This page describes common methods used to estimate how a specific project will affect traffic crash rates.
Many factors affect crash rates including roadway design, traffic speeds, traffic density, vehicle mix and speed variation. Crash rates also depend on how impacts are measured, including whether statistics reflect crashes, insurance claims, casualties (human injuries and deaths), or just fatalities, and whether they include motorists or all road users. Crash rates can be measured per unit of traffic volume (such as 100,000 vehicles traveling on the roadway), travel distance (per 100,000 million vehicle-miles), or per capita. For example, expanding congested highways tends to reduce total crash rates, but by increasing traffic speeds it can increase crash severity and therefore total casualties.
Crash rates tend to increase with traffic density (vehicles per lane-mile), so urban driving tends to have more crashes and insurance claims per vehicle-mile than rural driving, but rural crashes tend to be more severe because they occur at higher speeds and so have higher fatality rates (Janke 1991). As a result, per mile crash rates are three times higher for urban driving, but fatality rates per mile are about twice as high for rural driving. Crash rates tend to be lowest on moderately congested roads (V/C=0.6), and increase at lower and higher congestion levels, while fatalities decline at high levels of congestion, indicating a trade off between congestion and safety (Zhou and Sisiopiku 1997).
There is some debate about the impact traffic speeds and traffic speed control strategies have on crash risk. Some researchers argue that speed variation is a greater risk factor than speed itself. Although this may be true for the frequency of crashes, increased traffic speeds increases crash severity (Stuster and Coffman, 1998). A summary of research indicates that each 1-mph change in traffic speeds causes a 5% change in crash fatalities, with greater impacts on high-speed highways where a 1-mph change can change fatality rates 8-9% (Finch, et al. 1994). Because crash rates tend to increase with traffic density, large reductions in vehicle travel can provide proportionately larger reductions in crash rates. Since about 70% of crashes involve multiple vehicles, a 10% reduction in total vehicle mileage in an area should theoretically provide a 17% reduction in total crashes (Vickrey 1968). Empirical evidence suggests that the reduction is somewhat less, probably because drivers are more cautious as traffic volumes increase, but comprehensive studies indicate that per mile crash rates decline with reductions in total vehicle travel (Edlin and Mandic 2001). Transportation demand management (TDM, also called mobility management) tend to affect safety in various ways (Litman and Fitzroy 2008):
1. Strategies that reduce overall vehicle travel probably provide proportional or greater reductions in crashes. Available evidence suggests that a 10% reduction in mileage in an area provides a 10-14% reduction in crashes, all else being equal.
2. Distance-based vehicle insurance reduces total vehicle mileage and gives higher-risk drivers an extra incentive to reduce their mileage, and so can be particularly effective at reducing crash rates.
3. Strategies that shift travel from driving to transit or ridesharing tend to provide medium to large safety benefits, depending on specific conditions.
4. Strategies that shift automobile travel to nonmotorized modes (walking and cycling) may increase per-mile risk for the people who change mode, but tend to reduce total crashes in an area due to reduced trip length and reduced risk to other road users. Nonmotorized travel also provides health benefits that may more than offset any increased risk to users.
5. Strategies that reduce traffic congestion tend to reduce crash frequency but increase severity, because crashes occur at higher speeds. As a result, mobility management strategies that shift automobile travel time, route or destination but do not reduce total vehicle travel probably do little to increase road safety overall.
6. Strategies that reduce traffic speeds tend to reduce per-mile crash frequency and severity, particularly in congested urban areas with high pedestrian traffic.
7. Smart growth land use management strategies may increase crash rates per lane-mile (due to higher traffic density and congestion) but tend to reduce per capita casualties due to reduced vehicle travel, lower traffic speeds and more restrictions on higher-risk drivers.
8. Vehicle traffic restrictions may reduce crashes if they reduce total vehicle mileage, but may do little to improve safety overall if they simply shift vehicle travel to other times or routes.
First, look at the BTS data for motor vehicle fatalities by highway functional system. In 2001, the fatality rate for collector streets in urban areas was 0.73 per 100 million vehicle miles.
To find the number of fatal injuries predicted for the new roads, multiply the number of projected VMT, in hundreds of millions (0.3), by the number of fatal injuries per 100 million VMT (0.73). We estimate that the new capacity will cause 0.219 fatal injuries per year.
