Risk Software

It seems reasonable that in a recession, cost containment would become hot topic, and in the world of supply chain management, it's known as "spend control."

The U.S. Federal Reserve recently released the results of a comprehensive assessment of the financial conditions of the nation's 19 largest banks, which hold two-thirds of American economic assets. This “stress test” was designed to determine the capital buffers required for the banks to withstand losses and maintain lending even in worsening economic conditions. Officially called the Supervisory Capital Assessment Program (SCAP), the test identified the potential losses, resources available to absorb losses, and resulting capital buffer needed.

Monte Carlo simulation was used to determine the potential losses from further defaults on loans. According to Federal Reserve Chairman Ben Bernanke,  “The assessment program was a forward-looking, ‘what-if’ exercise.”

Monte Carlo simulation is one of the most widely used methods of stress testing for capital and operations risk,  according to Investopedia.  It takes into account variables such as interest rates, lending requirements, and unemployment. As any @RISK software user will tell you, this type of sophisticated simulation can be accomplished easily within the Microsoft Excel environment. The result of a Monte Carlo software simulation is a look at a whole range of possible outcomes, including the probabilities they will occur -- a valuable tool when stress testing.


Randy Heffernan
Vice President

This blog is a continuation of the series (started a couple of months ago) which discusses different reasons to perform a risk analysis.  In the first two blogs, I discussed the use of risk analysis to a) establish the average outcome, the average being a crucial reference quantity in decision-making in financial contexts, and b) establish the variability of the outcome, such variability being of high importance in both non-financial and financial decision making in practice. Risk analysis with Monte Carlo simulation can be accomplished directly in your Excel models with Palisade's @RISK software add-in.

Here I wish to make a third point: risk analysis will generally lead to a revised (and more correct) definition of the model.  It will at the same time allow any base (reference) case to be presented within the context of the full range of possible outcomes.

Examples of where differences arise in the two modelling philosophies include: 1. A situation in which there is an event risk with say 40% probability of occurrence would include this risk variable in the risk model, but typically not in a static model (its occurrence not being the most likely outcome). 2. A risk model would generally contain risk mitigation measures that make sense to conduct when the consequences of risky outcomes are considered; however, such line items may not form part of the static analysis.

The variables in the risk model will be an augmented set compared to the static mode, and ultimately, a risk model represents the residual uncertainty once all measures deemed to be appropriate have been implemented; in this sense risk modelling is inherently a process of optimization.

For more about optimization with Monte Carlo software in Excel, see RISKOptimizer.

Dr. Michael Rees
Director of Training and Consulting

In a second effort based right here in Ithaca, two Cornell professors, a climatologist and a horticulture specialist, have begun advising Congress and agricultural policy makers on how farmers can respond to local climate change to keep agriculture a sustainable part of the landscape.   Apparently looking up at the elephant as if it were standing right next to him horticulturalist David Wolfe commented, “We are the first generation of humans in modern history to face this kind of predicament, and it creates a serious problem for decision-makers of all kinds."

The elephant nods sympathetically and asks if the professor and his human relatives know about operational risk software.  

A few years ago, the online magazine for computer gaming Gamesutra published an article by a game developer Alan Carpenter extolling the virtues of using risk analysis to balance role-playing games.  In this case balance meant designing a game that was neither too difficult nor too easy, and it was the costly, time-consuming goal of game development companies--costly meaning an average of $3 million a title!

Carpenter had observed that by using the same operational risk software that the oil and gas industry uses to make decisions under uncertainty, a game designer could take advantage of the thousands of possible scenarios spun out by a Monte Carlo simulation to add a new element of reality to games where context and emotion may be big draws for the gamer but can't sustain entertaining play.

Carpenter advises game developers that many games could be designed using the same Monte Carlo Excel spreadsheet.  He offers a lengthy technical blueprint for games that are based on conflict--war, street fights, etc.--and in revisiting it what intrigues me is the idea that the same infrastructure of algorithms and probability functions that Carpenter lays out to capture events in an imaginary world could just as easily be applied to real-world political and military events.

If risk simulation is being used to help plan world events, it's not widely talked about.  But I suspect this is going on, and I would love to hear from anyone out there who knows more about this than I do.  

This blog is just a bit more fun … we talk about using Monte Carlo simulation to estimate the value of π (3.14159…).  This is of course not at all an efficient way of estimating its value, due to its slow convergence (Look for another post on the convergence of Monte Carlo simulation, coming soon.).

A circle of radius one will have area equal to π, and a square drawn around that circle will have area 4.  When centred at the origin the circle has the equation x^2+y^2=1. Therefore if one were to conduct a simulation in which:

• Random samples of x and y are drawn from a uniform continuous distribution between minus one and one
• At each draw it is tested whether the sum of the squares of the two variables is less than one, and a record kept as to whether this is the case
• At the end of the simulation the frequency with which this is the case will approximate π/4.

(When using @RISK software for decision making under uncertainty to do the above, the easiest way is to simply using the RiskMean Statistics function to calculate the mean of the result of the 1/0 logical check as to whether the draw is inside or outside the circle).

Some other trivia about π are:

• In 1896, the Indiana state legislature was in the process of passing a bill to decree that π should be equal to 3.2. Following a presentation to the lawmakers by a mathematics professor, it was decided to defer the final voting to a later date (which is still outstanding).
• The true figure is of course one that is infinite and without repetition or other known patterns.
• Some approximations include the ratios 377/120 (3.14166…developed in 2 B.C. and accurate to 4 decimal places), and 355/113 (3.1415929…, developed in 5 A.D. good to 6 places)
• The Excel Pi() function lists the value as 3.14159 26535 89790 which is inaccurate in the last listed decimal place (the true value being 3.14159 26535 89793 ….) i.e. to 15 places.
• The problem of calculating π to an arbitrarily large number of decimal places is essentially solved and only restricted by time and money. Already billions of places have been calculated and there seems little practical purpose in pursuing this further.
• Perhaps one of the most simple yet bizarre formula involving π and other fundamental quantities in mathematics is exp(iπ)+1=0, where i is the imaginary number corresponding to the square root of minus 1.

