View source for Do We Need Speed Limits on Freeways?

<!-- you can have any number of categories here -->
[[Category:Arthur van Benthem]]
[[Category:Speed Limits]]
<!-- 1 URL must be followed by >= 0 Other URL and Old URL and 1 End URL.-->
{{URL | url = http://www.stanford.edu/~arthurvb/JMP-van_Benthem.pdf}}
<!-- {{Other URL | url = }} -->
<!-- {{Old URL | url = }} -->
{{End URL}}
{{DES | des = [[Arthur van Benthem]] shows that private judgements of the cost of driving faster do not include extensive social costs.  Raising speed limits in the US would be socially costly if you value a human life at more than $0.9 million (well below current norms of about $6 million.)}}
{{Quotes}}

{{Text |
(Only the Abstract and Introduction have been copied here.)

Abstract
When choosing his speed, a driver faces a trade-off between private benefits (time savings) and private costs (fuel cost and own damage and injury). Driving faster also has external costs (pollution, adverse health impacts and injury to other drivers). This paper uses large-scale speed limit increases in the western United States in 1987 and 1996 to address three related questions. First, do the social benefits of raising speed limits exceed the social (private plus external) costs? Second, do the private benefits of driving faster as a result of higher speed limits exceed the private costs? Third, could completely eliminating speed limits improve efficiency? I find that a 10 mph speed limit increase on highways leads to a 3-4 mph increase in travel speed, 9-15% more accidents, 34-60% more fatal accidents, and elevated pollutant concentrations of 14-25% (carbon monoxide), 9-16% (nitrogen oxides), 1-11% (ozone) and 9% higher fetal death rates around the affected freeways. I use these estimates to calculate private and external benefits and costs, and find that the social costs of speed limit increases are three to ten times larger than the social benefits. In contrast, many individual drivers would enjoy a net private benefit from driving faster. Privately, a value of a statistical life (VSL) of $6.0 million or less justifies driving faster, but the social planner’s VSL would have to be below $0.9 million to justify higher speed limits. The substantial difference between private and social optimal speed choices provides a strong rationale for having speed limits. Although speed limits are blunt instruments that differ from an ideal Pigovian tax on speed, it is highly unlikely that any hidden administrative costs or unforeseen behavioral adjustments could make eliminating speed limits an efficiency-improving proposition.
∗This version: November 5, 2011. For the latest version, see www.stanford.edu/∼arthurvb/. I thank my advisors Ran Abramitzky, Larry Goulder, Matt Harding, Caroline Hoxby and Mark Jacobsen for their guidance and support. I am also grateful to Dan Bjo ̈rkegren, George Bulman, Kenneth Gillingham, Rick Hornbeck, Koichiro Ito, Theresa Kuchler, Alejandro Molnar, Nick Sanders, Paulo Somaini, Johannes Stro ̈bel, Reed Walker, and participants in various seminars at Stanford University for many helpful comments and suggestions. In addition, many individuals at various state agencies have been very generous with their time. This research was supported by the Leonard W. Ely and Shirley R. Ely Graduate Student Fund Fellowship through a grant to the Stanford Institute for Economic Policy Research, and by a Shultz Graduate Student Fellowship for data acquisition from the same organization.
†Department of Economics, Stanford University, 579 Serra Mall, Stanford, CA 94305, arthurvb@stanford.edu. 1
1 Introduction
An interesting and actively debated policy question is: should we leave speed choices up to individ- ual drivers? Drivers in many countries are so accustomed to highway speed limits that they take them for granted, but it is nevertheless not obvious that having such limits is socially desirable. Germany does not have speed limits for many of its freeways, and several U.S. states did not have them before 1974 or in the late 1990s. When choosing his speed, a driver faces a trade-off between private benefits (time savings) and private costs (increased fuel use, risk of personal injury, death or damage). It is thus an empirical question if driving faster than the current speed limit is rational. Besides private costs, there are external costs to driving faster: increased pollution, adverse health impacts and damage or injury to other drivers. To justify having speed limits, it is not sufficient to observe that driving faster has external costs as well as private costs. A speed limit is a crude rule that differs from an ideal Pigovian tax on speed. Uniform speed limits for all traffic and weather conditions take away discretion from drivers, and may establish a bad focal point during adverse driving conditions. Therefore, external costs must drive a substantial wedge between private and social optimal speed choices for speed limits to increase efficiency. This issue is particularly relevant given the recent debate about speed limits. Early in 2011, Spain reduced the freeway speed limit from 120 to 110 kilometers per hour (kph) to achieve gasoline reductions, while the Netherlands raised it from 120 to 130 kph to reduce travel time.1	In the United States, travel time reduction inspired Kentucky (2007), Utah (2009), Ohio (2011) and Texas (2011) to increase their posted maximum speed, with similar proposals underway in Illinois and South Carolina.
