The Art of the Probable
Most people associate forecasting with quantitative methods often lumped under the umbrella designation “number crunching”. But good forecasts can be made based on expert judgement. Asking the right questions, seeing a problem clearly and thinking logically through the range of possible outcomes can all result in the right decisions being made.
It is certainly true that econometric and technical analysis use complex mathematical and statistical techniques to measure relationships and predict outcomes more precisely. But you no longer need a Kray supercomputer to do the calculations. Predictive analytics can now be done on a laptop and in Excel, although there are also good econometric packages that supplement the analysis.
Art or science?
My experience is that forecasting is as much an art as it is a science. I vividly remember my first job in a forecasting team, when the model threw up a result that clearly made no sense in terms of “gut feeling”. My colleagues tweaked the model with what they jokingly called fudge factors, to get a result that was more intuitively believable. So much for econometric rigour!
It’s important to encourage a diverse approach to forecasting, as most good forecasts are the result of teamwork and even the best “super-forecaster” has periods where he or she loses their magic touch. That’s inevitable, as there is always a probabalistic element in forecasting performance; often, the forecast is accurate, but the timing is wrong, and I prefer to keep experience within a team rather than sacking staff like football managers. That said, it is important to monitor and measure the accuracy of forecasts on a regular basis.
There are three main types of forecasting used in energy analysis, often combined in different ways depending on whether forecasts are short-, medium- or long-term:
- Fundamental Analysis
- Econometric Analysis
- Technical Analysis
Fundamentals refer to the basic economic forces that drive prices: supply, demand, stocks, trade and competitors.
Fundamental analysis builds up a picture of whether the market is likely to be over- or under-supplied with a commodity in a particular time horizon, based on the various and often highly diverse economic and market drivers.
These often differ across time horizons: for example, demand may be driven in the short term by weather events, seasonal factors, trading positions, and do on; and in the longer term by GDP and population growth, wealth trends, government policy and trends affecting the price of competing commodities.
Similarly, supply may be affected in the short term by production shocks such as force majeure decisions, geopolitical events, cartel decisions, and unscheduled outages; but in the longer term, price trends, the availability of finance, investment decisions and the price of competitors may have a bigger effect.
Stocks are more difficult to predict because they depend closely on the time gradient of the market, which in turn is affected by the supply-demand fundamentals.
Competitors are also complex to incorporate in a forecast, partly because game-changing technological advances can be Black Swans, and because the price of competitors affect each other mutually.
Finally, imports and exports can be combined with the supply-demand analysis to create balances. The balance will affect prices and trade flows, defining whether a particular geographical market is long or short in certain timeframes.
Econometrics can be used to measure the relationships between variables, and these measurements can be used predictively to make forecasts – predictive analytivcs.
Econometrics is often associated with correlation and regression analysis, but there are a panoply of more complex techniques. Many of these are very useful.
The days when R2 was the only measure that mattered are long gone. Econometricians have refined their techniques to ensure that the relationships they identify are meaningful. Rigourous tests are made for the significance of an economic relationship, including causation testing.
That said, it is important that the analytical techniques are applied appropriately, and in ways that are consistent with economic theory.
It’s not uncommon for the mathematical wizardy of the Quants to be confused with a kind of super-intelligence — the means can all too easily become an end in itself, and you find that decisions are taken based on the mystique of little understood mathematical methods, based on little more than blind trust.
Econometrics is a useful toolbox, but it is still a toolbox. Quants make good and bad calls like anyone else.
They should not be accorded more prestige than any other member of the analytics team. Certainly it is reasonable for a manager to ask questions about the techniques used, and to expect a cogent explanation for any forecast.
My rule of thumb is that if the econometric analysis can’t be explained in simply language and is not easily understood by an intelligent manager, it should be taken with a pinch of salt.
Technical analysis refers to techniques that are based on the price of a commodity, rather than the supply-demand pressures and events affecting its value. A technical analyst examines how prices have moved in the past as the basis for a prediction about how they will move in the future.
The technique has been controversial. During their heyday in the heady days of the 1980s, major financial houses and trading firms paid fortunes for chartists’ predictions of where markets were heading. A series of academic analyses typically concluded that there was little evidence that the techniques used were statistically meaningful. Nevertheless, many traders still use the tools in trading decisions, particularly but not exclusively in taking short-term trading decisions.
The commonly used technical analysis tecniques include: drawing trendlines on charts; identifying chart patterns including continuation and reversal patterns; moving average analysis; momentum and stochastic analysis; volume and liquidity indicators and cyclical and wave analysis. But there are endless more varieties, some of which ressemble astrology or tea-leaf reading rather than quantiative analysis of price movements.
One of the problems with evaluating the success of technical analysis is the sheer variety of indicators that have been proposed. Chart patterns often “work” in the sense that they result in a series of good calls, but frequently this peters out after 3-4 runs. I diagree with those who say the techniques are self-fulfilling. A more serious criticism, I beleive, is that it often difficult to pin down whether technical forecasts are made with perfect hindsight, or whether the trading signals genuinely result in taking profitable positions ahead of time.
Pure chartists often refuse to engage with market fundamentals because they say this colours their analysis of the price signals. I believe this is sheer nonsense. It stikes me as about as daft as someone refusing to use a map to reach their destination because they want to find their way by instinct.