Tuesday, December 1, 2009

Driving is Extremely Inelastic


US economic output (GDP in chained 2005 dollars) per vehicle mile traveled 1960-2008 (left scale), together with annual average oil prices adjusted for inflation to 2008 dollars (right scale). Source: FHWA for VMT data, BEA for GDP, and BP for oil prices.

This is a critical fact that you must understand if you want to understand how the world will respond to a future of tight oil supplies.  People do not conserve on driving until they are forced to by a bad economy.  At least in the price ranges seen in recent decades, people do not act as though gas prices are a significant deterrent to driving as much as they feel they need to.

Recall this graph which I've posted several times now, showing the amount of GDP created in the US per mile driven.


US economic output (GDP in chained 2005 dollars) per vehicle mile traveled 1960-2008. Source: FHWA for VMT data, BEA for GDP. 
The first impression obviously is it doesn't vary much.  Both the economy and the amount of driving has expanded enormously since the 1960s, but the ratio is not much changed.  Now let's blow up the scale on the left hand side, so we can study the little fluctuations better, and add oil prices into the picture (on the right had scale).


US economic output (GDP in chained 2005 dollars) per vehicle mile traveled 1960-2008 (left scale), together with annual average oil prices adjusted for inflation to 2008 dollars (right scale). Source: FHWA for VMT data, BEA for GDP, and BP for oil prices.

The first thing to notice is there really is no correlation between these two lines across most of the period.  Formally, the R2 correlation coefficient is 2% - zero for all intents and purposes.  In particular, notice that in the big price hikes in 1974 and 1979, the GDP/mile ratio did not step up - which is what you'd expect if people were conserving on "economically frivolous" driving as a result of high oil prices.

More recently, both the GDP/mile and oil prices have been increasing.  But GDP/mile leads - it started rising several years before the oil prices, which makes it hard to see how the latter could be causing the former.  Again, my hypothesis is that the rise in GDP/mile is likely due to the effects of the Internet on driving - it's provided a substitute for a certain amount of face-to-face interaction previously required to create and deliver goods and services, and thus allowed the economy to get by on slightly less driving.  It may also have provided some users with a substitute activity - teenagers glued to computer games instead of driving to the mall.

Now, I'm not saying oil prices have no effect on driving behavior - they do.  But, at least in the realm of oil prices so far, what I am saying is that effects of oil prices on driving are entirely mediated via changes in the level of GDP.  People behave like the income elasticity of driving is about one, but the price elasticity of driving is about zero.  In other words, people don't say "Oh, oil prices are high, I better drive less", thus bringing oil prices down.  Instead, they keep driving, and then high oil prices trigger a recession, and then people drive less because some of them lost their jobs and don't need to go to work, and don't have any money to spend at restaurants and stores.

People really want to drive.

9 comments:

kjmclark said...

You made a very compelling argument for that in a post (two?) at the Oil Drum a while ago. It wouldn't hurt to link to that.

Datamunger said...

what I am saying is that effects of oil prices on driving are entirely mediated via changes in the level of GDP

That may be an exaggeration.

For instance: what if we look at VMT per capita? That peaked in spring 2005 despite the fact that GDP and various employment measures continued to grow until the end of 2007. In fact real GDP peaked in 2008.

To stop driving more is a change in behaviour.

But there is more....

The really sharp drops in employment didn't occur until 2nd half 2008, yet VMT per capita was already down some 3% in H1 form its 2005 peak.

Stuart Staniford said...

Datamunger - my point is that the improvement in GDP/mile has been doing on steadily since the early 1990s (the trough was 1992), even though oil prices were modest for the first ten years or so of that period. So it's pretty hard to argue the recent continuation of that trend is oil price driven (especially given there's no evidence of an oil price effect in the 1970s).

Datamunger said...

(especially given there's no evidence of an oil price effect in the 1970s).

Well there is since VMT per capita peaked in mid 1979 and by the time the recession began in early 1980, was already down 3%.

So why isn't there more of an upward move in your GDP/mile graph for 1979 through the early 80s? GDP fell faster than vmt in 1981.

Both then and now the drop in per cap vmt preceded the recession(s).

Not to deny that there is an additional internet related secular trend overlaid on this. Similarly, I agree that in the long run GDP is the biggest factor in mobility. But people do in fact react to prices and conserve before the economy forces them to.

Per capita measure are important because pop growth can obscure the effects of consumer response to prices.

KLR said...

For kjm and others, TOD articles from Stuart on this subject:

Why We Drive

The Auto Efficiency Wedge

An alternate hypothesis: the early 90s were when John Williams says GDP began to be heavily debased in its formulation: Inflation, Money Supply, GDP, Unemployment and the Dollar - Alternate Data Series. He shows it becoming slightly decoupled from the original tabulation in '85, increasing markedly around '96 - just when the VMT/GDP ratio began to take wing and fly. Perhaps not the whole story, but maybe an important factor.

There were correlations between rising oil prices and decreased VMT, though. We've all seen those charts. This didn't play out long term, since, as you'd expect, people responding temporarily to price shocks went back to driving solo when the coast was clear. You can see this in the slopes of the little hills in your VMT/GDP ratio, at the time of the price shocks.

Just found an interesting report on MT ridership, courtesy of the nice folks at the Heritage Foundation. It has a chart of public transit boardings going back to 1900; in your TOD papers all you had for reference was a few .pdfs from BTS etc showing 1980/1990/2000 relative percents, and I'd wondered if ridership showed any uptick before or during the slumps in VMT or recession. The Heritage chart shows that boardings peaked around the end of WWII and went into negative territory in the mid-50s, which nixes my theory; apparently by the 70s it was a very minor factor in travel. The chart says data is from the National Transit Database somewhere.

Perhaps increased shipping is delivering greater GDP than conventional trips in passenger cars to brick-and-mortar, too. Just found this study: Transportation Infrastructure, Freight Services Sector and Economic Growth: A Synopsis - Appendix A - FHWA Freight Management and Operations. Looks like an interesting read.

KLR said...

I assume you know about the BTS State Transportation Statistics. You get travel mode by state for 2000 and 2004-2007. Would be interesting to see if any of that VMT slump is showing up in a shift to MT/walking/biking/work-at-home.

Stuart Staniford said...

Datamunger - if you look at the ratio of GDP/capita and VMT/capita, the "capita" part cancels out :-)

Datamunger said...

Yes, but we are caputs and want to imagine we are part of the equation. :)

Which we are (or at least Europeans (in Europe)) are since fuel taxes there have been used to successfully constrain consumption.

But that canceling reminds me of the Garrett piece noted a few days back where he raises the possibility that CO2 emissions scenarios may be reduced to just consideration of real gross world product and carbonization of our energy sources. Be cool to get your views on it.

Datamunger said...

Would be cool there were recession bands on that GDP/VMT graph.

Just noticed that since the '70s all recessions are related to downward moves in that measure of efficiency except 2: the one in '80 which was minor and, it appears, the current one. This supports the notion that there is a new underlying secular trend at work.