Blog Post

Missing Fuel Cost Savings: Some Clues Emerge

Oct 9, 2018 | Joshua Linn

In a previous post, we tried to figure out why the National Highway Traffic Safety Administration (NHTSA) and Environmental Protection Agency (EPA) have reduced the estimated fuel cost savings from tighter fuel economy standards compared to their analysis from two years ago. By process of elimination, we concluded that the agencies’ assumptions on vehicle travel are a likely explanation. In this blog post, I take a closer look at those assumptions and conclude that they appear to play a big role in the agencies’ arguments for freezing the standards.

As we discussed a few weeks ago, the agencies assume that the amount each vehicle is driven over-depends on its own fuel costs. The agencies also assume that weaker standards increase scrappage and reduce the total size of the on-road fleet. Taken together, these assumptions imply that weakening the standards reduces total vehicle miles traveled (VMT) of the on-road fleet. To the extent that the two assumptions do not hold, the agencies would underestimate VMT and fuel consumption if they freeze the standards—and hence, overstate the societal net benefits (benefits minus costs) of freezing the standards. The question is, does the economics literature support these assumptions?

To answer that question, we’ll look at the rebound literature, where rebound is defined as the change in VMT caused by a change in driving costs. Much of the recent research on rebound asks how a given vehicle’s fuel costs affect its VMT, holding fixed the other vehicles owned by the household. In this blog post, I’ll consider a different notion of rebound, which I refer to as “aggregate rebound.” Aggregate rebound measures the change in total amount of driving across all vehicles on the road in response to a change in the average fuel costs of those vehicles. It is different from the first definition in that it includes changes in the vehicle stock, due to purchases and scrappage, caused by changes in fuel costs. This is the notion of rebound that Ken Small and various coauthors have studied in papers like this one. One of the results they report, which the agencies highlight, is an aggregate rebound estimate of 18 percent during the period 2000–2009, meaning that a 10 percent reduction in average fuel costs raises aggregate VMT by 1.8 percent. The agencies use that study to support their assumption of a 20 percent (non-aggregate) rebound effect.

We can compare the estimate of aggregate rebound from the literature with the aggregate rebound implied by the agencies’ two assumptions. To do this, I compare two scenarios that the agencies analyze: the current (Obama) standards and the proposed (Trump) standards. The horizontal axis shows the log of the ratio of average per-mile fuel costs for the Trump vs. the Obama standards. The log ratio is roughly equal to the fractional difference between the two scenarios. For example, a value of 0.1 implies that fuel costs are 10 percent higher with the Trump than the Obama standards. The vertical axis shows the log ratio of VMT for the Trump vs. Obama standards. Because the Trump standards are less stringent, fuel costs are higher and VMT is lower with the Trump standards. In this figure, I use the total VMT across the entire on-road fleet and the average cost of driving one mile across the entire on-road fleet (each x on the curve represents vehicles sold in a single model year).

The dashed line shows the VMT reduction if one assumes a 20 percent aggregate rebound effect. For example, if fuel costs are 10 percent higher with the Trump standards (that is, the log ratio is 0.1), VMT would be 2 percent lower (that is, a log ratio of 0.02). The solid line shows NHTSA’s modeling estimates for vehicles sold in each model year, and including model years 2016 through 2029. Importantly, the solid line lies below the dashed line, and the difference between the two is the effect of the two assumptions noted above (fixed VMT and scrappage). The blue line indicates that the aggregate rebound effect implied by their analysis is larger than 20 percent.

But wait, if the agencies assume a 20 percent rebound effect for all new vehicles, why does the solid line lie below the dashed line? The difference must come from the fact that the on-road fleet is smaller in the scenario with the Trump standards.

Because their model yields more rebound than the research supports, the agencies underestimate fuel costs and fatalities and overestimate net benefits of freezing the standards. By making some assumptions, I can provide a rough sense of how important this is. I begin with the agencies’ estimates of fleet-wide VMT and per-mile fuel costs for each calendar year in the Trump scenario. Then, I adjust fleet-wide VMT up until the aggregate rebound effect is 20 percent. Visually, this is like moving the blue curve up to the black line. Assuming that the VMT increase affects all vehicles in the fleet proportionately, increasing the VMT doesn’t affect average per-mile fuel costs. Consequently, total fuel costs scale up with VMT. This calculation yields a present discounted value of additional fuel costs equal to about $23.4 billion, as the first row of the table shows.

In NHTSA’s model, there is a close connection between traffic accidents and total VMT. According to the model, on average there are about 8 fatalities per billion miles of travel, although that number diminishes over time. If we increase the VMT to achieve a 20 percent aggregate rebound, this would increase the number of fatalities as shown in the second row of the table. In fact, there would be almost 3,000 additional fatalities. To put that number in context, the Trump administration claims that freezing the standards would eliminate almost 9,000 fatalities. In other words, almost a third of those fatalities can be attributed to their unsupported assumptions on vehicle travel. As the table shows, using NHTSA’s assumptions on the societal costs of fatal crashes, increasing VMT to achieve a 20 percent rebound effect raises fatal crash costs by about $23 billion.

Also, non-fatal crashes are closely related to total VMT in the NHTSA model. In their model, each mile of travel yields about 7–10 cents of costs from non-fatal crashes. As the table shows, adjusting VMT to achieve a 20 percent rebound raises non-fatal crash costs by about $36 billion.

The bottom row of the table shows the combined effect of increasing VMT on the net benefits of the Trump administration’s proposal to freeze the standards. The agencies claim that freezing the standards would create $176.3 billion of net benefits to society. Because fuel costs and accident costs would be higher than the agencies assume, these effects would reduce the net benefits of freezing the standards. A 20 percent aggregate rebound effect would eliminate about $82 billion of net benefits, or almost half of the net benefits that the agencies report.

The literature on aggregate rebound, including the paper mentioned above, suggests that in the coming decades, due to income growth and other factors, the rebound effect may be substantially lower than 20 percent. To allow for this possibility, I also calculate the changes in fuel costs and fatality costs under a 10 or 15 percent rebound. The bottom row shows that the net benefits of freezing standards disappear almost entirely assuming a 10 percent rebound.

Note that these are rough calculations, and they rely on the simplifying assumption that VMT increases proportionately for all vehicles. We need research on household travel choices to improve the analysis. Nonetheless, these calculations show that the unrealistic assumptions the agencies make on travel behavior play an important role in their conclusion that the benefits of freezing standards far exceed the costs.

The views expressed in RFF blog posts are those of the authors and should not be attributed to Resources for the Future.