Pedestrian Safety Relative to Traffic-Speed Management (2019)

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.

A-1 A P P E N D I X A Annotated Literature Review Aarts and Schagen 2006 Aarts and Schagen conducted a review of 11 studies on the relationship between speed and crash rate among crashes involving all road users. The authors reviewed studies that measured the relationship between crashes and speed on an individual vehicle level and at the road segment level. The authors found that existing studies support both the use of an exponential and a power function to relate speed and crash rate. All studies reviewed found that crash rates increased at a greater rate as the speed increased. This finding holds true regardless of road type (minor or major). The authors also found evidence to support the hypothesis that vehicles traveling much faster than the surrounding traffic have higher crash rates than vehicles moving at the speed of surrounding traffic. Garder 2004 Garder analyzed statewide crash data and pedestrian and vehicle volumes from 122 locations in Maine. The dataset included 1,589 pedestrian crashes that occurred between 1994 and 1998. Volume data were used in conjunction with a European crash prediction model to determine if certain factors, such as speed, may be associated with a higher frequency of crashes than would otherwise be expected. Garder divided the crashes into three categories based on the posted speed at the crash location and used these categories to evaluate the potential impact of speed on crashes. He found that high speeds and wide roads were associated with higher crash frequencies than other variables examined, and that there was a relatively strong relationship between the posted speed limit and a pedestrian’s risk of fatality. Garder found that low-speed locations had crash risks lower than predicted (p < 0.07), median-speed locations had higher risks than predicted (p < 0.0001), and high- speed locations had higher risks than predicted (p < 0.00001). There are some notable limitations with this study. Garder did not present the vehicle and pedestrian volume data used in the crash prediction model, so it is difficult to determine in which contexts the findings might be applicable. Similarly, he did not define the ranges of the three categories of speeds. For example, it is unclear whether 30 mph is considered a medium speed or high speed. The model the author used was based on English crashes at roundabouts, which may not be relevant to Maine, which may have a different road and walking environment compared to England. It is also worth noting that the speed data used in this analysis was posted speed, not travel speed. The findings from this study are also subject to sampling bias since fatal crashes are more likely to be reported in police crash databases. However, in general, the findings from this study support those of other studies on this topic. PEDESTRIAN-SAFETY AND TRAFFIC-SPEED REVIEW

A-2 Pedestrian Safety Relative to Traffic-Speed Management Henary et al. examined how pedestrian severe and fatal injuries are associated with pedestrian age and vehicle type. The authors used data from 1994 to 1998 from the Pedestrian Crash Data Study database. Only child (defined as aged 2 to 14 years) and adult pedestrians (19 to 50 years) were included in the study, resulting in a sample of 388 incidents. Vehicle body-type data were used to classify vehicles into passenger cars and light trucks and vans (LTV), the latter group also includes sport utility vehicles. Impact speeds ranged from 1 to 75 mph. Compared to passenger cars, pedestrians struck by LTV were more likely to have severe injuries (Odds Ratio = 1.31; 95% Confidence Interval: 0.88–1.94) or mortality (Odds Ratio = 1.40; 95% Confidence Interval: 0.84–2.34) for all pedestrians. Adjusting for pedestrian age, this association was more obvious and significant at lower impact speeds (≤ 20 mph). Adults hit by LTV had the highest risk of injury and mortality. Pitt et al. 1990 Pitt et al. used data from the Pedestrian Injury Causation Study and an analysis of variance to examine the relationship of pedestrian injury severity among pedestrians under age 20. In addition to age, the authors also examined the influence of other pedestrian, vehicle, driver characteristics and crash circumstances on pedestrian injury severity. The authors’ multivariate analyses indicated that vehicle travel speed greater than 30 mph, pedestrian age less than 5 years, time of day either early morning or late afternoon, residential zone, and type of road including collectors and major roads were associated with more severe injuries. The authors also found that on local streets in residential areas, nearly one- fifth of child pedestrians were struck by vehicles traveling faster than 30 mph, and these children experienced more severe injuries than children hit by slower moving vehicles. Leaf and Preusser 1999 This report included a review of existing literature and an analysis of NHTSA, Fatality Analysis Reporting System (FARS), and Florida crash data. All of the studies reviewed indicated that a pedestrian’s risk of fatal or severe injury increased with vehicle travel speed and posted speed. For example, Pitt et al. (1990) found that, compared to crashes with vehicle travel speeds of 10 to 19 mph, the risk of serious injury (or death) was 2.1 for speeds of 20 to 29 mph, 7.2 for speeds of 30 to 39 mph, and 30.7 for speeds of 40 mph or more. Pasean (1992) estimated that only 5% of pedestrians would die when struck by a vehicle traveling at 20 mph, compared with fatality rates of 40, 80, and nearly 100% for striking speeds of 30, 40, and 50 mph or more, respectively. The analysis of General Estimates System (GES) and FARS data found that fatalities rose from under 2% of struck pedestrians in crashes where the speed limits were below 25 mph to over 22% in crashes with speed limits of 50 mph or more. The analysis of Florida data revealed that the proportion of serious injuries and fatalities increased steadily with increasing vehicle speeds. For example, 36% of pedestrian crashes that occurred while vehicles were traveling 21 to 25 mph resulted in a fatal or severe injury, compared to 63% of pedestrian crashes that occurred while vehicles were traveling at 36 to 45 mph. Overall, this synthesis found a variety of studies and datasets which confirmed that higher vehicle speeds are associated with more severe pedestrian injuries. Henary et al. 2003

Annotated Literature Review A-3 The authors used the German In-Depth Accident Study database to examine the relationship between speed and pedestrian injury severity for pedestrians age 15 or older who were hit by the front of a passenger car from 1999 to 2007. In total, their dataset encompassed 490 pedestrians, including 36 fatalities. The data were weighted based on the injury severity distribution from the entire GIDAS dataset and national data. The authors used a logistic regression with the weighted sample to determine pedestrian fatality risk as a function of impact speed. The analysis revealed two key findings. First, the fatality risks presented in the study are quite a bit lower than fatality risks reported in similar, previous studies. Second, the authors found a strong, positive relationship between impact speed and fatality risk. The fatality risk at 50 km/h (31 mph) was more than twice as high as the risk at 40 km/h (25 mph) and more than five times higher than the risk at 30 km/h (19 mph). The risk of a pedestrian fatality at an impact speed of 48 km/h (30 mph) was approximately 7% and the risk of fatality at 64 km/h (40 mph) was approximately 25%. Richards 2010 Richards examined crashes in which the pedestrian was struck by the front of the car in Britain from 2000 to 2009. The data sample was weighted using national statistics. Richards also developed injury risk estimates using the Ashton (1980) dataset, but with data weighted to account for sampling bias. Richards found that at an impact speed of 30 mph, the risk of fatality for elderly pedestrians is 47%, compared with 5% for adults and 4% for children. The risk of pedestrian fatality at an impact speed of 30 mph was approximately 7%, and the fatality risk at an impact speed of 40 mph was approximately 31%. Based on Ashton (1980) data, the estimated risk of a pedestrian fatality is approximately 9% at 30 mph and 50% at 40 mph. Rosén and Sander 2011 Rosén and Sander searched for all research conducted on the topic of pedestrian fatality risk in relation to vehicle impact speed published in 2009 or earlier (one study from 2010 was included). Eleven studies were included in the literature review. Articles reviewed included Ashton (1980), Pasanen (1992), Davis (2001), Cuerden et al. (2007), Anderson et al. (1997), Yaksich (1964), Hannawald and Kauer (2004), Rosén and Sander (2009), Oh et al. (2008a and 2008b), and Kong and Yang (2010). The authors found that all studies reviewed as part of their research showed that increases in vehicle impact speed were associated with increases in a pedestrian’s risk of fatality. The risk estimates presented by studies varied considerably. In general, risk estimates developed before 2000 were much greater than estimates developed more recently. This difference is attributed to the fact that past research was based on datasets that were not adjusted for sample bias towards fatal and severe injuries, whereas more recent studies, in general, have accounted for this bias. The authors’ review suggests that Davis (2001) presents the most reliable fatality risks of the analyses based on the dataset used in Ashton (1980), however, the risk estimates are based on data from the 1960s and ‘70s and are likely not applicable today since there have been advances in medical and vehicle technology since then. Rosén and Sander (2009) present the only other risk estimates deemed reliable by the authors based on their methodology and data accuracy. Among the studies based on reliable and unbiased analyses, the estimated risk of pedestrian fatality at 50 km/h (31 mph) is about 10%. Most studies on this topic are based on European data. Since risk of fatality is strongly influenced by age and available medical treatment, the authors suggest that risks based on data from one country may not Rosén and Sander 2009

