Skip to content

Taking The Pedal Off The Metal

A new paper shows that people drive less and slow down when gasoline prices rise.

I drive an electric vehicle. So I am, of course, better than you and now get to say all sorts of snarky things about you gas guzzling large vehicle enthusiasts. So here we go. We energy economists spend most of our time complaining about the absence of a carbon tax and the rest of our time quantifying aspects of human behavior. Some of the most interesting questions are in the transportation sector, which is the main source of greenhouse gas emissions in California. More driving plus heavier, less fuel efficient and more quickly driven cars lead to more gas consumption, local and global pollution. More driving and bigger cars also lead to more congestion.


In the US we have tried to put downward pressure on gasoline consumption by regulating the fuel economy of the fleet. More efficient vehicles are of course more efficient, but also decrease the cost per mile driven, which provides an incentive to drive more. This is called the rebound effect. If you face a lower price of gasoline (either implicitly through more fuel efficiency or explicitly through lower prices at the pump), how does your behavior change?  Do you drive more? If so, how much? Do you drive faster, which burns more fuel per mile driven?

While most of us were passed out in a tryptophan or soy induced daze on our couch, Energy Institute alum (and now fancy MIT Sloan Professor) Chris Knittel and his coauthor Shinsuke Tanaka released a really interesting and pleasantly short new working paper providing some interesting new answers. They start with a jealousy-inducing new dataset from Japan, which via a mobile phone app collected detailed micro-level information on gasoline consumption, vehicle distance traveled, and gasoline prices paid for each fill up for over 90,000 drivers for 10 years!!!! This is next-level stuff. For comparison, most of the papers so far use monthly consumption data aggregated to the US state level, which does not allow you to study the finer aspects of consumer behavior. Specifically, they shine some light on three aspects of behavior.

  • They can tell us how each driver’s gasoline consumption responds to changes in gas price at the vehicle level. They know how much each driver and vehicle buys at the pump, the odometer reading at each visit, and they know the price!
  • They can decompose these changes in consumption into how many miles the car is driven versus changes in driving behavior. That is, do drivers turn on their inner Jeff Gordon or Danica Patrick when gas is cheap or ignore fuel saving maintenance actions since the actions now save less money?
  • Finally, they can take a look at whether consumers “learn” from past price changes, which helps them form habits.

The results are the best kind: interesting and policy relevant. First off, they find that drivers are much more price responsive than the majority of the existing literature finds. This is relevant, as government agencies use these numbers to project energy consumption at the national level. Is the difference small? No! An order of magnitude (that is 10x for you non-nerds) smaller. They estimate that a 10% increase in gas prices leads to a 3.7% decrease in gas consumption, when most of the literature suggests a 0.3% decrease. Wow.

Second, they decompose this response and show a significantly larger response of Vehicle Miles Travelled (VMT) to changes in gas prices than the remainder of the literature. Of the 3.7% price response, 3% can be attributed to decreases in VMT. The remaining 0.7% are attributed to changes in “driving behavior”. This could either be driving your car more slowly and getting better fuel economy, or increasing fuel saving maintenance such as properly inflating your tires. Why is this so cool? Nobody has been able to estimate this parameter before. This is nerd speak for “we have actually measured a hypothesized type of behavioral response in the wild”! When astronomers discover a black hole, the world gets excited! This is our version of actually seeing a black hole. Bam.

Third, they examine whether changes in prices a long time ago versus more recently have differential behavioral effects. Do we learn over time and form habits as we drive through life? The answer is no. There is no evidence of this. Drivers seem to react to the most recent fluctuations in prices.

Finally, and this is really cool, they show that humans are strange creatures. When we teach intermediate economics, we generally argue that price responses are symmetric. This means if price goes up by one percent and you consume 3 fewer marshmallows, we would assume that if price dropped by one percent, you would eat 3 more marshmallows. This paper shows that the consumers in the sample do not exhibit this symmetry. The price response is twice as high when prices go up compared to when they go down. I guess we are a “tank is half empty” type of society.

There are a bunch of caveats. This paper deals with very short run fluctuations in gas prices and hence characterizes very short responses of drivers. Also, this is Japanese drivers over the past decade who self-selected into signing up for the app. The road network, public transportation infrastructure and vehicle and gas pricing are very different from what they are in the US. But beyond the caveats, what the authors suggest is that gas taxes alone may have a transitory effect on driving behavior and may need to be supplemented by fuel efficiency standards and other policies such as feebate policies (which I am a HUGE fan of).  However, I am somewhat skeptical of that interpretation as gas taxes are permanent price changes, which consumers react to differently than to short run price fluctuations.

Overall, this is a major step forward in what we know about consumers’ response to short run gas price fluctuations – in Japan! I am hoping that someone will replicate this study in the US, Europe and possibly a lower income country to see whether these responses are similar.

Keep up with Energy Institute blogs, research, and events on Twitter @energyathaas.

