U.S. Storage Filling Up with Unaccounted-For Oil

This is a joint post with Matt Mushalik, a retired civil engineer and regional planner based in Sydney, Australia, and may be seen also on his website Crude Oil Peak.

U.S. crude oil storage is filling up with unaccounted-for oil. There is a lot more oil in storage than the amount that can be accounted for by domestic production and imports.

That’s a big problem since oil prices move up or down based on the U.S. crude oil storage report. Oil stocks in inventory represent surplus supply. Increasing or decreasing inventory levels generally push prices lower or higher because they indicate trends toward longer term over-supply or under-supply.

Why Inventories Matter

Inventory levels have reached record highs since the oil-price collapse in 2014. This surplus supply is a major factor keeping oil prices low.

Current inventories are 45 million barrels higher than 2015 levels, which were more than 100 million barrels higher than the average from 2010 through 2014 (Figure 1). Until the present surplus is reduced by almost 150 million barrels down to the 2010-2014 average, there is little technical possibility of a sustained oil-price recovery.

Figure 1. U.S. Crude Inventories Are ~150 Million Barrels Above Average Levels. Source: EIA, Crude Oil Peak and Labyrinth Consulting Services, Inc.

U.S. inventories are critical because stock levels are published every week by the U.S. EIA (Energy Information Administration). The IEA (International Energy Agency) publishes OECD inventories but that data is only published monthly and it measures liquids but not crude oil. It also largely parallels U.S. stock levels that account for almost half of its volume. Inventories for the rest of the world are more speculative.

Understanding U.S. Stock Levels

Understanding U.S. stock levels should be straight-forward. Every Wednesday, EIA publishes the Weekly Petroleum Status Report  which includes a table similar to Figure 2.

Figure 2. EIA publishes adjustments and defines them as “Unaccounted-for Oil.” Source: EIA U.S. Petroleum Status Weekly (Week Ending September 16, 2016), Crude Oil Peak and Labyrinth Consulting Services, Inc..

The calculation to determine the expected weekly stock change is fairly simple:

Stock Change = Domestic Production + Net  Imports – Crude Oil Input to Refineries 

Domestic production and net imports account for crude oil supply, and refinery inputs account for the volume of oil that is refined into petroleum products. If there is a surplus, it should show up as an addition to inventory and a deficit, as a withdrawal from inventory.

But that’s not how it works because EIA uses an adjustment in order to balance the books (Table 1).

Table 1. Calculation of Crude Oil Stock Change. Source: EIA Petroleum Status Weekly, Crude Oil Peak and Labyrinth Consulting Services, Inc..

The logic is that estimated stock levels in tank farms and underground storage are relatively dependable and that any imbalance must be from less reliable production, net import or refinery intake data.

There is nothing wrong with adjustment factors if they are small in comparison to what is to be balanced. In the Table 1 example from September 2016, however, the adjustment is 60% of the stock change–a bit too much.

A one-off perhaps? No, it’s a permanent problem that has gotten worse during the last several years.

Figure 3 shows that crude oil supply and refinery intake of oil vary considerably on a weekly basis. The balance is cumulatively negative over time beginning with a zero balance in January 1983. That suggests that crude oil stocks should be falling over time but instead, they have been rising.

Figure 3. Difference between U.S. crude oil supply and refinery intake. Source: EIA Petroleum Status Weekly.

The vertical bars show the weekly crude supply from production and net imports either exceeding the refinery input requirements (positive, green) or not reaching these requirements (negative, red). The solid red line is the cumulative.

Between 1991 and 2002, the deficit increased to a whopping 1.3 billion barrels.

Looking at only recent history, an additional gap of nearly 200 million barrels developed as refinery intake exceeded crude oil supply for most of 2010 through 2014 (Figure 4).

Figure 4. Difference between U.S. crude oil supply and refinery intake 2002-2016 (12-month moving average values). Source: EIA Petroleum Status Weekly, Crude Oil Peak and Labyrinth Consulting Services, Inc.

Adjustments were introduced in late 2001 so let’s look at the period starting January 2002 (Figure 5).

Figure 5. EIA adjustments to supply to reconcile stock changes. Source: EIA Petroleum Status Weekly, Crude Oil Peak and Labyrinth Consulting Services, Inc.

There are both upward (blue) and downward (red) adjustments. Upward adjustments resulted in a 420 million barrel stock increase over the period January 2002 through September 2016.

All together now

Expected or implied stock changes calculated from weekly crude oil balance indicate falling inventories from May 2009 through the present. Yet, EIA makes adjustments to that balance in order to match observed inventory levels. Rising inventories result after those adjustments are added to the physical balance or implied stock changes (Figure 6).

