Identification of Factors Contributing to the Decline of Traffic Fatalities in the United States from 2008 to 2012 (2020)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

76 10.1 Data Needs 10.1.1 Infrastructure There is a substantial need for more detailed information on the nature and condition of infrastructure in the United States. The Highway Safety Manual (AASHTO 2010) has detailed information on crash reduction factors for different infrastructure components. However, there is no detailed inventory of the distribution of infrastructure-based safety features (for example, road miles with rumble strips or installation of advanced crash cushions). Nor is there an evalu- ation of lives saved by different infrastructure features or changes, analogous to Kahane’s (2015) estimation of lives saved by improvements to occupant protection and crashworthiness of the passenger car fleet. The current study used state highway spending per highway mile as a surrogate measure of efforts to improve safety related to the roadway, but these are relatively crude measures. The safety impact of programs can vary significantly, so the relationship between dollars spent and lives saved is not linear. A comprehensive evaluation at the system level of the effect of infra- structure changes is needed. This might take the form of a comprehensive inventory of roadway features related to safety, such as median barriers, rumble strips (centerline and shoulder), and paved shoulders. In addition, there should be a systematic evaluation of the net contribution of infrastructure spending to safety. Currently, safety evaluations have focused on specific projects or specific types of activities in specific geographic areas. There is a need for a higher-level analysis to estimate the net impact of the roadway environment on traffic safety. 10.1.2 Drivers The relative dearth of exposure data is a continuing obstacle in traffic safety studies. The current study found evidence for the impact of variations in exposure for different populations. There is a substantial need for more granular data on exposure. Data for driver VMT by age are available only from the NHTS. The NHTS is conducted only every 5 to 8 years. It was fortunate that, within the time period covered by this study, one year was 2009, which was the depth of the recession, but the prior survey was in 2001, so it was impossible to discern any trends within the period. For example, the timing of the decline in teen and young adult VMT within the period would be of great interest, but was unavailable. Ideally, some estimate of VMT by driver age and road type could be collected. The period saw significant changes in travel patterns by road type and urban/rural. Teens and young adults may have restricted their discretion- ary and leisure travel more than older age groups, but the data to determine this was lacking. Because the distribution of who drives, how much they drive, and where they drive can have profound impacts on traffic safety, these characteristics—how they are influenced in the short C H A P T E R 1 0 Data Needs and Future Research

Data Needs and Future Research 77 term by economic factors and over the long term by social and demographic changes—are critical to fully understanding traffic safety and ways to improve it. Driver licensing rates by age and state were a plausible surrogate for driving exposure for different age groups. Licensing information would have been useful to document trends at the state level in the decision to drive. Delay in licensure has been found at the national level (Tefft, Williams et al. 2013), and this likely contributed to the long-term decline in teen traffic fatalities. However, the data at the state level were too unreliable to be usable in the current project (Curry, Kim et al. 2014). These data must be improved. The current study also identified the specific effect of changes in median household income on traffic fatalities, suggesting that the reduction in risk was related to changes in the driving of groups especially affected by the economic contraction. However, this could not be measured directly by, for example, examining changes in the income distribution of drivers involved in fatal crashes, or in travel. Gaining resolution on this question and building longitudinal data through repeated cross-sectional driver surveys would be very valuable. Linking socio- economic data such as mean household income by zip code to crash data would provide additional insights. There are mechanisms to achieve this, but pursuing them was beyond the scope of the current project. In addition, there is a substantial need for more granular data on driver behavior, particularly as vehicles deploy more technologies for communication and entertainment, and as the con- nected environment becomes more inescapable. Driver attention and distraction is inherently transient and difficult to capture, but is often used to explain significant increases in crashes and fatalities. Much better data are needed on driver behavior to be able to measure changes objectively and estimate their effect on traffic safety. 10.1.3 Vehicles Improvements in the crashworthiness of the vehicle fleet have been demonstrated by Kahane (2004, 2014, 2015) and Farmer and Lund (2006, 2014). However, researchers found it prob- lematical when they attempted to capture that concept in the data used for statistical modeling. The penetration of post-1991 model year vehicles into the overall U.S. fleet was used as a surro- gate. The usefulness of crash data would be considerably enhanced if they included information on the crash-avoidance technologies with which the vehicles were equipped. As technologies that permit increasing levels of autonomous driving are deployed, this will become increasingly salient. In some cases, information about available technologies can be derived directly from the vehicle identification number and in other cases can be inferred from the make, model, and model year. Linking the NCAP rating would also be useful in estimating the contribution of vehicle design to traffic safety. 10.2 Future Research Trends in recent years suggest that the decline in traffic fatalities after 2007 did not establish a “new normal.” After reaching a plateau of approximately 33,000 traffic fatalities per year from 2010 to 2014, the number of deaths increased in 2015 by about 7.4%, to approximately 35,092 (NHTSA 2016b), and again in 2016 by about 7.7% to an estimated 37,700 (projected from NHTSA (2016a). Either traffic fatalities are returning to trend or new factors (texting and cell phone use have been suggested) are driving up fatal crashes. A useful first step would be to apply the models developed for this study to the needed input data series from recent years to see how well the models predict recent trends. This may show

78 Identification of Factors Contributing to the Decline of Traffic Fatalities in the United States from 2008 to 2012 that current models are adequate or that other factors are influencing the result and should be added. Was the recent increase in traffic fatalities related to a recovery in teen and young adult employment? What was the impact of the decline in fuel prices in 2015? Has the effect of reduced household income diminished as households, economies, and governments adjust to reduced growth? Is the decline in teen and young adult driving part of long-term trends in culture and communication or due to lack of employment and resources? Given that the reduction in novice driver and youth fatalities were a big part of the decline, this area should get a lot of attention. How can this information help inform goal setting for safety improvements? Clearly, many elements out of the control of transportation agencies have great influence on the level of fatalities. The primary factors identified here were the impact of the economic recession on specific demographic groups. How can this be accounted for when setting goals? Estimating and tracking “lives saved” from actions (programs and projects) may be as important as track- ing actual fatalities. Moreover, tracking fatalities should be put into the context of these other influential factors. To do this, better and comprehensive data are needed on safety program measures (beyond dollars spent) and safety projects. Do these measures really have a small influence, or is there a need for better input data? The comparison of the MNCS and MCS models showed that states’ experience differed significantly in some regards. The variables included in the MNCS model did not capture all the salient differences between states. Moreover, some states performed substantially better than others during the recession period, meaning that their reduction in fatalities was greater than predicted by the models, beyond what was predicted from the impact of the recession in their state. Appendix D provides a set of panel graphs that compare the trend in actual traffic fatalities and the number of traffic fatalities predicted by the models. Some states did substantially better in reducing traffic fatalities over the period, and some did less well. A study comparing a set of states that did better than predicted with a set of states that did worse, with the objective of identifying the specific factors that contributed to the out- performance, would be very useful. These factors may be demographic or other factors that are not in the control of transportation agencies. However, there may also be specific programs that influenced the outcome. One limitation of the current study was the inability, because of the lack of data for all the states, to measure the effect of state-level safety programs, beyond the general surrogate of safety spending. A state-level study that focused on a manageable number of states would overcome that limitation. The goal would be to identify and measure the specific factors that led to the over- and underperformance, including factors related to economic performance, state demographics, and specific state traffic safety programs, both related to infrastructure and related to enforcement and education. The current study was unable to address these issues at a sufficient level of detail, but a study focused on a limited number of states should be able to. The current study has identified the major issues and a set of candidate states for an in-depth study.