In an (academic) brawl over sprawl, planners are debating whether compact development contributed to a recent decrease in vehicles miles traveled.
By Richard Florida, 6 April 2017
Since 2004, driving as measured by the number of vehicle miles traveled declined, reaching a two-decade low in 2012. Slumping figures for VMT per capita, number of vehicles on the road, and the average age of drivers all seemed to be telling us the same story: We’d passed “peak car” in the United States.
But then things got complicated. As the economy grew and cheap gas got cheaper, VMT per capita started rising anew—it’s gone up for five years straight. So, is there a real long-term decline? And if so, what is—or was—driving it?
Among urbanists, the argument tended to favor the shifting preferences for younger people away from driving and toward more dense living that’s closer to work, plus the availability of more and better alternatives such as transit or biking. Others saw the driving decline as a short-term effect of the economic downturn, fueled by a combination of stagnant incomes and wages and of the widening gap between rich and poor.
Well, buckle up, because the biggest experts in the field are fighting it out in the Journal of the American Planning Association. Just what’s going on behind America’s driving decline is the topic of a recent special issue, published in January. With contributions from some of the leading students of the subject, it suggests there are no easy answers and still considerable controversy around these issues.
A meta-analysis of the issues by Mark Stevens of University of British Columbia suggests that changes in the way we live—a shift from more sprawling to more compact development—explains at best just a small share of the driving decline.
To get at this, Stevens conducts a meta-analysis of 46 different studies from 1996 to the present in light of the “5Ds”—density, diversity, design, destination accessibility, and distance to transit—originally developed by Reid Ewing and Robert Cevero.
Stevens’s analysis seeks to create a common baseline for comparison by looking at the magnitude of the change in driving brought about by the key factors examined. Economists calls this an “elasticity”—a measure of the expected percent change in one variable (in this case the decline in driving) that stems from a change in another. The table below shows the key results.
Of the 5Ds, the most significant factor is distance to downtown. Its elasticity is -0.63. Or as Stevens points out, a household would drive 32 percent fewer miles if its distance to downtown decreased by 50 percent.
Density and job accessibility by vehicle also appear to be factors, albeit in a more limited manner. The elasticity for population density is -0.22, meaning that a 40 percent increase in density would produce a 9 percent decrease in driving. The elasticity for job accessibility by auto is -0.20. This likely stems from the relatively small number of big U.S. metros, such as New York City, that have central urban areas with low levels of car dependence.
The elasticities for the other variables are smaller. Intersection and street density—a measure of walkability—leads only to a modest decline in driving. Mixed land use (0.11) actually causes an increase in driving after adjusting the elasticity estimate for residential self-selection. Meanwhile, job accessibility by transit and jobs-housing balance—two key tenets of compact design—has virtually no effect on VMT.
Ultimately, Stevens concludes that the impact of compact development isn’t much. But other contributors take issue with this finding.
The original authors weigh in
Ewing and Cevero, the authors of the original study that Stevens’ builds upon, counter that compact development plays a far more substantial role in America’s driving decline. They argue that Stevens’ meta-analysis is at odds not just with their own seminal findings but with those of other studies. Some of this comes from studies that should be considered outliers, some of it comes from apples-to-oranges comparisons, and some comes from different weightings and statistical techniques. They conclude their critique by writing: “Many academic planners, in an effort to avoid the appearance of bias, throw up their hands when 90 [percent] of the evidence points in one direction and 10 [percent] in the other. We believe that Stevens has fallen into this trap, and that the practitioners who rely on our research deserve better.”
Instead, Ewing and Cervero argue that the effects of compact development on the driving decline remain substantial. “Increasing density by 25 percent to 2,500 persons or 1,000 homes per square mile—just 200 more residential units in a square mile—will reduce annual VMT by 550 miles for each of the 2,500 residents,” they write.
That difference isn’t trivial, argues Arthur Nelson, a professor of planning and real estate development at the University of Arizona. He writes in his article for JAPA that combining various factors, like population density, street density, and job accessibility, can have additive effects. Nelson uses the case of Tucson, Arizona, to demonstrate the power of these elasticity variables. If Tucson’s population grows from 988,000 in 2011 to its projected 1.5 million by 2050, compact development would dramatically reduce driving, he says. In 2011, Tucson’s VMT was 8,627 miles per capita, for a total of about 8.6 billion miles.
If the city’s development is 20 percent infill, that VMT number would only fall to 8,562 miles; at 100 percent infill development it would reach 4,952 miles. Assuming all variables held over time with population increases, the 20 percent infill plan would mean a total of 12.8 billion vehicle miles traveled—a 51 percent increase from 2011—while 100 percent infill would mean 7.4 billion miles—a 13 percent decrease from 2011.
So is the driving decline really happening?
In “The Driving Downturn,” Michael Manville, David King, and Michael J. Smart question the degree to which America has experienced a significant driving decline at all. They do not question the numbers themselves: Between 2004 and 2012, per capita vehicle miles traveled across America fell 8 or 9 percent—somewhere between 640 and 850 miles, depending on the source.
But they suggest this decline is quite modest. Have a look at the chart below from their analysis. It compares vehicles miles traveled to the change in GDP from 1936 to today. The lines seem to track each other closely, and suggest at best a very modest recent decline.
Moreover, there is very little evidence that Americans are trading in their cars for biking, walking, or using transit. Given that driving trips vastly outnumber transit trips, a small change in VMT would result in much larger transit ridership, one that does not appear on the chart below. For example, they note that American households take about 2,900 automobile trips per year, compared with about 66 transit trips.
For all the talk about the effects of urbanism on the driving decline, the biggest percentage decline has occurred in rural areas, as the chart below shows.
Vehicle miles traveled actually increased in urban areas, from 1.79 trillion miles in 2004 to 1.84 trillion miles in 212, an increase of 2.8 percent. This compares to a nearly 12 percent decline in rural areas, though urban areas did experience a larger decline in per capita terms.
The authors conclude that too much has been made of the driving decline and its causes. They find little evidence of a shift from cars to other modes of transit or that better kinds of urbanism are at play. Instead, they suggest that Americans are driving less simply because income and wage growth has stagnated and driving has become more expensive, especially for less advantaged people.
Urbanists may continue to cling to the narrative that Americans will move downtown, swapping their cars for transit, bikes, or walking. But whether that kind of development reduces driving in the aggregate is foggy at best.
CORRECTION: An earlier version of “U.S. urban and rural vehicle miles traveled, 2004 and 2012” flipped the numbers for urban and rural populations. The table has been updated