This is Why Refiners Didn't Act Like Gasoline Demand was Weak Last Week, Even if Traders Did

psw01 2016-2-18

It appears there was a very good reason refineries were ramped up 2 weeks ago, gasoline demand was huge this past week.  For all the talk of weak demand even for gasoline, the numbers just don't support that story for gasoline.  Gasoline demand appears to be surging in recent weeks.  Year to Date gasoline demand is up 1.3% YoY but if we eliminate the first week of the year that was likely down excessively due to covering both the Christmas and New Years holidays, gasoline demand is up 2.6% YoY.

psw01 2016-1-3

The above chart shows refinery inputs are up significantly YoY as well.  The data indicates refinery inputs are up 2.6% for the YTD period vs. the same 2015 period.  If global demand is weak, US refineries sure aren't indicating that is the case.  

psw01 2016-1-13

In recent weeks total petroleum supplied has been surging putting it back in line with 2015 demand.  The next several weeks of 2015 showed a noticeable drop off in demand, it will be interesting to see if the same thing happens in 2016 or if demand actually remains, much stronger, especially with prices as low as they are.  It is also becoming apparent that US production is falling, it is known that Canadian tar sands will soon be in seasonal maintenance, and OPEC/Russia is talking like they want to cap production/put a floor under prices.  This could mean current price levels are as good as they are going to get and it could spur demand in the next several weeks.  This is a dramatically different supply situation than this time 2015, since US production was still surging and tar sands didn't go into maintenance until summer.

My conclusion is it looks like it's time to by some gasoline and some refiners and probably crude as well.

EIA Overestimates US Production, Huge Production Drop in the Works #6

psw01 2016-2-24

I've been saying for quite awhile the EIA weekly production estimation model is way off.  It is also possible that since the US started exporting crude, an increase in exports could also be throwing off the model.  But from what I've read elsewhere, and what I've seen in well completion reporting I continue to believe the majority of the error is from overestimating lower 48 production.

This week Unaccounted for Oil was -320k bbl/day or 2.24MM bbl for the week.  This is the 4th week in a row for highly negative Unaccounted for Oil, which is totally unprecedented.  This adds to a string of 8 out of 12 weeks since Dec. 4 with highly negative Unaccounted for Oil.  Again very unprecedented.  But even more so when you consider Unaccounted for Oil has tended to run positive as detailed in my previous post EIA Overestimates US Production, .... #4.  In the last 4 weeks EIA models can't account for -11.7MM bbl of oil.  But they continue to tell everyone that US production is holding up pretty well even though Unaccounted for Oil is indicating has averaged -418k bbl/day over the last month, indicating US production could well be 4.6% lower than reported.  Do you think that might matter to some market participants?  The error looks a little less bad when you look at the 12 week period, which shows -14.5MM bbl that can't be explained or -173k bbl/day, a 2% error.

EIA monthly reports are just over a week away.  A prime opportunity for EIA to adjust their models.  Look out if they do, because this error will suddenly find its way into the weekly production estimates dragging them down very fast, potentially some weeks could show 75k-100k drops when this hits.  When it does, it will squeeze the shorts hard, because nobody expects this.  The speculative shorts are still at or near record short levels and the mainstream media continues to indicate US production is barely touched.  More importantly dropping US production will encourage OPEC to at least stick to a freeze and possibly redistribute quotas to enable Iran to increase production, ensuring they stick with the greater OPEC plan.

Many have called some of my predictions crazy, but my call on Unaccounted for Oil has been dead on for 3 months.  I expect my call on what is coming next for EIA weekly estimates to be correct as well.  Not to mention my call that when EIA estimated in January that February production would increase month over month was dead wrong.  After three weeks, the weekly data indicates February is down 0.89% over January, and that is using their overestimated numbers.  The real proof will be in the April monthly report, but it doesn't look very good thus far for the EIA estimate.

