How accurate are short-term oil price forecasts nowadays? I'm going to focus in this diary on the US Energy Information Administration and its highly respected Short-Term Energy Outlook, simply because I know EIA best (I worked there for 17 years and contributed to both short- and long-term oil price forecasts). By no means is any of this meant as a criticism of EIA, which I believe does an excellent job with the information and resources available to it. In addition, it is important to point out that EIA's short-term forecasts tend to be in line with those of the International Energy Agency and private sector forecasts. For that reason, the analysis below is basically applicable to those short-term oil price outlooks as well.Let's look at a few recent Short-Term Energy Outlook oil price forecasts and see how they turned out.
1. January 2007 Short-Term Energy Outlook (price for West Texas Intermediate crude oil at time of forecast about $60 per barrel)
Forecast WTI price for 2Q08: $65.67
Actual WTI price for 2Q08: $123.95
ACTUAL vs. FORECAST (% deviation): +89%
Forecast WTI price for 4Q08: $64.00
Actual WTI price for 4Q08: $58.35
ACTUAL vs. FORECAST (% deviation): -9%
When this forecast was made, oil prices had been floating in the $60-$70 range for about 1 1/2 years. Not surprisingly, the price forecast for 2Q08 and 4Q08 was in the mid-$60s-per-barrel range. Instead, what actually ended up happening was a sharp runup in oil prices to $123.95 for 2Q08, as the world economy (and oil demand) continued to grow, then a sharp decline with the economic meltdown (and consequent reduction in oil demand). EIA's macroeconomic assumptions (note: EIA doesn't do its own macroeconomic forecasting, but instead relies on Global Insight for this crucial input into its models) in January 2007 indicated steadily increasing economic growth through 2007 and 2008, from 2.0% in 1Q07 to 3.5% in 4Q08. Instead, what happened was that inflation-adjusted U.S. GDP growth fell to 0.7% in 3Q08 and -0.8% in 4Q08. The world economy also slowed sharply. Therein lies the bulk of the problem with EIA's - and pretty much everyone else's - oil price forecasts. Given that economic growth is an important, but exogenous*, input to the forecasting models, there's really not much that EIA or other oil price forecasters can do about the situation. End result: major problems in accurately forecasting oil prices during periods of sharp, sudden, unexpected economic contractions (or expansions, for that matter), like in this case.
*Exogenous is defined as an "[i]ndependent variable that affects a model without being affected by it, and whose qualitative characteristics and method of generation are not specified by the model builder."
2. June 2008 Short-Term Energy Outlook (price for West Texas Intermediate crude oil at time of forecast about $130 per barrel)
Forecast WTI price for 1Q09: $131.00
Actual WTI price for 1Q09: $42.90
ACTUAL vs. FORECAST (% deviation): -67%
Forecast WTI price for 2Q09: $128.00
Actual WTI price for 2Q09: $58.90
ACTUAL vs. FORECAST (% deviation): -54%
When this forecast was made, oil prices had been rising steadily for months, from the $60-per-barrel range in early 2007 to around $130 per barrel by late May 2008. Also, real U.S. economic growth had been humming along at about 2.5% for the previous two quarters. The assumption was for an economic slowdown by late 2008, but certainly no meltdown. Based largely on the assumed economic slowdown, EIA's short-term oil price forecast was for pretty much steady WTI oil prices, over $120 per barrel, through 2009. Instead, what we actually experienced was a collapse in oil prices to $42.90 for 1Q09 and $58.90 for 2Q09. Once again, the forecast was inaccurate largely because the (exogenous) economic growth forecasts were not correct.
