« The Case for Charisma | Main | Visualize This! »

May 11, 2005

Burnt CAFE

It's a rare treat to read a dry, technical report and--almost by accident--learn something surprising, counterintuitive, useful and (at least to me) genuinely new. 

Which is exactly what happened when I read this paper (beware, pdf) by Todd Litman at the Victoria Transportation Policy Institute.  The upshot:  raising vehicle fuel-economy standards, which always seemed to me like a good idea, may actually be counterproductive, even if they're truly successful at reducing the amount of gasoline the average vehicle consumes per mile.

Now, I'd long heard the argument that current fuel economy standards (also known as Corporate Average Fuel Economy or "CAFE" standards) were ineffective in practice, because of a big loophole:  current CAFE rules hold big pickups and SUVs to a lower standard than cars.  This has let manufacturers skirt CAFE standards by shifting production away from cars to big trucks, which in turn has led to a gradual decline of the overall fuel efficiency of the US vehicle fleet -- thwarting the purpose of the CAFE standards.

To me, that seemed to be an argument for fixing the big-truck loophole, rather than scrapping the standards outright.  Clearly, the techonology exists to produce more efficient vehicles; and our economy would be better off, and our roads safer, if there weren't as many huge gas guzzlers on the road.  So ratcheting up fuel economy standards for SUVs and pickups--though currently a political non-starter at the federal level--certainly seemed like it would be a wise move over the long term.

Or so I thought.  But Litman's article argues, fairly convincingly, that CAFE standards suffer from the law of unintended consequences.  Improving vehicle fuel economy would reduce the amount of gas a vehicle consumes per mile--which, as a consequence, makes driving cheaper.  And according to Econ. 101 (which I never took, sadly) if something's cheaper, we do more of it.  So, all else being equal, the cheaper it is to drive, the more driving we do.

Here are the numbers: a 10 percent improvement in fuel economy reduces fuel consumption by 6 to 8 percent (a good thing), but also increases driving by 2 to 4 percent.  The increase in driving increases congestion, parking costs, noise pollution, and traffic accidents.  Plus, making driving cheaper fosters sprawl, while an increase in vehicle traffic makes walking and biking more dangerous and less convenient.  Assigning a rough dollar value to these competing effects, it looks as though an increase in fuel economy standards is actually a net economic loss for society, because the costs of increased driving outweigh the benefits of fuel and pollution savings.  Not at all what I expected.

LItman gets similar results for alternative fuel subsidies -- by  reducing the cost of fuel, alternative fuel subsidies encourage driving.  The costs of increased driving largely offset any benefits in petroleum savings.

Fuel taxes are another matter.  On equity grounds, they don't do so well, since fuel taxes tend fall more heavily on the poor than on the well-off.  But otherwise they appear to do a fairly good job of reducing both fuel consumption and driving -- and all the associated costs for pollution, parking, etc.

But even better than fuel taxes, Litman finds, are strategies such as pay-as-you-drive car insurance (PAYD), which turn some of the fixed costs of driving (such as car insurance) into mileage-based fees.  Under PAYD people who drive less would pay less, which is a pretty good deal on equity grounds, and certainly gives people more control over their driving expenses.  But by discouraging low-value car trips, PAYD does well on just about any other measure one can think of.

I'm not quite to the point of thinking that improved CAFE standards are a waste of time.  But Litman's analysis certainly makes me a bigger fan of PAYD than I used to be.

Posted by ClarkWD | Permalink


Both fuel taxes and PAYD insurance need to be implemented and expanded (and quickly) to reduce petroleum consumption, or else the effects of Peak Oil will probably do it for us. The disadvantage of relying on Peak Oil to reduce consumption is that if we just wait around for it, consumption will not be reduced in an environmentally or socially rational manner (and perhaps not even economically rational, depending on whether military force gets involved). Economists like to talk about "demand destruction" when discussing things like oil shortages, but that really is just a euphemism for "he/she with the fewest dollars, loses".

