Monday, December 7, 2009

$300 Oil Needed to Reproduce '70s.

Year over year improvement in fuel economy of new light vehicles, left scale, together with oil price pain index, right scale.  See text for details.  Source: EPA for fuel economy statistics, BP for oil prices, with estimate for 2009 based on data through November, BEA for GDP.

Update: I fixed an error in the description of the price pain index from the first version of this post - apologies for any confusion.

I've been exploring the EPA data for the fuel economy of new light vehicles I mentioned on Friday, and have come up with a somewhat plausible story for what's going on.  I'd be interested in everyone's feedback.

Just to refresh our memory, the big question is why did the 2005-2008 oil shock cause a much poorer improvement in new vehicle fuel economy than the oil shocks in the seventies did?  To kick off, this first graph just shows the basic facts.  I plot the year on year improvement in new light vehicles (according to the EPA) on the left scale, and oil prices on the right scale.

Year over year improvement in fuel economy of new light vehicles, together with average annual oil price (expressed in $2008).  Source: EPA for fuel economy statistics, BP for oil prices, with estimate for 2009 based on data through November.

You can see the basic issue - whereas $100 oil in the 1970s caused our parents to start increasing light vehicle fuel economy by 4%, 6%, or even 16% per year, similar prices recently only caused us to make an anemic 0-3% increase in fuel economy.

The fuel economy data is remarkably noisy from year to year.  This noise appears unlikely to be correlated with anything of interest, and will reduce correlations.  In this next graph, I've shown a three year centered moving average, which I will use in subsequent analysis, just to reduce the short term noise.

Year over year improvement in fuel economy of new light vehicles, together with average annual oil price (expressed in $2008).  Source: EPA for fuel economy statistics, BP for oil prices, with estimate for 2009 based on data through November.

If we now make a scattergraph of these two things, we can see the relatively poor correlation:

Year over year improvement in fuel economy of new light vehicles (smoothed) versus average annual oil price (expressed in $2008).  Source: EPA for fuel economy statistics, BP for oil prices, with estimate for 2009 based on data through November.

The R2 is only 43%, largely reflecting the difference between the 1970s response and the recent response.

Thinking on the likely causes, I identified two important differences.  One thing is that we are wealthier (on average) than in the 1970s.  To the extent people have more income, they can afford to pay for the things they really want (like oil products).  Furthermore, we have more efficient vehicles than our parents (or grandparents!) did in the 1970s.  Obviously, if you have a car that only uses half as much oil to go the same distance, you can afford to pay twice as much per gallon and still be able to do whatever you do in your vehicle within the same overall budget.

To capture these effects, I constructed an "oil price pain index", in which I divided real oil price by real GDP (in billions of $2005), and divided by average fuel economy.  I then multipled by 10,000 just to place the index on a convenient scale.

Year over year improvement in fuel economy of new light vehicles versus oil price pain index.  See text for detail.  Source: EPA for fuel economy statistics, BP for oil prices, with estimate for 2009 based on data through November, BEA for GDP.

This has much more explanatory power - an R2 of 86% is pretty good for a single parameter fit in the social sciences!  For example, it's comparable to the agreement between the household survey and establishment survey estimates of employment, which are kinda-sorta measuring the same thing.

Finally, here are the same variables back in the time domain (the relationship between the two scales is not exactly that implied by the regression above, but a rough approximation of it).  

Year over year improvement in fuel economy of new light vehicles, together with oil price pain index. Source: EPA for fuel economy statistics, BP for oil prices, with estimate for 2009 based on data through November.

So this suggests that to get 6-8% annual average improvements in new vehicle fuel economy (which in turn led to 3-4% annual increases in whole deployed fleet vehicle fuel economy) we'd need an oil price index three times higher than in 2008, which would now correspond to an average annual oil price of about $300 (versus the $100 we actually saw across that year).


Steve From Virginia said...

If you are saying that fleet fuel economy has disconnected from fuel prices, you are right. I could have told you that and saved you a lot of tedious chart making!


The fuel economy of the USA fleet has more to do with internal dynamics of the auto industry, which saturated its market decades ago and needs to sell high- margin SUV's and giant pickup trucks in order to maintain cash flow.

That paradigm will continue until it cannot any more, supported by subsidies and newly available taxpayer credit. Once these are exhausted the industry components will crash again, the demand for high margin vehicles continues to shrink. It's not the input cost itself that matters as much as the productivity of the platform in its entirety measured against the input costs of the platform.

