Putting Up with Nonsense
Putting Up with Nonsense
There is a great deal of published and spoken nonsense about Monte Carlo simulation, much of which appears to come from critics who have never used it or are biased against it. Like any analysis, the accuracy of its outputs is no better than the accuracy of its inputs. The future is uncertain, and it is precisely when the uncertainties are greatest that Monte Carlo is most useful. It is risky to base an analysis on a single set of uncertain values that are essentially an analyst’s best guesses or are considered most probable. The risks are better recognized by doing scenario analyses, using worst-on-worst and best-on-best conditions in addition to whatever else is thought likely or possible.
Monte Carlo simulation adds an analysis of probabilities to scenario analyses. In the absence of knowing the type of distribution for input variables, triangular distributions can be used as approxima- tions. Though less than exact, they are better than basing an analysis on the single value regarded as most probable. Presenting the results in the form of downside risk charts and doing sensitivity analyses provides
a better understanding of the probabilities and magnitudes of the risks involved, and potential losses as well as potential gains. Don’t let the number of iterations scare anyone. With today’s computers, thousands of iterations are easily provided, even on laptops. Monte Carlo simulation is not new. It is being widely used in industry— and has been widely used for more than 50 years.
Excel has proven capable of large, complex Monte Carlo simulation models. The author has used Excel to create a successful multiyear, multiproduct financial plan for a corporation that included 35 input variables, each of which was modeled with most probable, minimum, and maximum estimates. The model’s file consisted of 17 worksheets and included statistical summaries, downside risk charts, and column charts for the overall plan and each of five product areas. Input values were entered on a separate worksheet, with the input cells linked to cells on the other worksheets in the file. Each random variable was simulated with 2,000 iterations to provide satisfactory accuracy, with a total of 288,000 random num- bers generated. The file used no techniques that are not covered in this chapter. No add-on software was used. Contrary to what some would have you believe, Monte Carlo simulation is not terribly difficult, and you should not be discouraged by their ill-formed misgivings.
Finally Yes, there are random events that occur outside the scope of even the best analysis. Wars, fires, terrorist
attacks, hurricanes, deaths, bankruptcies, and other disasters are a fact of life. After doing the best possible, you still need to cope with these things in your business as well as your personal life. That is why one buys life insurance to protect one’s family against the death of the breadwinner, looks in both directions before crossing a street to avoid being hit by an oncoming car, builds one’s home on ground that is not likely to be flooded or torn apart in an earthquake, avoids dangerous situations, prepares as best as pos- sible for disasters, and so forth. With luck, you and your business will survive. You can base your decisions on probabilities and risk analysis, but you still need to be prepared for the unknown random events that might hit you.
This page intentionally left blank