The Trump Administration Flunked Its Math Homework

The administration’s clean-cars rollback is riddled with errors. In one case, it forgot to divide by four. In another, it accidentally deleted 700 billion miles of driving.

Jorge Duenes / Reuters

One of the White House’s most consequential environmental rollbacks may be in self-inflicted legal danger.

In trying to freeze gas-mileage rules for new cars and trucks, Trump officials have hit an unusually damaging snag: They seem to have messed up their math homework.

The Trump administration’s official case for repealing car fuel-economy rules is riddled with calculation mistakes, indefensible assumptions, and broken computer models, according to economists, environmental groups, and a major automaker. The errors may seriously endanger the rule, hampering the White House’s ability to prove the proposal’s benefits exceed its costs and raising questions about whether it can survive an almost inevitable court challenge.

The mistakes range in scope from the comical to the bizarre, from the obviously accidental to the how-did-they-miss-that. In one case, federal employees have forgotten to divide a crucial figure by four. In another, officials have assumed that raising the cost of cars will lead more people to buy them, a violation of the principle of supply and demand. In a third case, the proposal asserts that freezing fuel-economy standards for new cars will lead the owners of old cars to drive their vehicles less.

Some of the most glaring errors described in this article were detected by EPA staff before the proposal’s publication. Others were affirmed in recent public comments from Honda, a major automaker. The Atlantic also confirmed some alleged errors with economists who were not involved in their discovery, who described them as “incontrovertible.”

Taken together, the errors artificially lower the rollback’s costs and boost its safety benefits, experts say. Every single error so far identified appears to tilt the analysis in Trump’s favor.

“There’s a systematic bias in all of these to inflate the crash fatalities [under the Obama-era rules]. We have not yet found any mistakes that work in the other direction,” says Kenneth Gillingham, an economics professor at Yale and a former senior economist in the Obama administration.

“Usually, in any rule-making, you can look and complain that there has been cherry-picking. But these errors seem to be more material. They move the dial more than is typical,” says Mark Jacobsen, a professor of economics at the University of California at San Diego.

The mistakes are not the only problem for the Trump proposal. On Friday, Honda and General Motors broke ranks with other automakers and signaled unease about the rollback, with Honda openly opposing it. The car industry had previously stood united in supporting the Trump administration.

But it’s the math errors, and not the loss of corporate support, that could prove most damaging long-term for the rollback. Jonathan Adler, a law professor at Case Western Reserve University and a conservative commentator, told me that simple errors can imperil a regulation in court.

The errors that appear to exist in the Trump proposal are “the kind of thing that an agency would clearly be under a legal obligation to address, and what I think any reviewing judge would raise an eyebrow about,” he said. “The easiest way to make a court pay attention is to raise this sort of question.”

The rollback’s fate will hinge on how the Environmental Protection Agency and the Department of Transportation, the federal agencies overseeing the rollback, respond to the errors and fix them, he said.

But this may prove difficult. In some cases, the mistakes are so large—and so central to the rule’s legal justification—that remedying them may destabilize the entire argument for the proposal. Public documents also make it clear that the Trump administration knew about some of the errors before the rollback was published.

Neither the EPA nor the Department of Transportation responded to a request for comment.

Trump officials have made one overarching pitch for their fuel-economy rollback. They’ve insisted that the new rules are much safer than what’s come before—so much safer, in fact, they named their proposal SAFE. But under scrutiny, many of these safety benefits vanish into air.

Nearly since he took office, President Trump has sought to throw a wrench in the Corporate Average Fuel Economy standard, or CAFE, a set of gas-mileage rules recently strengthened by former President Barack Obama. Under the Obama-era program, new cars, trucks, and SUVs had to get a little more fuel-efficient every year, an ever-ascending upward march of miles-per-gallon. Trump proposed putting an end to that parade. Under his plan, carmakers would no longer be required to improve their fuel efficiency after 2020.

While Trump’s proposal will lower costs for automakers, it makes for a tough sell to the public. Most economists say the rollback will cost Americans at the pump: The Rhodium Group, an energy research firm, estimates it could cost Americans $236 billion in extra gasoline by 2035. Even the plan’s supporters admit that it will warm the planet: The Trump administration says freezing the rules would increase U.S. carbon-dioxide emissions 9 percent by 2035.

So Trump officials instead argued that their rollback will make Americans safer. By lowering costs for automakers, SAFE would make cars cheaper overall, they said. This would encourage consumers to buy new vehicles and take their old, dangerous clunkers off the road. To underscore the point, Trump officials renamed the Obama-era rules: dropping its old title, CAFE, and christening it SAFE, the Safer Affordable Fuel Efficient Vehicle rule.

