Please Note: Blog posts are not selected, edited or screened by Seeking Alpha editors.

Fundamental Oil Data, On Balance, Suggest Lower Crude Oil Prices In H1 2017, But Show Likely Higher Prices By Year-End

Fundamental oil data, on balance, suggest lower crude oil prices in H1 2017, but show likely higher prices by year-end

  • The difference between CO&LF inventory net withdrawals (NYSEMKT:EIA) and existing supplies has predictive properties. The implied price profile is of firmer price during Q1, then lower oil prices into mid-year.
  • EIA's production data were overstated in recent months; the next data adjustments over the next months should tilt towards the weaker side (positive), then a tilt to stronger side (negative).
  • By juxtaposing the non-linear sum of output-consumption versus Brent price, advancing the oil prices by 5 months to match the inflection points of the data, we derive an implied price forecast. The forecast suggest that oil prices could be subdued for another 5 months.
  • Fundamental data suggests that the oil price may still rise over the next few weeks. But the over-all tenor is one of weakening price structure; indications show an oil price trough by mid-year 2017, followed by a significant rise into year-end 2017, and H1 2018.

So far, we see that rising inflation in the US has been driven by the rapid rise in energy costs during the past four quarters of 2016. But this trend may not last long -- we see some indications of lower energy prices during the latter part of H1 2017. One basis is the extrapolation of Crude Oil and Liquid Fuel Inventory net withdrawals by the US Energy Information Administration , and projected difference between net withdrawals and existing supplies. The implied price profile is of range-bound to firmer price during the rest of Q1, but followed by lower oil prices going into mid-year, and then sharply rising prices during H2 2017 (see chart below). Oil inventories are lagging function of the oil price (prices lead by 15 weeks on average), and that may provide some positive spin on prices over the next two months, from a market sentiment point of view. But from that point on, it may be downhill for oil prices until mid-year.

Live link:

product.datastream.com/dscharting/gatewa...

What could also support oil prices into March is the likely adjustment lower of US domestic production data based on the history of US crude oil supply adjustments. The EIA oil supply model has historically overstated or understated output projections, which were then corrected or modified by a series of subsequent monthly adjustments, as shown by the model below. It appears that oil production data have been slightly overstated over the past several months, and so the next few month-on-month output data corrections will tilt towards the weak side, creating positive output surprises which will tend to help oil prices in the very near-term (see chart below). However, this will be followed by a tilt towards the opposite direction (stronger output), hurting prices going toward mid-year.

Live link:

product.datastream.com/dscharting/gatewa...

Keen observers will note that output adjustments have seasonality signatures for all to see - and that is what it really is. These adjustments stem from the periodic and highly seasonal character of US oil production, and is underlined by the current US output forecasts from the EIA (see chart below).

Live link:

product.datastream.com/dscharting/gatewa...

Clearing the deck: the relationship between inventories, Cushing stocks and oil price

The algorithms of crude oil price models can be as elaborate as you want to make them, but you are setting yourself for failure if the intention is make point price forecasts. Many analysts will tell you that it is difficult to do such thing. We will tell you that it is impossible to design such a model - it can't be done. However, if the intention is to make projections of where the oil price trend is likely going in the near-term, then the task becomes more tractable. Indeed, even very simple oil output-consumption models, properly designed, have proven useful to us in the past, and we are confident that will they continue to provide valuable input to trading decisions in the future. But there are caveats, and the primary one is getting the relationships/correlations between oil data sets correctly at the outset before any modeling work begins, even if we are going for simple ones.

Let us begin the exercise by laying a conventional wisdom to rest - the power of the oil inventories to predict the future level of oil prices. Conventional wisdom says that the higher oil inventories go, the weaker oil prices should be. Why should that be the case? This notion is wrong, and the error starts with oil inventories being conflated with total oil supplies. Investopedia suggests that " Inventory levels affect the price of oil, with higher inventories leading to lower prices," but offers to proof to support the claim. The notion is that as inventories rise, total supplies rise as well, and we all know that increased market supplies of any commodity tends to lower the market price per unit of that commodity.