Posted by: Roy Smith | May 11, 2005 2:14:54 PM

I think that anyone can adapt to higher fuel taxes, and that the "poor" are more likely to do so than "middles" who have the ability to ignore them.

Posted by: odograph | May 11, 2005 3:05:11 PM

What about the minimum wage earner who commutes to his graveyard shift in downtown Seattle (no bus available, because he works the graveyard shift) in his 20 year old car that gets 24 mpg (at best, and that is all he can afford) from Marysville (because that is the closest he can get to downtown where he can afford to house his family)?

How do you propose that he adapt to higher fuel taxes or high gas prices in general? Although I support high gas prices in principal as a means to reduce petroleum consumption overall, the fact of the matter is that the working poor get hit very hard (and are often unable to adapt) whenever gas prices go up.

Posted by: Roy Smith | May 11, 2005 3:33:28 PM


I haven't read the ful VTPI paper you cite, but I've got to admit I'm skeptical of their conclusions. Most of the analysis I've read about gasoline prices and driving suggests that it takes a substantial increase in gasoline prices before people reduce their driving habits. (E.g., this Christian Science Monitor article: http://www.csmonitor.com/2004/0413/p02s02-usec.html ) That is, higher fuel costs (which could come from inefficient cars or expensive gasoline) are not a significant deterrent to driving. For instance, a poll cited in the Monitor suggests that less than 10% of people would "drive somewhat less" if gas prices rose from ca. $1.50 (where they were in April 2004) to $2.75 per gallon. Conversely, cheaper fuel costs would be unlikely to provoke a stampede of people into their cars.

If the VTPI conclusions were accurate, we would see gasoline consumption and annual miles driven respond to fluctuations in gasoline prices in recent years. Might be worth using those data to test their hypothesis and their figures for price elasticity of demand.

Posted by: Seth Zuckerman | May 11, 2005 4:25:24 PM

I know what I see around me. The genuinely poor take rapid transit, ride bikes, carpool, and drive the bare-bones little cars that other people cast off.

Now, you make a compelling hypothetical with your poor man in Marysville ... but do you seriously want policy based on hypotheticals? Where are your studies of transportation costs?

I'd love to see those studies, and I'd be willing to flip my opinion when faced with fact, but as my first comment sort of implies ... I'm tired of the untested assumption.

Posted by: odograph | May 11, 2005 4:47:06 PM

Correction: gas prices in April 2004, when the poll was published, averaged $1.83 per gallon (Monthly Energy Review, US Dept of Energy). In March 2004, when that poll was taken, they were $1.76.

Posted by: Seth Zuckerman | May 11, 2005 5:01:20 PM

Seth --

Yeah, you always have to temper readings of studies like this with some skepticism -- which is one reason that I'm not convinced (yet) that CAFE standards are counterproductive.

But on the relation between gas prices and driving: the econ. geeks say that there's a different reaction for short-term vs. long-term price changes. Over the short term, a 10% increase in gas prices leads to a 1-2% decrease in consumption. That's not much -- in fact, it's barely noticeable. And it mostly happens by vehicle switching--a person with access to 2 vehicles may start driving the sedan rather than the SUV for some trips.

Over the long term -- that is, if price increases are sustained, and consumers come to expect that that prices will continue to stay high -- consumption drops much more, through a variety of mechanisms. Some people change their commuting habits. Some take gas prices & travel distances into consideration when choosing new homes. Others combine some trips, or reduce low-value trips. People buing new cars tend to buy slightly more efficient vehicles, etc. Those effects take longer to materialize, since they mostly rely on changes in habits (people take some time to adapt) and, more importantly, big decisions that only come around every few years (new cars, new homes, new jobs, etc.)

Still, I think you're right to be skeptical. The econ geeks may be underestimating the effects of fuel prices on people's behavior, given how wealthy we've become. And they don't all agree on the actual degree of fuel price elasticity.

If you really want to geek out on this stuff, go here: http://www.vtpi.org/tdm/tdm11.htm#_Toc68662038
Todd Litman has enough on this topic to keep you puzzling for days.