“This rule promises to save lives,” said Heidi King, the acting chief of the National Highway Traffic Safety Administration, when SAFE was announced in August. “It could save up to a thousand lives annually by reducing these barriers that prevent consumers from getting into newer, safer cars.”

All those saved lives provide important legal justification for the rollback. SAFE trumpets them in its preamble: “Today’s proposed rule is anticipated to prevent more than 12,700 on-road fatalities and significantly more injuries ... as more newer, safer vehicles are purchased,” it says. “A large portion of these safety benefits will come from improved fleet turnover as more consumers will be able to afford newer and safer vehicles.”

It’s not clear that these claims are true. The government is not actually sure how many lives SAFE will save. The fine print of the proposal says that “newer, safer cars” will prevent 30 deaths every year at most—a far cry from the claimed 1,000. But an EPA memo included in the proposal warns that SAFE will increase traffic fatalities, leading to 17 more deaths every year.

In any case, “a large portion” of the lives that safe claims to save do not come from “newer and safer vehicles,” as the proposal says. They arise instead from an erroneous calculation spat out by a broken computer model; and an assumption—never advertised by the Trump administration—that Americans will drive less when forced to buy less fuel-efficient cars.

How did the Trump administration botch its safety facts? Many of the most glaring errors track back to one piece of software, an economic model written to satisfy a car-industry talking point.

For years, automakers have insisted that fuel-economy rules make new cars too expensive, dragging down sales and crowding the roads with old vehicles. Under Trump, the Department of Transportation sought to turn this contention into official regulatory analysis. It created what economists call a “scrappage model”: a mini-simulation of the U.S. economy that calculates how many people would scrap their old cars if new cars were cheaper. It then used this model to simulate the American car market under both the Obama-issued rules and the Trump-era rollback.

On its face, there’s nothing inherently wrong with this technique. It springs from solid economic insight, researchers told me. But the scrappage model created by the Trump administration is often sloppy and haphazard, and its bad math overwhelms the rest of the rule’s reasoning.

Its most ruinous problem is the case of the “phantom miles.” At a crucial moment in the scrappage model’s analysis, it mysteriously deletes roughly 700 billion miles of nationwide driving from its simulation of the Trump rollback. It does so due to a nonsensical assumption that owners of old cars—cars built between 1977 and today—will drive much, much less under the rollback than they would under the Obama-era rules.

This makes little sense. The Trump rollback only affects new cars sold after 2020. Yet the scrappage model projects that the owner of an old car—say, a pickup truck built in 2016—will restrict her driving in 2022 because new cars, none of which she drives, get worse fuel economy. The model then extrapolates this effect across the United States.

Remarkably, this bizarre situation emerges from a deeper problem: The Trump administration’s model is so poorly constructed that it violates basic economic principles.

The most rudimentary macroeconomic model would assert that as car prices increase, fewer people will choose to own cars, opting instead for other modes of transportation. The Trump administration argues that the Obama rules made cars more expensive. So it stands to reason that—when simulating the world of the Obama-era rules—it will show fewer people choosing to buy cars, and fewer cars on the road at all.

“But the model works the opposite way,” Gillingham, the Yale economist, told me. The Trump scrappage model says that the Obama rules increase the price of new and used cars—and then it bizarrely asserts that more people will decide to buy those cars anyway.

The model, in other words, seems to believe that as the price of a good increases, more people will purchase it. “In economics, that is an incredibly rare case that basically doesn’t happen,” Gillingham told me. “You can have it with status goods like wine, for example, but for cars? It’s a crazy assumption.”

“That [finding], in some sense, is impossible from basic economic theory,” Jacobsen, the economics professor at UC San Diego, told me. “It is incompatible with any notion of how people choose their mode of transportation.” Jacobsen’s condemnation is especially damning: His economic research first suggested the existence of a scrappage effect, and it is cited in SAFE five different times.

These assumptions all add up to one big claim: Americans will drive more cars for more miles under Obama’s rules than under Trump’s rollback. Since there are generally more car accidents when more drivers are on the road, the model can then conclude that Obama’s rules are more dangerous than Trump’s rollback. The “phantom miles” from these cars account for roughly half of the saved lives that Trump officials claim their rollback will bring.

Economists have been unsparing in their criticism of the scrappage model. Eleven of them, including Gillingham and Jacobsen, have prepared a major peer-reviewed paper condemning its findings. Indeed, I could scarcely find an economist or expert aware of the problem who had not already publicly criticized it. Honda has also lambasted it at length. On Friday, Honda told the Trump administration that the scrappage model is so “flawed that it should not be considered in the cost-benefit analysis at this time.” It called use of the model “premature and ill-advised.”

