If this is a system invariant and not a spurious correlation the implications are stark. With say 2% annual decline in fuel supply hence miles travelled then the US economy will be halved by year 2030. It would be interesting to compare the series for China for the last 10 years to see if other automotive economies follow a similar pattern. There are wake up calls all around but some aren't listening.

"Compared to 1973, it now takes one-third less energy to produce a unit of GDP in IEA economies. An important reason for this development is the considerable energy savings that have taken place in the various branches of manufacturing, in different end-uses in households and commercial buildings, and for different modes of passenger and freight transportation. The IEA analysis shows that without the savings achieved since 1973, IEA energy use at the end of the 1990s would have been 50% higher than what it actually was."

http://www.iea.org/Textbase/press/pressdetail.asp?PRESS_REL_ID=123

"Between 1970 and 2002, U.S. real GDP, measured in 2000 chain-type dollars, grew 167 percent. During the same time period, U.S. freight transportation, measured in ton-miles, grew only 73 percent. Consequently, the freight transportation intensity of the U.S. economy decreased from 0.59 ton-miles per dollar of GDP (measured in 2000 dollars) to 0.38 ton-miles per dollar of GDP (see box).

Freight transportation intensity declined even within the goods producing sector. In 1970, it took 2.1 ton-miles of freight transportation to produce $1 of goods GDP. In 2002, it took only half that amount, 1.1 ton-miles, to produce the same value of goods GDP (in real terms). This trend reflects two underlying changes in the U.S. economy:

*    the downsizing of products towards lighter weight products (such as computers, cell phones, and hand-held digital devices), and
*    improvement in the efficiency of the freight transportation system, not only in terms of faster and timelier delivery, but also higher direct accessibility."

http://www.bts.gov/programs/freight_transportation/html/transportation.html

Paraphrasing Jevon's Paradox here: When we are able to do more with less, we eventually consume the difference anyway.

I think what you've graphed is the result of monetary policy consistantly tuned over the years to accomodate our consumption of energy... AKA goods and services. Plot various cost-of-capital series (bond rates, etc...) along with energy prices against this data. What you have is fascinating.

Will

Stuart - very interesting analysis.

There's some question in my mind about the averaging though - given the large differences between the different groups of vehicles, are you getting the right numbers for the improvement from the "weighted average"? And is the averaging within a sector even giving the right sort of numbers?

Even if the weighting is by highway miles driven, it still might not give you the right improvement ratio.

For an example, let's say cars get 20 mpg, light trucks 15 mpg, and other trucks 5 mpg. Let's say of every highway mile driven, 40% are cars, 40% light trucks, and 20% heavy trucks. Then, with those percentages you get a "weighted average" mpg of 15 (0.4*20 + 0.4*15 + 0.2*5). However, the actual number of gallons of oil used for one "average" mile is 0.08666 (0.2/5 + 0.4/15 + 0.4/20) giving a true average mpg of 11.5, not 15.

I.e. the real improvement may be much less than you would get from a simple weighted average of sector mpg's.

Also, the numbers for "transportation" energy use in your final graph presumably include air transportation (which has gone up considerably) and railway transportation (which has gone down), which would further shift the numbers to use of more energy per mile traveled. Given that transportation is using a greater fraction of the economy's energy that probably agrees.

Off topic for the content of this post, but a style comment....

This post, like several others lately, uses a graph as a teaser for further content that you have to click through to.  However, the way the click-through works (at least on my browser), is that after I click through, the graph is scrolled up off the screen.  The result is that I must first click through and then scroll back up halfway to bring the graph back on the screen.

I'd like to suggest that people who do put a graph (or other graphic) in the post before the click-through, put at least the first paragraph of discussion of it before the click-through as well.  Doing that would mean that people would have the graphic in front of them as they read the initial discussion of it.  Admittedly the user may still have to scroll back after clicking through, but at least it won't necessarily have to be the first thing done after clicking through.

As already clearly pointed out, many things that are included in the GDP use little, in any energy, and the energy consumed in the manufacture of imported goods doesn't show up as US energy consumption.   Therefore, the connection between GDP and energy consumption has to be a fairly weak one. But I think it might be worthwhile to briefly address just what IS the GDP.

Perhaps an economist out there can answer these question:

1. Is the sale of real estate and the commissions derived thereof included in the GDP?

2. Are the commissions for financial transfer activities, such as from the sale of stocks and bonds, included in the GDP?

3. Does a lawsuit contribute to the GDP?

4.  What type of economic activities are NOT included in the GDP?  

Really astonishing graph, but when you think of ot it is quite logical to be like that. Every economic activity in USA is connected with travelling certain amount of miles.

Recreational travels create make the business of the cinemas, theaters, restaurants, parking lots, lodging services etc. This is a two way correlation - to create a product/value you need to travel some miles to produce it and to consume it you also need to travel.
I'd expect this to change only if the avarege miles needed for a product to be created and consumed drop and/or the consumers shift their preferences from "high-miles" to "low-miles" products. Both are unlikely to happen unless energy prices go really high, I'd expect the economy to react with improvements in efficiency and/or contraction of energy-intensive sectors. The other way would require major restructuring of the economy,the infrastructure and the consumer behaviour, which can only be expected in the very long term.