Posted by: Clark Williams-Derry | May 12, 2005 8:26:02 AM

I agree with Seth. If the price of beer dropped 25%, would people consume significantly more of it? Driving consumption (miles driven) in my area is fairly inelastic. There is little bus service and people have to get to work, and to the grocery store.

Posted by: sf | May 12, 2005 8:40:51 AM


Just did some noodling with numbers. Between 1999 and 2004, per capita gas consumption in the northwest states fell by 7 percent, as close as I can tell. Based on short-term fuel elasticities, that's about what you'd expect from a short-term elasticity of -10% (that is, every 10% increase in gas prices results in a 1% decrease in consumption). Which suggests pretty strongly that the long-term elasticities haven't really taken hold yet. Declining sales of SUVs -- reported pretty commonly in the press over the past few weeks -- may suggest that some of the longer-term effects may be visible soon. But they're not yet -- at least, not in the numbers I've got.

Posted by: Clark Williams-Derry | May 12, 2005 8:54:35 AM

The arguments in this thread, IMHO, illuminate the issue but don't quite get there, but Roy smith is closest:

Many people have a fixed amount of miles that they MUST drive each week.

Total Vehicle Miles Traveled = (fixed + discretionary driving).

Raising fuel prices will not reduce fixed miles driven. It will reduce discretionary driving. It is difficult to carpool because of our separation of live-work space, as Roy implies.

Lowering fuel prices (by increasing mileage) will likely increase discretionary driving, which is a percentage of total miles driven. It will not increase fixed miles driven.

If the fixed miles driven is still greater than discretionary miles driven (after the increase from efficiencies), there likely will still be a net plus.



Posted by: Dano | May 12, 2005 9:04:27 AM

Re: fixed, non-discretionary driving - the big problem lower income people have is that the nature of real estate prices around Seattle forces the burden of driving onto poorer people. I don't drive very much, but I pay very handsomely in terms of my mortgage (for a house in a convenient location) for that privilege.

So, when gas prices go up (for whatever reason, whether supply and demand or taxes) the poor suffer very disproportionately.

Posted by: Roy Smith | May 12, 2005 10:11:45 AM

That is an interesting chain of logic Roy ... that one must have money for an expensive house in a convenient location ... or one suffers "very disproportionately" from fuel taxes.

The most interesting thing is that it doesn't rely on any numbers. We don't need to know what "the poor" spend each month on fuel, or what a X% change would do to their bottom line ...

Posted by: odograph | May 13, 2005 8:07:20 AM

We already know these things, odograph.

Land economics tells us about land rents and proximity to downtown. Neighborhood plans in Seattle have goals to try to set aside affordable housing, even though land rents there preclude the poor from home ownership.

The lengthening Murrican commute is a direct result of land rents - rent seekers go far afield and maximize their utility in land rents, foregoing their time for cheap rent. Look at the Bay Area: the affordable housing development up in the Altamont hills got bid up out of reach of the lower class because of its proximity to the peninsula.

Roy is correct, and thanks for teasing out what I left out of my earlier post.



Posted by: Dano | May 13, 2005 9:49:01 AM

Maybe it's just because I'm a scientist/engineer. I'm perfectly OK with someone positing a relationship ... if they test it, and put numbers on it.

With my background, I find it laughable that someone would be declared "correct" without testing ... but maybe that's just me, and my version of a fact based reality.

Posted by: odograph | May 13, 2005 10:09:13 AM

You're right - but it's been tested, odo. We're just writing in shorthand here. I'm not sure I have the time today to run one down, but there are graphs of land rents showing the fall-off as you go away from downtown (the reason why you have high-rises downtown and few in the suburbs is the land rents). It is self-evident that lower-income people have a harder time living in areas of high rent. Hence, Roy's distance from employment centers, income, and driving.

There are also numerous studies showing how they tested these relationships. Perhaps someone with more time than me today can cite some, say, Glaeser or Carruthers papers that detail these relationships. I know of at least one current PhD planning student who has quantified these relationships in a thesis paper as well (I attended the presentation - a transportation demand forecast using wage, income, demographic data).



