US defends its oil-output data
What do the EIA's weekly production figures tell the market—and what don't they?
Every Wednesday at 10:30am Eastern Time oily eyes turn to the Energy Information Administration's website for a weekly insight into how much oil the US is producing, exporting, refining and storing. The numbers, especially when they're unexpected, move markets.
They're also not well understood. That is especially true of US output, which has become an increasingly important market indicator as the country marches towards becoming the world's largest oil producer.
The EIA shed some much-needed light on how it compiles those weekly numbers in a presentation this week, which should at least clear up some confusion. The short explanation is that the weekly production figure is an extrapolation of the monthly forecast in the Short-Term Energy Outlook (STEO), another closely-watched EIA report, with some tweaks based on weather, rig counts and other indicators. It is a far less precise figure than the monthly production data the EIA compiles through surveys of operators across the country, but which comes with a two-month lag—an eternity in today's world.
That the figure is derived from a forecast is not ideal for oil-market transparency. In a perfect world, all 1.2m wells across the US would be kitted out with sensors feeding a central system that could spit out a daily tally of production. But that isn't going to happen. As the EIA's director of petroleum statistics Robert Merriam points out, some operators even have trouble reporting their monthly figures on time.
Monthly data muddles
Still, for its shortcomings, the weekly production figures have proven mostly accurate and reliable when compared to the actual production recorded in the monthly surveys. According to the EIA, the weekly figures were on average off by a little more than 1% and in 16 of the last 21 months it was correct on the trend of production, whether output would rise or fall.
Many in the market have accused the EIA of systematically overstating US production in the weekly figures—cue much conspiracy-mongering on oil twitter. The reality is that the EIA has gone through stretches where it consistently overestimated and underestimated production. In 2016, the clear trend was to understate production as the STEO forecasts didn't keep up with the industry's recovery. In 2017, as many have charged, there was a strong bias towards overstating production—on average by around 100,000 barrels a day. But this doesn't appear systematic. More the result of forecasters getting ahead of reality. The grassy knoll it was not.
The major outlier in recent data is October, when the industry was rocked by a major hurricane. It highlights a serious shortcoming of the weekly data. In that month, the EIA’s weekly figures understated output by 4% because it overestimated how long production would be down due to storm-related outages. Because there was no mechanism in place to track real-time output in storm-hit oil patches like the Eagle Ford, the EIA like the rest of the market was in the dark and mostly guessing.
In short, use the weekly production figures at your own peril. If its precision you're after, wait for the survey data.
* Correction: the original article stated that the EIA initially overstated October’s output, which was heavily affected by extreme weather. It has been corrected to say the EIA understated production after output recovered faster than expected.