Dr. Michael Rees
Director of Training and Consulting

In a recent comment, I considered the role of speculators in the current volatility of oil prices (and, with the help of a savvy columnist, found them innocent of price manipulation).  At the time, I was under the impression that a speculator was the same kind of critter as a hedger or an arbitrageur.  Not so.  I have been corrected, and now realize these three creatures of the financial world are distinct individuals.  What lumps them together is risk and risk assessment.

While a speculator is someone who accepts a risk of loss in return for a possible reward, the hedger is a more conservative creature.  He or she--or, in the case of a business, it--makes an investment as insurance against loss.  A hedger calculates the value-at-risk in an investment in one market  and then makes a corresponding, opposite investment in a different market. The hedger swaps the probability of big win for the increased probability of a smaller win.

I, however, am a true fiscal chicken and 99 percent risk averse.  So I believe that if risk is what the speculator and the hedger have in common, what they should also have in common is some good operational risk software.   The same is true of the arbitrageur, whose modus operandi I will consider in my next column.

Neural networks--by now a household term, at least in our IT house--are kind of ambidextrous.  They can use historical data to either predict or classify.  Sometimes both hands work together, and a neural network will make predictions based on classification.  One of the more interesting classification applications of this technology that made it into the press this week is software that analyzes the human traffic that makes its way through a particular room.  

The 'intelligence audience measurement' software can count people.  It can recognize and categorize gestures.  It can recognize people according to their gender or their generation.  Isafetek, the company that developed the software, calls its ability to count and classify people in a group "pattern recognition."   According to the developer the platform has promise not only as operational risk software for public spaces such as airports but for advertising and marketing efforts, where statistical analysis of the data collected on a particular group of passersby in a particular place could be useful.

If this introduction brings up fears about Big Brother, don't worry.   You won't be an identifiable face in the crowd.  You won't even be a face in the crowd.  You'll be a type of face in a group in another group in a crowd.

Real options are the flexibilities that are inherent in general business or other decision situations. In general, a real option is present in any decision situation involving a decision-chance-decision sequence; the possibility to (at the second decision) select from a range of different decision possibilities after the occurrence of the chance event may alter the choice of the decision earlier on in the sequence (and/or increase its value). The extra value created by this flexibility is sometimes described as a real options value.

Real options analysis concerns itself with analysing such flexibilities. On some occasions it may be desired to value such flexibilities explicitly. On others, the valuation is not explicitly required and the analysis concerns itself mostly with making the correct decisions and planning risk response or mitigation actions. The topic has links to financial market options, as well as to traditional net present value analysis.

A more detailed description of this topic, with example models using Excel, @RISK (software for risk analysis using Monte Carlo simulation) and PrecisionTree (decision trees in Microsoft Excel) can be found in Chapter 5 of my book Financial Modelling in Practice (John Wiley & Sons, 2008. ISBN-13: 978-0470997444).

Dr. Michael Rees
Director of Training and Consulting

A while ago I planned to take the last aeroplane of the evening to fly out for a client meeting the next day. Unfortunately the plane was cancelled and I had to miss the meeting. On speaking to the client, he said dryly: “But you are supposed to be a risk manager!”

I smiled but did not respond. However, as he was saying this (and indeed well before in anticipation), the following thoughts went through my head: First, the reason that I planned to take the last plane was that the client was particularly unwilling to pay my normal charge out rate, and after protracted negotiations we had just about come to a satisfactory solution. As a result to make the whole situation worthwhile, I had to do a full day’s chargeable work on the day before visiting the client. Second, in any case the only way of guaranteeing with close to 100% certainty that I would have been at the meeting would have been to take a flight two days earlier, and lose even more productive work time. Third, the client and I had prepared well for the meeting in advance, and he was well able to represent our joint project to his colleagues without my presence.

In other words, risk management is not really about risk elimination at all costs, but rather about balancing the costs of risk mitigation (e.g. high cost associated with travelling to a meeting two days early) with the consequences if risks materialise (costs associated with my not being at the meeting). Effective risk management is about finding the optimal balance between these. Even if all risks could be eliminated, generally doing so would be too costly.

In many everyday situations, we already optimise our actions to reflect this balance. When crossing the road, we look both ways, perhaps several times (an action which costs little but reduces the probability of an accident), but we do not build a bridge (an action that could eliminate the risk, but only at great cost). We intuitively find the optimum point and accept residual risk.  More formally one can approach optimisation problems by building quantitative risk models (e.g. using Monte Carlo simulation or decision tree analysis). Often such analysis is most easily implemented by the use of risk software such as @RISK, PrecisionTree or RISKOptimizer (which uses a genetic algorithm to optimise within the context of a Monte Carlo simulation).

The topic of residual risk (and optimisation) is an interesting one, as it does pose challenges from a communication perspective. Management may wish rather to hear that all risks have been eliminated (rather than their being a residual which is too costly to eliminate). In addition, the notion of an optimum does open the discussion of which criteria are used to assess that (e.g. maximising the total average profit after including the cost of risk mitigation measures, or minimising the losses in the worst 10% of cases etc.). In more complex cases (such as in some agricultural, veterinary, or environmental risk analysis situations), the costs of risk mitigation and the benefits associated with this are borne by different parties. Such externalities can make an appropriate assessment even more complex; perhaps there will be more on this in a later blog.

Dr. Michael Rees
Director of Training and Consulting