U.S. Inventory Reductions Probably Not Sustainable

The decline in U.S. comparative inventories since February is the most significant oil market development since prices collapsed three years ago. It means that U.S. demand has exceeded supply for most of the last 5 months. The main cause is lower net imports, not higher domestic consumption, and that is probably not sustainable.

Comparative Inventories: The Key To Understanding Oil Prices

Comparative inventory (C.I.) is the difference between current storage levels of crude oil plus a select group of refined products, and their 5-year average for the same weekly time period (Figure 1). It is an indicator that normalizes seasonal variations in production, consumption and refinery utilization.

Figure 1. Comparative Inventory Is The Difference Between Stock Levels & Their 5-Year Average. Source: EIA and Labyrinth Consulting Services, Inc.

C.I. is the key to understanding oil prices yet few analysts use or even discuss it. Instead they try to explain price fluctuations by events in the daily news cycle or by simple year-over-year comparisons.

The negative correlation between C.I. and WTI price is strong. The 121 million barrel (mmb) increase in C.I. that began in June 2015 corresponded with a decrease in oil prices from $60 to $28 per barrel (Figure 2). The subsequent decrease in C.I. from April to July 2016 corresponded to an increase in oil prices from $28 to $50 per barrel.

Figure 2. Strong correlation between Comparative Inventory and WTI Prices. Source: EIA and Labyrinth Consulting Services, Inc.

U.S. comparative inventories have fallen more than 104 million barrels since mid-February 2017. Average weekly withdrawals of 4.3 mmb of crude oil and refined products indicate that demand has exceeded supply by almost 600,000 barrels per day (b/d) over the past 24 weeks.

Figure 3 shows the same C.I. vs. price data as a cross-plot (with the time dimension suggested by the light blue connecting lines). The resulting “yield curve” (Bodell, 2009) offers a structure for organizing seemingly random variations in oil prices.

Figure 3. Comparative Inventory Provides a Framework and Context For Oil Prices. Source: EIA and Labyrinth Consulting Services, Inc.

The yield curve does not provide a precise solution to comparative inventory vs. price trends.  Nor does it represent a simple regression fit although the data correlate systematically in time. Interpretation based on experience is required because much of the apparent data scatter is due to sentiment-based fluctuations in price.

Nevertheless, the C.I. vs. price yield curve presents a unique framework and context for prices and price trends. Because it reflects movement of oil volumes in and out of storage, it integrates true demand and supply variations with price. It also places probabilistic constraints on future price movements.

The yield curve shows that the March-June 2015 “false” price rally to more than $60 per barrel was a significant over-shoot. It reflected unwarranted optimism soon after oil prices collapsed that a return to $100 per barrel prices was likely. A short-lived fall in comparative inventory supported that optimism (Figure 2).

Similarly, the price collapse to less than $30 per barrel in late 2015-early 2016 represented a substantial under-shoot. Yet, it corresponded to the second phase of the largest increase in comparative inventory in history (Figure 2). Cushing storage capacity was greater than 80% during that period and it is a factor that generally puts downward pressure on prices.

The yield curve further shows the artificial uplift in prices from November 2016 through February 2017. This was based on unrealistic optimism that OPEC production cuts would result in a meaningful oil-price recovery to perhaps the $60-$70 range. Again, the anticipated mechanism was expected to be inventory reduction.

These price outliers reflect the daily decisions of oil traders who must trade, and the influence of sentiment and analyst narratives on short-term prices. Their causes were recognized as probable deviations from the yield curve norm at the time they were occurring.

The yield curve furthermore imposes constraints on the potential limits to future price trends. Its intersection with the y-axis is called the mid-cycle price, the price that the market deems necessary to maintain supply for a complete price cycle.

That price for the current cycle is approximately $75 per barrel. It represents the most likely price when stock levels are equal to their 5-year average. Barring fear premiums because of geo-political events, the mid-cycle price tempers overly optimistic price expectations as long as comparative inventories remain elevated.

I used July STEO forecasts to estimate a possible range of comparative inventory values and corresponding WTI prices for December 2017. These suggest C.I. in the range of 40 to 60 mmb above the 5-year average, and corresponding WTI prices between $48 and $53 per barrel. These may be conservative based on recent large inventory withdrawals—if those continue, oil prices could be in the mid-to-high $50 range by year-end.