Figure 6. Unaccounted-for oil in U.S. storage: the result of adjustments to the supply balance. Source: EIA Petroleum Status Weekly, Crude Oil Peak and Labyrinth Consulting Services, Inc.

The green area represents the physical balance (crude production plus net crude imports minus crude refinery intake). The gray area shows the unaccounted-for (adjusted) stocks.

The adjustment for unaccounted-for oil averaged about 15% from 2002 through 2010. In 2016, almost 80% of reported stocks are from unaccounted-for oil.

When You Have Eliminated The Impossible

There is no obvious solution for the mystery of unaccounted-for oil in U.S. inventories. Possible explanations, however, include:

  1. Crude field production is underestimated
  2. Net crude oil imports are underestimated
  3. Refinery inputs are over-reported
  4. Crude oil stocks are over-reported

or any combination of those possibilities.

Production, imports and refinery inputs are taxable transactions. It is likely that reporting errors are largely self-correcting over time because of the financial incentive for government to collect its due.

State regulatory agencies are the source of production data. Their principal objective is to assess production taxes. It is unlikely that states would consistently under-estimate production and forego substantial tax revenue.

Also, producers must state crude oil production in their SEC (U.S. Securities and Exchange Commission) filings and pay federal income tax on revenues from oil sales. It seems improbable that the SEC and U.S. Treasury would consistently accept under-reported production and associated lower tax payments.

Crude oil imports are subject to both tariffs and excise taxes so it seems unlikely that the U.S. government would consistently fail to identify under-payment of those revenues.

Similarly, taxes are involved when refiners buy crude oil and sell refined products. It seems improbable that they would over-state those transactions and consistently over-pay associated taxes.

The principal components of supply balance—production, imports and refinery intake—are shown in Figure 7. In a general way, increased production and decreased imports tend to cancel each other out. Refinery intake has increased since about 2010.

Those trends determine the physical balance or implied stocks. The inescapable conclusion is that implied stocks (in light blue) are substantially less than reported stocks (in gray).

Adjustments for unaccounted-for oil are unreasonable and out of proportion to the underlying factors that determine crude oil stock levels.

Figure 7. Components of unaccounted-for oil in U.S. storage. Source: EIA Petroleum Status Weekly, Crude Oil Peak and Labyrinth Consulting Services, Inc.

It would be speculation to blame anyone for this apparent statistical disaster. Nevertheless, there is a problem that has major implications for oil price and the reliability of reported data.

In several of his Sherlock Holmes mystery stories, Arthur Conan-Doyle wrote, “When you have eliminated the impossible, whatever remains, however improbable, must be the truth.”

We have not eliminated any impossible explanations. We have, however, eliminated the three most improbable explanations for unaccounted-for oil.

The truth—however improbable—is that inventories are probably much lower than what is reported.


Some readers noted that EIA Weekly Petroleum Status Report (PSW) data that we used in our evaluation includes estimates of production, net imports and refinery inputs. They questioned whether EIA monthly Supply and Deposition data might resolve the disparity between implied and reported stocks described in this post.

Figure 8 shows the monthly averaged data from the PSW and implied stocks calculated from monthly Supply and Disposition data (the red line). The difference between implied and reported stocks before 2010 is less than in our original evaluation. The difference from 2010 through 2012 is approximately the same, and for 2013 through the present, the differences are actually greater.

Figure 8. Comparison of Implied Stocks From EIA Petroleum Status Weekly & Supply and Disposition Monthly Data. Source: EIA Petroleum Status Weekly, EIA Supply and Deposition, Crude Oil Peak and Labyrinth Consulting Services, Inc.


  • PAOil

    Another thoughtful post on an important topic. Thank you for sharing your work. Do any of the inventory reports involve actual measurement of storage tanks? For example, how is the API estimate produced–similar formula as EIA or any field work involved? Does the EIA periodically reconcile its formula back to actual field data as a business would do in an annual audit or inventory? In retrospect what you and Matt have done is one of those things that is so obvious you shake your head why it wasn’t thought of before. But that’s what we say about most clever inventions.

    • Arthur Berman


      I do not know how API estimates storage but I assume that they use pipeline flow data similar to what Bentek provides.

      EIA explains its methods in the appendices of the PSWR. It mostly involves surveys of companies that dominate the production, import, export, storage and refining of crude oil. It is, therefore, a model that must improve over time like any model.

      I have little doubt that the method is fundamentally sound but it is based on certain assumptions. Those assumptions become increasingly uncertain through time especially when there are basic changes to the system dynamics including the increased production of very light oil from tight reservoirs beginning around 2010, and the price collapse that began in mid-2014. Those events are apparent in the disparity between implied and reported stocks in Figures 6, 7 and 8 (recently added as a postscript).

