West Texas Intermediate crude oil, which had been selling for $105 a barrel at the end of March, fell to $80 a barrel last week, while Brent has come from $125 down to near $90. These price declines will translate into substantial savings for U.S. consumers in the weeks ahead.
Since Brent and WTI diverged, it has been Brent that matters for U.S. retail gasoline prices; this fact and the reasons for it were discussed here. A regression of the average U.S. retail gasoline price on the price of Brent over 2000-2012 captures the close relation (OLS standard errors in parentheses):
The price of gasoline and price of Brent turn out to be cointegrated, meaning that any permanent change in the price of Brent eventually shows up as a permanent change in the price of gasoline. The coefficients of the above relation are very much what you’d expect. A barrel holds 42 gallons, and the estimated coefficient (0.025) is 1/40. The intercept (0.84) captures an average state and federal tax of 50 cents per gallon plus a bit over 30 cents in markups and other costs.
 |
With Brent on Friday at $91.50 and an average retail gasoline price about $3.47, we’d thus expect gasoline prices to come down another 35 cents a gallon or so from where they were on Friday. Historically those adjustments usually come pretty quickly. For example, last December U.S. gasoline prices temporarily fell about 25 cents/gallon below the long-run relation, but by March they were right back on track.
If gasoline prices do fall from their value in April near $3.92 to $3.12, that would be an 80 cents/gallon swing. With Americans buying about 140 billion gallons of gasoline each year, that translates into an extra $112 billion over the course of a year that consumers would have available to spend on other things besides gasoline.
So should we be celebrating? I’m afraid not. The primary reason that oil prices have come down is because of growing signs of weakness in the world economy. I am very concerned about where events in Europe are going to lead, and recent U.S. data indicate some weakening. Cheap gas helps, but not so much if you don’t have a job.
But I will offer this advice: wait another week or two if you can before filling up the tank.
As someone on the west coast, I wonder just how much savings we will see. I remember a few weeks ago reading somewhere (possibly here?) about a lot of refineries breaking down or closed for maintenance?
You note that sometimes it takes some time for shifts in the crude market to appear in the refined market. And the reverse is presumably also true. So if there’s this sort of limited passthrough in the short-run, doesn’t that suggest that the recent decline in crude prices ahead of a decline in refined prices indicates a supply rather than a demand shock? If it were all demand, it seems like refined prices should have fallen first. (This is all from Ahn and Kogan…)
The price of gasoline and price of Brent turn out to be cointegrated,
Which implies that individually the price of gasoline and the price of Brent each follow a nonstationary path. It also means that your OLS regression is a representation of the long run relationship. Do you know how many months or weeks it takes to settle into the long run equilibrium?
The primary reason that oil prices have come down is because of growing signs of weakness in the world economy.
The other day the NY Times had a story about how low level managers in China are faking economic data in order to hide the slowdown in Chinese economic activity. For example, coal is piling up at distribution points at record levels. Electricity generation appears to be way down despite official statistics that try to hide the problem.
strongly suggests PCE will print at about 1.5% YoY in the coming months.
dual mandate watch epic fail. thank god we are keeping import prices low through massive unemployment. If 12 Million people had a job, they could afford the gas to commute to it.
I agree with this analysis.
But it was a long time coming. China overshot its own carrying capacity, and is now adjusting accordingly.
It could just be a period of softness of a few months. Or it go be a harbinger of a more serious global slowdown.
Here’s what I noticed eyeballing the eia and fhwa data and graphs:
The last few times oil was this low gasoline was about $2.75. We drove more and used ~5% more gasoline. Demand for gasoline from foreign countries must be amazingly high.
Where are all the left wing ideologues who turn lachrymose when the price of oil is galloping higher? At those times they are busy railing against speculators and speculating themselves on the plots hatched by oil companies to keep prices elevated?
So have the cunning and crafty folks who allegedly kept the price elevated now working to keep it down? And if other forces are driving prices lower then why cant these very smart folk step in and levitate the price higher?
Maybe the market is bigger than individual players and maybe for the moment sellers are in control.
Take that Senator Schumer!!
