Adam Czyzewski

Random forecast

Not a day goes by without me being asked about future oil prices. And all discussions that follow end with the inescapable conclusion that oil prices cannot be predicted. This is a complete reversal in the way of thinking about the future from 8–10 years back, when most business decisions in the energy sector hinged on the assumption that oil prices were very much predictable. Which is it then? Can we look into the future or not?

As an economist I tend to answer: it depends on how we think about the future and, consequently, what we want to predict.

I was compelled to write this commentary by a publication of Roland Berger, a renowned consulting firm, comparing the accuracy of oil price predictions assumed in the budgets of states which are the largest oil exporters, as well as by international institutions publishing oil price projections (the International Energy Agency, the US Energy Information Administration and NYMEX). Roland Berger have made such comparisons since 2007.

An analysis of the accuracy of annual price projections shows that in the forecast for 2015 (made in 2014) none of the forecasters was able to predict such a pronounced decline in the average annual prices as the one that actually materialised. Let me remind you that in 2015 the average price of WTI (the US benchmark, equivalent to Brent) dipped to USD 49 per barrel down from USD 93 per barrel in 2014. In the case of oil-exporting countries, the average from forecasts of the three countries with the best record of accurate predictions so far amounted to USD 95 per barrel. According to the same countries, in 2016 one barrel of WTI will cost an average of USD 53. International institutions on the other hand, which have shown better accuracy than oil exporting countries in the last couple of years, put the price at USD 46 per barrel.

Let me start by saying that, on a global scale, a whole army of people is tasked with predicting the prices of oil. The trajectories of future oil prices are a part of financial plans and strategies of oil companies, the budgets of oil exporting and importing countries, projections of international institutions, projections of central and commercial banks and futures contracts for oil on commodity exchanges. How do we justify the existence of this army given that even the best of its soldiers seem unable to hit the mark?

Some may find the answer rather surprising. Forecasting is not about being right on target, though it is nice to actually be accurate. Why? Because when we formulate the price of crude oil for the next year, the forecasted value does not yet exist and thus cannot be guessed.

Before I justify this approach to forecasting, which prevails among experts, let us take a look at it from an entirely different viewpoint. Let us assume that the future is determined, it exists in unchanged form regardless of what we do. Given this assumption, which is an expression of a certain world-view, attempts to guess such specifics from the future as oil prices or the PLN/USD exchange rate make perfect sense. Unfortunately, this cannot be achieved through scientific methods, so the only thing I can say on that score is that there are competitions for the most accurate forecasts you can sign up for. There is nothing to be afraid of − all you have to do is give the numbers and no one will ask how you have arrived at them.

By contrast, the scientific approach to forecasting, especially long-term forecasting, starts with the assumption that future is undetermined, random and is in part influenced by human actions. Going by our example, the specific numerical value of WTI prices at the end of 2016 will largely depend on consumer and producer behaviour. The collective outcome of individual decisions will be manifested through supply and demand changes over the year and beyond, and the distribution of effects over time will depend on the inertness of supply and demand, which can be empirically measured (estimated). As far as the supply of oil goes, we know that the effects of production decisions taken by US shale field operators today will be visible in a couple of months, whereas similar decisions in respect of conventional reserves will materialise in a couple of years.

In the horizon until the end 2016, we may expect completely random factors to emerge, such as the forest fire in the vicinity of Canada’s producing oil sands, which, by altering supply, will affect end-of-year WTI prices. All we know of such factors is that they will happen, but until they actually come, we are unable to predict their date or effect on prices. These random factors also randomise oil price predictions, and random predictions are never expressed as specific numbers. They are always a range, whose span depends on our reluctance to make a mistake, i.e. on how badly we want to have the actual future number lie within the forecast range, and overall uncertainty. The more we want it, the broader the range will be. For example, we can say today with near 100% certainty that on December 31st 2016 the WTI price will be in a range between USD 10 and USD 80 per barrel. If we narrow the range down to USD 30-50 per barrel, the odds of capturing the actual figure within the range we bet on will be significantly lower. As to prevailing uncertainty on the market, it is markedly higher now than it was 3-4 years ago, when it was widely believed that OPEC would not allow prices to change to any significant extent. Accordingly, if we wish to maintain equally high odds of being on target with the next year’s oil price predictions as a couple of years ago, we would have to bet on a markedly wider range.

Future values are easier to capture than to hit. We should bear in mind that a forecast is a range, especially when using ‘scenarios’ that differ only in respect of the positioning of oil price trajectories. It is very likely that our high and low price trajectories are alternative visualisations of one and the same random forecast.

Author :

Leave a Reply