      We do not imply any criticism of EIA in this post. Our purpose is to point out that there are considerable uncertainties in inventory data that may have significant implications for oil price.

      All the best,


  • Jeff

    Art – Why don’t exports factor into the supply equation? If oil is exported it would lessen the volume of oil needed to be stored.

    Is there any incentive for producers to over report inventories (say in lease storage tanks) to defer paying taxes or royalties on production or are taxes and royalties
    paid as oil is produced even if stored. Do producers self report inventory stocks?

    thanks for the post.

  • Jeff

    Art- meant to ask if the API data has a better correlation than the EIA data for inventory levels. Seems since they are not a government entity, they may have a better correlation in what they report each week.
    Best regards, Jeff

  • GV

    An astonishing fact that screams for explanation. Have you contacted the EIA?
    I suppose (because I hardly find this comment enlightening http://www.eia.gov/todayinenergy/detail.php?id=21472) that the adjustments represent the difference between the formula and the real inventories provided to the EIA which are obtained from the different PADD’s.
    In concreto the information based on the physical presence of the oil should be correct but raises the question what accounting item is wrong. Production must be higher or refinery input lower or a combination, but it’s almost impossible.
    A true riddle for Sherlock Holmes!

    • Arthur Berman


      We have added a postscript to that shows how monthly EIA Supply and Disposition data changes the picture. Basically, it lessens the disparity before 2010, is about the same for 2010-2013 and is worse for 2014 to the present. Hardly a solution and I believe that is EIA’s final word on input data.

      All the best,


  • Jonathan

    Hi Art,

    I have to thank you for all of the time and effort you put into these posts. It is really appreciated and enlightening for someone in the upstream part of the industry. I have much better insight into the bigger picture than I did before I started reading your blog!

    With regards to the current article, how is it possible that the inventory measurements could be off by ~400Mbbl? I agree with you that the companies and regulators are not going to allow that significant an oversight on production, let alone imports/exports and refinery inputs. It seems that, nearly all are going to be incentivized to report accurately.

    However, inventories can be physically measured, providing some indication of actual storage levels (companies such as Genscape etc.). Wouldn’t these independent estimates of at least the major tank farms quickly refute a +/-400Mbbl accounting error? Do you think that there is sufficient storage that oil could ‘hide’ in a distributed fashion throughout the US, possibly as a result of cumulative inventory over reporting? Are storage companies in anyway incentivized to over report?

    Alternatively, do you think that it is possible that cumulative uncertainties in flows from producers and flows into/out of refineries on such a massive scale could contribute in a significant way?

    Kind Regards,

  • Anders Sørensen

    Dear Mr.Berman. As always you seems to have made a profound piece of work. I have some questions. Who reports the inventories? Who would benefit from reporting bigger inventories, than they actually have?

  • Tita

    Are not production numbers from “Weekly petroleum status report” estimates, rather than accurate (which is not possible to account in less than a week)? If we look at the production from the “Monthly report”, there is a substantial difference. From the weekly report of week ending 09/18/2015, domestic production is at 9’136 Mb/d, while the monthly report (more accurate) state that september 2015 was at 9’415 Mb/d.

    We see that in february 2016, the numbers seems consistent (no gap). And indeed, the weekly numbers (9’100) are almost the same as the monthly report (9’145).

    I went a little bit earlier, in october 2013, where the gap is decreasing. Weekly report 7’800 Mb/d while monthly report 7’700 Mb/d.

    Weekly reports make very poor estimates of production. While production was increasing during the shale boom, weekly estimates may have been constantly lagging behind the real numbers. I didn’t check the whole numbers, but it could be easy to check if the difference between weekly and monthly report correlate with the cumulative adjustments.

    • Arthur Berman


      Please see the postscipt that I put up yesterday.

      Some readers noted that EIA Weekly Petroleum Status Report (PSW) data that we used in our evaluation includes estimates of production, net imports and refinery inputs. They questioned whether EIA monthly Supply and Deposition data might resolve the disparity between implied and reported stocks described in this post.
      Figure 8 shows the monthly averaged data from the PSW and implied stocks calculated from monthly Supply and Disposition data (the red line). The difference between implied and reported stocks before 2010 is less than in our original evaluation. The difference from 2010 through 2012 is approximately the same, and for 2013 through the present, the differences are actually greater.


      All the best,


  • […] U.S. Storage Filling Up with Unaccounted-For Oil […]

  • Tita

    Thank you Art for your answer. I didn’t saw the postscript when I made my post.

    Yet I was not entirely satisfied, and I searched for inconsistencies between the petroleum status report monthly (PSM) and weekly (PSW), for production / refinery inputs / net imports. Basically, I just summed up data from eia for the same period and compared them. I choosed two period, from jan 2005 to dec 2010 and jan 2010 to july 2016.

