Recent events have shown us that forecasting the future involves a significant degree of uncertainty. Over the past 12 months or so, the decision of the UK public to leave the EU, the election of Donald Trump as president of the United States, and the UK election resulting in a hung Parliament, are all examples of times where the vast majority of financial markets, statistical models, economic and political experts, and bookmakers, have not been able to predict the correct outcomes.
When modelling investments for Defined Benefit pension schemes, adopting a holistic approach, that models assets and liabilities in tandem, provides a more robust framework for decision making. Pension schemes are among the largest institutional investors therefore having a robust modelling technique in place is imperative. Necessary input parameters such as future expected inflation, interest rates, GDP and asset class returns need to allow for uncertainty; there are numerous techniques that can be implemented to achieve this.
“Technological progress has merely provided us with more efficient means for going backwards.” – Aldous Huxley
The most common technique “the survey method” is the simplest. Relying on expert opinions, this method utilises professionals who are immersed in the financial markets. Assessing the fundamentals of markets regularly gives a unique human feel that stochastic programming can potentially overlook. However, this method has the disadvantage of subjectivity and “herding”, where people are more likely to sit on the fence or follow the opinions of others than have a contrarian opinion.
“The future influences the present, just as much as the past” – Friedrich Nietzsche
A second method is to use historical data as a proxy for future events. This eliminates subjectivity and personal bias, and provides a sound basis, but relies on a very heavy assumption that the future will be reflective of the past.
Since the 2007/08 financial crisis, the global economy is said to be heading towards the “new normal”, which is expected to be very different to economic environments witnessed before. In Developed Markets, interest rates are at record lows – negative in some cases, inflation is muted, dampened by low energy prices and equity markets are at record highs bolstered by loose Monetary Policy. This environment is unlike any experienced before, resulting in data derived from the past becoming less representative of the future.
“Facts are stubborn, but statistics are more pliable” – Mark Twain
Econometric and stochastic models use statistical techniques to try and capture uncertainty. Using this as a foundation, these models use complex algorithms to predict a range of future economic variables. By analysing historical data, these models can look for relationships between variables and use random simulations to build up a large number of plausible future economic scenarios. This technique allows a user to quantify a very subjective area of finance, and assess the probability of different events occurring.
A common measure is Value at Risk (“VaR”) which in the pension industry can be used to measure downside risk. For example, a “1 year 5% VaR estimate of £10m” implies that over a year, a pension scheme deficit has a 95% probability of not increasing by more than £10m. Despite these advantageous characteristics, the effectiveness of the model to predict future variables is limited by the choice of statistical distributions and correlations derived from past data. Furthermore, these inputs are assumed to be static and therefore do not allow for any change in correlation during market stress, which may result in underestimations of downside risk measures such as VaR.
“In this world, nothing can be said to be certain, except death and taxes” – Benjamin Franklin
Results from modelling assets and liabilities will be highly sensitive to the inputs of the model, hence the choice of method to derive these inputs is critical.
Quantum uses a combined approach when considering future economic forecasts. We are able to use a combination of statistical methods, past data and a qualitative overlay to help us paint a picture of various future economic environments. A scheme’s assets and liabilities can then be modelled against these environments, allowing for the construction of an investment strategy which best suits a scheme’s objectives.
Jordan Griffiths
jordan.griffiths@quantumadvisory.co.uk