A-4 Pedestrian Safety Relative to Traffic-Speed Management be applicable to other countries, unless the available medical care and population distributions are similar. Tefft 2013 Tefft used U.S. data (from NHTSA’s National Automotive Sampling System 1994–1998). The dataset included 315 crashes, including 46 fatalities, but excluded crashes involving pedestrians under age 15. The data were weighted to correct for oversampling of pedestrians who were severely injured or killed. Tefft included all crashes involving a forward-moving car, light truck, van, or sport utility vehicle. A logistic regression was used to adjust for potential confounding pedestrian and vehicle characteristics. Risks were standardized to represent the average risk for a pedestrian struck by a car or light truck in the United States in 2007–2009. Tefft examined pedestrian age, sex, and type of striking vehicle (weight, height, and bumper height) as potential confounding factors. Tefft found that the average risk of severe injury reached approximately 10% at an impact speed of 17 mph, 25% at 25 mph, 50% at 33 mph, 75% at 41 mph, and 90% at 48 mph. The average adjusted, standardized risk of death reached 10% at an approximate impact speed of 24 mph, 25% at 33 mph, 50% at 41 mph, 75% at 48 mph, and 90% at 55 mph. Similar to previous studies, this study highlights the fact that the risk of fatality varies by age. In addition, this study quantified the difference in risk by age. For example, the average risk of death for a 70-year-old pedestrian struck at any given speed was similar to the average risk of death for a 30-year- old pedestrian struck at a speed approximately 12 mph faster. NTSB 2017 This NTSB report discusses the prevalence of speed-related crashes, risk of speeding, and emerging strategies that communities are using to reduce speeding. The reviews completed for this report found that speed increases crash risk by increasing the likelihood of being involved in a crash and the severity of the injuries sustained by users involved in the crash. The report explains the procedure for setting speed limits using the 85th percentile and discusses how following this approach can be problematic and other approaches may improve roadway safety, especially in areas where there are pedestrians. The report also discusses the need for enforcing speed limits and the use of ASE. Kroyer 2015 Kroyer analyzed Swedish crash data from 2004 to 2008 using a multinomial logit model to measure the relationship between average vehicle speed and pedestrian injury severity by pedestrian age. Kroyer used two datasets in the analysis. The first dataset included all injury crashes in Sweden between 2004 and 2008 for which the age of the pedestrian was known. The second dataset included randomly selected crashes from the first dataset within each injury severity group that occurred in Scania and met six other criteria, including the crash having occurred during dry, lit road conditions and where the injury was deemed not associated with speed; and crash locations with very low traffic volumes were excluded as were locations with major traffic changes during the study period. Spot speed measurements were taken at crash locations from the second dataset. The number of speed measurements taken per crash site ranged from 59 to 100. The datasets were weighted before analyzed. Kroyer found that even though fatal crashes are rare in locations where the mean travel speed is below 40 km/h (25 mph) and severe injuries are rare below 25 km/h (16 mph), over 30% of severe injury crashes occur in speed environments below 35 km/h (22 mph). This indicates that 30 km/h (19 mph) speed limits might not be as safe as previously believed.

Annotated Literature Review A-5 The mean speed at accident locations with fatal injuries differs from those at locations with minor injuries (p = 0.003) and severe injuries (p = 0.031). While it is clear that there is an increased risk of a fatal or severe injury for pedestrians ages 65 or older, there may also be an increased risk for pedestrians ages 45 to 64. Young pedestrians also have an increased risk of severe injury. Martin and Wu 2017 The authors used a weighted sample of pedestrian crashes in France from 2011. A clog-log model was used to optimize risk adjustment for high collision speeds. Impact speed was derived from pedestrian throw-distance using the average distance derived from various formulae. There were 205 fatal and 227 nonfatal crashes in the dataset before it was weighted. The authors found that the risk of death was nearly 100% when impact speeds exceeded 80 km/h (50 mph). For impact speeds below 60 km/h (37 mph), the shape of the curve relating probability of death to impact speed was very similar to those reported in recent studies. The authors stated that the use of the log-log model was better than models used by other studies because it resulted in a better fit to the data at higher impact speeds. The risk of death doubled when speeds rose from 30 to 40 km/h (19 to 25 mph), and increased by a factor of six between 30 and 50 km/h (19 and 31 mph). Davis 2001 Davis analyzed crash data from the United Kingdom from 1966 to 1969 and from 1973 to 1979 (N = 358). There were 81 fatalities included in the dataset. Davis used a logistic regression to analyze the relationship between pedestrian injury severity and vehicle impact speed. He found that a pedestrian’s risk of fatality was 10% at 53 mph, 25% at 61 mph, and 50% at 69 mph. Roudsari et al. 2004 Roudsari et al. reviewed crash data from 1994 to 1998 to measure the difference in injury severity and mortality rates among pedestrians struck by light trucks and vans. Their review of 542 pedestrians revealed that pedestrians struck by light trucks and vans (LTV) had a higher risk (29%) of severe injuries compared to those struck by passenger vehicles (18%) (p < 0.02). After adjusting for pedestrian age and impact speed, LTVs were associated with 3.0 times higher risk of severe injuries (95% confidence interval (CI) 1.26 to 7.29, p = 0.013). Mortality rate for pedestrians struck by LTVs (25%) was two times higher than that for passenger vehicles (12%) (p < 0.001). After adjusting for age and impact speed, a pedestrian’s risk of death for crashes involving LTV was 3.4 times higher than that for passenger vehicles (95% CI 1.45 to 7.81, p < 0.005). Lefler and Hampton 2004 Lefler and Hampton reviewed U.S. crash data from the Fatality Analysis Reporting System (FARS), the General Estimates System (GES), and the Pedestrian Crash Data Study (PCDS) for a 5-year period from 1994 to 1998. The authors found that large vans, large pickups, and large utility trucks were associated with more pedestrian fatalities per 1,000 single vehicle–pedestrian impacts compared to compact pickups, compact utility trucks, minivans, and cars. In fact, pedestrians were found to have a two to three times greater likelihood of being fatality injured when struck by a light truck or van than when struck by a car. In addition, a pedestrian’s risk of serious injury in the head or thoracic region was also much greater when struck by a light truck or van compared to a passenger car.

A-6 Pedestrian Safety Relative to Traffic-Speed Management Kim et al. 2008 The authors used a heteroskedastic generalized extreme value model to explore the relationship between age and pedestrian injury severity among pedestrians involved in crashes with motor vehicles. The dataset used for the analysis included more than 5,000 crashes from North Carolina that involved one vehicle and one pedestrian and occurred between 1997 and 2000. The authors found that as age increases, a pedestrian’s risk of being involved in a serious injury when in a crash with a vehicle also increases. COUNTERMEASURES REVIEW Engineering Countermeasures Treatments with Vertical Deflections FHWA 2014 The authors reviewed 100 studies on traffic-calming measures in urban areas, including speed humps, speed cushions, speed tables, raised intersections, curb extensions, neck downs, chicanes, center islands, reduced lane widths, road diets, “SLOW” pavement markings, speed-activated speed limit signs, speed feedback signs, traffic circles, and combinations of chokers and speed humps. They found reductions in 85th percentile speeds and mean speeds in nearly all studies reviewed from 1997 to 2013. Roundabouts, speed humps, and combinations of speed humps and neck downs were associated with the greatest reductions in 85th percentile speeds, followed by speed cushions, speed tables, chicanes, speed-activated speed limit signs, and speed feedback signs, which all had similar levels of effectiveness. Note that reductions in speeds varied by study; some changes were as small as 1 mph. Speed humps, speed tables, and speed feedback signs were the most commonly studied treatments in urban areas. Mountain et al. 2005 The authors reviewed 150 speed management treatments along 30 mph roads (posted speed) in the United Kingdom. The treatments included speed cameras, speed-activated signs, and engineering treatments. Crash data were reviewed for 3 years before and after the installation of the treatments. Where possible, mean speed, 85th percentile, standard deviation, percentage exceeding the speed limit and the mean speed of speeders were collected at treatment sites. The study controls for general trends in accidents, regression-to-mean effects and migration, separately estimating the accident changes attributable to the impact of the schemes on traffic speed and on traffic volume. The authors found that all types of speed management efforts tend to reduce injury crashes at a rate of 1 crash per km per year. When comparing the percentage of injury crashes attributable to the treatment by treatment type, engineering treatments with vertical deflections tend to be the most effective (44% reduction of all injury crashes), and are twice as effective as sites with speed cameras (22% reduction of all injury crashes). Engineering treatments with vertical deflections were considerably more effective at reducing fatal and severe injury crashes. Engineering treatments without vertical deflections were less effective than those with vertical deflections, but were slightly more effective than cameras (29% reduction of all injury crashes). Among fatal and serious injury crashes, the average reduction for engineering treatments with vertical deflections (35%) is over three times that of cameras (11%) and over twice that of treatments with horizontal features (14%). When comparing the impact of the treatments on speed, engineering treatments with vertical deflections were more effective than those with horizontal deflections, which were more effective than cameras for all injury crashes and fatal and