Below I'm including an update to my projection for Production + Unaccounted for Oil.  As predicted it continues to track in the lower portion of the channels and is tracking Projection 1 well.  Using polynomial curve fitting to project the future is also beginning to confirm this value will track the lower projection line as well.  I also noted these projections are based on WTI staying sub $50, the lower WTI stays the steeper the fall off.  At this point it may not matter much, US producers are in a semi-panic dropping appx. 30 rigs/week and any brief spike is unlikely to change their production plans anytime soon.  I expect very low well completion rates in February and March and rigs to be shutdown quickly until less than 300 are working in the oil fields.  I believe every shale CEO now knows to not jump back in too prematurely, even if they hedge.  It will just drive down/hold down prices for longer when they need to get back to turning a profit.

psw01 2016-2-24

Refiners Aren't Acting Like Gasoline Demand is Weak, Even if Traders Are

Traders have been beating down the price of gasoline and gasoline crack spreads as well.  While refiners run at levels significantly higher than the same time in 2015 helping traders hold down gasoline prices and crack spreads.  With a huge inventory of crude, gasoline inventories seasonably high and a situation where WTI should supposedly stay low for a long time, why would refiners be pushing so hard, hurting their near term profits.

It seems there really can be only one answer.  Crude production is rapidly falling, and will continue to rapidly fall, and refiners know it.  Therefore they are sacrificing near term profits to create a hedge for the longer term by taking as much of what is currently ultra cheap crude, refining it and building stores of gasoline, distillates, etc. at a very low cost.

EIA Overestimates US Production, Huge Production Drop in the Works #5

The EIA weekly estimate of production is still too high as indicated by Unaccounted for Oil and production estimates based on well completion rates.  However, the weekly estimate is starting to move down in a big way.  Last week down 30k bbl/day, this week 50k bbl/day.  50k bbl/day is huge, that is 200k bbl/day/month  The reality is US production has been falling 20k-30k bbl/day for weeks or even months already, now it is likely falling 30-37k bbl/day/week not 50k bbl/day/week.  50k bbl/day/week would imply almost zero well completions, when around half of each months shale decline is currently being replaced monthly.  But EIA is so far behind, with likely production levels already below 9 MM bbl/day compared to the official estimate of 9.135 MM bbl/day, that they are being forced and will continue to be forced to show huge weekly declines for awhile to catch up.  Won't be surprised to see 100k bbl/day cut reported next week and potentially the week after as well.

Unaccounted for oil has been running dramatically negative since Dec. 4.  We have now just seen 3 back to back weeks with negative unaccounted for oil.  This week it was -500k bbl/day or 3.5MM bbl for the week.  Imagine what happens when -500k bbl/day in Unaccounted is zeroed out and put in production instead.

EIA Overestimates US Production, Huge Production Drop in the Works #4

psw01 2016-2-10

It seems that EIA is doubling down on their faulty model for weekly US oil production estimates.  EIA continues to drive monthly production estimates up, at least through February, and in turn this is supporting weekly overestimates of US production.  In fact their latest monthly report estimates production in the major US shale basins will exceed their previous January production estimate by 68k bbl/day.  This is in direct conflict with dropping rig counts and more importantly falling well completion rates reported by the states of Texas and North Dakota.

More telling is the huge negative error in the EIA production estimate that is showing up consistently in the Unaccounted for Oil category in the weekly reports since Dec. 4, 2015.  For all of 2015 Unaccounted for Oil averaged 192k bbl/day each week, in 2014 the average error was 150k bbl/day, but since Dec. 4 the error has averaged -126k bbl/day and in the last six weeks it has averaged -166k bbl/day.  Not only is it alarming that Unaccounted for Oil is suddenly averaging a large negative value, which is a major deviation from the norm of averaging positive, the magnitude of some of the recent values has been extreme.  This week Unaccounted for Oil was -509k bbl/day or 3.56MM bbl/week which goes along with the week of Dec. 4 which printed -598k bbl/day and Jan 8 which printed -459k bbl/day.

What is possibly more alarming is when the 10 week averages of Production + Unaccounted for Oil are looked at.  This exposes a trend in US production that is clearly masked by the official weekly estimates.  The 10 week average was 9.672MM bbl/day for the week of Aug 28, 2015,  For the week of Feb, 5, 2016 the average is 9.077MM bbl/day which implies at least a 595k bbl/day or 119k bbl/day/month decline over that period, which is a significant rate of decline.  However, the official weekly production estimate shows a decline from 9.22MM bbl/day to 9.19MM bbl/day which amounts to a measly decline of 32k bbl/day over that period or a 6.4k bbl/day/month rate.  Hence the erroneous belief that US production is holding up well.  But the recent decline since December is far greater than the 5 month data implies.  Since Dec. 4, 2015 to Feb 5, 2015 the 10 week average has fallen from 9.585MM bbl/day to 9.077MM bbl/day or a 508k bbl/day decline.  This indicates US production has been falling about 225k bbl/day/month over the last two months.  Which is double the rate implied by a 5 month view.