“One has very little confidence that this scrappage model ... can distinguish significant data from background noise,” it added.

Worse still for the Trump administration, federal employees appear to have identified many of these problems before the rollback was first published. In an EPA memo included in the proposal, William Charmley, a senior official in the agency’s Office of Transportation and Air Quality, says that the agency has significant concerns with many of the scrappage model’s conclusions. The scrappage model “is not consistent with the basic principle of supply and demand,” he says. And he frets that the model will “erroneously inflate” national miles driven overall, “increasing the estimated fatalities due to the [existing] standard by many hundreds of lives.”

It was a spot-on description of the scrappage model’s problems. Charmley submitted those concerns in June 2018. The Trump administration seems to have done little about them.

What about the other thousands of saved lives? They also arise from mathematic sleight of hand.

Economists have long recognized that when a technology becomes more energy efficient, people will probably use it more: a phenomenon called the “rebound effect.” Historically, the EPA and the National Highway Traffic Safety Administration have assumed a rebound effect of 10 percent: If cars become 100 percent more fuel efficient, most people will drive them about 10 percent more, so actual gasoline use will only fall by 45 percent—rather than the 50-percent savings that would otherwise have been expected.

But under the Trump administration, NHTSA has doubled this effect. It assumes that if fuel economy increases by 100 percent, Americans will drive 20 percent more, so actual gasoline use will only fall by about 40 percent. This allows it to assume people will drive their cars even more under the Obama standards than under the Trump rollback, which in turn inflates the number of overall fatalities under the Obama rules.

This rebound effect is responsible for the other roughly half of the avoided fatalities. And while changing the rebound rate isn’t quite a math error—more like a dubious statistical contention—it flies in the face of economic consensus. Just this summer, the nonpartisan economic research firm the Analysis Group reported that U.S. fuel-economy obeys a rebound rate of about 10 percent.

These are not the only math errors or unsupported assumptions in the Trump administration proposal. At one point in the proposal, an economic model of new car sales attempts to calculate how increasing the cost of cars by $1,000 changed vehicle sales overall. It is supposed to show that boosting the cost of cars by just a few thousand dollars—as the Obama rules are alleged to do—imposes high costs on automakers at scale.

But here it commits another math error. When calculating year-by-year vehicle sales, the proposal’s authors appear to have confused an annualized rate of change with a quarterly rate. “It’s a forgetting-to-divide-by-four issue,” said Gillingham, who worked with James Stock, a professor of economics at Harvard, to identify the mistake.

Using these flawed rates, the Trump administration projected that increasing the price of a new car by $1,000 led to 600,000 fewer sales over the next 10 years. But once the problem was fixed, Gillingham and Stock found that raising the cost of a new car actually decreased sales by only 120,000 vehicles—many fewer than once projected. All but 5,000 of those lost sales are concentrated in the first year. The mechanism turned out to be yet another way that federal employees overestimated the economic benefits of the Trump rules.

“I think that’s just a plain old-fashioned mistake, done by people who are rushing, probably,” says James Sallee, a professor of economics at the University of California at Berkeley. But he told me that evidence of the error in the data was “incontrovertible.” (Sallee has worked with Gillingham on researching SAFE but was not involved in finding this error.)

These several mistakes and unsupported assumptions—in the scrappage model, the rebound effect, and new car sales—are not the only potential problems with the proposal’s analysis. Environmental groups have criticized the rollback’s projection of technology costs. Honda takes issue with its blunt projections of mass deduction and how it calculates a social cost of carbon.

But it’s the mistakes that could very well prove the most damaging to its survival. Most of the rule’s economic analysis turns on the scrappage model and the sales model. If both models need to be heavily revised, then the rule’s costs may very well exceed its benefits. “The assumptions about new car sales are core to everything,” Sallee told me. He called them “the most fundamental part” of the analysis.

Jacobsen agreed. He has already warned the Trump administration that fixing the scrappage model could send the rollback’s costs skyrocketing. He told me that once both errors were revised, he was “skeptical [the administration] could pull out a top-line number that would stay positive,” that is, that would keep the rule’s benefits above its costs.

Government moves slowly. The Trump administration will now have several months to revise its math and address the problems in its proposal. After that, if it can sufficiently finagle the math to work, it will announce a final rule. Then a court case will certainly follow, and judges may very well approach the rollback more skeptically because of the scale of the errors that came at the beginning.