Annotated Literature Review A-7 serious injury crashes. Engineering treatments with vertical deflections have the greatest average impact on the mean, 85th percentile speed and the percentage of drivers speeding (p < 0.05). The authors could not account for regression-to-the-mean effects for pedestrian and bike crashes, so caution is needed when interpreting the results associated with these crashes. Among bike crashes, engineering treatments were associated with a much larger percentage reduction in crashes than cameras (34% versus 6%). Among pedestrian crashes, engineering treatments were more effective at reducing crashes than cameras (54% versus 27%). A similar trend was observed for child pedestrian crashes (62% for engineering treatments versus 13% for safety cameras). Gonzalo-Orden et al. 2016 The authors reviewed speed data from 22 street sections with traffic calming treatments. The speeds at each treated location were compared to a similar location without the treatment. The final dataset included speed data from over 10,000 vehicles, and measurements included traffic volume, 50th percentile speeds, and 85th percentile speeds. Treatments tested included raised crosswalks, speed warning signs, lane narrowing, and radar speed cameras. Results for each treatment varied. The raised crosswalk and lane narrowing were the most effective at reducing vehicle speeds. The radar speed camera and the radar speed warning signs were only effective at reducing traffic speeds at individual points and did not reduce the travel speed of the entire studied section. Speed Humps, Lumps, and Cushions Gulden and Ewing 2009 The authors examined the impact of speed lumps on different types of vehicles, including fire–rescue vehicles and passenger vehicles, and compared their effectiveness to that of speed humps. Based on the original research and literature reviews conducted as part of this study, speed lumps were found to reduce speeds and be well suited for use on streets along emergency vehicle routes. Speed lumps reduced the 85th percentile travel speed at all study locations; on average, 85th percentile speeds were reduced by 25%, or 9 mph. The speed reduction with lumps is comparable to that of speed humps, which reduce speeds by 23% or 8 mph. The authors suggest that the width of the speed lump is integral to its effectiveness. Ideally, speed lumps will have an approximately 6-ft-wide center lump. Outside lump widths are less critical and are typically wider than 6 ft. City of Alexandria n.d. Alexandria has completed neighborhood traffic calming installations and collected at least 85th percentile speeds before and after installations. At least a brief evaluation was conducted at three sites. Russell Road—85th percentile speeds decreased by 21% in the southbound direction and 18% in the northbound direction. Curb extensions and neck downs or speed cushions and crosswalks were installed on Russell Road. West Abingdon Drive—85th percentile speeds decreased by 30% on this roadway from 35.2 to 24.7 mph. Two speed cushions and a crosswalk were installed as part of this project. Martha Custis Drive—85th percentile speeds decreased by 21% on this roadway from 30.4 to 23.9 mph. Two speed cushions were installed as part of this project.

A-8 Pedestrian Safety Relative to Traffic-Speed Management Agerholm et al. 2017 The authors used a before–after approach to study the effect of sinus speed humps of height 10 cm (approximately 4 in.) and length 950 cm (about 31 ft) and chicanes [with a free carriageway width of 5.3 cm (2 in.) and a length between the 0.5 m (1.6 ft)-wide obstacles of 18 m (59 ft)] to determine the impact of the two treatments on vehicle travel speed. The authors used GNSS data from 3,216 vehicle trips passing one or more traffic-calming measures in their analysis. Overall, the implementation of the countermeasures reduced mean travel speed from 53.5 to 49.4 km/h (33.2 to 30.7 mph) and the 85th percentile reduced from 59.9 to 55.6 km/h (37.2 to 34.5 mph). The authors observed a decrease in the effect of the treatments as the distance from the treatments increased. The speed humps and chicanes were both associated with reductions in the mean travel speeds, but the variation in speeds was slightly greater near chicanes. City and County of San Francisco 2016 SFMTA conducted before and after traffic counts and tracked vehicle speeds for eight nonarterial residential streets where residential traffic-calming measures were installed between 2011 and 2015. The traffic-calming measures included a median island, rubber speed hump, speed humps, and speed cushions. Key findings from the study include Residential traffic-calming measures were successful at significantly decreasing speeds traveled, particularly the incidence of vehicles traveling at high rates of speed. Overall, changes in ADT on streets where traffic-calming installations occurred were negligible, while the 85th percentile speeds decreased by 18% and vehicles traveling over 30 mph decreased by 78%. At locations where speed humps were installed, the 85th percentile speed decreased by 23%, and vehicles traveling over 30 mph decreased by 87%. Although the data show reductions in speed on residential streets where traffic-calming measures were installed, SFMTA did not include data on the number of collisions or collision severity before and after the traffic-calming measures were installed because collisions on the residential streets were infrequent and speeds at the time of collision were generally low. In general, vehicle speeds at the time of collision were associated with severity of injury or likelihood of fatality. Ponnaluri and Groce 2005 Five speed humps were installed along a residential street with approximately 500 ft in between each hump. Two other segments with pre-existing speed humps were also included in the analysis. The authors collected volume and speed data one month before installation (3 years prior for the two streets with pre-existing speed humps), and the post data were collected one month after installation for all three sites. Data were collected in 15-min intervals over a 48-h period on a typical weekday. The treatments reduced 85th percentile speeds by 22% to 29% and increased the number of vehicles in the 10-mph pace by nine to 31% and decreased speed limit violations. The number of vehicles exceeding the posted speed limit decreased by 43% to 61%. The findings also suggest that installing more than one speed hump within 1,000 ft of another hump may not yield any additional benefit.

Annotated Literature Review A-9 Rothman et al. 2015 Speed humps were mapped along with police-reported pedestrian crashes from 2000–2011. A quasiexperimental study identified crash counts before and after speed hump installation and modeled the data using a repeated-measures Poisson regression adjusted for season and roadway characteristics. Stratified analyses were conducted by age group and injury severity. There were 27,827 vehicle– pedestrian crashes with 1,344 crashes along 409 roadways with speed humps. A 25-m (82-ft) buffer around each speed hump was used to capture crashes that occurred close to the road. Crashes were excluded if 1. the crash occurred beyond the 25-m roadway width and 2. occurred on the same day of speed hump installation. There was a 26% reduction in vehicle–pedestrian crash rates (0.74; 95% CI 0.62, 0.89) on local roads after speed hump installation after controlling for winter and built environment roadway characteristics. The association between speed humps and vehicle–pedestrian crash rates decreased more for children (ages 0 to 15) (44% reduction; IRR 0.57; 95% CI 0.41, 0.79) than for adults (20% reduction; IRR 0.80; 95% CI 0.68, 0.95). Note that the decrease in crashes associated with roads with speed humps was substantially greater than the overall trend in decreasing crashes during the study period. Tester et al. 2004 The authors used a matched case–control study to measure the impact of speed humps on child pedestrian injuries. The study examined injuries among children (ages 5 to 15) seen in a pediatric emergency department after being struck by an automobile between 1995 and 2000. Over the 5-year study period, Oakland installed about 1,600 speed humps on residential streets. Cases were matched to children of the same age and gender who visited the emergency department on the same day from within the city boundary but visited for a reason other than being struck by a vehicle. Living on a street within one block of a speed hump was used as the predictor variable. The final study sample included 100 individuals. A multivariate conditional logistic regression analysis showed that speed humps were associated with a 53% to 60% reduction in the odds of injury or death among children struck by a vehicle in their neighborhood (adjusted odds ratio = 0.47), after controlling for race and ethnicity. Ewing 1999 This report presents an overview of the design, implementation, and impacts of engineering traffic- calming measures in the United States. The report includes a review of different countermeasures’ impacts on speed based on a number of previous studies from around the country. The table below presents a summary of the impact of a variety of traffic-calming measures on 85th percentile speeds and average speeds. Speed humps were found to have the greatest impact on 85th percentile speeds, reducing them by an average of more than 7 mph, or 20%. Raised intersections, long speed tables (longer than 22 ft), and lane narrowing and diagonal diverters were found to have the least impact.