The platform is fuel for vehicles, the vehicles themselves, roads to drive them on, destinations along the roads and services at the destinations. The real estate components of destinations and services have been been rendered unaffordable by the trend rise in oil prices since 1998. Parts of the platform might appear affordable in isolation - such as fuel guzzling cars - but the platform as a whole is fatally compromised by the failure of real estate components.

That the platform hasn't collapsed already is due to trillions in credibility expended in support of it by the central banks and governments. Once that is exhausted - as full faith and credit at some elevated level morphs into skepticism - the avalanche will continue from where it left off ... IMHO ...

Datamunger said...

Impressive work, Stuart. Thanks for publishing it.

The version of this post I received via google reader confused me but you've fixed it.

I wonder what correlation you'd get if you used real median household income? Or even Real personal disposable income (aggregate).

Stuart Staniford said...

Datamunger - yeah, sorry about the confusion. I was down in the cafe at the bottom of the hill writing it up from memory, and then realized I'd gotten it a bit wrong when I double checked the spreadsheet. Bad Stuart - won't do it that way again.

Stuart Staniford said...

Steve - the suggestion is that it fuel economy *isn't* disconnected from prices, just that the connection needs to take into account ability to pay the prices, and once you do that, it all makes sense again.

Stuart Staniford said...

Datamunger - I doubt using PDI versus GDP would make a big difference, since the ratio of them only changes by a few percent over time. However, there'd be some real advantage to using a per-capita measure rather than aggregate measures because then we'd have some hope of doing cross-country comparisons. A future post maybe...

KLR said...

Passenger car CAFE standards only rise to 30.2 MPG in 2011, whereas '78/'79 moved the bar up 5.26% and 5.00%
respectively. Corporate Average Fuel Economy - DDWiki

kjmclark said...

Really, that is interesting. It would be better if you pushed it back to the oil embargo earlier in the 70s, since changes there would be before CAFE. There should be a cleaner signal back then. The common story is that it took the earlier shock, combined with continuing high prices, to prime consumers for higher fuel economy when the peak prices hit due to the Iranian embargo.

Related to that, the CAFE changes should skew this data somewhat, since it forced the manufacturers to either increase their fuel economy or pay fines. That just means that the signal is mangled a bit by the mandates. Still, what people were willing to buy is a much stronger signal than the requirements, since CAFE was largely following what people were already doing.

But it's hard to argue against that high a correlation. Of course, that means prices will have to get painful indeed before we get real change in fuel economy again. Prof. Hamilton figures it would take $130/barrel oil to start doing serious damage to the economy, or 6% or so of average consumer budgets. This makes it look like we'll keep crashing before we deal with the problem.

Stuart Staniford said...

Kjm - the data on the EPA website only go back to 1975, which is why I started there. Do you know of older data offhand?

KLR said...

CAFE was only enacted by Congress in 1975. An NAS report actually concluded mileage would be higher in its absence: Effectiveness and Impact of Corporate Average Fuel Economy (CAFE) Standards (2002) I've seen mileage estimates for pre-1975 somewhere, just a chart I think.

This correlation you're showing for 30 years ago is a complete chimera, right? In the absence of the 1979 fuel price spike the vehicles for sale would've been of substantially increased mileage anyway; consumers' shift towards smaller cars had that much more an impact, but that should be analyzed outside of mandated increases.

Stuart Staniford said...

KLR - my mental model is that CAFE more or less just validated what people were doing anyway on account of price, and it's the latter dynamic that is being captured by the "price pain index".

kjmclark said...

Stuart - I'm sure I can get those data, but I'm not in the office today. I work at the lab that produces those data, so I'm sure it's around somewhere. I'll head down to the library tomorrow.

Stuart Staniford said...

KJM - cool! Thanks...

Stuart Staniford said...

I just checked and the Transportation Energy Data Book looks to have the data back to 1970 which helps a bit. But going back further would be better (the other series used go back almost to WWII)

kjmclark said...

That turns out to be a good question. It stumped our librarian for a bit. He came back a few minutes later and said we have old copies of ... the Transportation Energy Data Book that might have that information. Apparently we have dusty old copies of that and some DOT books that might have those data. I'll head over to the library and dig through those a little later.

kjmclark said...

