (John Mauldin’s “Outside the Box”) newsletter:

During World War II, [Nobel laureate, Ken] Arrow was assigned to a team of statisticians to produce long-range weather forecasts. After a time, Arrow and his team determined that their forecasts were not much better than pulling predictions out of a hat. They wrote their superiors, asking to be relieved of the duty. They received the following reply, and I quote “The Commanding General is well aware that the forecasts are no good. However, he needs them for planning purposes.” 

In my white paper on Forecasting under Uncertainty, I made the point that there are no accurate forecasts about the future, whether made by banks, governments, businesses or other entities. Arrow’s quote would seem to apply to many business forecasts today.

If you need forecasts for planning purposes, it is more useful to consider a range of values around each factor that make up the projection. For instance, when projecting the revenue to be generated from a new product, there will be uncertainties around price, market share, market size, competitors, margins and other factors. The range may be considerable, particularly if you are entering a new market and less so where you have deep experience. Even in the latter case, one can never be 100% sure.

Since there is no such thing as an accurate forecast, the best you can do — and should do — is to provide a forecast that reflects the impact of uncertainty, indicating that your project NPV “could be as low as ….” or “could be as high as …,” clearly indicating that you have addressed unknowns and, if prudent, have developed contingency plans.

As addressed in the blog “Resolving Conflict and Confusion with Objectivity and Evidence” by Somik Raha, the “tornado chart” is a most useful tool in dealing with uncertainty.



  1. Don Creswell says:

    I am pleased to append this comment on my blog, from Peter McNamee,my colleague at SmartOrg. Peter is a highly experienced decision analyst and co-author of the book Decision Analysis for the Professional.

    I liked this article, but there is one point it missed — perhaps well encapsulated in the quote from Arrow. “However, he needs them for planning purposes.”

    Sometimes a deterministic forecast is needed to provide consistency to evaluations and to coordinate efforts, whatever the quality of the forecast.

    For instance, oil price. Imagine we have six projects that depend on oil price. If they all use different forecasts for oil price, we will not be able to compare their evaluations. If they all use the same forecast for oil price, we will have achieved comparability.

    This benefit comes almost independent of the quality of the oil price forecast, assuming that the forecast is not ridiculous. It enables discussions on a common basis. Of course we can improve management and conversations and decisions by improving the quality of the oil price forecast (perhaps including the effects of uncertainty).

    This is, I believe, the reason why some companies have “official” forecasts of key variables.

    A real problem comes up if a group believes the “official” forecast is “wrong” (outside their assessed range). If decisions depend on which forecast is used, at least a real problem has been surfaced and identified for discussion and resolution.

  2. Having worked with oil & gas companies for many years, I have often heard the argument presented by McNamee (who’s book is brilliant) on the benefits of working with a single path (deterministic) oil price. I have several issues with this argument. First, the uncertainty which has the largest impact on most oil & gas projects is the commodity price. Ignoring this uncertainty is not providing a very realistic picture of the overall uncertainty in the value of the project. In a static few of the project, and with the assumption of a risk-neutral; i.e. expected-value, corporate decision maker, this may not be very important. However, the dynamic nature of oil & gas investments makes it very important to include a good model for the uncertainty in the cash flows which will be a strong function of the uncertainty in the commodity price.
    Secondly, many oil & gas executives will argue, and rightly so, that the main purpose of valuation is not to assess the absolute value of any given project but to get the relative ranking of the projects. As the uncertainty in the commodity price has the same impact on all of the projects, we can keep this uncertainty “outside” the model by using a deterministic price path. This may sound reasonable but is incorrect. Most oil & gas cash flow models are non-linear (e.g. due to tax regimes or non-linear functions in the software implementation of the models). One result of this non-linearity is that the ranking of the project will often differ when using a deterministic price model versus a stochastic price model.
    Finally, in any valuation exercise it is important to define “value”. A useful definition (benchmark) for a public company is to define value as the market value of the risky cash flows from the project. With this definition, we should develop a price model which is consistent with the market’s view on the future oil prices. This approach has been advocated by prominent researchers such as Jim Smith at Duke University and David Laughton at the U of Alberta.
    So … I very much agree with the sentiment that our forecasting ability is very poor – including our ability to forecast oil prices. However, is assessing the value of projects for a public company, we should not try to forecast the “true” (actual outcome) price path but, rather, the market’s view on the future prices. This can be done by using forward information (futures, options and forwards) readily available in the market combined with a consistent price process. Doing this will provide the consistency many executives are looking for.

Leave a Reply

Leave a Reply