This paper aims to answer three related questions. First, should we raise speed limits? A social planner would do so only if the social benefits of speed limit increases exceed the social (private plus external) costs. Second, are speed limits binding? That is, would individuals enjoy private net benefits from driving faster if speed limits were raised? Third, could completely eliminating (as opposed to raising) speed limits be an efficiency-improving proposition? In order to answer these questions, I estimate the effect of speed limit increases on a wide range of outcome variables: travel time, accidents, air pollution and health. I use these estimates to calculate the private and external benefits and costs summarized in Figure 1.
I use a unique setting and rich data to address these questions. The 1987 amendment and 1995 repeal of the National Maximum Speed Law in the United States provide quasi-experimental variation in speed limits. Between 1974 and 1987, this law prescribed a maximum speed limit of 55 mph across the entire United States. In 1987, states were allowed to raise the speed limits to 65 mph on rural interstates, but not on other similar urban or rural highways. In 1996, speed limit
1Spain’s deputy prime minister Alfredo P ́erez Rubalcaba expressed it as follows: “We are going to go a bit slower and in exchange for that we are going to consume less gasoline and therefore pay less money.” (http://www.guardian. co.uk/environment/2011/feb/25/spain-speed-limit-oil-prices). Dutch transport minister Melanie Schultz van Haegen defended her decision by claiming that “a higher speed limit leads to a travel time reduction of up to eight percent.” (http://www.rijksoverheid.nl/ministeries/ienm/nieuws/2011/02/28/130-km-u-van-start-op-afsluitdijk.html). Other governments proposed to decrease speed limits to reduce traffic accidents (United Kingdom, 2009) or pollution and associated adverse health effects (Texas, 1992; green parties in Europe).
2
Figure 1: An Overview of the Costs and Benefits of Speed Limit Changes
Notes: Higher speed limits may lead to a higher average travel speed. This higher speed has a direct benefit (reducing travel time), but also three direct costs: higher accident rates, increased pollution and increased fuel expenditures. The pollution channel has indirect negative effects on infant and respiratory health, and climate change. Time savings benefits are private, while some of the costs are externalities.
authority was returned to the states, which decided to raise speed limits on a variety of highways.
This provides a rare opportunity to use difference-in-differences and ratio-in-ratios (count data)
methods to identify the effect of speed limit changes on travel speed, accidents, pollution and
health. To account for potential statewide trends in these outcome variables, I need to predict what
would have happened on the affected highways in absence of the changes in the law. I therefore
A P
Speed Limit	Speed
construct control highways or areas that are unaffected by the speed limit changes, but otherwise very similar to the affected highways or areas. Also, I exploit geographically precise micro data to make within-state difference-in-differences comparisons while holding constant weather, daylight, hour-of-day, traffic density, road construction, and much else.
I explicitly address whether control highways are indirectly affected by the speed limit changes through traffic substitution towards roads with higher speeds. I find that this is only a minor concern. I also demonstrate that my results stand up against potential identification challenges. One such challenge is that governments are unlikely to assign speed limits randomly. Fortunately, the 1987 speed limit changes provide a source of almost random variation as states were only permitted to raise speed limits on their rural interstates but not on other – otherwise very similar – major highways. I show that there are no differential pre-existing trends for the treatment and control highways, and present robustness checks with subsamples of control highways that match, e.g., the accident rate of treatment highways. The results are robust to such specifications.
My paper uses a detailed data set to evaluate the effects of speed limit changes. First, I use location descriptions of speed limit changes in California, Oregon and Washington. These states are selected because of superior data quality and availability. Second, I collect hourly measurements of actual traffic speed. Third, I use a data set of all highway accidents. Fourth, I use daily air pollution
3
measurements at various monitoring stations. Fifth, I requested all birth records in California to estimate the effect on infant health. Finally, I use geographical mapping techniques to augment these data sets with meteorological and geographic information wherever applicable.
In terms of the specific outcome variables, I find that a 10 mph speed limit increase leads to a 3-4 mph increase in travel speed, 9-15% more accidents, 34-60% more fatal accidents, a shift towards more severe accidents, and elevated pollution concentrations of 14-25% (carbon monoxide), 9-16% (nitrogen oxides) and 1-11% (ozone) around the affected freeways. The increased pollution leads to a 0.07 percentage point (9%) increase in the probability of a third trimester fetal death, and a positive but small and statistically insignificant increase in the probability of infant death. I use these estimates to calculate the time saving benefits and the private and external costs from accidents and deteriorated infant health. Moreover, I combine the travel speed estimates with engineering data to compute the increase in fuel consumption at higher speed. Similarly, I combine the air pollution estimates with epidemiology research to compute adverse health effects for adults.