To judge the significance of the graph it would be helpful to know how much variability there had been in the growth of GDP and miles travelled. If both have been relatively steady then it is no surprise that their ratio would be flat. On the other hand if the two have varied significantly, but always together in lock-step, that would be a strong hint that they are in fact correlated.

Here is a graph where the dark blue line is total vehicle miles travelled (1987=1.0):

http://www.fhwa.dot.gov/policy/ohim/hs02/mvfvm.htm

As you can see, the line is almost completely straight, with only a few tiny jigs and jags. I assume therefore that GDP growth is also straight (since their ratio is constant). Dividing one form of straight growth by another produces a flat ratio.

This means that the correlation is likely spurious. Any other commodity whose consumption has growth steadily over the same period would also show a flat curve when divided into GDP. Perhaps lawn furniture purchases per year or bicycle miles travelled per year might show similar curves.

The whole notion of "productivity of vehicles" implies a degree of causation which is entirely unfounded. The mere fact that vehicle miles travelled has grown at a steady rate just like GDP does not imply that these vehicle miles are producing GDP growth, any more than we would speak of "productivity of lawn chairs" if their sales also happened to be growing steadily.

In short I think this analysis is misguided and is based on a spurious correlation due to the fact that vehicle miles travelled has grown at a very steady rate over the past several decades.

This is a time series modeling issue, and they get muddled. I think there are two statistically challenging things going on:

Multicollinearity, with GDP and miles jointly dependent on each other:
     More highway miles cause more GDP (because we sell more gas, oil, cars, trucks, mechanical service, road maintenance, insurance, radio advertising, road food, etc.) and
     More GDP causes more highway miles (more trucks of commodities, more salesmen, more new suburbs create more commuters ...)
     This is a feedback loop, and it gets hard to see what is causing what.

Both miles and GDP also show the incredibly aptly-named autocorrelation, (or series correlation) where the numbers from this year are highly correlated with last year. We don't start each year's GDP or mileage with a zero base: we do what we did last year, plus or minus a little.

Unfortunately, without a lot of data, it's hard to tease out the exact relationships. It might work if one data series is transformed. It is a generic problem. I've been looking at energy per-capita figures, and they are a bit clearer but not without statistical problems. More people cause more energy consumption. Does more energy consumption cause more people? Worldwide, long-term, absolutely. In America, 1970-2005? Probably not.

Re: "there must be some deep reason for this, but it's not obvious what it is"

I'm sure I'll get lambasted for this, but why not take the $GDP/miles result at face value? The correlation is expected if we simply consider the geographical expansion of the area in which economic activity takes place (by which I mean suburban and exurban sprawl and the concomitantly larger network of physical economic connections - driving - between places). Now, what would it take for the graph to not be flat? Since 1970,

1. $GDP/mile is rising over time. Driving less and getting more done. Geographically, the network area over which goods & services are delivered is flat or has slow growth.
2. $GDP/mile is falling over time. Driving more and getting less done. Geographically, the network area over which goods & services are delivered is expanding faster than the amount of goods & services delivered.

Plainly, in American culture #1 is impossible as stated. But, I'll have to think about #2. It seems implausible but I can't pin down why.

I'm just throwing this out there for discussion.... but think literally of the physical network graph of highways, roads, streets, etc. over which goods & services are delivered.

The chart of average fuel efficiency of vehicles show no improvment in mpg for trucks. The reason may be that the average weight of the trucks is increased. Physics tell that the mpg depends on the weight of the vehicle. To measure the real fuel efficiency changes - the technological advance - we should measure mpg / ton.

The problem of measuring energy used per GDP dollar is in the GDP. Structural changes and the way GDP is calculated have quite significant effects. Foreign trade can have significant effect also. As I estimated in a previous post, the foreign trade deficit may represent 10% of the energy usage in the US economy and even more.

I am not quite ready to accept the idea of significant and continuing technology-based energy efficiency increase. No doubt the technological energy efficiency has gone up somewhat but it has a price. This price is investments. Adopting new technology means using new equipment. This means that increased energy efficiency means using more energy. So we get a little bit more with more.

We speak about increasing energy efficiency but see growing overall energy consumption. Jevons' paradox works, but this might tell also that increasing efficiency is really tied to increasing consumption. One explanation could also be the learning curve. Doing something long and widely enough teaches to do it better and more efficiently. But doing means usually using energy. So energy gets technological advance. Besides, learning curve leads to optimization, not to continuously increasing efficiency.

I have even speculated with the hypothesis that all seeming advances in energy efficiency are generated by more energy use. The world may be more efficient but still uses more and more energy. If this hypothesis is true decreasing energy supply would also slow down energy efficiency improvment. I think that there is some evidence to support this view.

We get less with less, may be even a little bit more less with less.