OK, so here's what I found. The October 76 copy of the Transportation Energy Conservation Data Book (ORNL-5198 Special) had a nice table on page 88 (Table It lists Average fleet fuel economy (calendar year basis) and New car fuel economy (model year basis), for 1950 - 1973. Unfortunately, the new car fuel economy data only run from 1957 to 1973. Here they are:
1957, 13.67
1958, 14.07
1959, 13.85
1960, 13.36
1961, 13.55
1962, 13.96
1963, 12.62
1964, 13.49
1965, 12.98
1966, 12.95
1967, 12.86
1968, 12.44
1969, 12.21
1970, 12.51
1971, 12.21
1972, 12.03
1973, 11.67

That book cites US EPA "A Report on Fuel Economy", October 1973 as its source. I looked that up on microfiche. It was actually titled "A Report on Automobile Fuel Economy" from October 73. Those data don't go any farther back. It does make clear that the values are actually sales weighted miles per gallon, which you'd expect, but are from EPA testing(!) Of course, the EPA didn't exist in 1957. I talked to our librarian about that, and he pointed out that OTAQ's ancestor was a Public Health Service lab in Cincinnati, OH that was doing testing at the time.

The other interesting thing is how many old fuel economy reports there are from the early 70s. No surprise, I suppose, but as I was only 5 in 1973, they're pretty interesting reading. I can see why many baby boomers would have a "been there, done that, not a problem" attitude toward peak oil. They're probably expecting a President Reagan to come along and solve the problem for them.

Mike Aucott said...


This is a thought provoking analysis as usual, but in my view there are some other important variables that should be considered before concluding that the price of oil will have to be $300/bbl before we see real improvement in motor vehicle mpg.

As it happens, I also have some historical data on motor vehicle efficiency. A while back, I collected a bunch of historical mpg and vehicle weight data and compared it with more recent data. I developed a plot of ton-miles per gallon for several different periods of history. (A ton-mile per gal represents the weight of a vehicle divided by mpg. It's a measure of efficiency that normalizes for differing weights of vehicles.) From the 40s through the 60s, the U.S. automobile fleet averaged about 27 to 29 ton-miles/gal. (A car that weighed 1.5 tons and got 18 mpg would have an efficiency of 1.5 x 18 = 27 ton-miles/gal.) By the mid-70s, this efficiency had declined to about 24 ton-miles/gal. The main reason for this decline was almost certainly air pollution controls. By the early 70s regulations were in place that limited emissions of hydrocarbons, carbon monoxide and nitrogen oxides. To do this with carbureted engines of the day manufacturers had to resort to a numbing array of pumps, diaphragms, and de-tunings. The result was that engine efficiency - and drivability - declined. Those of us who drove the cars of those days will remember what dogs the cars of the mid- to late-70s were compared to cars of the late 60s and earlier.

In 1975, the law requiring CAFE standards passed, and by the late 70s, the CAFE regulations were requiring manufacturers to increase mpg. The view that CAFE just put into law what was already happening is simplistic. Yes, Congress does what people want - to a degree, but it is extremely unlikely that manufacturers would have increased mpg on their own at the rapid pace that CAFÉ required.

Also at that time inexpensive printed circuits and micro-computers came along (1980 was the first PC) and these allowed computerized engine controls, including fuel injection, which allowed a major increase in engine efficiency. The result was that ton-miles per gallon went up to about 31 by 1980, and then to 36 or so by the mid-80s, by which time most cars had fuel injection. This allowed manufacturers to meet the CAFE standards without having to sacrifice weight and acceleration. Ton-miles/gal stayed at about 36 through 2001. (I haven't carried the analysis beyond that.)

It was very likely largely the improvement in engine technology made possible by the microcomputer coupled with CAFE standards that brought about the big up tick in mpg in the early 80s.

Also, while the "petroleum pain" indicator and its correlation with the change in mpg is interesting, things can appear to be connected when they are not. The petroleum pain factor is influenced both by the petroleum intensity of the economy and the mpg of the time. Either more use of petroleum per GDP, or lower mpg, or both, will lead to a higher "pain" value. In the late 70s to early 80s, the U.S. got much of its energy from the combustion of residual fuel oil and heating oil. The country's industrial base was also dominated more by heavy industry, such as steel smelting, much of which has now moved elsewhere.

This leads me to a prediction that can soon be tested. In 2007, Congress passed the Energy Independence and Securities Act, which requires substantial increases in fuel economy. If there is not a fortuitous intrusion of new technology as there was in the late 70s and early 80s, the manufacturers will have to meet the higher mpg by reducing weight and performance. They'll do this if they have to. Some way, they will meet the new requirements, and we will see another significant increase in mpg of new vehicles in a few years irrespective of the price of oil.