This week's weekly report showed a 30k bbl/day drop in US production which isn't terribly far off from the 50k bbl/day/week decline rate implied when Unaccounted for Oil is factored in.  However the weekly production estimate is at least 100k-120k bbl/day higher than what is indicated when Unaccounted for Oil is included.  This weeks' drop is not an anomaly.  Is it the beginning of EIA showing a 30k-50k bbl/day drop every week is unclear given past history?  What is clear is EIA will catch up at some point, and the longer this goes on the more dramatic it will be.

The conclusion is US shale production is far less robust than the world of oil traders currently believes.  If we have now moved into a mode where Unaccounted for Oil tends to be negative, say -150k bbl/day every week, opposed to the normal +150k bbl/day every week, and this is just week 2 of a much longer streak of back to back negative values for Unaccounted for Oil, US production could already be well under 9MM bbl/day and very rapidly falling toward 8MM bbl/day.

Below I'm including my updated 2016 US production + Unaccounted for Oil projection.  I've included a couple of my own linear projections as well as a couple polynomial trend lines.  My belief is US production is declining at a rate best described by Projection 2.  This means the decline is trending in the lower channel and possibly the lower half of the upper channel.  The less steep poly trend line also likely describes the situation well.

January 2016 150k NonFarm Jobs Gains is Totally Bogus

Total non farm 2016-2.xlsx

Looks like the BLS pulled a repeat of the March 2015 jobs report.  I really don't understand how these guys can continuously botch the seasonal adjust of the employment data like they do.  First in January they report 292k jobs gained which was totally bogus to the high side.  Basic YoY comparisons for Dec 2015 vs Dec 2014 showed a 232k gain which is in line with the same comparison for the previous several months.  Then it appears BLS immediately made up for the error in the January 2016 report by revising December down to 262k but putting January at 150k.  The 150k number totally freaks out the markets when the average has been running around 220k for the year.

What is the January number really.  YoY shows it to be 217k, which on its surface looks terrible considering 2015 job gains averaged 242k/month which would constitute a 10% drop in a month.  However, one must also consider December 2014 and January 2015 were blowout employment months with each month showing the best gains for their respective years at 252k or about 2.2% annual rate for the months.  Job gains have been trending down slightly throughout 2015 with the year starting with a 2.2% annual growth rate and ending at 2%.  But January 2016 coming in at 1.9% growth when compared to a previous years blowout month is not that alarming.

However if you do further correction for the blowout January 2015 be using January 2013 data one comes up with a 234k gain for January 2016 which is in line with the appx. 232k monthly gains in previous months.

In the end we have a government agency that one months puts out 292k when 232k is more real a 25.9% error, which they already corrected by 10 percentage points, and then they follow it up with 150k when again 232k is more real a -35% error.  

Below a 3 year non farms employment chart shows the seasonal employment levels channel.  January is always the low point due to end of the year retirements, seasonal retail layoffs and corporate end of year house cleaning.  It is clear January 2016 is directly in line with the previous January employment levels.

Total non farm 2016-2.xlsx

Construction Employment Remains a Bright Spot, Even if it is Slowing a Bit

construct 2016-2-xlsx

Construction employment even though slowing slightly remains a bright spot in the US economy.  Construction employment growth was approximately 3.7%, 5% and 4.8% for 2013, 2014 and 2015 respectively.  Over the three year window construction employment growth grew about 4.9% per year.  That is the good news, the somewhat less good news is construction employment is slowing.

Year over year construction growth peaked in Dec 2014/January 2015 which is abnormal given the typical winter season construction slowdown in the US.  January 2016 showed 4.45% YoY growth in construction employment, which is pretty great given an economy that is growing at around a 2% annual rate and manufacturing employment growth of under 0.5% for January 2015 to January 2016.  4.45% YoY growth for January 2016 looks worse than it really is when you consider how strong employment was in early 2015.  For the period of July 2014 to May 2015 construction employment growth averaged about 5.5% YoY, and Dec 2014 to Feb 2015 averaged 6.2%.  So January 2016 is being compared to a huge January 2015.  While January 2016 shows 4.45% YoY which is slightly less than the 2015 YoY average of 4.84% growth, January growth exceeds the growth average of the prior 6 months which was a bit lower at 4.16%.

