Explaining the Evolution of Passenger Vehicle Miles Traveled in the United States.

• Author: Leard, Benjamin
• Position: Report - Statistical table

1. INTRODUCTION

How much people drive their vehicles will play a central role in determining future U.S. oil consumption and pollution. Passenger vehicles account for almost half of U.S. oil consumption, about 15 percent of greenhouse gas emissions, and much of the pollution that reduces local air quality. Mechanically, passenger vehicle emissions depend on vehicle fuels, vehicle emissions rates, and the total vehicle miles traveled (VMT). The United States, like many other countries, sets standards for new vehicles' fuel economy and emissions rates, which determine emissions rates across the entire fleet in the long run. (1) Via the Renewable Fuel Standard program, federal policy also determines the carbon content of vehicle fuel. VMT depends on policies such as fuel taxes, congestion charges, and decisions of individual drivers. In short, the future path of VMT has implications for energy security, greenhouse gas emissions, and local air quality.

Recent developments have attracted media and public attention to how much people drive their vehicles. After decades of growing at about 2 percent per year, VMT suddenly leveled off in the mid-2000s and by some estimates decreased. The popular media offered a range of hypotheses, in-eluding household demographics (such as an aging population, since older households tend to drive less) and economic characteristics (such as declining household incomes, rising unemployment attributable to the recession, and rising or volatile gasoline prices). (2) We refer to such developments as changes in household demographics and economic characteristics. Another set of hypotheses involves changes in how much households drive, conditional on demographics and economic characteristics. For example, the Amazon hypothesis contends that online shopping reduces driving. Another hypothesis is that current younger households (i.e., the millennials, or individuals born after 1980) drive less than younger households in previous generations because of a stronger preference for public transit, virtual connectivity, or other reasons. We refer to these developments as changes in household driving habits, which are defined as the average number of miles driven by groups of households with common demographics and economic characteristics. For example, low-income elderly households typically drive about half as much as high-income elderly households, reflecting a difference in driving habits between these groups.

These possible explanations have differing long-term implications for VMT, oil consumption, and emissions. For example, a persistent change in household driving habits implies that VMT will grow more slowly in the future than it did in the years prior to the 2000s. On the other hand, if the recession was the main factor, expected future economic growth would imply that VMT will rise roughly at historical rates. Whether VMT will be flat or grow at historical rates will have profound effects on U.S. oil consumption and emissions and on the costs of meeting particular energy security or environmental objectives. Comparing these two hypothetical cases, rising VMT would eliminate about half of the savings attributed to U.S. fuel economy standards--that is, an effect much greater than the rebound effect, which has been the focus of the literature (e.g., Gillingham 2014).

This paper explains the slowdown in VMT growth after 2000 and the subsequent recovery and draws implications for future VMT growth and transportation policy. Although a vast body of literature has characterized the effects of income and fuel prices on VMT and gasoline consumption (e.g., Hughes et al. 2008), most of this literature has used aggregated data and assumed linear or log-log relationships among national VMT and average income, fuel prices, and demographics. Such assumptions are necessary given the limited number of observations and the limited data available at the aggregate level. However, as we explain in Section 3, using aggregated data makes it challenging to distinguish changes in driving habits over time from nonlinear relationships among household-level VMT, demographics, and economic characteristics. A few recent studies (e.g., Blumenberg et al. 2012) have focused on possible changes in driving habits among certain demographic groups, such as millennials, but they have not quantified the overall importance of changes in driving habits or the implications of any such changes for future VMT growth. In short, the literature offers little insight on whether changes in demographics and economic characteristics or changes in household driving habits explain the recent slowdown and subsequent recovery in VMT growth.

In this paper, we distinguish between the contributions of demographics and economic characteristics and the contributions of household-level driving habits to changes in national VMT. We begin by estimating the relationships among household VMT, demographics, and economic characteristics in a base year before the VMT slowdown. The large sample of households in our data enables us to estimate nonlinear relationships between VMT and variables such as age or income. Subsequently, we use the Oaxaca-Blinder methodology to decompose changes in total VMT between the base year and any subsequent year into two classes: (a) changes in demographics and economic characteristics; and (b) changes in household driving habits, conditional on demographics and economic characteristics. We estimate habits in the base year as the coefficients from a linear regression of household VMT on explanatory variables. The predictions of VMT in years following the base year derive from changes in explanatory variables after the base year. As we show below, the change in predicted VMT between the base year and a subsequent year reflects changes in demographics and economic characteristics. The difference between the actual and predicted change in VMT in any year following the base year reflects the contribution of changes in household driving habits for particular demographic or economic groups--for example, including the differences in typical miles traveled by millennials compared with other cohorts of young adults.

Our first result is that changes in demographics and economic characteristics, rather than changes in household driving habits, largely explain changes in VMT between 1995 and 2015. The microdata that underlie the decompositions suggest the existence of nonlinear relationships between household VMT and the explanatory variables. Aging of the population made a negative but relatively small contribution to changes in VMT between 1995 and 2015, whereas income and the number of workers per household contributed to the VMT changes. Moreover, the increase in number of workers per household after 2010 explains the increase in VMT per household in the 2010s.

Based on our first result, we predict future VMT assuming that demographics and economic characteristics continue to explain VMT and that driving habits of each household group remain persistent. Future VMT will therefore reflect the offsetting contributions of rising income, which increases VMT, and the aging population, which reduces VMT.

Our second result is that we predict average annual VMT growth of about 0.9 percent between 2015 and 2025, which is lower than historical averages although higher than the growth observed during the 2000s. Our predicted growth rate implies that 2025 oil consumption and greenhouse gas emissions will be about 10 percent higher than if VMT were to remain at 2015 levels. If VMT grows at the predicted rate rather than remaining constant, VMT growth will offset nearly half of the reductions in oil consumption and greenhouse gas emissions in the year 2025 caused by fuel economy standards for passenger vehicles.

The results have implications for policies attempting to reduce the greenhouse gas and local air quality problems that passenger vehicles cause. Our analysis implies that VMT growth will increase the challenges of reducing greenhouse gas emissions, local air pollution, and traffic congestion. The projected VMT growth implies that tighter vehicle greenhouse gas emissions standards will be needed to meet climate objectives than if VMT growth were lower. Federal regulations determine air quality standards for various forms of air pollution, such as carbon monoxide, and many urban areas exceed current standards. Because vehicles are major contributors to these problems and federal policy sets emissions rates for new vehicles, state policy makers can target VMT and the emissions rates of older vehicles. Therefore, the conclusion that VMT will grow at nearly historical rates means that it will be harder to achieve local air quality standards.

2. THE RECENT DECLINE IN VMT GROWTH AND LITERATURE REVIEW

2.1 The Recent Decline in VMT Growth

Between 1975 and 2000, VMT of U.S. passenger vehicles grew steadily and was strongly correlated with income and employment. Figure 1 shows data on VMT per licensed driver, income per capita, and nonfarm employment from EIA (2014), with all variables normalized to equal one in 1975 for comparability. Between 1975 and 2000, VMT per driver grew at an annual rate of about 1.5 percent, and the graph indicates that periods of rising income and employment are accompanied by rising VMT per driver. Likewise, periods of falling income and employment are accompanied by falling VMT per driver. Before 2000, VMT per driver was positively correlated with income and employment, but after 2000, the correlations between VMT per driver and income and employment were approximately zero. (3) Although income continued to rise in the 2000s, VMT per household leveled off and then declined after 2007. Declining employment growth during the 2000s could at least partly explain this decoupling, but after 2000, VMT per household and employment were weakly correlated with one another. This graph suggests that the relationship...