    PSM report 50 Mb more production than PSW. It represent 0.5% of the total production of this period.
    PSM report 63 Mb more refinery inputs than PSW. It represent 0.2% of the total inputs for this period.
    PSM report 9 Mb more imports than PSW. It represents 0.05% of the total imports for this period.

    PSM report 175Mb more production than PSW. It represent 1% of the total production for this period.
    PSM report 67Mb more refinery inputs than PSW. It represent 0.2% of the total inputs for this period.
    PSM report 152Mb less imports than PSW. It represent 0.8% of the total imports for this period.

    But then it struck me that we are tracking an error of 400 millions barrel over the account of 60 billions barrels over the period 2010-2016 (when we add production, refinery inputs and imports). Isn’t it just a case where we can blame accuracy? And I also hope that eia report inventories with the help of people reading the gauge on their tanks, not just accounting barrels going in and out.

    Nonetheless, they have become less accurate for some reason, with the same mistake repeating over years.

    Kind regards, and thank you for the work you do

    • Arthur Berman


      The way I think about this is not the sum of (Production + Net Imports) + Refinery Inputs but as the difference between (Production + Net Imports) – Refinery Inputs. That is not a trivial failure of accuracy but a shocking deficit at a time of unprecedented production.

      We are not suggesting that over-supply is not real. We are merely saying that we cannot come close to balancing the equation. We cannot understand how production or imports could be under-stated enough or refinery intakes over-stated enough to account for the imbalance.

      All the best,


  • Art, a very nice piece of analysis. A few questions:#1. Who benefits from an under-reported inflow of crude? #2. Why is there a benefit for-under-reported crude? #3.What is the benefit? 4.When and where will the benefit be “collected”?

    This is simply too big to be accidental over the time frame you are suggesting.

    Best Regards,


    • Arthur Berman


      You ask some very big questions. I don’t believe that crude production is under-reported, as we stated in the post. There is too much tax revenue at stake for state and federal tax collectors to consistently miss revenues. Also, field production is near the all-time high of almost 10 mmbpd. If current production of 9+ mmbpd is in fact much higher, that seems inconsistent with historical reality.

      As we said in the post, there is something else going on. Oil is showing up in storage but EIA public data does not allow full accounting for some reason. EIA called me today and left a voice message saying that we had “overlooked” something. I will relate what I learn when I connect with EIA.

      Thanks for your comments,


  • JV


    My question is how are NGL’s, casinghead etc. accounted for in the reports? My gut is that the blurring of reporting with BOE production numbers and the associated incentive to imply a potential revenue stream based on oil price vs nat gas or NGL prices has something to do with this disparity.

    Thanks for the great work you do.

    • Arthur Berman


      Oil and lease condensate are the only components of oil field production. NGLs are separated from the gas either at a plant or using a stripping unit on the lease but those liquids are not counted as oil production–they are included in gas production as a volume of gas before shrinkage.

      Some NGLs may end up in storage but not as commercial crude oil and that is what we are tracking

      All the best,


  • Anonymous

    Hi. You have misunderstood the data.

    Inventories are exactly that: physical hydrocarbon in tank as measured. The adjustment factor does not = “unaccounted for inventory”. Quite the opposite; all inventory is accounted for.

    The adjustment factor is the delta between the supply-disposition equation and the identified stock change. Put simply, it is the combined error term. If the adjustment factor is positive then the EIAs production number could be low, or imports could be low, or runs could be high or exports could be high, or even the inventory number itself could be incorrect – although you can see these inventory revisions are always small in the monthly data.

  • Tita


    I understand that you basically made the math “Production + Imports – Refinery inputs”. So, let’s put some numbers so you understand what I mean with accuracy. From Jan 2010 to july 2016, I add every barrel accounted by eia, and get this:
    17’703 Mbbl + 19’099 Mbbl – 36’997.5 Mbbl = -195 Mbbl change in inventory.

    I have no idea of the methods the eia uses to do their reports. But I see that a 0.5% error on the estimates makes a rather big change in the expected inventory change. Both Weekly and Monthly reports are estimations. So, I ask you again: why dismiss an accuracy problem?

    Kind regards

    • Arthur Berman


      We do not dismiss an accuracy problem. The purpose of this post is to clearly document the problem. The task of correcting the problem is well beyond our power. The EIA must decide to re-evaluate its estimation survey methods if it cares to address what we have shown in our post.

      EIA has contacted us and we have replied to their critique of our post. I will inform readers of any resolution.

      All the best,


  • […] dass das durch beispiellos hohe US-Lagerbestände bewiesen werde. Diese Leute sollten sich einmal den Aufsatz eines US-Erdölanalysten […]

Leave a Reply

This website uses cookies and asks your personal data to enhance your browsing experience.