Suggested citation: Auffhammer, Maximilian. “Taking The Pedal Off The Metal” Energy Institute Blog, UC Berkeley, December 2, 2019,

Maximilian Auffhammer View All

Maximilian Auffhammer is the George Pardee Professor of International Sustainable Development at the University of California Berkeley. His fields of expertise are environmental and energy economics, with a specific focus on the impacts and regulation of climate change and air pollution.

11 thoughts on “Taking The Pedal Off The Metal Leave a comment

  1. When I drive my HEV [even when the battery is fully discharged] i average about 33mpg [in city]. When my wife does the average is about 25mpg. I will coast to a stop and accelerate gently, she goes full-speed to the red light, and goes off at escape velocity when the light turns green. So it is not just the price of gas, it is also personality?!

  2. Max,
    The more things change, the more they stay the same. We used to say that the price elasticity of the demand for water and energy was between -0.3 and -0.4. Then the identification police said that that’s too “high”. Now we’re back where we started. Sometimes two wrongs do make a right.
    P.S. You needn’t have bothered to buy an electric car to prove you are better than us. We already knew.

  3. Thank you for bringing up Japan. Essentially all energy used in Japan must be imported. I worked in Japan in intervals between 1986-1990 for a Japanese joint venture with a U.S. firm. I was struck by the difference in gasoline prices relative to the U.S. It cost three to four times as much to purchase gasoline in Japan relative to Ohio. Personal vehicles tended to be significantly smaller than those sold in the U.S. Large, less fuel-efficient vehicles such as trucks had green “enforcement lights” atop the cab. The first came on at 40 km/hr, the second at 60 km/hr, and the third at 100 km/hr. The enforcement lights made it easy to determine if a large truck was speeding. Thus, high gasoline prices lead to more efficient use of gasoline. . . . . . . .
    Now let’s consider the case in 2019 California in-state electricity generation. Natural gas is the main energy source. I believe inflation-corrected prices for natural gas are at historic lows. Per the World Bank, the U.S. wastefully flared an average of 1 billion cubic feet (BCF) per day of natural gas between 2013-2017. To put that figure in perspective, on a really cold day in California, natural gas usage is about 8 BCF, considerably higher than average. Those low natural gas prices are also hiding inefficient and intermittent dispatch patterns for California natural-gas-fired generators. The surprising reason is the large amount of intermittent solar (12,000 MW) and intermittent wind (6,000 MW) now deployed in California. The capacity factor (percent on time) of solar or wind for the half -year ending on January 31, 2017 was close to 20%. Thus, the natural-gas-fired generation that “firms” these intermittent power generation means must run about 80% of the time. This pair of articles explain why fast-acting natural-gas-fired generation is preferred for firming. “Turns out wind and solar have a secret friend: Natural gas,” by Chris Mooney, August 11, 2016, The Washington Post, serves as an introduction to: the problem. A more technical paper “Bridging the gap: Do fast-reacting fossil technologies facilitate renewable energy diffusion?” by Elena Verdolini, Francesco Vonab, and David Popp, Energy Policy 116 (2018) 242–256, . . . . . .
    Using a dragster as a metaphor, California in-state natural gas generators are now typically operating in “pedal to the metal” mode. Using the “Renewables Watch” net generation data published by the California Independent System Operator (CAISO) on the “high ramp day” June 11, 2019, the mostly fossil-fired “ramp” began at 9:30 AM and ended at 7:30 PM. The ramp height was almost 20,000 MW. That means fossil-fired generation equal to 1 Hoover Dam (2,078 MW) was turned up the first hour, 2 Hoover Dams the second hour, until 10 Hoover Dams of fossil-fired generation was turned up for the last hour. These “high ramp” days have become much more frequent. CAISO even boasts about the large ramp rates. . . . . .
    Since when is wasteful dispatch of generation a virtue, particularly in California? Eventually, the U.S. will be able to export via tanker large quantities of liquefied natural gas (LNG) to Asia and Europe. California natural gas prices will then rise to more typical values, making California electricity prices absurdly high.

    • The AMOUNT of natural gas used for dispatching is relatively small, and much more than offset by the renewable generation that occurs in most of the hours of the day. The loss of efficiency is relatively small. Not sure how your point is relevant given the dramatic displacement of gas generation over the last decade in California.

  4. Super neat paper to reinforce the idea that we ought to have higher gas taxes. The big longer-run behavioral change, which I suspect is larger than VMT or micro-adjustments in driving, is that people drive different cars. That was a pretty clear response to the gas price spikes around 1980 and 2008, as well as the subsequent declines in gas prices. These spikes also seemed to influence policy, like CAFE standards, and probably innovation. Sadly, events seem to drive policy more than economists do.

    If possible, they ought to use their data to see how gas price changes affect decisions to replace cars.