The covariation with gas price may attract more buying interest, less neutral, less inert and will most probably be more volatile. The IEA report shows the industrial consumption on the upside and so is the domestic consumption, the agency confirms the increased stocks and the upside on price.
Looking at the Nymex charts double bottom on the July 12 and volume on the longer duration are good, it could be rated a buy. It should help delicate cash flows balances in the LBO business as “nothing is lost, nothing is created everything is transformed”
http://www.eia.gov/forecasts/steo/pdf/steo_full.pdf
Professor,
Spot on analysis. The drop in oil prices coincided with Bernanke’s announcment that he would essentially do nothing. Traders were hoping for another QE. Most still believe the FED keeping QE in its pocket for when it really gets bad – HELLO!!
The global economy is in serious trouble. The relative stability of the dollar over the past 6 months has allowed commodities to take a breath. Oil is not reacting to lack of demand rather than QE.
Aaron –
According to the BP Statistical Review, annual growth of world consumption for the last five years for coal was 3.4%; for natural gas, 2.7%; for oil, 0.7%. Certeris paribus, we would expect oil consumption to be growing at about the same pace as natural gas. We have for years pegged the inherent growth rate for world oil consumption at 2.4-2.8% per annum, figure 2.4 mbpd / year. But, of course, we haven’t even gotten close to increasing the oil supply at this pace, so we have to re-allocate the existing supply between OECD and non-OECD consumers. And indeed, OECD consumers are providing 85% of the incremental consumption of non-OECD consumers since Dec. 2007 (the beginning of the Great Recession). Thus, the purpose of the price mechanism must be to re-allocate this consumption and therefore be above the carrying capacity of the OECD economies and below the carrying capacity of the non-OECD economies. Carrying capacity is thus defined as the price at which oil consumption in a country begins to fall.
Now, if we look at which countries are in financial crisis, they belong seemingly exclusively to the OECD: Greece, Italy, Spain, Netherlands, UK and the such. None of the countries in financial crisis are in the non-OECD and exclude the usual suspects like the Phillippines, Indonesia, Argentia, Mexico, etc. And none of the countries in crisis are major commodity exporters. Thus, Australia, Canada and Norway are not in crisis; and as its oil and gas production increases, the US seems to be holding up better than Europe as well. Countries who are net donors of oil are struggling; countries who are net recipients of oil are doing better; countries who are net exporters of oil are doing well.
So, oil demand seems to be out-stripping supply, which gives rise to supply-constrained forecasting, which is characterized by the recurring themes of inherent demand and carrying capacity. When oil prices are above a country’s carrying capacity, it will tend to struggle.
Steve, I’m not sure that explains the difference in gasoline price from the price it was when oil demand was this low.
Gas is still ~$1 more (2.75 was a high end estimate).
All I can think of is that gasoline demand is stickier than oil demand, consumers are still demanding as much gas even though industrial activity is down.
My comment on foreign demand was kind of tongue in cheek. I didn’t really think gasoling demand abroad could stay so high relative to oil demand, but I guess it is plausible.
Hypothesis, the US economy is doing faily well. Our demand for gasoline is much higher than the global demand for gasoline and oil. We are driving a whole lot more this months than last.
Aaron –
The US doesn’t set the price of oil. China, or arguably the “Other non-OECD”, does.
See my article on the topic: http://www.europeanenergyreview.eu/site/pagina.php?id=3761
Here are a couple of charts on miles driven, This one’s through April.
http://www.advisorperspectives.com/dshort/charts/indicators/miles-driven.html?miles-driven.gif
This is VMT through April adjusted for potential drivers, which really speaks to a catastrophic loss of physical mobility:
http://www.advisorperspectives.com/dshort/charts/indicators/miles-driven.html?miles-driven-CNP16OV-adjusted.gif
May vehicles miles traveled is -1.1% on April, up 1.7% for a year earlier. Year on year, on a rolling 12 mma basis, VMT is -0.5%, but showing an improving trend. US oil consumption for the three months ending May is -2.0% off the same period the year earlier.
And here’s a chart on real personal income. Statistics this bad are usually assoicated with recessions:
http://www.advisorperspectives.com/dshort/charts/index.html?indicators/Real-Personal-Incomes-YoY.gif
The US economy is not an abject disaster, to be sure. And there are positive signs. But in the broader context, whether it is real income, employment, or physical mobility, the US continues to struggle.
