Oil Prices: Oil Futures as Predictors

Menzie Chinn has a great post on how well futures prices predict spot prices:

... A relevant question is whether futures are actually good predictors of future spot oil prices. The answer is not obvious -- for instance, for currencies futures aren't good predictors.

It turns out that in a horse race against a random walk and a simple time series (ARIMA) model, futures do not have the smallest mean error.

... On the other hand, futures do have the smallest mean squared errors. Hence, one could do worse than using futures to assess prospects for oil prices. (See also Wu and McCallum, 2005.) On the other hand, the differences aren't too great between the futures and a random walk (at 3 month horizon) and an ARIMA(1,1,1) -- close to a random walk -- at the 6 month and one year horizons, thus consistent with this post by James Hamilton on the time series characteristics of oil prices.

One caveat to this conclusion is that our results only apply to horizons of up to one year. The markets further out on the maturity spectrum are very thin, and the characteristics of these futures are unknown.