The dangers of mixing forecasts and forward curves30 Mar, 2015
Forward market liquidity has steadily developed with the evolution of traded gas and power markets in Europe. This has supported traders and asset managers to hedge forward asset exposures, reducing portfolio earnings volatility.
Energy price forecasting on the other hand has not undergone the same evolution. The track record of industry price forecasters today is no better than it was a decade ago. Perhaps in recognition of this fact, price forecasts have become cheaper to obtain. But we are still not aware of any forecasters who are prepared to publish a ‘mark to market’ back test of their historical accuracy.
The US Energy Information Administration (EIA) does deserve an honourable mention in this regard, given it allows access to its past forecasts. The EIA price forecasts have been as consistently and strikingly erroneous as any other, as they freely acknowledge. But the EIA also regularly compares its price predictions to those of other forecasters, showing that they rarely fall much outside of a reasonably tight ‘consensus’ band. This neatly demonstrates the poor track-record of the price-forecasting community. It is understandable given these circumstances that the energy industry looks elsewhere for a view on future pricing outcomes.
At first glance, forward curves offer an attractive alternative, given they provide a transparent and objective view of market price. As a result many companies (and price forecasters) have adopted the forward curve as a spot price forecast, on the basis that it represents the market’s consensus view of future spot price outturn.
In our view this logic is a capital error, and in a short series of articles we set out to explain why. We also address the question – if the forward curve is not to be seen as a forecast of spot prices, how are we to interpret it? In this article we use the animation of Brent crude pricing history that we published recently to illustrate some of the dangers of using the forward curve as a spot forecast.
The differences between forward curves and forecasts
The two terms ‘forward curve’ and ‘forecast’ certainly both start with the letters F-O-R, but that is just about where the similarity ends. One might also remark that the axes on which forward curves (FC) and forecasts are plotted look the same. Certainly, the y-axis in both cases is price: but closer examination of the x-axes reveals a critical difference.
While both are denominated in units of future time, the x-axis of a forecast is a time-series continuum, generally of daily granularity across the chosen forecast period; while the x-axis of a FC is contiguous ‘time-buckets’, often starting with months, then becoming perhaps quarters, or seasons (winter / summer), and eventually years. It may be common practice to interpolate a baseline FC into common, more granular time divisions to give a smooth curve, but it should always be borne in mind that such a transformation signals a modelling-based step away from the raw input of forward prices coming from the market.
These time-buckets match the currently-traded forward instruments available in the market being reported upon: and the corresponding prices recorded on the FC are a snapshot of the various prices currently reflecting that traded market, as of the date on the curve. To be even more precise, they are prices from a series of related but individual markets: the forward market for month n, the forward market for month n+1 etc. With several months till delivery, ‘August gas’ (for example) is a different commodity to ‘September gas’.
To make the point even more strongly: in April the ‘August gas market’ is a market for pieces of paper on which appear obligations relating to deliveries of gas in August. ‘August gas’ only becomes the same commodity as ‘September gas’ if and when both have been delivered and are sitting in the same gas storage facility. Some August gas contracts will be liquidated before August arrives, and thus never go to delivery – they were never gas, and they were never cashed in for gas. Some August contracts were always to be financially, rather than physically settled, and were never destined to turn into gas at all – just a difference-payment, i.e. cash.
And there could be many more August gas contracts by volume than there will ever be physical gas either produced or consumed in August. August gas contracts have laws of supply and demand of their own – the supply and demand for pieces of paper with obligations written on them – which we discuss in terms of the liquidity of the forward market. For the most part these are quite independent of the physical realities of what will happen in the spot market in August which will be driven by physical supply-and-demand characteristics.
The spot versus forward price relationship
Of course there are connections between spot and forward markets, but some are fairly trivial. For example:
- Spot prices are generally used to settle forward contracts. This follows directly from the original purpose of forwards as hedges against uncertain future spot prices.
- The price of the nearest forward contract converges with spot price. This is generally true, but not a remarkable phenomenon because over short periods the spot and near-forward markets are readily arbitraged via storage and other means of temporal exchange.
Other connections are more interesting. For example:
- As neatly illustrated by the Brent animation (shown again in Chart 1), from day to day the biggest influence on the forward curve along its entire length is the spot price.
- Indeed, the most common phenomenon in forward curve dynamics is the ‘parallel shift’, with the entire curve moving up or down almost in parallel, even if over the medium-term the gradient of the FC itself gradually changes, sometimes significantly.
Chart 1: Evolution of Brent crude spot and forward prices
Source: Timera Energy (based on ICE Brent Futures settlement prices)
These phenomena clearly undermine the logic that ‘the forward curve is a forecast of spot prices’. Why would a three-year forecast change every day, and by the same amount across the whole forward period, and by the same amount as today’s spot price change? Who changes their long-term forecasts every day? Who thinks that an increase in today’s spot price, perhaps caused by the weather or a well-understood physical event in the supply chain, has any bearing whatsoever on prices three years from now? What kind of implicit fundamental modelling would give rise to such a result?
The Brent price animation illustrates a very simple test: has the FC been a good predictor of Brent spot prices? The answer is that it has clearly been a very poor predictor indeed.
In the articles to follow in this series we will consider a number of econometric theories as to how the FC relates to physical markets. We will see that they often bear very little relationship to reality: and that the dynamics of FCs is a subject best studied empirically.
This article was written by Nick Perry (Senior Advisor).