My eyeball was bad. The data show exactly what JDH wrote. Gasoline should come down to about 3.10 if oil stays about 90.
It just hasn’t had time to come down. We should probably see it in the next week or so.
Steven Kopits Now, if we look at which countries are in financial crisis, they belong seemingly exclusively to the OECD: Greece, Italy, Spain, Netherlands, UK and the such.
Hmmm….”financial crisis” is not quite the same thing as being in a recession. The UK is in a recession because its idiotic leaders have decided that it must be in a recession in order to purge the rottenness. Germany is an oil importer and part of the OECD, but I don’t see any evidence that Germany is suffering. And thanks to a horrible disaster Japan is seeing strong economic growth. Japan is also an oil importer and an OECD country.
So, oil demand seems to be out-stripping supply, which gives rise to supply-constrained forecasting
Can you explain what you mean here? I’m not following you. JDH‘s point was that oil prices are coming down because of weakening global demand.
Slugs –
In 2002, if I brought you the following list and said, four of these countries are in debt or fiscal crisis, which would you have picked?
– Nigeria
– Iceland
– United Kingdom
– United States
– Indonesia
– Thailand
– Mexico
– Ireland
If I had further refined the query, and said, Slugs, they’re either all in the OECD or all out of the OECD, then you would have, without hesitation, said, “Well, obviously, it’s the non-OECD countries.” At least I would have said that.
Nope, it’s the OECD countries, and virtually only the OECD countries. Why? Are the emerging economies now immune to hot capital inflows and the OECD countries not? Why was governance apparently so bad in the OECD, and apparently so good in the non-OECD? Why this split?
***************
I agree entirely with Jim about weakening global demand. But why is it weakening?
World demand is the sum of OECD demand and non-OECD demand, for which we proxy in the US and China for simplicity’s sake.
US demand, as I have stated before, started falling in August 2010 at $85 Brent. So US demand has been weak–indeed, negative–for some time. Consequently, the US could not have been sustaining high prices in the absence of a falling oil supply.
Therefore, prices must have been sustained by China. And they were. China’s demand was growing, even in the face of $125 Brent. However, the pace of China demand growth peaked about two years ago and has been falling since. Indeed, by April of this year, it was effectively zero.
Therefore, if US demand is falling and Chinese demand growth is zero, then aggregate demand must be falling–thus, weak oil prices.
However, the question you’re begging is how prices could exceed the carrying capacity of China to begin with. If China is setting the price, how can it set the price above its own carrying capacity? For this, you need either “speculators” or price inelastic demand, such that demand does not react quickly enough to increased prices, thus allowing them to rise above long run sustainable levels. This is what I argue in the article “High Oil Prices are Caused by Consumers, Not Speculators” (http://www.europeanenergyreview.eu/site/pagina.php?id=3761).
If you analyze China’s carrying capacity, it’s probably between $110-120 Brent, most likely in the $112-117 range. Thus, $125 Brent was likely to undo Chinese demand as well, and it did.
Now, normally, when inelastic demand is broken, it’s done so through a recession. In China’s case, because its growth rate is so high, it may just be a slowdown. But we very much do see such a slowdown, and it is consistent with the a model combining price inelasticity and an oil price exceeding that country’s carrying capacity.
So to answer you question: In oil, a supply-constrained market will tend to create unsustainable price spikes which result endogenously from price inelastic demand. This ends in a price correction, normally a downside overshoot related to a recession.
At least that’s the theory.
The price of gasoline and price of Brent turn out to be cointegrated
OK. Makes good intuitive sense. However, I was just going to say that the DW stat for that simple OLS equation probably suggest that the regression residuals are highly auto-correlated. Might be nice to throw in a diagnostics statistic or two now that you have opened the door to this level of technical complexity.
Besides, how robust is the intercept estimate? Does the t-stat have any value?
westslope However, I was just going to say that the DW stat for that simple OLS equation probably suggest that the regression residuals are highly auto-correlated.