Consumption and Net Imports

Ordinarily, higher demand is because of increased consumption but data suggest that year-to-date U.S. consumption (product supplied) is flat compared with 2016 (Figure 4). Record levels were reached in late June and July but first-half consumption was actually about 150,000 b/d less than for the same period in 2016.

Figure 4. Late June-July Consumption Has Reached Record Levels But Year-to-Date 2017 Consumption Is Flat With 2016. Source: EIA and Labyrinth Consulting Services, Inc.

Although the last 5 weeks of data indicate approximately 440,000 b/d of additional consumption, comparative inventory began to fall in mid-February and this recent increase, therefore, cannot account for most of the C.I. reduction.

Net oil imports provide a more consistent but still incomplete explanation for increased U.S. demand and C.I. reductions. Since mid-February, average net imports of crude oil and refined products have decreased about 334,000 b/d (Figure 5).

Figure 5. Lower Crude & Refined Product Net Imports Are Responsible For ~55% of Comparative Inventory Decline. Source: EIA and Labyrinth Consulting Services, Inc.

That translates to 2.4 mmb/week or about 55% of the average C.I. decline of 4.34 mmb/week. Lower crude oil net imports account for about 65% of that total decrease. Year-to-date crude exports have averaged 757 kb/d compared to 445 kb/d in 2016.

Unaccounted-For Oil

Lower net imports and higher consumption account for all of the comparative inventory reductions since late June but only for a little more than one-half of reductions for the previous 4 months. It is disturbing that this important development cannot be more fully understood.

Last October, my colleague Matt Mushalik and I wrote a post about the large and growing volume of unaccounted-for oil in U.S. storage. We showed that implied stock changes based on input and output volumes published weekly and monthly by the EIA could not be reconciled with reported stock changes to U.S. crude oil storage.

That situation has not changed. Implied stock levels (field production + net imports – refinery intakes) are consistently less than reported stock levels. The cumulative difference is now 137 million barrels from a common starting value in January 2015 (Figure 6). That is the amount of unaccounted-for crude oil in U.S. storage.

Figure 6. 137 Million Barrels of Unaccounted-For Oil in Storage (28% of total). Source: EIA and Labyrinth Consulting Services, Inc.

An EIA representative initiated dialogue about the points raised in our post just after its publication. The EIA contends that its methods for estimating stock changes are more reliable than its published flows we used to calculate implied stock changes. Although this may be valid for relatively recent data, it seems that revisions should largely cancel those differences after some reasonable number of months have passed.

The representative also noted relative agreement between EIA and API stock estimates as validation of its methodology. It seems, however, that differences between what EIA and API report are common and sometimes large.

The main point is that these stock levels are estimates. Storage volumes are not directly measured by the EIA. Circumstances change and algorithms that may have been reliable in the past have become progressively undependable.

Although I agree with the EIA that absolute stock levels are probably more correct than the underlying flows, unaccounted-for oil is a problem that the EIA has chosen not to acknowledge. Either inventory reductions are less than reported or the underlying flows are not accurate enough to account for those reductions.

Lacking unambiguous percentages, most of increased demand is because of exports. The potential to continue reducing inventories in this way is, therefore, limited by world capacity to absorb excess U.S. supply. Much of the increase in U.S. exports was possible because of temporary outages in places like Nigeria.

Meanwhile, the outlook for meaningful reductions in world over-supply seems questionable. OPEC’s output jumped almost 1 mmb/d above its target in July. Wood Mackenzie anticipates that global supply may increase almost 2 mmb/d in 2018 if the agreement does not hold.

Figure 7 is a synthesis of supply-production and demand-consumption data from the principal international reporting organizations. It suggests that the world supply surplus is likely to increase during the rest of 2017 and the first half of 2018. It further indicates that the second quarter of 2017 may be the only period of supply deficit for the next 17 months—that the last quarter was as good as it gets for quite awhile.

Figure 7. World Market Balance Suggests Increasing Over-Supply Going Forward. The Second Quarter of 2017 May Be The Only Under-Balanced Quarter Through 2018. Source: IEA, EIA, OPEC, BP and Labyrinth Consulting Services, Inc.

None of this is good news for continued high U.S. export levels. The last 5 weeks of data suggest that increased consumption may become more important going forward. It is more likely, however, that these elevated levels reflect temporary, higher seasonal consumption after lower-than-normal usage in the first half of the year.