Unfortunately, that is not how it works - in the past decade at least, changes in oil inventories have risen and fell inversely as a delayed function of the changes in oil prices, and the response lag averages 15 weeks. Put another way, inventories rise 15 weeks after a fall in oil prices, and vice versa: inventories tend to fall as oil prices rise, after a 15-week lag (see two charts below).

Live link:

product.datastream.com/dscharting/gatewa...

Live link:

product.datastream.com/dscharting/gatewa...

Subjecting the relationship to a coefficient of determination test (R-Squared) yields a respectable R^2 of 0.505, suggesting that more than half of the changes in Oil Inventories can be explained by the changes in WTI oil prices (see chart below).

Live link:

product.datastream.com/dscharting/gatewa...

The other half of inventory changes can probably be explained by supply-demand fundamentals that are being reflected in the oil term structure, primarily the WTI front spread. The chart below shows the lead impact of the front spread on both oil price and inventories, especially the latter. The C2-C5 spread has a 3-month lead over inventory, and a good 6-week lead over the direction of oil price (see chart below).

Live link:

product.datastream.com/dscharting/gatewa...

A regression of the spread versus inventory yields a significant R^2 of 2.58, suggesting that more than half of the 40% of changes in oil inventories not explained by the evolution of oil prices may be influenced by the storage decisions of oil users and speculators who are guided by the evolution of the front (C2-C5) spread (see chart below). We actually believe that investors' reaction function to both term spread and price changes primarily drive the course of oil inventory builds.

Live link:

product.datastream.com/dscharting/gatewa...

What makes this relationship even harder to accept for some economists is that it violates the law of supply. The law of supply is a fundamental principle of economic theory which states that there is a direct relationship between price and quantity: quantities respond in the same direction as price changes. Put another way, all else equal, an increase in price results in an increase in quantity supplied. If this economic principle applies in the case of inventories vs price, then inventories should be increasing with the rise in oil prices. That the inventory-price relationship conundrum is shown to work the opposite way is a tell that oil inventories (as defined) should not be conflated with total oil supplies (as defined).

In fairness, oil inventories have been classified as "crude stockpiles," which were defined by Investopedia as " reserves of unrefined petroleum, measured in numbers of barrels. Oil producers use crude stockpiles to smooth out the impact of changes in supply and demand." Let's see if we get better clarity with this narrative. The major storage center for crude stockpiles in the United States is in Cushing, Oklahoma. If crude stockpiles have indeed the same functionality as inventories, then there should be good, if not excellent, comovement (or correlation, if you will) between the two data sets - and indeed, that is the case (see chart below).

Live link:

product.datastream.com/dscharting/gatewa...

And here is what we are getting at . . . if oil inventories comove or correlate with Cushing stocks, then changes in Cushing builds should also follow the inverse lead of the changes in oil prices after a lag, as is the case with inventories. And we definitely see that to be true in the juxtaposition of the changes in the data sets, as shown in the two charts below.

Live link:

product.datastream.com/dscharting/gatewa...

Live link:

product.datastream.com/dscharting/gatewa...

Having figuratively cleared the deck, we can proceed with more productive analyses without worrying about the state of oil inventories or about the "specter" of an "impending glut" at Cushing, with stocks just off the historic high levels during early 2016. Some analysts are starting to go atwitter about storage tanks topping out at the Hub ("and where will all that excess production go!?"), but as we discussed in an earlier Seeking Alpha article, tanks reaching 100% of practical capacity does not signify a likely collapse of oil prices (see the article here : "What Will Happen If Or When Oil Storage Capacity At Cushing Is Reached?"). In fact, a top-out at the Hub signifies nothing much, except that oil prices were in fact relatively lower 16 weeks prior.

Simple Oil Output-Consumption Models

The other pitfall to watch out for is the oft-made error of making a linear comparison between oil output and oil consumption; e.g., today's oil consumption is deducted from today's oil production to derive a delta which could provide a data set which may have a bearing on the evolution of oil prices. The effort is laudable, but the methodology will likely provide an erroneous vector because global (as well as US) changes in oil consumption are, on average, 5 months ahead of the changes in the response from the output side. Said another way, changes in oil output lag behind consumption changes by about 5 months on average (see chart below).

