Is your portfolio modelling as good as you think?

So we’ve been in a bear market for three years, and to compensate, you no longer talk about averaging high double-digit future returns.

Dropping back to say, a more moderate eight per cent should solve the problem of being more realistic, shouldn’t it? Don’t bet on it! If you don’t want to find yourself facing litigation down the track, you might want to rethink the way you explain risk and return.

In researching the material for our annual national roadshow in February and March, we rapidly became aware of the fact that the use of simple mean or average returns, could be very misleading.

In the fallout from the buoyant 80s and 90s, it is becoming apparent that the financial planning industry must move forward into more mathematically robust ways of explaining risk and return to clients.

Clients want to know how much their investment will be worth when they retire or reach their investment time frame, and then they want to know if it will last the distance.

If an average rate of return is used, we assume we will achieve the average rate of return, and this can create problems. This is because average returns are exactly that — the average of a larger number of returns.

But in only focusing on one point in what could be the range of returns a client could experience, our clients anchor onto this return, and may be set up for disappointment down the track. Unfortunately, the advisers can also be setting themselves up for future litigation.

Some of the problems of using mean return or deterministic analysis stem from the assumption that we are achieving a fixed rate of return. Is real life as predictable as a rate of return of, say, eight per cent every year?

If we use that magic eight per cent average rate of return for a portfolio, based on historical data, we are assuming that the portfolio will grow exactly at a fixed rate of eight per cent per year. So if the starting value of the portfolio is $100,000, the projected mean value in 10 years will be $215,892 (that is $100,000 x (1+0.08) x 10 years = $215,892.)

But in reality, the rate of return will fluctuate and the asset value in 10 years time could be any one of a range of values, depending on the actual investment environment experienced by the portfolio.

The $215,892 is the mean value, which is just one of the possible values in the range of possible outcomes. Yet an average return of eight per cent implies a nice steady rise in the value of the investment. If only that were true!

It would be nice, wouldn’t it? But it is unrealistic. This brings us to the opposite of deterministic analysis — probability based analysis.

Most clients, if they thought about it, are looking for a more realistic expectation of the chances of them meeting their goals than that provided by a deterministic analysis. For instance, in retirement planning, a deterministic analysis may suggest a client with $1,000,000 at retirement can withdraw, say, $68,000 per year through a 30-year retirement.

What it doesn’t tell you is there is a 60 per cent chance they will run out of money. A probability based analysis, on the other hand, would suggest that with, say, a $45,000 draw down per year, there is an 85 per cent confidence that the client will not run out of money — a much more realistic way of addressing a client’s retirement spending affordability.

Probability based analysis takes the volatility of the portfolio into consideration, by running simulations of possible outcomes the portfolio might produce. It’s almost like spinning a roulette wheel numerous times to get the range of outcomes, which is where the analysis dubbed Monte Carlo gets its name.

As a popular way of computing the statistics of results, Monte Carlo simulation is one type of spreadsheet simulation, which randomly generates hundreds or thousands of values for random variables conforming to the statistics of the input rates of return over and over, to simulate a model.

These forms of modelling are known as probabilistic or stochastic forms, which consider volatility better than mean or deterministic analysis, leading to a range of results with different confidence levels. Each of the results can then be plotted either into a distribution curve, or to give a visual picture of the likelihood of outcomes.

But Monte Carlo simulation is still not a ‘guaranteed’ way of predicting outcomes — as with any prediction, it can be wrong.

Don’t forget that in any mathematical model, the assumptions, be they averages, or ranges based on historical data, are just the instructions telling the calculation engine what we want it to calculate. In other words, maths engines (like Monte Carlo simulations) are not intelligent and they unquestioningly calculate exactly what you tell them to calculate.

And the precision of the results often lulls us into a false sense of security that we have done a thorough analysis, when in fact the engine analysed nothing and only did what it was told to do.

It is still up to us to check our assumptions are robust, which will in our opinion mean looking at historical returns longer than 10-20 years.

Taking a view on the future of markets also needs to consider these issues. If the results of our simulations are too aggressive, we can put our clients at too great a risk of failing to achieve their financial goals.

If they are too conservative, we can cause them to make unnecessary sacrifices in the only life they have. But at the end of the day, Monte Carlo and other probability based methods of analysis provide more guidance in ranges as a likely outcome, as opposed to a definitive average, which by being too ‘precise’ can leave an adviser open to litigation.

So how do we tie all this into everyday practice management? The good news is that there are some very good Australian based software systems that provide simple to use and affordable methods to access some of these issues.

Firms such as ProQuest Limited provide a subscription-based package that enables an adviser to scientifically establish a client’s risk profile, in such a way that litigation is far less likely down the track. And the financial planning software team at Xplan Technology have taken their software one step further by introducing Monte Carlo theory into their very affordable modelling package.

Please note this is not a recommendation of these services but rather an indication that we can expect further competition in this area over time, as the industry moves away from being too simplistic, and rather adopts a more robust approach of ranges and confidence levels.

After all, do we want our clients to anchor to a definitive average, and then complain if they are one of the many people that fall short of that average — as mathematical statistics dictate?

Or do we want to really work with our clients in partnership to help them understand a fuller range of outcomes before making more informed choices?

Jacinda Green is a business development manager withCredit Suisse AssetManagement .