The OLS regression is the long run equilibrium. If you took a simple Engle-Granger error correction model (which would probably be okay since we’re only talking about two variables), then you would expect to see a high R-square in the long run OLS regression if the residuals are nonstationary. But you would then take a regression of the differences of the residuals (which should be stationary) and those would become the dynamic short run error correction. This is more along JDH‘s line of work and he could certainly explain it better than I could, but it’s precisely because the residuals are autocorrelated that opens up the possibility of a cointegrating relationship. If the OLS residuals did not have a unit root then there could not be a cointegrating relationship.
westslope: You are correct that the serial correlation of residuals means that the standard errors are not appropriate. I reported them only because some people are used to having them reported. But, as 2slugbaits notes, it is because the raw series themselves are much more correlated (indeed, are I(1)) while the residuals are stationary that makes the regression a meaningful cointegrating relationship.
By the way, one could also have a cointegrating relationship in which the errors have no serial correlation.
JDH By the way, one could also have a cointegrating relationship in which the errors have no serial correlation.
Learn something new every day. I did not know that. I’ve never seen it presented that way. Thanks.
I’m pretty poor at demand side analysis; my primary focus is on the supply side, but it’s interesting to look recent progression in higher annual highs and higher annual lows in oil prices.
Following are the recent annual high Brent prices preceding year over year annual price declines, showing the progression in higher annual highs and higher annual lows.
While it seems a virtual certainty that we will see an annual decline from 2011 to 2012, it also seems almost impossible for the 2012 annual price not to exceed the previous annual year over decline level ($62 in 2009).
EIA price data: http://205.254.135.7/dnav/pet/hist/LeafHandler.ashx?n=PET&s=RBRTE&f=A
1997 to 1998: $19/$13
2000 to 2001: $29/$24
2008 to 2009: $97/$62
2011 to 2012: $111/?
Here is a copy of a missive I sent to the ASPO-USA discussion group that I think explains the progression in higher annual highs:
The Few Becoming The Fewer
Following is a recollection of a scene from the movie “The Longest Day,” (about the Allied invasion of Normandy in 1944). Richard Burton played a British pilot, and a veteran of Battle of Britain. From memory, the dialogue went something like this, “What concerns me about being one of The Few (a reference to Churchill’s Battle of Britain Speech) is how we keep becoming the fewer.”
Our net export data base, which calculates the individual and combined net exports from the top 33 net oil exporters in 2005, has been updated to incorporate the 2011 BP data base. I define Global Net Exports* (GNE) as the combined net exports from the Top 33 in 2005. The updated 2011 data base showed that 22 of the 33 had flat or declining net exports from 2010 to 2011, as Brent averaged $111 for 2011. Of these 22, three of them (14%)–Vietnam, Argentina and Malaysia–slipped into net importer status in 2011, i.e., “The few becoming the fewer.”
There was a small year over year increase from 2010 to 2011 in GNE, from 43.3 mbpd in 2010 to 43.7 mbpd in 2011 (versus 45.6 in 2005). Note that this can be solely attributed to the year over year increase in Saudi net exports (up from 7.2 mbpd in 2010 to 8.3 mbpd in 2011). This was above my upper end estimate of 8.1 mbpd, but they were still well below their 2005 annual net export rate of 9.1 mbpd. In other words, the Saudi post-2005 net export decline rate slowed from 4.7%/year (2005 to 2010) to 1.5%/year (2005 to 2011).
We have seen six straight year over year declines in Available Net Exports (ANE, or GNE less Chindia’s combined net imports), with ANE falling from 40.4 mbpd in 2005 to 35.4 mbpd in 2011 (because of data revisions, the volumetric decline didn’t change much versus 2005 to 2010).
Given a steady decline in the ratio of total petroleum liquids production to domestic liquids consumption in an oil exporting country (P/C), one can extrapolate and estimate when the P/C ratio (which I have renamed ECI, or Export Capacity Index), hits 1.0, which means production = consumption, which of course means zero net oil exports. Given net exports at peak, we can do some simple integration and estimate total post-peak Cumulative Net Exports (CNE). Using this approach and extrapolating the initial three year rate of decline in the ECI for Indonesia, UK and Egypt (IUKE), I estimated that the combined post-peak CNE for IUKE would be 4.0 GB. The actual combined post-peak CNE number for IUKE was 4.0 Gb.