Posted by: Dano | May 13, 2005 12:15:25 PM

This phenomenon has been known for 140 years!

Among Peak Oil cognoscenti, this is known as "Jevons' Paradox": when use of a resource becomes more efficient, use of that resource goes up. William Stanley Jevons observed the effect on coal consumption when more efficient steam engines were developed.


I don't think Jevons' Paradox is a good reason to not improve efficiency, but it does demonstrate that efficiency alone is not the answer. It needs to be combined with other, more direct reduction strategies.

Also, Jevons was describing a world with expanding coal supplies, in other words, based on classical economics. His paradox may not hold true in a world where petroleum is in permanent, irrevocable decline.

Posted by: Jan Steinman | May 13, 2005 4:16:41 PM

Jan - I think this is related to Jevons Paradox (or, more generally, to rebound effects, of which Jevons is an extreme example). But it's not quite the same thing.

By way of explanation -- according to the literature that Litman's reviews, if CAFE standards reduce the cost of driving a mile by 10%, then people drive a bit more -- I can't recall offhand exactly how much more, but let's call it 3-4%.

This "rebound effect" reduces, but does not eliminate, the fuel savings of more efficient vehicles. On net, with stricter CAFE standards in place, people would still use less fuel than they would with looser standards. (A true Jevons' paradox example would be if a 10% efficiency gain made people increase their driving by, say, 20% -- overwhelming the fuel efficiency gains, and resulting in more total fuel consumption despite the increase in efficiency.)

So CAFE standards still reduce fuel consumption -- just a little less they would without the rebound effect caused by the decreased cost of driving a mile.

However, driving has *other* external costs besides the fuel & pollution/GHG costs. In fact, Litman argues that the fuel costs are a small share of the external costs of driving, compared with accident risk, congestion costs, noise pollution, the pedestrian barrier effect, road maintenance costs that aren't included in fuel taxes, etc.

So the efficiency gains resulting from CAFE standards are good for fuel savings and reduce pollution some; but the *other* external costs of increased driving overwhelm those fuel and pollution benefits.

So what I hadn't fully considered before is that -- duh -- there's more to driving than burning gas. Much more. So even if a policy substantially reduces fuel consumption, the net effect on social welfare could well be negative if that policy also increases driving, even by a much smaller amount.

Of course, this is probably perfectly obvious to other folks, but it wasn't to me.

Posted by: Clark Williams-Derry | May 13, 2005 8:55:45 PM

By the way, Jan -- re-reading my post above, I realize I sounded sort of didactic. Sorry. Obviously, if you know about Jevons' Paradox you probably know about rebound effects, and a whole lot else. So I apologize if I sound lecture-y -- I just wanted to be clear, and I seem to have a hard time doing that without being long-winded.

Posted by: Clark Williams-Derry | May 13, 2005 10:57:48 PM

I'm going to exit this thread ... but I think the evidence you should look for is the direct kind (monthly fuel expenditures by the "poor") rather than the indirect kind (relations between property values and distance traveled).

After all, you've asked me to take distance as a measure of transportation cost, and then named commute time as second variable. That undermines your own arguement.

To make the home cost thing work, you'd have to argue that housing prices are directly proportional to the fuel-cost of transportation from city center to each home.

I can't think of a city in the world where other factors (desireable neighborhoods, commute time, culture, etc.) do not weigh higher than actual fuel costs.

Posted by: odograph | May 14, 2005 6:58:42 AM

Clark, the didactic method works for me!
Helps me follow along and understand everyone's thoughts on the subject, here.
Ya'll are teaching me a lot!

My own 2-cents on what I've learned over the years:
One does have to "pay" for convenience. Whether that convenience means living closer to downtown so you reduce your commuting time to work; or buying a hybrid vehicle so you don't have to fuel-up as often...
Gets kinda pricey, ay?

Posted by: Michelle Parker | May 14, 2005 5:17:23 PM