3. January 2009 Short-Term Energy Outlook (price for West Texas Intermediate crude oil at time of forecast about $40 per barrel)
Forecast WTI price for 2Q09: $42.67
Actual WTI price for 2Q09: $58.90
ACTUAL vs. FORECAST (% deviation): +38%
This forecast was produced with oil prices at about $40 per barrel and after the U.S. and world economies had already begun to melt down. As a result, the macroeconomic forecast was for -2.0% real GDP "growth" in 2009 and the oil price forecast was for continued $40-per-barrel (approximately) through the rest of the year. Again, the forecast was off - undershooting in this case - although for reasons I'm not particularly confident about. Perhaps the actual oil price ended up higher than expected because of more-optimistic-than-expected market assumptions regarding a recovery in world economic and oil demand growth by late 2009. Perhaps it had to do with OPEC oil production cutbacks. Perhaps it had to do with currency fluctuations and/or speculative trading in oil. Perhaps it was random "noise." It's hard to say, but the bottom line is that the actual WTI oil price ended up 38% higher for 2Q09 ($58.90 vs. $42.67) than had been forecast in January 2009. Not horrible, but not great either.
Are there any overall conclusions we can draw from all these inaccurate short-term oil price forecasts the past couple of years? One obvious conclusion is that oil price forecasts are only going to be as good as the assumptions, economic and otherwise, that go into the forecasting models. If the assumptions and/or data are off, then the price forecast will almost certainly be off as well, sometimes by wide margins. That certainly has been the case over the past couple of years, with almost nobody predicting the runup to $130 per barrel, and almost nobody predicting the collapse to $40 per barrel. It's also worth noting that almost nobody predicted the Asian economic crisis in 1997, which meant that almost nobody predicted the collapse in oil prices to around $10 per barrel in 1998. Another example: almost nobody predicted the Yom Kippur War/Arab Oil Embargo of 1973/74, or the Iranian Revolution of 1978/79, which meant that almost nobody predicted the sharp spike in oil prices that resulted. Also, almost nobody predicted the damage Hurricanes Katrina and Rita would do to Gulf of Mexico oil production and refining facilities. We can go on and on with this, but you get the idea; economic, political and meteorological disruptions are extremely difficult, if not impossible, to predict, yet they often have an enormous impact on oil prices.
In addition to the problems with exogenous inputs - economics, politics, war, weather, etc. - there tends to be a strong bias towards forecasting prices at about the levels they've been, and also in the direction they've been trending. Thus, back in the mid 1980's when oil prices had been high (and rising) for years, oil price forecasts tended to be for oil prices to - shocker - remain high (and rising) for many more years. During the 1990s, when oil prices had been low (and fairly steady) for years, oil price forecasts tended to be for oil prices to - again, shocker - remain low (and fairly steady) for many more years. Sensing a pattern here? :)
Meanwhile, as if all this doesn't make short-term oil price forecasting difficult enough, how about Deutsche Bank's view that "oil prices appear to have been divorced from the underlying fundamentals of weak demand, ample supply and high inventories." Then there's the view of former Shell CEO Jeroen van der Veer, who says that "prices are increasingly dictated by long-term assessments of supply and demand, rather than current market fundamentals," and "advised taking a long-term view of the market." Got that? :)
Last but not least, there's the issue of "speculation", with "a growing number of critics [having] blamed those who are betting on the direction of energy prices for some of the extreme volatility." Partly in reaction to this volatility, "federal regulators announced on Tuesday that they were considering new restrictions on 'speculative' traders in markets for oil, natural gas and other energy products." How can anyone create (and run) a model that will consistently forecast short-term oil prices correctly, and preferably for the "right reasons" (as opposed to pure luck or even the "wrong reasons")? It ain't easy, that's for sure.
In the end, given the economic, political, and other exogenous uncertainties in the world, the accuracy of short-term oil price forecasting is likely to be sketchy at best, highly inaccurate and misleading at worst. This is part of the reason why I believe that instead of doing point estimates - or even price range forecasts - we'd be better off switching to scenario analysis ("a process of analyzing possible future events by considering alternative possible outcomes"). A major potential advantage of scenario analysis is that it can help policymakers understand the consequences of various policy decisions or shifts in other variables. A disadvantage of scenario analysis is that politicians, policy makers, the public, etc. are likely to continue demanding short- and long-term oil price forecasts, and are not as likely to want to wade through detailed, complex scenarios in order to truly understand the potential advantages and disadvantages of various courses of action. In my view, that's unfortunate.
