Tracking the Price of Oil

I have been running models to see which security tracks the price of oil more closely. The most widely used would of course be the USO ETF, but it suffers from some issues such as contango and backwardation. So in this preliminary round, I will model with not only USO, but also the following ETFs - XLE and OIH. 


Keep in mind that these models are not perfect. They seek only to find a relationship in the form of an equation between the prices of one security and the prices of another. The software relies on neural network to find the closest equation. 


Below are the results of the testing. Let's start with USO. This model runs over the last 1000 trading days. I created the model based on the first 800 days and tested it on out of sample data of 200 days.

 Stastically speaking, the results are not bad. However, the last 400 or so days don't match the model anymore. The big oil run to $150 and its subsequent correction are modeled correctly, but since then, there has been a big divergence which is probably explained by contango. And it's clearer when you look at the 2010 relative performance between the futures and USO:

USO is lagging badly ever since June 2010.


OIH should not suffer from the same ills as USO, however, being a limited basket of stocks, it might be dependent on other factors. Here are the results of the model - same parameters as above.

Based on the stats, OIH doesn't track the oil price as well as USO. But it is somewhat accurate. The model seems to indicate that OIH is underpriced in reference to the price of oil (or oil is too expensive). There is clearly something there, but it's not perfect. And actually, as far as relative performances as concerned, OIH and oil are close this year. 

How about XLE then? It holds not only service companies like OIH, but also integrated and refiners. 

Statstically speaking, it tracks as well as USO but doesn't suffer from the contango and backwardation issues. Being a basket of companies, we can hardly expect perfection, but it seems close enough to be the preferred ETF to track oil prices! The model seems to suggest that it is overpriced relative to the price of oil (or oil is too cheap right!). And the relative performance chart for 2010 seems to confirm that as well. The ETF tracked oil prices very closely during 2010 in any case. 

So there you have it - my winner in this round would be the XLE ETF. 


In future analysis, I will seek to model relationships between oil and companies such as ExxonMobil and Schlumberger.