A-10 Pedestrian Safety Relative to Traffic-Speed Management Gitelman et al. 2017 The authors analyzed the impact of installing raised pedestrian crosswalks and speed humps at nonsignalized midblock crossings on urban arterial and collector roads on vehicle travel speeds. The controlled study included eight sites. Travel speeds were captured using free-flow speed measurements. All sites had a raised center median with two travel lanes on either side; two marked, raised crosswalks; a speed hump preceding the crosswalk; speed limits of 50 km/h (31 mph); 85th percentile speeds above 50 km/h (31 mph) in at least one direction; and at least 25 pedestrian crossings per hour. Speed humps were installed 15 to 20 m (49 to 66 ft) before the crosswalk. Immediately after installation, all sites were associated with significant decreases in vehicle speeds (p < 0.05). At five of the sites, the decreased vehicle speeds persisted over time (measured as 2 months post installation). At three sites, vehicle speeds initially decreased, but after 2 months, vehicle speeds had increased above the initial decrease—in all but one case, the speeds remained below the speeds measured before the treatments were installed. These trends apply to both mean speeds and 85th percentile speeds. A decrease in pedestrian–vehicle conflicts was observed at more than half of the sites; however, the decrease was only statistically significant at three of the sites (p < 0.05). The percentage of drivers who yielded to pedestrians increased at all sites with driver yielding below 100% during preinstallation conditions; the increases in driver yielding were statistically significant at four of the sites (p < 0.05). See also FHWA 2014. Speed Tables/Raised Crossings Huang and Cynecki 2001 Huang and Cynecki used before and after data and treatment and control sites to analyze the impact of curb extension, raised crosswalks, raised intersections, and refuge islands on pedestrian and motor vehicle behavior. The authors collected data on vehicle speeds, driver yielding behavior, and average

Annotated Literature Review A-11 pedestrian wait time. In terms of vehicle speeds at raised crossings, the authors found that 50th percentile speeds were lower at the treatment sites compared to the control sites. The lower 50th percentile speeds at the treatment sites were only statistically significant at two of the three sites. The differences in 50th percentile speeds at the treatment sites compared to the control sites ranged from 2.5 to 12 mph. Chen et al. 2013 The effectiveness of 13 countermeasures and street redesigns installed in New York City between 1990 and 2008 were evaluated using a two-group pretest–posttest design to determine impact on crash frequency. The countermeasures directly related to speed included road diets (460 locations), speed humps (601 locations), and speed-limit reductions (270 locations). Speed humps and speed-limit reductions were found to have a minimal impact on vehicle–pedestrian crashes and injury crashes. A Crash Modification Factor (CMF) of 0.45 for pedestrian crashes was developed for speed humps along segments and a CMF of 0.59 for pedestrian crashes was developed for road diets along segments. Neither of the CMFs was significant at the 5% level. Road diets were found to be very effective at reducing segment-based injury and fatal crashes. See also FHWA 2014, Gonzalo-Orden et al. 2016, and Gitelman et al. 2017. Engineering Countermeasures Treatments with Horizontal Deflection See FHWA 2014 and Mountain et al. 2005. Chicanes Traffic Advisory Unit 1997 This document presented information about the use of chicanes in the United Kingdom. The information included findings from a review of 49 chicane schemes. When combing the findings from all chicanes studied, the authors found the chicanes to be associated with a reduction in mean speed and 85th percentile speed of 12 mph. The average mean speed was found to be 23 mph and the 85th percentile speed 28 mph. See also FHWA 2014, Ewing 1999, and Agerholm et al. 2017. Mini Traffic Circles Zegeer et al. 2013 PEDSAFE includes case studies of various traffic-calming treatments. Some of the case studies include speed and crash data. The treatments included on the website include raised crossings, curb extensions, raised medians, neck downs, chicanes, mini circles, speed humps, and speed tables. Many of the data points cited are from city or county evaluations, not academic studies, however, occasionally academic sources are cited. FHWA 2010 This FHWA technical summary presents a summary of design and operational considerations for mini roundabouts. The document indicates that this treatment is best suited for low-speed roads (under 35 mph) and that it is an effective treatment to reduce vehicle speeds and enhance the pedestrian environment.

A-12 Pedestrian Safety Relative to Traffic-Speed Management U.S. Roads 1998 This article presents a summary of information compiled by the City of Seattle to evaluate the impact of their neighborhood traffic circle program. The article presents treatment design information as well as crash trends from before and after the program was implemented. The article also presents evidence of public support for the program. Out of more than 600 circles that have been installed, only two have been removed. Residents frequently petition the city to install traffic circles in new places (about 700 requests per year). The article states that neighborhood traffic circles effectively reduce vehicle speeds without reducing vehicle volume. No speed data were presented in the article. See also Ewing 1999. Engineering Countermeasures Lane and Road Reconfiguration Road Reconfiguration (Road Diets) Corkle et al. 2001 The authors conducted a literature review, local before–after studies, and driver simulation study of before–after driver behavior on streets where traffic-calming treatments were implemented. The local before–after studies consisted of measuring speed data before and after the implementation of neck downs, lane reductions, pavement marking and signage, and a raised crosswalk. The authors found that traffic calming had a limited impact on average driver speed, but had a greater impact on reducing the number and speed of drivers traveling above the 85th percentile speed. More specific findings include Words painted on the pavement along with chevrons and high-visibility speed limit signs reduced speeds. Neck downs significantly reduced mean speeds and 85th percentile speeds (95% CI), however, the extent of their impact appears to be localized. The impact of raised crosswalks extended only several hundred feet from the crosswalk. Road diets reduced speeds only when there was sufficient traffic volume. Speed humps and speed tables (which incorporate changes in vertical alignment) showed greater reductions in vehicle speeds than devices that incorporated horizontal shifts (traffic circles and one-lane slow points) or roadway narrowing. Pavement markings such as “Slow” are not effective when used alone. Pavement markings and signage reduced mean and 85th percentile speeds; however, it is unclear if the impact was statistically significant. The speed reductions did not persist 2 years after; however, after the street was resurfaced and repainted the speeds did reduce again, suggesting maintenance is important for traffic-calming treatments that involve pavement markings. The driver simulations conducted as part of this study indicated that the use of a combination of medians and neck downs was more effective at reducing vehicle speeds than the use of either of the two treatments individually.