The U.S. oil industry is justifiably proud of the ingenuity it used to survive the last 3 years of reduced commodity values. Technology, efficiency and lower oil-field service costs enabled increased production since September 2016 despite low oil prices.

Production growth, however, does not solve the underlying cause of low prices namely, over-supply. It makes it worse and prolongs lower prices.

I have no illusions that tight oil producers will willingly resort to business discipline. Only reduced access to capital will impose that necessary change on their behavior.



  • Edmund

    Hi Art,
    Thank you – great analysis that non specialists can follow!
    One question (probably stupid but …..) are you implying that implied stocks are higher, for whatever reason, or that reported stocks are actually lower?


    • Arthur Berman


      Implied stocks are lower than reported stocks (Figure 6). The difference is “unaccounted-for oil.”

      All the best,


  • Capt. David Williams

    Hindsight is alway 20/20! Thanks Art for a beautiful article!

    • Arthur Berman

      Captain David,

      Hindsight is all the we know is true assuming, of course, that the data is reliable!

      All the best,


  • ART, your analysis is EXCELLENT. The two missing piece’s of the puzzle for the public are many of the current plays in the USA ( and world ) have at least a 5% decline rate each year and very little exploration has taken place over the past two years .
    I believe this will result in a large SPIKE because one day we will wake up ( soon 0 and not have what we think we have . I also believe SAUDI ARABIA does not have what it claims to have for many reasons, larger decline rates, heavy water flooding, pushing larger production rates, etc

    • Arthur Berman


      I have said in many posts and webcasts that a price spike is coming some time in the next several years.



  • Jeff

    “The main cause is lower net imports, not higher domestic consumption, and that is probably not sustainable.”

    Can you elaborate on that? US consumptions look much stronger now than 2012-2014. Almost like 2007. The story used to be that US inventory will be the last to draw, since the cost of floating storage and on-shore storage in other countries is higher. If so, aren’t we now witnessing the last part of the “rebalancing”? If a major exporter (e.g. Venezuela) can’t continue to export then this could go much faster than most folks expect.

    • Arthur Berman


      I am comparing the more recent past since the price collapse because that was when storage and C.I. transformed into another dimension that may have forever changed the mechanics of oil price formation. Pre- and Post-Financial Collapse was a different watershed that affected oil markets with zero interest rates afterward.

      For those reasons, comparing 2016 and 2017 consumption is relevant and because the C.I. reductions I am trying to understand happened this year but did not last year.

      For net imports, the world before tight oil was different. Until crude exports were legal again in 2016, the balance of net imports was a one-way street of import substitution. Because tight oil is ultra-light, light oil substitution was the limit. That is part of why imports increased in 2016. Now that exports are an option, net imports are again dynamic.

      I hope this helps.



  • Good info. My feeling ‘in the field’ is that operators are shaving costs and one of the troubling aspects of that is charging royalty owners (and likely working interests) very high “post-production” costs. I just finished an appraisal for a charity that made about $30,000 of a property that has a bunch of wells across the 60 acres as it is in two units. Out of the dozen wells I think all but one is a cross-unit well. But 3 of those wells posted a NEGATIVE income meaning post-production expenses exceeded the income. How can that happen to a ROYALTY interest? I know it is only $40 or so, but that isn’t right.

    • Arthur Berman


      Thanks for another interesting field case history! It sounds like the royalty owner is being cheated.

      All the best,


  • Ken Johnson

    Regarding unaccounted for oil, is it possible the implied stock levels are less because the field production is less than what’s being accounted for? This would be due to the flat futures contango taking away the incentive to put the put the oil in storage.
    Maybe much of the field production is now actually shut in due to a lack of buyer but is actually being counted as production?

    • Arthur Berman


      It is possible that field production is under-reported but national levels are fairly stable within monthly ranges that don’t have big effects on stocks. On the other hand, EIA stopped including lease storage in their weekly stock reports in 2017. Field observers tell me that there is a huge volume of lease storage in the Permian basin. I’m not completely certain of the accounting but it is reasonable that the oil counts as production but is not included as storage in the new EIA reporting scheme.

      I suspect that under-reported exports are a big part of the problem.

      All the best,


  • Timo Puts

    Hello Art,
    Great analysis! For figure 7, is this based on those extra 2mmb/d anticipation from Wood Mackenzie? Or is it the combined “predictions” from the eia, opec and bp? This means that oil prices for the coming year are indeed dependant on what opec does.

