Over the years I have been involved in many projects that have produced or used long term forecasts of prices for one or more commodities – power, oil, coal, fish, and more. Unfortunately, I have also witnessed many situations where long term forecasts are misused or misunderstood – often with value-destroying results (Drax, anyone?)
Forecasts are comforting – they give an impression of certainty, of being able to manage the future and cope with unknowns, of being a step ahead. They are used to justify many, many investment decisions (and in the financing thereof). They are often presented by experts as “the answer” – and an expensive answer at that. But are they actually any good, or are they just simply delusional?
Forecasts of crude oil are the classic example of long term forecasts. Here I use oil forecasts to illustrate the danger in long term forecasting, and in particular in basing business decisions on point forecasts alone. Everyone knows the cliché that the forecast will be wrong. The question is how wrong, and why?
We can start by looking at historical forecasts from the IEA and EIA. These are not the only organisations to offer such forecasts; however they are pretty representative of the type of forecasts available, and how these have developed over the years. They are often used as benchmarks for oil (and other) long term price forecasts throughout the commodity sector. So they can comfortably play the role of “industry forecast” here.
The first thing that is apparent is that the forecasts are very different even when made only a few years apart, and that the forecast in any given year seems to be heavily dependent on the spot price of oil at around the time the forecast was made. Forecasts made around 2000, when the oil price was (in real 2010 terms) 20-30 USD/bbl, and following a decade of similarly low prices, were also around 20-30 USD/bbl. Jump forward 5 years and the forecasts followed the spot price higher to the 50-60 USD/bbl mark. The price spike in 2008 saw also a spike in the forecasts to well over 100 USD/bbl (for the IEA forecast – followers of the EIA forecast had to wait until 2009 for the big step up to the 100$++ range). Zooming in on the IEA forecasts shows this even more clearly.
The other aspect that is also pretty clear is that all the forecasts are upward sloping – that is, they forecast a continually increasing price of oil. This Malthusian belief – that prices are destined to rise as the good (oil) gets more scarce – seems to be common throughout the forecasting industry. Often this rise is justified via a pseudo-equilibrium argument based on the long run marginal cost of new sources of oil (tar sands being one of the recent favourites). Whatever the precise analytical justification, the argumentation rests on the idea that as the cheaper stuff gets used up, the more expensive stuff will have to be used, the price will rise and equilibrium maintained.
Well, you may ask, so what? Is there anything really wrong with these forecasts? After all, just as the forward curve shows a similar anchoring to the spot price, the industry forecasts simply reflect the market view at the time they were made.
The problem with this argument is that the spot price reflects short term supply and demand (ignoring any ability or otherwise to manipulate spot indexes for the moment). I have physical oil to sell now. I need to buy physical oil now. It has very little (or nothing) to do with the LRMC of Canadian tar sands, or even short run costs of extraction. Short term changes in supply or demand, and expected near term supply and demand, drive the substantial daily, monthly, and yearly shifts in the market price for oil. The price crash following the financial crisis of 2007/2008 was not anticipated in the forecasts immediately prior, in the same way as the price jump to 2008 was also not anticipated. The fact that the forecasts seem unduly driven by spot price levels introduces a significant level of randomness into proceedings – the overall price forecast level is essentially an accident, a result of how the short term market just happened to be behaving at the time of the forecast.
The other problem is the straight line, upward slope of the forecasts. The fundamental assumption behind this Malthusian result is that tomorrow will be essentially the same as today. We will all be doing the same things, buying and selling the same stuff, transporting it in the same way, perhaps just a little more efficiently. And oil will be getting scarcer whilst demand for it slowly increases. Yes, progress is nice, incremental, linear, and, above all, highly predictable.
Unfortunately (or fortunately, depending on your outlook and attitude to change), the real world shows this for the fallacy it is. The financial crisis in 2007/2008 was not predicted by the market until it was well underway. Low oil prices in the 1990s were not forecast in the 1980s at all. Instead, there is a tendency in forecasting to adopt an “end of history” approach, which assumes that all that could happen, has happened, and the future will be just like today, just a little bit faster and more efficient. Humans are short term creatures at heart, and when we think of “the past”, we essentially think of, basically, yesterday. Today is just like yesterday, and tomorrow will be, too.
In reality, tomorrow will not be like yesterday. I don’t know how different it will be, and in what ways, but it will be different. I may have some expectations, and my expectations of the near future may even turn out to be pretty good (although that is generally only because things haven’t had much time to change, so it is easier to think that a short term prediction was pretty good). I can try to assess what different possible futures may mean for me or my business, and try to adjust my activities with this in mind. However, I don’t know, and simply forecasting a nice upward progression based on an assumption that everything will remain basically as it is today is limited at best. Disruption changes markets, creates and destroys industries, and drives economic development. Whatever else, a forecast that ignores disruption ignores reality.
Interestingly, none of the baseline oil forecasts shown above assume a decline in prices. The potential impact of technological change on demand, for example, is not considered beyond small incremental developments (like a small growth in electric car use, for example). Standard economic theory states that the price of a scarce good will rise until it is more expensive than a substitute, upon which it will mean revert on the substitute price level. However, we have no way of knowing now what this substitute may be, and there is no reason to assume it will be more expensive that today’s oil price. Whilst we cannot assume that technologies around today will be the ultimate oil substitute, we do know they are getting cheaper all the time. Oil’s eventual substitute may also exhibit the same pattern, and be cheaper than 100 USD/bbl.
So, if basing oil price forecasts on today’s (short term) spot oil price, and assuming increasing scarcity and an upward sloping price curve, is at best a limited approach, what can we do instead? Examining the historical price behavior, we can see a specific cyclical pattern appearing, as shown below.
Thus we can forecast a declining oil price from 2011, bottoming out in the early part of the 2020’s, before increasing again to around 100 USD/bbl in real terms by 2035 or so.
Or can we… clearly, this forecast is, well, pretty unbelievable – or at least it offers very little evidence for its efficacy. It is in fact simply a sine curve superimposed on prices from 1960 to 2010. However, it is fundamentally no worse than the forecasts standard in the industry, even where these forecasts are justified by long explanatory texts describing future supply and demand developments. A reasonable sounding justification could easily be constructed for this forecast, built around a fundamental business or economic cycle argumentation, driven by supply side cycles (high prices driving innovation, low prices promoting existing resource use only until these are used up) with a lagging demand side response. However, it would still be rubbish.
Long term forecasting should instead focus on examining the range of possibilities, why these might happen and what else this implies about the markets you operate in, and what extremes could arise and what these could mean for your business and assets. This can still be quantitative, and indeed quantifying such potentials is valuable in evaluating portfolio strategies, future business outturns and so on. Long term forecasting is dangerous when such point forecasts are adopted as gospel, as if the forecaster somehow “knows” what the price will be tomorrow.