Annotated Literature Review A-13 Knapp and Giese 2001 Knapp and Giese report on the results of converting a number of four-lane undivided roadways to three- lane roadways with two-way left-turn lanes. The road diets reviewed included several in Iowa, Montana, Minnesota, California, and Washington. Many of the road diets reviewed were associated with improved safety, measured as either a reduction in crashes or decreases in speed, or both. A road diet at the Sioux Center in Iowa on a 30-mph road was associated with a reduction in average travel speeds from 28 or 29 mph to 21 mph and a reduction in average free-flow speeds from 35 to 32 mph. The percentage of vehicles traveling more than 5 mph above the speed limit decreased from 43% to 13%. In total, travel speeds reduced at 3 of the 10 sites with speed data. Thomas 2013 Thomas conducted a review of existing literature on rechannelization (road diets) projects to assess their impact on traffic safety. After reviewing six studies on the topic, she concluded that, in general, road diets effectively reduce vehicle speeds, and in many cases, reduce crashes. For example, four-to- three lane conversions may be expected to reduce crashes by nearly 30%; however, crash reductions are likely to be larger in rural areas and may be slightly lower in urban areas. Thomas found only one study that examined the impacts of road diets specifically on crashes involving pedestrians. Her research indicates that the largest safety benefits may occur on roadways with a high density of driveways, multilane crossings, and a history of severe crashes and/or vehicle–pedestrian crashes, and speeding. She also found that, in many situations, vehicle volumes are not diminished by road reconfiguration projects. See Chen et al. 2013. Raised Medians and Pedestrian Crossing Islands NYCDOT 2004 The NYCDOT collected before and after data to examine the effectiveness of various traffic-calming treatments, including a raised center island. Data were collected at one site, on a local street in Brooklyn, NY. The speed data revealed that after the treatment was installed, the average speed reduced by 2 mph and the 85th percentile speed reduced by 12 mph. Data were collected 12 months after the treatment was installed. Gan et al. 2005 The authors used before and after data and cross-sectional studies to develop crash reduction factors (CRF) for a variety of engineering treatments. Gan et al. established a CRF of 0.75 for vehicle–pedestrian crashes for raised medians. Zegeer et al. 2002 In this report, Zegeer et al. evaluated 5 years of pedestrian crashes at 1,000 marked crosswalks and 1,000 unmarked comparison sites. The authors collected data on variables including traffic volume, pedestrian volume, number of lanes, median type, and speed limit. The authors used Poisson and negative binomial regressive models to evaluate the safety impacts of marked crosswalks, with and without other improvements, compared to unmarked crossing locations. Zegeer et al. found that multilane sites with raised medians were associated with significantly fewer pedestrian crashes than

A-14 Pedestrian Safety Relative to Traffic-Speed Management comparison sites, regardless of whether there was a marked or unmarked crosswalk. A CMF for a raised median at a marked crosswalk at an uncontrolled intersection is estimated to be 0.54 for vehicle– pedestrian crashes. Similarly, a CMF for a raised median at an unmarked crosswalk at an uncontrolled intersection is 0.61. See also Ewing 1999. Neckdowns/Bulb-Outs/Curb Extensions/Chokers See Corkle et al. 2001. Bella and Silvestri 2015 Three safety countermeasures at pedestrian crossings (curb extensions, parking restrictions, and advanced yield markings) and the condition of no treatment (baseline condition) were designed on a two-lane urban road and implemented in an advanced driving simulator. Forty-two drivers completed the simulation. The authors used a multivariate variance analysis (MANOVA) to assess the impact of each treatment. In total, approximately 110 to 120 vehicle–pedestrian interactions were recorded for each of the three treatments and baseline conditions. The impact of the tested treatments may differ from real life. Only 42 drivers were included in the analysis. The following were identified as key findings from the analysis. The driver’s speed behavior was affected by conditions of vehicle–pedestrian interaction and was consistent with previous findings in the literature. Curb extensions induced the most appropriate driver’s speed behavior while approaching the zebra crossing compared to the other treatments and baseline conditions. More than 80% of the drivers perceived that the curb extensions were effective, which indicates that when this countermeasure was present, the drivers were more willing to yield and that the visibility of the pedestrian crossing was better. Curb extensions were associated with the highest percentage of drivers yielding to pedestrians. City of Cambridge 2012 Brattle Street in Cambridge, Massachusetts , was reconstructed in 2010 to reduce vehicle speeds and improve pedestrian safety. The street reconstruction project included the addition of curb extensions, chicanes, and crossing islands at intersections. Speed studies were conducted before and after the installation of the traffic-calming measures. Both the percentage of vehicles exceeding the speed limit and the 85th percentile speed were used to evaluate changes in traffic speed. The changes in the 85th percentile speed did not change significantly, but there was a significant reduction in the percentage of drivers traveling 25 to 30 mph, and an increase in the number of drivers traveling at speeds below 25 mph. The City also conducted a survey among residents 2 years after the project was constructed. The City found that 55% of survey participants responded favorably to the project, 70% believed the overall look of the street had improved, and 58% of respondents felt that the changes improved pedestrian safety (N = 417). See also Ewing 1999.

Annotated Literature Review A-15 Lane Narrowing with Pavement Markings Thomas et al. 2016 The authors of this synthesis reviewed existing studies of 25 pedestrian crossing treatments and surveyed transportation agencies from 40 states to determine the impact of the 25 crossing treatments on pedestrian safety. Based on the research conducted for this project, the authors found that a number of crossing treatments were associated with reduced vehicle speeds in some situations or were expected to reduce speeds based on expert opinions. These treatments included raised crosswalks or speed tables, reducing a corner radius, road diets, corridorwide traffic calming, mini traffic circles, speed humps, and chicanes. VHB 2008 VHB conducted a study for the FHWA on low-cost strategies to improve safety at intersections in rural areas. The authors found that there was very limited research on the impact of lane narrowing on safety. At test sites, the lane widths prior to the narrowing treatment ranged from 11 to 12 ft. The lane- narrowing treatments reduced the widths to 9 or 10 ft. At some sites, 4-ft painted medians were added to the treated roadway segment. Speed data, including mean and 85th percentile speeds, were collected at nine sites with the narrower lanes. Speed limits at all sites were 50 or 55 mph. The average mean and 85th percentile speeds across all sites decreased by nearly 4 mph and 5 mph, respectively, after the treatment was installed. The reductions in mean speeds ranged from 2 to 5 mph and the reductions in 85th percentile speeds ranged from 4 to 5 mph. See also Ewing 1999. Curb Radius Reductions Zangenehpour et al. 2017 Zangenehpour et al. studied the impact of curb radius reductions on pedestrian safety using a before– after methodology in Toronto, Canada. The curb radius reductions were installed specifically to reduce vehicle turning speeds and improve pedestrian safety. The authors’ analysis measured safety via dangerous conflicts, rather than crash history. Because dangerous conflicts occur more frequently than crashes, it is easier to obtain a sufficiently-sized dataset in a short period. In addition to vehicle– pedestrian conflicts, the authors also measured changes in traffic speed. Conflicts were measured using post encroachment time (PET) between a pedestrian and a turning vehicle and were classified into three categories, including low, medium, and high risk. High-risk conflicts included those with a PET of less than 1 s, medium-risk conflicts had a PET between 1 and 3 s, and low-risk conflicts had a PET between 3 and 5 s. The authors collected 3 days of video data from before and after the curb radius reductions were implemented. This resulted in 144 hours of data for two intersections. At the first intersection, the low-, medium-, and high-risk conflict rates decreased by 72%, 38%, and 30%, respectively. At the second intersection, the low-, medium-, and high-risk conflict rates decreased by 90%, 100%, and 100%, respectively. It is worth noting that the number of conflicts during the before period was much greater at the first intersection than at the second intersection. While the percentage of conflicts decreased at both intersections, there was no significant change in vehicle turning speeds.

A-16 Pedestrian Safety Relative to Traffic-Speed Management Signs and Signals Speed Feedback Signs and Speed-Activated Speed-Limit Signs Hallmark et al. 2013 Five different traffic-calming treatments were evaluated along the main roads through small communities. Speed and volume data were collected using road tubes. Volume and speed data from before treatment installation were compared to data 1 and 12 months after installation. The treatments included transverse speed bars, colored pavement with speed limit, temporary center islands, radar- activated speed limit signs, and speed feedback signs. Data were typically collected for 48 h on a Monday through Friday under mostly dry weather conditions. Transverse speed bars and colored pavement with the speed limit were both associated with only a 1.6- mph and 1.3-mph mean speed decrease compared to baseline conditions, respectively. Both of these treatments were associated with large reductions in the percentage of vehicles traveling 5, 10, or 15 mph over the speed limit 1 month and 12 months after the treatments were installed. Temporary center islands were associated with a 2.2-mph decrease in mean speeds and 3-mph decrease in 85th percentile speeds, again, significant (at least 30%) reductions in the percentage of vehicles traveling 5, 10, or 15 mph over the speed limit. Speed feedback signs were associated with an 8-mph decrease in mean speeds, and 9-mph decrease in 85th percentile speeds. Similar to the other treatments, the percentage of vehicles traveling 5, 10, or 15 mph over the limit decreased by 33, 53, and 71%, respectively. Radar- activated speed limit signs were associated with a 6-mph decrease in mean speeds and 7-mph decrease in 85th percentile speeds. The decrease in the fraction of vehicles traveling 5, 10, or 15 or more mph over the posted speed limit was 25, 40, and nearly 53%, respectively. Speed reductions persisted 12 months after installation for all of the aforementioned treatments except for temporary center islands, although in most cases, the reductions were not as dramatic. See also FHWA 2014. In-road Signs Van Houten and Hochmuth 2017 A Gateway treatment (three R1-6 signs placed in the road) was installed on 10 different streets in three cities in Michigan. The posted speeds on the study streets were 25 mph. The authors measured vehicle speed using a laser speed measuring system before the gateway treatment was installed, and at one month, three months, and five months after the installation. Data were collected from 400 vehicles during each month’s speed measurement. Data were collected whether or not pedestrians were present at a crosswalk. Speed reductions at both the dilemma zone (a location beyond which a driver can easily yield if a pedestrian enters the crosswalk) and crosswalk occurred after the installation of all 10 treatments. The data show that the speed reductions at the study sites averaged around 4 mph and remained consistent over the 5-month measurement period. The average speed reductions ranged from 2.7 to 8.7 mph after the first month. At a previous study site, the average speed decreased from 26.8 to 23.1 mph at the dilemma zone and from 28.3 to 18.1 mph at the crosswalk, a 10-mph drop. The changes in speed were associated with large shifts from the high end of the speed distribution.