    I haven’t done the numbers but shouldn’t we be starting to see the fall in production in (non-state run) oil from the massive levels of underinvestment in the 30% coming from offshore and high cost offshore? The u.s. rig count is plateauing now too… But maybe I am too early again…

    • Arthur Berman


      Figure 7 is the synthesized supply and demand estimates and has nothing to do with Wood Mackenzie’s guesses about OPEC. I expect that under-investment will result in tighter supply and higher oil prices in a few years but probably not before 2019 or 2020. U.S. rig count rate of increase is slowing but spare capacity is high with uncompleted wells.

      All the best,


  • Timo Puts

    *high cost onshore

  • SK

    Hi Art,

    Great analysis! Quite different from most other stuff we get to read in the name of analysis. This one is the most convincing to me especially the way you have used EIA data to explain the correlation between comparative inventory and oil prices (Figs. 2 & 3).

    Another thing which I found very interesting is the “unaccounted-for-oil”. At 137 mmbls it’s a big number and for some reason if the EIA’s reported stocks starts converging towards implied stocks in future, it can lead to a big change in the markert structure.


  • Conrad E Maher

    Another very interesting look at oil price by Art. Thank you for your work.

    CI shows a strong negative correlation with price decreases leading to CI increases over the period until the the decrease in net imports have a greater influence on CI and a lesser influence on price.

    The Unaccounted-for-Oil is fascinating, It increases with time, but not uniformly. In Feb 2017 the 213 million barrels of oil would have had a value of approximately $10.6 Billion. The question is who would benefit from having this unaccounted for oil on the books? Junk bond holders? Producers producing oil at a loss? Institutions who loaned the companies the money to produce oil at a loss?

    There is for me a nagging feeling that I know too little about all the things that effect the price of oil to feel comfortable with conclusions I might reach from using the data that Art has once again presented with such clarity. That stock levels are estimates raises doubts because I do not know by how much these estimates might vary from the actual level.

    I will watch with interest to see how this plays out over the next few years.


  • Scott


    Another great piece of analysis. Somewhat related and was wondering if you could shed some light. There is a large gap now being reported for Texas C&C production between the Texas RRC and the EIA. Normally, the gap was smaller and merged over time with the RRC data typically moving a bit higher and the EIA moving lower. However, for the latest month of data (April), the gap is over 345,000 bopd. Any thoughts on how to reconcile this difference would be appreciated. And, if the EIA is overstating production, then your inventory gap is even wider.

  • Scott


    Another great piece of analysis. Somewhat related and was wondering if you could shed some light. There is a large gap now being reported for Texas C&C production between the Texas RRC and the EIA. Normally, the gap was smaller and merged over time with the RRC data typically moving a bit higher and the EIA moving lower. However, for the latest month of data (April), the gap is over 345,000 bopd. Any thoughts on how to reconcile this difference would be appreciated. And, if the EIA is overstating production, then your inventory gap is even wider.

  • Anonymous

    Of course inventory decline (or growth) is not sustainable. Eventually you would run the tanks dry or spill over the top. But of course, price adjusts instead.

    What is happening is just US inventories shrinking to a more normal days of inventory level.

    The more interesting trends are those in price or production itself. Inventory is over discussed in the news cycle. It is the tail, not the dog.

  • Anonymous

    Of course inventory decline (or growth) is not sustainable. Eventually you would run the tanks dry or spill over the top. But of course, price adjusts instead before that happens.

    What is happening is just US inventories shrinking to a more normal days of inventory level.

    The more interesting trends are those in price or production itself. Inventory is over discussed in the news cycle. It is the tail, not the dog.

  • Tita

    Hi Art, thank you for your great work.

    As you said, the EIA uses a lot of estimations for it’s report. Although they are quite accurate, the cumulative differences over a year can be quite high.

    One thing I check is the differences between monthly and weekly petroleum status report. Last data for monthly estimates is from may (9’169 kbbls/d), and we can see that there is quite a gap with the weekly report (avg of 9’320 kbbls/d).

    Best regards

  • Heinrich Leopold


    There is a substantial rise of the depletion (legacy) rate over the latest few months. Combined Eagle Ford, Permian, Bakken legacy rate reached 93.7 % in August up from around 45% last year. Do you have any studies about this trend? In my view this is important as this will seriously impact the financial position of shale companies – and thus reduce their access to capital for future production.

Leave a Reply

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