Using these estimates and a wide array of plausible values of a statistical life (VSL) and values of time routinely adopted by governments, I find that the social costs of raising the speed limit from 55 to 65 mph are three to ten times larger than the social benefits. My social cost estimates are two to four times larger than in previous studies, in large part due to the greater comprehensiveness of my approach: I not only consider travel time and fatal accidents, but also non-fatal accidents, climate damages, fuel costs and health. Perhaps surprisingly, the costs from pollution-induced adverse health impacts are about as large as the costs from traffic fatalities. I further find that if travel time is valued only slightly below the average after-tax wage, the net social benefits are negative even when road fatalities and pollution-induced mortality are ignored. In contrast, I find that many individual drivers would enjoy a net private benefit from driving faster as a result of the higher speed limit. This reflects the substantial external costs not considered by a typical individual. Privately, a VSL of $6.0 million or less justifies driving faster, but the social planner’s VSL would have to be below $0.9 million to justify higher speed limits. While $6.0 million is within the conventional range of VSL estimates, $0.9 million falls well below it. These results suggest a surprisingly large difference between the social and private optimal speed choices. Although speed limits are a blunt instrument for dealing with the various externalities involved, it seems highly unlikely that any hidden administrative costs or unforeseen behavioral adjustments could make completely eliminating speed limits an efficiency-improving proposition.
A seminal paper in this literature is Ashenfelter and Greenstone (AG; 2004), who use the 1987 speed limit changes to estimate the value of a statistical life based only on travel time and fatal accidents (see Section 2 for details). My paper’s main contributions are threefold.
First, I explicitly distinguish between private and external costs and benefits, and show a stark contrast between the net benefits from the perspective of an individual driver and a social planner. Second, because I employ an unusually rich data set, I am able to estimate the effect of speed limit changes on additional outcome variables such as non-fatal injuries, property damage from accidents, air pollution and the health of infants and others who live near freeways. To the best
4
of my knowledge, these results are new to the literature.2	These estimates allow me to perform a more complete cost-benefit analysis. They are also interesting in their own right. For instance, estimating the effect of speed on pollution informs us about how higher tailpipe emissions translate into higher atmospheric pollutant concentrations. This could help guide environmental legislation such as engine technology standards.
Third, by exploiting within-state variation in speed limits and a wide range of control variables, my approach mitigates potential omitted variable bias. Using within-state speed limit variation is a useful and necessary complementary approach to existing cross-state regressions for speed and fatal accidents. AG use the fact that not all states raised their rural interstate speed limits in 1987 and compare highways of the same type (e.g., rural interstates), but across states with potentially dissimilar driving conditions (e.g., New Jersey vs. Iowa). I compare similar but differently classified highways (e.g., rural interstates and “rural principal arterial”), but within one state so that driving conditions are more similar. I present evidence that this choice of treatment and control highways is reasonable. Moreover, due to data limitations, AG do not include any control variables. This raises omitted variable bias as a possible concern. States differ in weather, climate, highway construction activity, and trends in vehicles types, vehicle safety, and other laws that affect driving (e.g., drunk drivingandseatbeltlaws).3 Mypaperaddressesthisconcernbyincludingmanycontrolvariables, such as weather, road and driver conditions, and demographics.
This research is relevant for today’s policy makers, since past experiences can inform them about whether speed limits are desirable and if raising speed limits further would be in society’s best interest. Using engineering and epidemiological evidence on the past and current relationships between speed, pollution and health, many results in this paper can be extrapolated to obtain pollution and health effects for current speed limit changes. The large difference between private and social net benefits is likely to persist, especially because governments currently consider speed limits in the 75-90 mph range where the strong upward sloping relationship between speed and emissions remains even for today’s new vehicles. This paper also demonstrates that the common approach to evaluating speed limits, based on a single cost or benefit or a single trade-off between travel time and fatal accidents, may lead to incorrect conclusions.
The paper starts by discussing relevant literature. Section 3 lays out the empirical strategy. In Section 4, I briefly discuss the various data sources. Section 5 provides summary statistics on quantities and trends to explore the plausibility of the identification assumptions. Sections 6-9 present the econometric framework and estimation results of the various outcome variables: speed, accidents, pollution and infant health. Section 10 combines these estimates in a cost-benefit analysis that also includes impacts on fuel consumption, climate change and adult health, and contrasts the private and social costs. Sections 11 and 12 offer further discussion and conclusions.
2Keller et al. (2008) study the effect of a temporary five day speed limit reduction on Swiss freeways in 2003 using an air pollutant dispersion simulation model, and predict that NOx traffic-related emissions decreased by 4%. 3For example, consider the weather. AG’s seven control states in the Northeast are far away from many treatment states, and a bad winter could hit the Midwest but not the Northeast. Since most accidents occur in the winter, and if the safety of rural interstates was particularly compromised by winter conditions, the treatment effect would
absorb this weather-induced change in accidents.}}