Saw an oil price forecast? Junk it

THE price of oil has fallen by half in the past two years, to just over $10 a barrel. It may fall further-and the effects will not be as good as you might hope," predictedEconomist newspaper in March 1999, and we all know what happened next. Crude prices rose more or less steadily thereafter, reaching a peak of close to $130 a barrel just before the Great Recession. The road to predicting oil prices is fraught with pitfalls and epitaphs of forecasters. But that hasn't stopped anyone from trying to predict future prices of a commodity that is often at the heart of geopolitical tussles and can contribute to inflation or trigger deflation. While there has been some modicum of success, generally, it is the big moves in the markets that we want to predict-and fail to do so.

There is little hope that such an algorithm which can predict these big moves can be found soon. Research by Christiane Baumeister (University of Notre Dame) and Professor Lutz Kilian of University of Michigan reveals that even the most sophisticated econometric models often fail to predict big swings. To be fair, such econometric models generally fare better than other ways of forecasting crude oil prices. Their paper characterizes forecasts of oil prices into four categories: One, economists using econometric models; two, financial markets, using futures data and adjusting for risk premium; three, consumer expectations based on inflation expectations; and four, policymakers' forecasts often based simply on futures markets. Data shows that the models of economists have fared better than others in predicting crude oil prices, in the period between 1988 and 2014 (CHART 1). However, when it comes to predicting big moves like the sharp drop in oil price from around $120 per barrel in July-September 2008 to average $60 per barrel in October-December 2008, almost all methods failed by quite big margins (CHART 2). For example, chart 2C shows that the actual oil price in the December quarter of 2014 was more than 20% below what the markets, consumers and policymakers estimated at the beginning of the quarter. Hence, the oil price shock is shown to be around negative 20% when compared to such forecasts. Of course, the oil price shock seems a bit smaller (at negative 10% for October-December 2014) when compared to the econometrician's forecast, thereby suggesting that economists fared slightly better in forecasting. However, the inaccuracy of forecast by the econometric model only worsened in January-March 2015 despite the fact that the direction of actual oil price movement did not change in the quarter.