Using State and Local Accident Data
State and/or local accident rates may provide more accurate estimates for the facility under analysis. The Cal-B/C model estimates the accident rate of the pre-improvement scenario using historical averages for the existing facility. For the post-improvement scenario, Cal-B/C uses statewide averages for the new facility classification adjusted by the degree to which the accident rate of the current facility differs from the state average for the current facility classification. In other words, Cal-B/C assumes that a facility with historical accident rates greater than the state average will continue this trend after an improvement.
If safety improvements are proposed for a facility, such as a road or a rail corridor, the accident rates on the existing facility can be compared with the accident rates on a similar facility that has already been improved. For example, a railroad crossing at grade in an urban area could be compared with a grade-separated crossing in an area with similar traffic patterns. This comparison helps predict how a proposed grade separation could reduce accident rates. The results could be presented as a reduction in the number of crashes per 100 million VMT or as a reduction in the number of crashes on a facility per year. In the case of an intersection improvement, the estimated number of crashes per 100 million vehicles may be the appropriate measure.
However, not all differences in accident rates are caused by safety improvements. There may be other physical or geographic factors, such as proximity to a residential neighborhood, that cause one facility to have lower accident rates than another.
A more sophisticated way to estimate how a project will affect accident rates is to create a statistical model that can estimate the number of accidents on a facility per a set unit of VMT. The model would incorporate many different characteristics of the facility; for example, a model for a road might include variables representing how sharply it curves and how steep its grades are.
The advantage of using a model is that it becomes possible to estimate how changes to a specific characteristic of a facility would affect accident rates. The disadvantage is that creating a good statistical model requires access to a large amount of detailed data about existing facilities. This data may be nonexistent or hard to obtain for some geographic areas and types of projects. Also, for new facilities, it is unlikely that their curves and grades will be finalized before conducting a benefit-cost analysis. Agencies usually do not make these engineering decisions until they have chosen which alternative to build.
ADB (2005), Accident Costing Reports, Arrive Alive, Regional Road Safety Program, Asian Development Bank (www.adb.org); at www.adb.org/Documents/Reports/Arrive-Alive/Costing-Reports/default.asp.
Bureau of Transportation Statistics (Annual Reports), National Transportation Statistics. U.S. Department of Transportation. Available at: www.bts.gov/publications/national_transportation_statistics.
D.J. Finch, P. Kompfner, C.R. Lockwood and G. Maycock (1994), Speed, Speed Limits and Crash, Transport Research Laboratory (Crowthorne, UK).
GRSP (2003), Estimating Crash Costs, Global Road Safety Partnership (www.grsproadsafety.org). Available at: www.grsproadsafety.org/themes/default/pdfs/Estimating%20crash%20costs.pdf
Paul F. Hanley (2004), Using Crash Costs in Safety Analysis, Public Policy Center, University of Iowa (http://ppc.uiowa.edu/dnn4/PublicPolicybrCenter/tabid/36/Default.aspx). Available at: http://ir.uiowa.edu/ppc_transportation/15.
Mary Janke (1991), “Accidents, Mileage, and the Exaggeration of Risk,” Accident Analysis and Prevention, Vol. 23, No. 3 (www.elsevier.com/locate/inca/336), pp. 183-188.
Todd Litman (2009), "Safety and Health Impacts," Transportation Cost and Benefit Analysis, Victoria Transport Policy Institute (www.vtpi.org). Available at www.vtpi.org/tca/tca0503.pdf.
Bhagwant Persaud (2000), Crash Reductions Following Installation of Roundabouts in the United States, Insurance Institute for Highway Safety (www.iihs.org).
TRISP (2005), “Valuation of Accident Reduction,” Economic Evaluation Notes, UK Department for International Development and the World Bank (www.worldbank.org). Available at: http://go.worldbank.org/ME49C4XOH0.
UKDfT (2009), Transport Analysis Guidance, UK Department for Transport (www.dft.gov.uk). Available at: www.dft.gov.uk/webtag/webdocuments/3_Expert/4_Safety_Objective/pdf/3.4.1.pdf.
Vicky Feng Wei and Gord Lovegrove (2010), “Sustainable Road Safety: A New (?) Neighbourhood Road Pattern That Saves VRU (Vulnerable Road Users) Lives,” Accident Analysis & Prevention (www.sciencedirect.com/science/journal/00014575).