US crude production rose by 6 thousand b/d m/m and 1.14 mb/d y/y in January to 9.97 mb/d, compared to our forecast of a 0.12 mb/d m/m decline. Gulf of Mexico production unexpectedly rose by 79 thousand b/d m/m, suggesting Shell’s offshore infrastructure returned from unplanned outages sooner than expected. We had forecast a m/m output decline in the Permian of 80 thousand b/d, as our short-term model picked up a marked drop in gas output. The EIA matched that, reporting that total Texas and New Mexico (a proxy for the Permian and Eagle Ford) production fell by 73 thousand b/d m/m. This was the largest monthly drop in production from the two combined states since December 2015 and reversed six months of consecutive m/m production increases. Total liquids output underperformed our forecasts by 0.27 mb/d to come in at 14.93 mb/d, higher y/y by 1.6 mb/d, as NGLs output fell m/m.
Analysing H2 17 well-level data from the Permian show that output grew by around 60 thousand b/d each month. However, the EIA reported combined growth in Texas and New Mexico of 0.1 mb/d each month. Indeed, in September and October (for which the EIA reported growth of 0.24 mb/d and 0.23 mb/d m/m), well data for the Permian showed growth of 0.11 mb/d and 0.14 mb/d m/m respectively. And well-level data for the Eagle Ford indicates average growth of 12 thousand b/d each month in H2 17. So, either areas outside of the Eagle Ford and Permian drove production growth in H2 17, or the EIA is over-reporting production. The former is more likely. This is because, after making adjustments to import and export data, the EIA’s adjustment factor (which accounts for the delta between supply-demand and stock change) is not consistently negative. Still, the large difference between well data and EIA data warrants close inspection.
Other areas of outperformance vs our model in January include the Anadarko basin and the Niobrara. According to the EIA, combined production in Oklahoma and Kansas (a proxy for the Anadarko) grew by 28 thousand b/d m/m (we had expected 5 thousand b/d m/m). Combined output in Wyoming and Colorado (a proxy for the Niobrara) fell by 5 thousand b/d m/m (we had expected a 25 thousand b/d m/m decline). Interestingly, while oil output in the Permian and Eagle Ford follows gas output more closely than others, all basins see a breakdown in the relationship between the two when one-off factors cause large swings in output. Now that the freeze-offs are behind us, short-term gyrations in gas pipeline scrape data should once again prove to be an accurate indicator of oil output. Thus, we expect total crude output to have risen by some 0.1 mb/d m/m in February, driven by gains in Texas and New Mexico. Finally, given the outperformance of January production data, we have raised our 2018 estimate of US crude production growth by 54 thousand b/d, to 1.16 mb/d y/y.