Annotated Literature Review A-17 Kamyab et al. 2003 Kamyab et al. examined the impact of various traffic-calming treatments in rural areas with high pedestrian volumes in Minnesota. The authors collected speed data, including average speeds, 85th percentile speeds, and speed-limit compliance, before and after the treatments were installed. In one area, removable pedestrian islands, Yield to Pedestrian in-street crossing signs, and multiple crossing signs (without islands) were installed. In a different location, a variable message sign that sent single- word messages (“SLOW”) to drivers traveling over the speed limit was installed in conjunction with short-term increased enforcement. After collecting the speed data, the authors found that less than half of the vehicles complied with the speed limit during the pretreatment conditions. At sites with the Yield to Pedestrian signs, speed-limit compliance increased from 30% to an average of 55%, and the mean speed decreased by approximately 5 mph. Engineering Countermeasure Combinations City of Alexandria n.d. Alexandria has completed neighborhood traffic-calming installations and collected at least 85th percentile speeds before and after installations. At least a brief evaluation was conducted at three sites. Russell Road—85th percentile speeds decreased by 21% in the southbound direction and 18% in the northbound direction. Curb extensions and neck downs or speed cushions and crosswalks were installed on Russell Road. West Abingdon Drive—85th percentile speeds decreased by 30% on this roadway from 35.2 to 24.7 mph. Two speed cushions and a crosswalk were installed as part of this project. Martha Custis Drive—85th percentile speeds decreased by 21% on this roadway from 30.4 to 23.9 mph. Two speed cushions were installed as part of this project. SDOT 2017a Seattle’s Department of Transportation has conducted a number of other before–after evaluations of rechannelization projects completed to reduce vehicle speeds and improve pedestrian safety. Four distinct 30-mph roadway conversions of four travel lanes to two travel lanes with one center two-way left turn lane, a bike lane, and improved pedestrian crossings all resulted in reduced speeds and reductions in crashes. Stone Way N: 85th percentile speeds decreased slightly (about 3 mph), and the percentage of motor vehicles exceeding the speed limit by at least 10 mph dropped by 75%. A comparison of crash data (2 years pre- and postrechannelization) revealed that there was an 80% drop in pedestrian crashes, a 14% drop in total crashes, and 33% drop in injury crashes. There was no change in the percentage of bike crashes; however, bicyclist volume increased by 35%. It also appears that traffic was not diverted to parallel neighborhood streets, which saw a reduction in ADT. NE 125th Street: 85th percentile speeds reduced by 8%. The percentage of people driving faster than 30 mph decreased by an average of 11%, driving faster than 35 mph decreased by 44%, and driving faster than 40 mph decreased by nearly 70%. It is unclear how long after the installation

A-18 Pedestrian Safety Relative to Traffic-Speed Management the post data were collected. The crash rate for all crashes and injury crashes per million vehicles declined by 10% and 17%, respectively, after the project was completed. NE 75th Street: 85th percentile speeds decreased by about 10% 1 year after the project. After the redesign, the percentage of speeders driving over the speed limit reduced significantly (56% in one direction, 64% in the other). The percentage of top end speeders (10+ mph over the speed limit) also declined significantly (75% and 79%). A pre- and postinstallation crash analysis revealed that pedestrian crashes decreased by 100% and total crashes by 45%. Note that preinstallation crash data included 3 years of data and postinstallation crash data included only 1 year. Nickerson Street: Approximately 6 months after installation, the 85th percentile declined from above 40 to 33 mph, the percentage of speeders declined by 63%, and the percentage of drivers speeding by more than 10 mph declined by more than 90%. Total crashes declined by more than 20% (5-year average predata compared to 1-year post). SFMTA 2015 SFMTA implemented its first “home zone” program near Marshall Elementary School (referred to as the “Minna-Natoma Home Zone Pilot Program”). The program included the installation of traffic-calming measures to improve the safety of all road users. The traffic-calming measures were installed in a two- by-two block radius around Marshall Elementary School. The treatments included speed humps, edgelines, two raised crosswalks, curb extensions, and a road diet for one street. Construction was completed in 2014, at which point SFMTA collected speed and volume data and distributed resident surveys. The data collected for the report (in 2010 and in 2014) included 24-hr vehicle speeds and volumes at five locations, Peak-hour bicycle volume counts, Peak-hour pedestrian volume counts, Resident surveys distributed to 460 area residents, Surveys distributed to 250 parents of Marshall Elementary School students, and Five-year vehicle collision data from before project implementation. Key findings from the study include On average, motor vehicle speeds decreased to below 20 mph throughout the home zone area. Perception of pedestrian safety with regard to vehicle yielding/stopping improved. Pedestrian volumes in the project area increased by an average of 20%. Bicycle volume on 15th Street between South Van Ness and Mission Street increased 6%. Motor vehicle volumes increased significantly, but increases are likely associated with residential developments concurrent with the home zone project. Perception of safety was strongly influenced by social characteristics of the neighborhood, including homelessness and illicit activity. The prevailing speeds on Minna and Natoma alleys, previously over 20 mph, decreased to below the 15- mph speed limit. Capp Street speeds reduced by 9 mph (from 26 to 17 mph), and speeds reduced from 25 to 21 mph on 15th Street. Resident and parent survey results show that more people feel that motor vehicles stop more frequently for pedestrians and pedestrian counts show more people are walking in

Annotated Literature Review A-19 the neighborhood. However, surveys also show continuing public health concerns in the area that diminish overall feelings of safety and likely affect the transportation modes that people choose. A collision analysis has not yet been conducted; post installation crash data were unavailable when the report was written. Additional details from the parent survey are available in the report. See also Corkle et al. 2001, Gitelman et al. 2017, and City of Cambridge 2012. Enforcement Countermeasures Goodwin et al. 2015 This document presents a summary of evidence-based countermeasures that improve roadway safety. The section of this document on speed-related countermeasures that improve pedestrian safety explains that there is evidence to support the reduction of speed limits as an effective method for reducing travel speed. However, the authors indicate that the reductions in travel speeds associated with speed-limit reductions alone are small (1 to 2 mph for every 5-mph speed-limit reduction; Leaf & Preusser, 1999). Combining speed-limit reductions with media and outreach campaigns and highly visible enforcement programs may lead to larger reductions in vehicle speeds. The authors suggest that speed-limit programs will be most effective when implemented in a small area as part of a highly visible areawide change, for example, the creation of a pedestrian safety zone accompanied by a speed-limit reduction, media coverage, increased enforcement, streetscaping, and increased lighting. Cunningham et al. 2008 Cunningham et al. examined the impact of an ASE camera campaign in Charlotte, North Carolina. The campaign included the use of three mobile cameras along 14 corridors from 2000 to 2005. The authors examined changes in the number of crashes, mean speeds, median speeds, 85th percentile speeds, and top-end speeders by comparing results from the sites with cameras to comparison sites without cameras. Speed data were collected before the program started, just after it began, and more than one year after it began. The findings from the analysis suggest that the cameras reduced crashes along the corridors where they were used, and top-end speeders and mean, median, and 85th percentile speeds all decreased at sites with cameras. The decreases in speeds at sites with cameras were statistically significantly lower during both “after” measurement periods compared to the before period (p < 0.0003 for all periods and speed measurements except the second after measurement for top-end speeders which was p < 0.03). Among mean speeds, median speeds, and 85th percentile speeds, the second “after” measurement period had lower average speeds than first “after” measurement period, but the differences were not statistically significant. Gains et al. 2004 and 2005 These reports summarize findings from the installation of speed and red-light cameras in London. Only locations where the cameras were operating for at least one year were included in the analysis. After the cameras were installed, average vehicle speeds at camera sites decreased by 6%. At sites with stationary cameras, there was a 70% reduction in vehicles breaking the speed limit, and at mobile sites the percentage of vehicles breaking the speed limit reduced by 18%. The percentage of vehicles exceeding the speed limit by at least 15 mph decreased by 91% at fixed camera sites and 36% at mobile camera sites. Overall, fixed camera types were associated with much greater decreases in average and 85th percentile speeds and numbers of vehicles breaking the speed limit or excessively speeding than