Live link:

product.datastream.com/dscharting/gatewa...

Off the bat, this relationship tells us that focusing solely on global and US oil output (which, by correlation and by logic, merely follow the trend in global consumption) is particularly counter-productive, as in many cases (as in the past few months) consumption has already been falling but output is still on the rise (see chart below).

Live link:

product.datastream.com/dscharting/gatewa...

In this case, a linear summation to arrive at the delta between output and consumption will likely provide misleading values. What is more useful (and procedurally correct) method is to advance the consumption data by 5 months before summing up the production-consumption equation. The delta derived this way has very good predictive properties over the direction of oil price in the short term (see chart below)

Live link:

product.datastream.com/dscharting/gatewa...

We see from the charts above that consumption peaked in August year-on-year (yoy) and has been falling but may have bottomed in December, according to forecasts of the EIA, which sees consumption subsequently picking up until June this year. Global and US output have been rising as a delayed response to the rise in consumption, which peaked in August, 6 months ago. Indeed, global and US output may be peaking and should be declining over the next four months. Note that over that same period, consumption should be rising yoy. It is a very simple model which could reasonably indicate where oil prices should be going over the next three to five months (thanks to the leading function of consumption).

It also effectively explains the seeming divergences between output and consumption every once in a while; see the two charts below. In the following two charts, we juxtaposed the non-linear sum of output and consumption versus Brent oil prices, taking care to advance the Brent oil prices by 5 months, to match the inflection points of the data sets (in rates of change). The two charts suggest that oil prices could be lower soon and would be so for another 5 months.

Live link:

product.datastream.com/dscharting/gatewa...

Live link:

product.datastream.com/dscharting/gatewa...

The oil term structure suggests a 3-place pivot for the year (up-down-up)

A weaker oil price profile for H1 2017, as suggested by the interplay between supply and consumption as shown in the earlier charts shown above, is also supported by the month-on-month oil term structure for 2017. This construct (see chart below) shows narrowing contango coming close to backwardation into late Q1 2017 (encouraging firmer prices), followed by sharply wider contango into mid-year 2017 (see July-Aug/2017 spread) -- another hint of lower prices going into mid-year. This is also pretty much in consonance with the weaker price profile obtained from the simple consumption-supply models shown above. However, from mid-year up to year-end, the contango narrows considerably and the Jul-Aug spread goes into backwardation by year-end. This could lead to sharply higher oil prices going into 2017 year-end (see chart below).

Live link:

product.datastream.com/dscharting/gatewa...

The 2018 time spreads are distinctly oil price-friendly. By Q4 2017, all 2018 month-on-month time spreads are firmly backwardated, especially at the far end. The market seems to believe that by then, over-supply issues will be less dire, and improved demand and/or more moderate inventory builds could underpin oil prices. Those are speculations on our part, but the fact remains that H1 2018 time spreads are a whole lot more constructive than those of H1 2017 (see chart below).

Live link:

product.datastream.com/dscharting/gatewa...

Something global, next

The foregoing exposition has been limited to US oil fundamental data, just because the most granular oil data the EIA provides is on US domestic production and consumption. We are aware of the limitations this narrow focus brings. But we plan to have a two-part series, and the second installment will go global. For now, it is sufficient to provide a first look at global oil fundamentals, and the tableau is not much different from what we have seen in the US-centric analyses -- lower oil prices towards mid-year, then higher during H2 2017. See that in the two charts below:

Live link:

product.datastream.com/dscharting/gatewa...

Live link:

product.datastream.com/dscharting/gatewa...

Conclusion: On balance, the fundamental data suggests that the oil price may still rise over the next few weeks. But the over-all tenor is one of weakening price discovery structure, and indications show a trough in oil prices by mid-year 2017. Nonetheless, several positive factors should kick in at that time, which should set up a significant rise in oil prices going into year-end 2017, and higher oil prices will very likely cascade into H1 2018, as well.

Disclosure: I/we have no positions in any stocks mentioned, and no plans to initiate any positions within the next 72 hours.