I plan to do this same exercise for the three new members AFPEC (Association of Former Petroleum Exporting Countries)–Vietnam, Argentina and Malaysia (the VAM case history).
In any case, if we extrapolate the 2005 to 2008 rate of decline in the ECI for GNE and for Saudi Arabia, the predicted and actual results for 2011 are shown below:
Actual 2005 Data:
GNE ECI: 3.75
Saudi ECI: 5.60
Predicted and Actual 2011 Data:
(Predicted values based on 2005 to 2008 rate of change)
Predicted GNE ECI: 3.29
Actual GNE ECI: 3.24
Predicted Saudi ECI: 3.80
Actual Saudi ECI: 3.91
So, the 2005 to 2011 rates of change indicated that the GNE Export Capacity Index fell slightly faster than predicted, and the Saudi Export Capacity Index fell slightly slower than predicted, relative to 2005 values (again, predictions based on 2005 to 2008 rates of change). As noted above, at a 1.0 ratio, net exports = zero.
The ratio of GNE to Chindia’s Net Imports (or GNE/CNI) is a related metric. The GNE/CNI ratio fell from 11.03 in 2002, to 8.88 in 2005 and to 5.28 in 2011.
Based on the 2005 to 2008 rate of change in the GNE/CNI ratio, the predicted value in 2011 would be 5.57. The actual ratio in 2011 was 5.28, so the actual ratio of Global Net Exports to Chindia’s Net Imports in 2011 was 5% below the predicted value, based on the 2005 to 2008 rate of change. At a 1.0 ratio, the Chindia region would consume 100% of GNE.
GNE/CNI chart:
http://i1095.photobucket.com/albums/i475/westexas/GNEoverCNI.jpg
*Top 33 net exporters in 2005 (countries with 100,000 bpd or more of net exports in 2005), BP data base + Minor EIA input, total petroleum liquids
Incidentally, I don’t think that China & India will actually be consuming 100% of GNE (Global Net Exports of oil) in 2030, but on the other hand, it sure is one heck of a trend line, and it looks like China’s oil production may be peaking. After US crude oil production peaked in 1970, our net imported doubled in about five years.
Note that at the 2005 to 2008 rate of decline in the GNE/CNI ratio, the Chindia region would be at a 1.0 ratio (consuming 100% of GNE) in 2033.
At the 2005 to 2011 rate of decline in the GNE/CNI ratio, the Chindia region would be at a 1.0 ratio (consuming 100% of GNE) in 2030.
Gasoline prices are almost always a function of actual versus expected supply. It’s a supply story.
Question: if you know that there is a lag between changes in the price of Brent and the price of gas, then why wouldn’t you model it that way?
I’m marvelled at the simplicity of this model. Typically when there are regressions of price on an input price, doesn’t that imply a fixed demand curve? Or does the cointegration solve the problem of having to have simultaneous equations?
Would the markup really be constant? Fixed costs would stay the same but aren’t there important variable costs? Is it too tough to model or not enough data? What do the omitted variables do to the OLS estimators?
I suppose I’m asking the same questions as 2slugbaits now that I think of it.
It appears you used nominal prices. Do you assume that the prices of Brent and gas rise at the same rate of inflation?
Not criticizing here, just trying to learn some econometrics.
Rien DeRien: Ordinary least squares minimizes the sum of squared values of the residual e_t. For a nonstationary series like gasoline prices, the average squared value of the price of gasoline itself g_t would tend to infinity as your sample size gets bigger. When one claims, as I have here, that a particular regression is a cointegrating relation, the claim is that, for those particular values for the regression coefficients, the average squared residual is finite. There are some auxiliary statistical procedures for testing whether such a claim is valid that I conducted but didn’t report, because I try not to let this blog get too technical.
One neat thing about such relations is that the difference between a finite number and an element of a sequence that is headed to infinity might be expected to be pretty big for a sample of this size, meaning one can get a good estimate of that long run relation even if there is a lot of other stuff going on. Yes, one could also model other details such as adjustment lags, but it’s not strictly necessary for purposes of coming up with a reasonable estimate of the cointegrating relationship itself.