A-20 Pedestrian Safety Relative to Traffic-Speed Management mobile cameras. In urban areas, the percentage of drivers exceeding the speed limit decreased by 33%, and the percentage of drivers exceeding the speed limit by at least 15 mph decreased by 56%. Fixed urban cameras were associated with an immediate and sustained decrease in vehicle speeds. The report indicates that fixed urban cameras reduced average speeds by 18% over the course of five postinstallation speed measurements. Crash data from 4,100 sites were obtained to evaluate the impact of the cameras on crashes. There was a 22% reduction in personal injury crashes at sites after cameras were installed, although this does not account for potential regression-to-the-mean effects. In addition, there was a 32% reduction in the number of children killed or seriously injured and a 29% reduction in the number of pedestrians killed or seriously injured across all camera sites. More than 3,500 sites included locations with pedestrian crashes. Among those sites, fixed cameras were associated with a 34% decrease in pedestrian fatalities and serious injuries, compared to 25% among mobile camera sites. There was a 22% reduction in all pedestrian crashes that resulted in a pedestrian injury at fixed camera sites, compared to 24% among mobile camera sites. The authors indicate that the automated enforcement program received a large amount of public support, with 82% of people surveyed indicating that they agree that safety cameras should be used to reduce fatalities. Tang 2017 Tang reviewed more than 10 studies of speed enforcement cameras conducted between 1997 and 2016 in the United States, Europe, and Australia and found that speed enforcement cameras effectively reduced vehicle speeds. Three European studies found that speed cameras were associated with reductions in average vehicle speeds ranging from 4 to 6 mph (Li et al. 2015 citing ARRB Group Project Team 2005; Mountain et al. 2004; and Gains et al. 2004 and 2005). Using Poisson regression, Tang observed that ASE cameras are effective across roads with all speed limits, but they are generally more effective at improving safety on higher-speed roads. For example, the impact is greater on 30-mph roads compared to 20-mph roads, and it is greater on 40-mph roads compared to 20- or 30-mph roads. After reviewing data from more than 3,000 sites with cameras, Tang (2017) found that the effectiveness of speed cameras is localized and does not spill over into other areas. Specifically, Tang found that the effects of the speed cameras were strongest close to the cameras and were limited to distances of up to 1,640 ft (500 m) from the cameras. Beyond 2,300 ft (700 m), the speed cameras had no effect. Li et al. 2015 Li et al. used 8 years of crash and traffic count data and a before–after Empirical Bayes methodology to measure the impact of ASE cameras in Edmonton, Canada. The study examined only midblock crash data and was able to account for potential regression-to-the-mean effects. The results showed consistent reductions in different crash severities, ranging from 14% to 20%, with the highest reductions observed for severe collisions. Speed-related crashes reduced by nearly 19% (p < 0.05). The study also compared the effects of continuous and discontinuous enforcement strategies on different arterials and found that continuous enforcement was associated with a greater reduction in crashes of all types than discontinuous enforcement. The findings also suggest a spillover effect of enforcement on other approaches that are not experiencing enforcement.

Annotated Literature Review A-21 Freedman et al. 2006 Portland conducted an evaluation of ASE cameras installed in five school zones during a 2-month period. During the study, ASE was deployed at each school zone an average of two to three times per week. The program was well publicized through a public information and education campaign that was conducted before and during the demonstration. Unlike many of these types of studies, Portland compared results of the five test sites to five comparison sites where ASE was not implemented. Speed reductions at the test school zones were deemed attributable to the ASE program because speeds at the comparison sites remained unchanged throughout the study period. Key findings from the study include Mean and 85th percentile speeds at test school zones were reduced by approximately 5 mph when ASE was present, and ASE still had an effect (although reduced to 1 to 2 mph) when it was not present. The proportion of traffic that exceeded the speed limit by more than 10 mph was reduced by about two-thirds when ASE was present and by about one-quarter when ASE was not present. Maximum speed reduction was obtained with the combination of ASE and a flashing beacon, which is used during certain hours in many Portland school zones. The speed-reduction effects observed at the demonstration school zones were still present one month after ASE operations ended (May 2005). Hu and McCartt 2016 The authors compared travel speeds at sites with cameras from 6 months before the program began to 7.5 years after. In addition to the pre–post comparison, speeds from sites with cameras were also compared to control sites in Fairfax and Arlington Counties. A telephone survey of Montgomery County drivers was conducted in Fall 2014 to examine attitudes and experiences related to ASE. A logistical regression of crashes from 2004 to 2013 was used to examine the program’s effects on the likelihood that a crash involved an incapacitating or fatal injury on camera-eligible roads and on potential spillover roads in Montgomery County, using the crashes in Fairfax County on similar roads as controls. Nineteen sites were included in the final analysis, all of which were on residential streets with speed limits of 25 to 35 mph. About 7.5 years after the program began, speed cameras were associated with a 10% reduction in mean speeds and a 62% reduction in the likelihood that a vehicle was traveling more than 10 mph above the speed limit at camera sites. Speed cameras alone were associated with a 19% reduction in the likelihood that a crash resulted in a fatal or severe injury. The corridor approach was associated with an additional 30% reduction over and above the individual cameras. Mean speeds decreased by 13% from 2006 to 2014, compared to 5% and 4% at potential spillover sites and control sites, respectively. The percentage of vehicles exceeding the speed limit by more than 10 mph decreased by 64% compared to 39% and 43% at potential spillover sites and control sites, respectively. The results of the driver survey indicated that more than half (62%) supported the use of the speed cameras. Declines in the mean speed and the proportion of speeding vehicles at the camera sites were much larger than declines at potential spillover sites. The study findings indicate that the program was associated with spillover benefits; however, additional research is needed to confirm the magnitude of the potential spillover benefits due to lack of baseline speed data.

A-22 Pedestrian Safety Relative to Traffic-Speed Management Soole et al. 2013 Soole et al. conducted a literature review of average speed enforcement programs. The authors found that despite the variety of study types and methodologies, there is a large body of evidence to support the use of average speed enforcement as a means of reducing vehicle speeds and increasing roadway safety. It is well documented that average speed enforcement programs can reduce average vehicle speeds, 85th percentile speeds, and the proportion of speeding vehicles. Average speed enforcement programs also appear to be politically feasible, with many drivers reporting that it is a fair approach to enforcement. While less researched, there are other benefits that can be associated with average speed enforcement programs, such as improved traffic flow and increased roadway capacity. There is little evidence to suggest that the benefits of speed enforcement spill over into areas outside of the immediate vicinity of the enforced area. High-Visibility Enforcement See Goodwin et al. 2015. Walter et al. 2011 Walter et al. evaluated the impact of Operation Radar, a 4-week high-visibility enforcement campaign that took place along a 6-mi corridor in London. The campaign was conducted during the daytime on weekdays from May 6th to May 30th in 2008. During the campaign, two shifts of six police officers and one sergeant were deployed per day. The media campaign consisted of press releases from the police department, and two newspaper stories, and images and slogans related to the campaign were placed on billboards in the area. The billboards also explained that there would be increased police presence. Signs were posted to tell drivers about the number of offenders already caught during the campaign. After analyzing the speed data collected during the campaign, Walter et al. concluded that the changes in speeds were minimal. Eighty-fifth percentile speeds decreased by less than 2 mph during the campaign and less than 1.5 mph up to two weeks after the campaign, relative to baseline speeds. The percentage of drivers traveling above the speed limit decreased from 57% to 48% in one area, 76% to 72% in another area, and 37% to 30% in another area. Policies, Programs, and Legislative Efforts Speed-Limit Reduction Jurewicz et al. 2016 The authors reviewed a few studies on the relationship between impact speed and pedestrian and vehicle occupant risk of injury and determine critical impact speeds for a Safe System approach. The authors suggest that 20 km/h (12 mph) is a critical impact speed for vehicle–pedestrian crashes. The term critical impact speed refers to the point at which 10% of road users would suffer an MAIS3+ injury. See also Leaf and Preusser 1999, Kroyer, et al. 2015, Tefft 2013, and Rosén and Sander 2011. Slower Speed Zones Wegman et al. 2005