Given the high bar set in 2015 there is no indication that US construction is coming off the rails.  Even though manufacturing and energy are slumping badly, construction employment is an indicator that a recession may be a ways off.  February and March will provide a much better picture of whether a construction slump is also in the offing for 2016.

Below I'm including another view of US construction employment which does show a bit of slowdown in 2015 over 2014, but has yet to show any indication that employment is about to fall out of the growth channel in a significant way.

construct 2016-2-xlsx

US Manufacturing is NOT Doing Well

manufacturing 2016-2.xlsx

Just read an article stating that US manufacturing is doing great and US employment was at a 7 year high.  So first thing I did is pulled the non-seasonally adjusted data from BLS to see if this was really true or if it was just a byproduct of the seasonal adjustment.  I've known for months that all data has been pointing towards a US manufacturing slowdown, so I was wondering if we had actually made a significant new high.  The data is shown above, plotted YoY for seasonal clarity.

For a true 7 year YoY comparison I compared Jan 2016 to Jan 2009.   What did I find?  12.45M jobs in Jan 2009 and only 12.25M jobs in Jan 2016, so 200k down.   From the beginning I knew the author likely used seasonally adjusted data, but I also discovered he compared Feb 2009 to Jan 2016.  The Jan 2009/Jan 2016 comparison still shows a 200k drop using the seasonally adjusted data, but only a 24k drop when compared to Feb 2009.  With rounding Feb 2009 equals Jan. 2016 at 12.4MM.

Still the article is clearly misleading.  If you look at 2014 vs 2015 employment data it is clear to see since August US manufacturing employment levels have been contracting towards 2014 levels.  Using non-adjusted data, YoY data shows November and December 2015 employment levels were only 0.285 and 0.21% higher than in the same 2014 months.  This is after showing 1.4% to 1.7% YoY gains each month in the period June 2014 to May 2015.  The only glimmer of hope for manufacturing in the latest jobs report is that January shows 0.42% gain YoY compared to an average of 0.25% YoY the previous two months.  Yet 0.42% is still below the 0.54% shown in October 2014.  Regardless, there is clear contraction with little indication that manufacturing jobs won't cross below 2015 levels in 2016.  So far the most optimistic estimation would be that 2016 employment levels closely match 2015 levels.

Of course the article does not even begin to talk about how manufacturing employment is still down 2MM jobs from the 2005/2006 peak.  That is about a 14% loss in manufacturing jobs in a decade while the US working age population grew 6.6% over the same period.  

2/9/2016 Update:  Even the fundamental claim that 29,000 manufacturing jobs were added in January is totally due to the magic of seasonal adjustment.  The reality is that in absolute, non-adjusted terms, manufacturing jobs are almost always lost in January.  January 2016 was no different with 66k lost.  This is a small loss by January standards.  But when a wider view is taken total manufacturing job losses from August thru January were 155k, compared to a 93k loss last year, and 120k loss the previous year.  A 155k loss exceeds the average loss  (-134k) over the last 6 years of the Aug-Jan period.

Below I'm including a plot of seasonally adjusted data which clearly shows the same trend as the non-seasonally adjusted data set.

manufacturing, adj 2016-2.xlsx

US Demand for Crude is Not Weak

Plot of Year over Year US Refinery Inputs  (psw01 2016-2-3)

If demand for petroleum products and therefore crude is weak, someone better tell US refiners.  As can be seen in the graph above, US refinery inputs are running much higher than this time last year.  In fact US refinery inputs are averaging 3% higher for the month of January 2016 than January 2015.  For all the hype that much cheaper crude is on the way, US refiners are signalling that they do not believe that to be true.  Given refiners already have high storage levels for gasoline and very high storage levels for distillates why would they be running at such a feverish pace when they could slow down earlier and more dramatically that previous years.  Allowing crude inventories to build and refined product inventories to fall, putting more pressure on crude and propping up gasoline and distillate prices.  Just like last year, they have more orders on the books or expect more orders than the average bear.

Current US Production Situation is Not the same as Jan 2015

psw01 2016-1-21

Recently I saw a published analyst state that US production is now essentially the same as it was in January 2015 since for the week of January 23, 2015 US production was reported by EIA as 9.213MM bbl/day and for the week of January 29, 2016 it was 9.214MM bbl/day.  However, nothing could be further from the truth.  For starters when Unaccounted for Oil is included the January 2015 weekly average was 9.55MM bbl/day compared to 9.13MM bbl/day for January 2016.  This is a 426K bbl/day decrease year over year, a 4.5% drop.

