Forecasting Oil Prices: It’s Easy to Beat the Experts

Most economists are used to being button-holed at parties and asked about some specific feature of the economy. And the more distant the topic is from your research, the more likely it is that you will be asked about it. Right now, I’m getting plenty of questions about what is happening to oil prices.

I recently dug into the recent oil literature and discovered something amazing: It is easy to do better than the experts. At least this is what I learned from a recent working paper, “What Do We Learn From the Price of Crude Oil Futures?” by Michigan economists Ron Alquist and Lutz Kilian.

The authors compare a whole range of different ways of forecasting oil prices: they look up the Consensus Forecast (from a survey of expert economic forecasters), oil futures, the difference between the oil price in futures and spot markets, and also a range of more or less complicated econometric models that take account of recent trends, as well as variables like the interest rate.

And it turns out that they all do worse than one simple forecast: the current oil price. That’s right: the most accurate forecast of oil prices over the next month, year, or quarter is the current oil price. We call this the no-change forecast.

The Alquist-Kilian finding was the subject of my latest commentary for American Public Media’s Marketplace (available here; or here for the audio version). Here’s the highlight:

Amazingly, this simple rule did better than the average of dozens of professional forecasters! In fact, the no-change forecast was 34 percent more accurate at predicting oil prices in 3 months time, and 18 percent more accurate at predicting prices in a year’s time. While professional prognosticators might argue that this difference isn’t statistically significant, it sure is embarrassing.

It turns out that the so-called experts add too much variance to their forecasts, leading them to do worse than the no-change rule.

I was more surprised by the poor performance of the oil futures prices as forecasts. My usual response to being asked about the future of oil simply involved looking up the futures market and calling those prices a forecast. However this isn’t quite right: the spot price embeds in it an option value or “convenience yield” that isn’t factored into futures prices. And if this convenience yield moves about (and it tends to move up and down with economic uncertainty), then this causes the futures price to fluctuate in ways unrelated to the future oil price.

If you are interested in learning more about oil prices, Kilian’s recent papers are a useful starting point. The longer Marketplace story is available here.

Leave A Comment

Comments are moderated and generally will be posted if they are on-topic and not abusive.



View All Comments »
  1. Hal says:

    Futures markets are not forecasts. They incorporate forecasts of price moves, but also storage, interest, delivery and other constraints. It bugs me when alleged experts point to the futures market as the market’s prediction of future prices. No, it’s the price today for future delivery, not the future spot price.

    Thumb up 0 Thumb down 0
  2. Larry says:

    I learned about the superior accuracy of the “no change” forecast over other methods when I first poked my nose into forecasting decades ago.

    I learned that its relevance extends to so many areas that the successful use of other techniques is the exception, not the rule.

    Thumb up 0 Thumb down 0
  3. Carl says:

    The huge thing about the oil markets that no one talks about is the moral hazard that is currently taking place with Iran and Russia.

    Iran and Russia both have governments that depend on high oil prices to keep them afloat.

    Every time the price of oil starts to dip, Iran’s government does something crazy to drive the price back up. Iran has a balancing game of keeping the world on edge enough to keep the price of oil high, but not so much that they’ll be the recipient of military action. Russia then opposes many sanctions against Iran and supplies them arms so that this can continue and then they both have been benefiting from the high oil prices.

    I’m all for a free market and don’t think speculation is bad, but in this instance I think it could drag us into World War III.

    Thumb up 0 Thumb down 0
  4. Charlie says:

    Many years ago, they ran regressions on things like class rank, SAT scores, etc. to try to predict students’ rank list (a number between 1 & 7). They announced that they were able to predict it +/-1 in 60% of the cases. A professor pointed out that 70% of students were in ranks 2-4. By not predicting 3 in all cases, he said as he walked out, they had achieved a net loss of accuracy just by opening the students’ folders.

    Thumb up 0 Thumb down 0
  5. mao says:

    Efficient market hypothesis suggest that prices are martingales (i.e best forecast for tomorrow is today’s price or differently said speculative returns equal zero), what could be simpler……

    Thumb up 0 Thumb down 0
  6. brian says:

    reminds me of the most accurate weather forecast, if it’s more then three days out is will not rain…

    Thumb up 0 Thumb down 0
  7. brian says:

    Microeconomics 101: Price=Marginal Cost in the long run. The highest estimate I have seen for the marginal cost for a barrel is around $75.

    I am not saying that it will quickly approach a number close to $75, but I don’t have to make that call. All I’m saying is lower, probably much lower over a year or two. I’ll take that bet over any over-engineered econometric model any day.

    Thumb up 0 Thumb down 0
  8. PDB says:

    Predictions in anything (finance, politics, sports), whether made with sophisticated algorithms and lots of data or pulled from somebody’s ass, tend to do worse than rules-of-thumb. For example, in pro football, picking the team with the better record going into a game, and if have the same record, than picking the home team, has outperformed all the predictions over the last few years, whether made by computer or football commentator.

    Thumb up 0 Thumb down 0