I recently debunked the myth that crude oil time spreads are a good signal to predict where oil prices are headed. In my regression analysis, I showed that simply knowing the current price provided a much better forecast for crude prices up to five weeks out.

I am occasionally asked what the crack spreads imply for future crude prices. Crack spreads are the differences between crude oil prices and petroleum product prices. The most common ones are gasoline cracks, heating oil cracks, and 2-1-1 cracks (i.e., 2 barrels of crude v. 1 barrel each of gas and heating oil). Below I show a long-term and short-term graphs of crude prices v. the 2-1-1 crack spreads.

These crack spreads represent the gross refining margin of buying crude and selling products. And so oil refiners use them to hedge their refining margins.

Decades ago, before NYMEX began trading oil futures contracts, upstream producers bought refineries as a means of hedging their risk. The theory was that crude prices and refinery margins were uncorrelated and so refining profits provided a natural hedge to upstream crude profits.

**Study Results**

I conducted a study to assess the predictive value of crack spreads as they relate to forecasting crude oil prices. In my study, I used weekly data for the 10-years beginning January 2007 and ending January 2017. I performed a regression analysis with the crude price being predicted by the time spread (Month 1 - Month 4). I ran 6 regressions, one being the same week (coincident) and the others with lags of 1, 2, 3, 4, and 5 weeks.

The r-squared is shown in the table below. The forecast 1 week out has an r-squared of just 11.1%.

The r-squared drops as the forecast timeframe is increased. For example, predicting out five weeks, the r-squared falls to 10.5%.

Crack Spreads | Price | |

# weeks | r^2 | r^2 |

1 | 11.1% | 98.4% |

2 | 11.1% | 96.2% |

3 | 10.9% | 93.6% |

4 | 10.7% | 91.0% |

5 | 10.5% | 88.1% |

I compared that signal to simply using the crude price to predict the crude price the next week, and the r-squared is 98.4% due to the autocorrelation of prices. Using the time spread as a price signal is far less information than simply knowing the price!

Using the price as a predictor of itself deteriorates as the time horizon increases. The r-squared forecasting out 5 weeks, drops to 88.1%.

**Conclusions**

Crack spreads are not a good signal for predicting where crude prices are likely to go. Knowing what the price is today is more informative for predicting the future than knowing the crack spreads.

**Disclosure:** I/we have no positions in any stocks mentioned, and no plans to initiate any positions within the next 72 hours.

I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it (other than from Seeking Alpha). I have no business relationship with any company whose stock is mentioned in this article.