  5. Max,This is fascinating and important.
    In 975 Congress tried to set fuel economy goals for automobiles. A special interest group lobbied to have an exclusion for small vehicles that small businesses and farmers used. In 1980, when Chrysler was nearly bankrupt, they use thsi loophole to build minivans Before you could say miles per gallon all the automobile manufactures were making SUVs.
    I am finding it increasingly difficult to park my little Toyota because all the C permit parking spaces, clearly labelled “COMPACT”, are filled with giant SUVs. As you suggest, we are a reasonable well informed group of car owners, fully aware of global arming. I suggest that the campus increases the cost of parking permits fro SUVs by 40% over compact vehicle, unless the driver is handicapped or has three or more children. The campus needs to money and the planet needs fewer SUVs
    Malcolm Potts

  6. “When we teach intermediate economics, we generally argue that price responses are symmetric.”

    Didn’t Duesenberry show that was wrong in 1949? (Harvard University Press)

    Houthaker and Taylor did as well, in 1966. (Harvard University Press)

    Consumption is habit forming.

  7. “However, I am somewhat skeptical of that interpretation as gas taxes are permanent price changes, which consumers react to differently than to short run price fluctuations.”

    Not sure how to interpret this statement. In the face of a long-term price change, consumers can change their vehicle. Knittel had an earlier paper that showed the CAFE standards had no influence on fleet fuel economy until gas prices fell in 1984 because consumers were buying more fuel efficient cars due to higher gas prices up to the point.

  8. This is indeed interesting and useful data. But it’s missing at least two important dimensions.

    First, does a 10% increase in gasoline price actually induce a 3% decrease in VMT in households with more than one vehicle, as was observed for the Japanese drivers, or does it result in vehicle substitution which may reduce gasoline consumption in the short run while holding VMT constant? Second, does a 10% increase in gasoline price influence vehicle purchasing, creating a long-run shift in the vehicle fleet.

    We did observe some substitution of vehicles — in the short run — in response to price spikes during the second oil embargo in 1978-80. I was working for the Washington State Senate Transportation Committee at the time. However, we also observed a surge in carpool and transit usage.

    We had photographs of employee parking lots, taken as part of a planned repaving and repainting project, from just before the embargo, in March, 1979. We went over those, and counted over 500 state employee vehicles, and categorized them into “efficient”, “average”, and “gas-guzzler” categories. This was really small, medium, and large; we had an interesting time categorizing an AC Cobra, a small car with a Ford 427 engine.. We then surveyed the exact same parking lots a year later, in April of 1980, when gasoline prices were more than twice as high. The vehicle mix had shifted to about 10% fewer gas-guzzlers and 10% more efficient vehicles. People did not buy new vehicles — they just took the more efficient one on the longer drive. When the repaving and repainting occurred, after the effect of the embargo, we striped in “compact” spaces, and actually added about 10% to the parking space count.

    When gas prices declined, I less formally observed people stuffing big cars into compact spaces, because it appeared that the fleet mix regressed towards the previous mean. I didn’t do a formal count in 1982; I just had a bit of time to kill because the Carryall and F-100 parked next to my VW in a “compact” space made it impossible for me to get into my car until one of the two vehicles that had sandwiched me in returned. So I walked around, and found that about one in four “compact” spaces had a decidedly non-compact vehicle in them. We’d actually striped the lot based on actual data — in 1980. At that time, we thought gasoline prices would stay high and go higher. But, of course, early-arriving compacts often took the big spaces, to avoid the problem I’d encountered, and I did not count those, so there’s no science whatsoever to my post-embargo survey.

    Karen and I do that today. We have a Kia Niro plug-in (50 mpg gasoline; 3.3 m/kWh on electric) and a Honda Odyssey (about 20 mpg). We routinely decide who is driving which car based on our distance, unless one of us has a need for the capacity of the Odyssey. We are probably motivated as much by Greta-induced guilt as by pragmatic cost savings. Factor that into your felicific calculus.

    The point is that a 10% increase in gas price to a driver with a choice of taking the F-150 or the Prius may produce substitution of the more efficient vehicle for some of the miles driven. The VMT may be unchanged, while the gasoline consumed drops by 3.7%. This is good from a balance of payments and emission perspective, but it does little for congestion, little for the amount of land devoted to parking, and little for the psychological stress that people exhibit from driving in congestion.

    As for the second issue, the long-run effect on the vehicle fleet, we have observed large car and light truck sales decline, relative to efficient vehicles, during gasoline price spikes. And the reverse — the collapse of efficient vehicle sales when gasoline prices decline. This is one reason I continue to advocate that electricity prices be regulated based on long-run marginal costs, so that consumers make housing and appliance purchases, and businesses make building and equipment investment with an eye to minimizing their long-run cost and societal long-run cost.

    We must plan for the long run, for if we do not, in the long run we will all be dead. Including our kids and grand-kids.

%d bloggers like this: