Debunking Crack Spreads As A Predictor Of Crude Oil Futures Prices

|
Includes: BNO, DBO, DTO, DWT, NRGD, NRGO, NRGU, NRGZ, OIL, OILD, OILK, OILU, OILX, OLEM, OLO, SCO, SZO, UCO, USAI, USL, USO, USOD, USOI, USOU, UWT, WTID, WTIU, YGRN
by: Robert Boslego
Summary

Debunked myth time spreads are predictors of crude prices.

Correlation of crack spreads to prices confuses people.

Draw incorrect conclusion that crack spreads are predictors.

Correlation is not causality.

Zero predictive value of crack spreads to crude.

I recently debunked the myth that crude oil futures price backwardation (or contango) - time spreads - is a predictor of crude futures prices in my article, Crude Oil Backwardation: Theory, Facts and Myths. In this article, I address another myth, which is that crack spreads are a predictor of future crude prices.

Crack spreads are a proxy for refinery margins because refiners buy crude and make products. Their profits are based on the differences between input costs and output revenues, not the outright price of crude.

The gas crack spread is the price of gasoline minus the price of crude. Traders also monitor the heating oil crack spread and combinations of gasoline and heating oil v. crude.

The myth that high crack spreads “support” higher crude prices, and vice versa, stems from the fact that there is some correlation between crack spread levels and crude price levels.

Correlation = 32%

Correlation = 45%

Correlation = 43 %

Note: The 2-1-1 spread uses 2 crude contracts, 1 heating oil contract, and 1 gasoline contract.

Correlation Is Not Causality

In statistics, there is an often-used phrase that correlation is not causality. In this case, I tested for whether there is a statistical relationship between crack spread levels and future changes in crude oil prices. My test covered weekly data over the period beginning in 2010 up through the week ending June 7, 2019.

I performed a regression between the crack spreads and future WTI crude oil futures price changes, in the periods trailing 1 week, 2 weeks, 3 weeks, and 4 weeks. The results of each regression showed an r-squared (explanatory value) of zero percent in each case. The T-ratio in each case showed the spread to be statistically insignificant as a predictor of future crude price changes.

Conclusions

The crack spreads logically correlate to some degree with crude price levels because they are a proxy for refinery margins. And, for refiners to earn the same percentage rate of return at higher prices, the margins must be higher in dollar terms.

However, I have observed that this correlation is confused as a predictor of future price changes. My statistical testing shows they have no predictive value.

This result is logical to me based on how futures markets operate. It would be too simplistic for traders simply to observe crack spreads and have a good idea where prices are headed. If that were the case initially, the advantage would be arbitraged out of the market. It does not exist.

The next time you read how crack spreads are supportive of crude prices, ask the author for statistical support.

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.