The EIA has once again undercut its previous estimates for U.S. oil production, offering further evidence that the U.S. shale industry is not producing as much as everyone thinks. 

The monthly EIA oil production figures tend to be more accurate than the weekly estimates, although they are published on several months after the fact. The EIA just released the latest monthly oil production figures for June, for example. Meanwhile, the agency releases production figures on a weekly basis that are only a week old – the latest figures run up right through August. 

The weekly figures are more like guestimates though, less solid, but the best we can do in nearly real-time. It is not surprising that they are subsequently revised as time passes and the agency gets more accurate data. 

But the problem is that for several months now, the monthly and the weekly data have diverged by non-trivial amounts. The weekly figures have been much higher than what the monthly data reveal only later. And remember, it is the monthly data that tends to be more accurate. 

Let’s take a look. A month ago, I wrote about how the EIA’s monthly data for May put U.S. oil production at 9.169 million barrels per day (mb/d). But back in May, the EIA’s weekly figures told a different story. The agency thought at the time that the U.S. was producing nearly 200,000 bpd more than turned out to be the case. Here were the weekly estimates at the time: 

• May 5: 9.314 mb/d
• May 12: 9.305 mb/d
• May 19: 9.320 mb/d
• May 26: 9.342 mb/d

But two months later, the EIA published its final estimate for May and put the figure at 9.169 mb/d. So, as it turns out, the U.S. was producing much less in May than we thought at the time. 

Now, the EIA has once again skewered its own weekly estimates. On August 31, it released monthly figures for June, and the discrepancy is even larger than the month before. The EIA says oil production in the U.S. actually declined in June, falling by 73,000 bpd to just 9.097 mb/d. Compare that to what the agency thought at the time with its weekly estimates: