Building a super calculator
A previous article, Not-so-super calculators, highlighted some of the benefits to financial advisers, and their customers, of using a stochastic simulation approach when projecting future benefits. This article examines the main hurdle that has prevented this modelling technique being adopted at a retail level and draws on our own experience in building the stochastic simulation tool used for the PFS Standard Risk Measure Model available on our website.
|
Hitting the Speed Bump Stochastic projection techniques require a lot more computing power than a traditional deterministic approach. This is because the modelling has firstly to go through the process of:
As a result it can take minutes or even hours to produce just one outcome. When face to face with a customer the time it takes to produce the outcome becomes a significant factor - 10 seconds can be an eternity, let alone hours! A web based tool must allow multiple users simultaneous access without the speed being compromised. As an example, a 40 year projection for a single investment option with 10 asset classes requires 400 possible investment returns to be produced. We typically do 20,000 simulations to deliver a “stable” result; this means that 8 million possible investment returns are required just for one outcome. Another way to look at this is if there are 100 different users of our simulation tool, then this is equivalent to 2 million people, or 10% of the Australian population using a typical deterministic model at once!Overcoming this Hurdle Improving the statistical process is one way to reduce the time it takes to do the simulations. However, our experience shows that a technology oriented approach yields more significant improvements. The first aspect to address is the program that generates the investment returns. Some statistical programs or spread sheets are able to produce stochastic investment returns. However, they are encumbered with more functions and features than needed and this slows down the simulation. By building a program that is custom designed to produce only investment returns, we have found that simulation times have fallen dramatically. As a comparison we timed how long it took for one of our Excel stochastic tools to do 1,000 simulations and found that it took just over 20 seconds on a laptop. Using our own custom built simulator, 20,000 simulations can be completed on the same laptop in less than 1 second: an improvement of over 400 times. Technological solutions are also available to help with multiple users. In particular, scalable server facilities are available that bring more computers online when usage levels increase. These facilities now make providing the stochastic tools cost effective when compared to existing products. The PFS Standard Risk Measure Model, described here, is a working example of how this hurdle has been overcome and illustrates the benefits of using a stochastic technique to calculate and show risk.
Thach Huynh
|
For more information, contact: |