Annotated Literature Review A-23 This report summarizes the impacts of The Netherland’s sustainable safety program, with particular attention paid to the 30-km/h (19-mph) zone program. The program began in the 1990s and by 2003 there were about 30,000 km of 30-km/h (19-mph) roads. The program included the installation of speed humps, speed tables, and speed cushions. An evaluation of crash data revealed a reduction of 10% for the fatality rate per 1,000 km (621 mi) of road length and 60% for the in-patients rate per 1,000 km (621 mi) of road length. Grundy et al. 2009 Grundy et al. analyzed the impact of 20-mph zones in London of traffic safety. The analysis reviewed traffic injury data from 1986 to 2006 and nearly 120,000 road segments with at least one injury. Injuries in treatment areas were compared to adjacent, untreated areas using conditional fixed-effects Poisson models. The treated segments were associated with a 42% reduction in total traffic-related fatalities (95% confidence interval, 36.0% to 47.8%). The percentage reduction was largest in young children (46% in 20-mph zones compared to 5% in adjacent areas) and fatal or seriously injured children (44% in 20-mph zones compared to an increase of 5% in adjacent areas). Injuries to pedestrians were reduced by approximately 32% in 20-mph zones compared to 4% in adjacent areas. Fatal and serious injuries to pedestrians decreased by 35% in 20-mph zones compared to a 2% increase in adjacent areas. The authors found no evidence that injuries were displaced to other areas of the city. Gill 2007 This report evaluates the home zones for London project and discusses the impacts of the program in more than 30 different neighborhoods. The design of home zones varies, from woonerf-style treatments to traditional traffic-calming treatments. All seven initial home zone projects evaluated as part of this effort show that traffic speeds and volumes decreased after the projects were implemented. The impacts of an additional 39 design schemes were also evaluated. At more than half of the sites, speeds reduced by 10 to 15 mph, and three sites had average speeds below 10 mph. Hagan 2018 Hagan compared before–after data at neighborhood slow zones and control sites in London and New York City. He found that neighborhood slow zones did not reduce vulnerable-user fatalities. Hagan used only 2 years of postinstallation crash data and did not integrate volume data for pedestrians or vehicles. Hagan examined differences in fatalities, only; it is possible that New York City’s program affected pedestrian injury severity or the number of vehicle–pedestrian conflicts. Hagan also compared results from New York City and London, finding that London’s programs were strongly influenced by the use of a greater number and diversity of engineering treatments. New York City uses only speed humps, whereas London uses 12 types of physical treatments, including speed humps, speed cushions, raised crosswalks, curb extensions, and pedestrian refuge islands. In addition, London installed approximately 16 traffic-calming devices per mile, whereas New York City installed three devices per mile of slow-zone street.

A-24 Pedestrian Safety Relative to Traffic-Speed Management Li and Graham 2016 Li and Graham estimate the effect of 20-mph zones on traffic fatalities in London. Their sample included data from 1989 to 2007, including 234 treated zones and 2,844 randomly selected control zones. Zones within 150 m (492 ft) of the 20-mph zones were excluded from both groups to avoid including zones that may be outside of the 20-mph zones but still within their sphere of influence. The authors used propensity scores and outcome regressions to assess the impact of the 20-mph zones. Li and Graham found that the 20 mph consistently had statistically significantly reduced casualties, both in terms of absolute numbers and in terms of percentages. The number of pedestrian-related casualties decreased by 21% for all four methods used (p > 0.01). SYSTEMIC APPROACHES TO REDUCING SPEED LIMITS Automated Enforcement Legislation See Goodwin et al. 2015. Traffic-Calming Programs See City of Cambridge 2012 and City of Alexandria n.d. Comprehensive Programs Blomberg and Cleven 2006 A comprehensive traffic-calming campaign was implemented in six neighborhoods for 3 to 6 months. The program included the distribution of yard signs, pamphlets, and other educational materials to local residents. Enforcement patrols, including “warning stops” and tickets for speeding violations were also temporarily increased as part of the program. Speed tables or speed humps were added in two of the neighborhoods during the campaign. In three of the neighborhoods, pavement markings that created the illusion of physical traffic-calming treatments we also installed. The program was evaluated using police data, speed data, and a before–after survey of residents. Residents felt that the program effectively reduced speeds in their neighborhoods. Data from the speed measurements indicated that speeds decreased during the campaign on all but one street. The one street with no observed reduction in speeds had a low vehicle volume and a pre-existing speed hump and baseline condition speeds were well below 25 mph. On all but the one street, there was a significant reduction in mean speed and in the percentage of vehicles traveling 7 mph or more above the speed limit. Speed-limit compliance increased from 17% to over 117%. Blomberg et al. 2012 This study reviewed the impact of a Heed the Speed program in Philadelphia. The program included the installation of engineering treatments; increased enforcement, including the use of mobile speed trackers; and public education. Travel speeds were reduced at 17 of 24 measurement locations within the program area. A public survey found that few members of the community were aware of the increased speed-related enforcement or education campaign. Retting et al. 2008 Retting et al. reviewed speed data collected 6 months before and after radar speed cameras were installed in school zones in Montgomery County, MD. The authors also selected a set of control sites to compare with speed data collected at treatment sites. Retting et al. found that the percentage of drivers

Annotated Literature Review A-25 traveling at least 10 mph above the speed limit decreased by 70% at locations with both warning signs and speed cameras, by 39% at locations with warning signs, and by 16% on residential streets with neither speed cameras nor warning signs. Vision Zero Fleisher et al. 2016 Fleisher et al. discussed Vision Zero, a Safe System approach, and reviewed best practices from Vision Zero and other safety-related planning efforts to develop a Traffic Safety Best Practices matrix. The matrix presents a variety of strategies jurisdictions can use to help achieve their Vision Zero goals. The information in the matrix draws on strategies implemented both domestically and abroad and at the local and national level. Fleisher et al. use information from Countermeasures that Work, NCHRP 500 Report: Guidance for Implementation of the AASHTO Strategic Highway Safety Plan, and the Crash Modification Factors Clearinghouse to illustrate the effectiveness of each strategy, presented as a three- tier rating of either proven, recommended, or unknown. The matrix presents a number of planning, engineering, education, enforcement, policy, technology, and evaluation-related strategies and indicates whether specific cities or countries are using them. The cities and countries included in the matrix are San Francisco, New York City, Chicago, Portland, Seattle, Washington, D.C., Boston, Los Angeles, Sweden, the Netherlands, and London. The information in the matrix indicates that there is widespread adoption of many engineering strategies that focus on reducing speeds. Some potential Vision Zero strategies are being implemented by less than 40% of cities and countries included in the review. These include vehicle technology, vehicle safety, automated enforcement, publishing websites with safety data, and conducting routine evaluations of traffic safety initiatives. The latter is more common in the international community than in the United States. The authors conclude their review with a set of recommendations for United States Vision Zero efforts based on their review of effective international strategies. Greedipally et al. 2018 Greedipally et al. explored factors associated with the decline in traffic fatalities in the United States from 2008 to 2012. The authors used a log-change regression model to analyze the impacts of the different road characteristics on fatalities in the following year. The authors found that many of the factors associated with the decline in traffic fatalities had nothing to do with road or road user conditions—many of these factors were related to changes in the economy. For example, the largest contributing factors to the drop in fatalities was a large increase in teen unemployment, a reduction in beer consumption, and a reduction in gross domestic product/capita income. This study highlights the importance of looking at traffic fatality data in the context of larger demographic and economic trends.