Then we should factor in the 2000 well completions/month in Texas during the summer of 2014 and 1500 well completions per month in the last several months of 2014.  North Dakota was running over 250 completions/month during the summer of 2014 and finished the year running about 140 completions/month.  Currently Texas is doing less than 800 completions/month and ND is doing about 20-30/month.

While US production surged into the beginning of 2015 and was set to surge for many months more, albeit slower than in late 2014 and early 2015,  US production is has been falling for months and is set to accelerate that decline in coming months.

With Canadian tar sands set to go into maintenance in March, possibly in February, and US declines set to hit in a big way at the same time, US inventory will not build at anywhere close to the rate that bears expect.  Some are talking of topping out inventory.  They will be shocked.

EIA Overestimates US Production, Huge Production Drop in the Works #3

                                                                                                                           psw01 2016-1-21

I've determined that I should have started a blog series on this topic months ago and called it EIA Misstates US Production.  I've added the US production data from the monthly reports.  It shows the weekly production estimate made a wide divergence to the low side in the July-November 2015 window.  So a false rise in production has been shown in the last 3-4 months to account for a drastic understatement earlier.

Still the drastic negative values for Unaccounted for Oil and therefore the very low values for US Production + Unaccounted for Oil point to a major fall in the works or one that already exists.  If one looks at the monthly data, which lags 2 months, you can see that US production fell 65k bbl/day per month from September to November.  Is it really believable then with the very low prices in November, December and January that US production is rising, especially in the lower 48?  When you look at well completion rates in Texas and ND the answer is clearly no.  The most recent data for Texas oil well completions is shown in the above chart and the most recent ND well completion data can be found here ND Oil Well Completions.  A generous nod to the EIA estimate using the monthly data for a guide and a 65k bbl/day/month rate of decline, puts current production at 9.19MM bbl/day compared to the EIA 9.21MM bbl/day estimate.  Which makes the EIA estimate close.  However, the EIA monthly data shows a 79k bbl/day drop from Sept. - Oct. while completion rates were much higher than in recent months and the 5% decline rate for shale wells leads one to believe recent declines are likely more like 100k bbl/day/month.  This would put US production at 9.02MM bbl/day on average for January, and would project about 8.9MM bbl/day for February.  This also means current production levels could actually be about 8.95MM bbl/day.

Since Dec. 4, US Production + Unaccounted for Oil has come in under 8.8MM bbl/day 5 weeks out of 9.  The frequency and magnitude of the negative Unaccounted for Oil swings below the reported US production number points to something very major in the works for US production.

Since Alaskan production typically increases in the winter, and shale wells are in the lower 48, not Alaska, the brunt of any declines is really in lower 48 production.  This really calls into question the lower 48 production number which has been estimated at exactly 8.703MM bbl/day for each of the last 3 weeks.  The fact that lower 48 production is estimated to be exactly the same for 3 weeks running in the face of drastic rig count cuts, lower prices and few completions, points to an EIA weekly estimation model that currently has little clue what is really happening.

What does this all mean?  Currently EIA shows 8.703MM bbl/day production for the lower 48.  Realistically lower 48 production could be in the 8.35-8.45MM bbl/day range.  My feeling is the huge surge in Canadian imports has confused the US production estimation models since it is causing unexpectedly large US inventory builds for the season.  So if we see Canadian tar sands production drop off in the coming months--maintenance is scheduled to start in March, but given the low price environment many producers want to move up schedules--we should also start to see massive drop offs in reported US production.  Until the models catch up we will continue to see frequent and large negative values for Unaccounted for Oil.

It remains my belief that computerized trading systems pick up on and trade based on the Unaccounted for Oil data.  This weeks move up post EIA report was entirely credited to hopes for a Russian + OPEC cut, but if that is the case why did futures immediately drop when the data was reported and then surge rapidly in the following minutes, while the Russian reports had been out for hours with little reaction.  The CFTC COT reports also show that swap dealers, refiners and some other professionally managed money is growing heavily long now.  Could it be that those with the closest ties to the physical oil, know something the rest of us don't and the Russian/OPEC production cut talks are just a distraction from where the real cuts are coming from and the magnitude.