Midland, Cushing and the Story-of-the-Year


Back in August, this blog shortlisted the Permian Basin bottleneck as a candidate for story-of-the-year in the oil market. With 2018 drawing to a close, we think this prediction has held up pretty well.

A major issue lurking in the background came to the fore, which was the pipeline capacity — or lack thereof — connecting the Permian with Cushing, Oklahoma and Houston.

Midstream companies are building more pipelines, anticipated to come online in 2019. Until then, the situation will remain tense. Just look at Western Canada to see how bad things can get when production overwhelms pipeline capacity.

One consequence of this delicate situation has been extreme volatility in the spread between crude prices in Midland, Texas and Cushing (See Figure 1). Betting on the wrong side of these movements can even blow a hole in traders’ pockets.

The spread falling into deeply negative territory over the summer meant the oil price in Midland was more than $15/b below Cushing.


Figure 1
Source: NYMEX via MarketView

As the name implies, this spread reflects the conditions in Midland versus Cushing. One way of capturing the conditions is examining the amount of crude in storage at both locations.

Ursa measures inventories at 150 sites around the world on a weekly basis using synthetic aperture radar (SAR). That list includes Cushing and Midland.

We calculated the difference between Midland and Cushing inventories, and then correlated that figure to the Midland/Cushing spread.

As expected, there was a negative correlation (-0.55), which you can see in the graph below (Figure 2). When Midland inventories increased by more than Cushing inventories, the Midland price fell relative to the Cushing price.


Figure 2
Sources: NYMEX via MarketView, Ursa

We used this data to create a simple linear regression model to predict the Midland/Cushing spread (Figure 3). The result (R-squared = 0.31), although low, was statistically significant, and allowed us to continue this analysis with confidence.



Figure 3
Source: NYMEX via MarketView, Ursa

The model above could predict 31% of the Midland/Cushing spread movement on a weekly basis.

To improve, we added a variable — Permian rig count — as a proxy for oil production. We included a seven-week lag to account for the time between drilling and flows starting.

This multi-linear regression model resulted in an R-squared of 0.59. In other words, our model could predict about 60% of the Midland/Cushing spread movement on a weekly basis.


Figure 4
Source: NMEX via MarketView, Ursa

That’s not a bad start.

The Midland/Cushing spread matters to anyone with direct price exposure, as well as the broader market.

Think of how these oilfields in West Texas and New Mexico have transformed the global oil market. Permian production has tripled over the last six years, expected to average 3.8 million barrels per day next month.

If the Permian were an OPEC member, it would rank as the third largest producer, behind only Saudi Arabia and Iraq.

The price of oil in Midland will be critical for determining whether the Permian can maintain its newfound status.




Live count of Ursa Space’s SAR + Optical imagery/data catalog

Learn more about our imagery as a service.