In the U.S. during 1995, there were about 84,000 pedestrian injuries and 5,585 pedestrian fatalities (NHTSA, 1996), for an overall ratio of 15.0 injured pedestrians for every fatality. This ratio varied substantially as a function of posted speed limits. On roadways with posted limits of 25 miles per hour or less, there were 57.1 injuries for every fatality. The number of injuries per fatality dropped sharply as speed limits rose, showing an increase in typical crash severities. For posted limits of 30 - 35 mph, there were 19.3 injuries per fatality; for posted limits of 40 - 45 mph, there were 10.1 injuries per fatality; the ratio dropped to 3.0 injuries per fatality for posted limits of 50 - 55 mph and to just 0.3 injuries per fatality for posted speed limits of 60 mph or higher.

While posted speeds are not necessarily the same as travel speeds or impact speeds, the data clearly suggest a strong relationship between speed and the severity of resulting personal injury. Several foreign studies suggest that lowered speeds result in less severe pedestrian injuries and fewer injuries and fatalities – both through reduced collision intensity and through reduced numbers of collisions (slower-moving motorists can avoid entirely crashes that would have occurred if they were driving faster).

Although there is some U.S. literature available, most of the available literature on the relationship between speed, pedestrian crashes, and resulting injuries is from other countries. The foreign literature has not been organized into one data set, nor has it been reviewed from the point of view of applicability to the U.S.

Moreover, there remain the questions of how to reduce speeds and how to do it cost effectively. Techniques include reducing speed limits, increasing police enforcement, and re-engineering streets to make traffic move more slowly. The measures can be taken city-wide, in selected neighborhoods, or at selected times of the day.

This project had three objectives. First, to reaffirm and quantify the relationship between vehicle speeds and pedestrian crash severities through literature review and data analysis. Second, to describe techniques that have been used for reducing vehicle speeds and review their effectiveness. Third, to synthesize these results into recommendations for countermeasure programs to be tested in this country.

The results of the project are presented in the following four sections:

1. (This introduction)
2. Methods used in the crash data analyses and in the literature acquisition and review.
3. Vehicle Speed and Pedestrian Injury, a brief review of the literature relating vehicle speeds to injury severity and a review of U.S. crash data from the General Estimates System (GES) and from the Fatality Analysis Reporting System (FARS), as well as data from the state of Florida, which records vehicle travel speeds on their crash reports.
4. Speed Control Literature, divided into three broad topics: speed management including legislation, enforcement, and education; specific engineering techniques used to control speed, and general engineering approaches including traffic calming and other wide-area approaches.
5. Recommendations, for countermeasure approaches likely to be effective in this country.

The work on this project consisted of two distinct efforts, each with its own methodology.

First, analyses were conducted of existing crash record datasets. Three datasets were studied:

1. The General Estimates System (GES) is a nationwide probability sample of police-reported crashes on trafficways with all levels of severity (fatalities, injuries, and property damage only). GES is part of NHTSA’s National Automotive Sampling System (NASS). Each crash in the GES database is weighted based on its probability of having been sampled, and the combined weighted crash data form estimates of national crash figures.
   
   For these analyses, data from the years 1994 - 1996 were included. In those years, there were 6,171 pedestrians in 5,921 sampled crashes which, when weighted to account for sampling probability, represented an estimated 283,828 crash-involved pedestrians – nearly 95,000 per year – from 273,440 crashes.
1. State of Florida pedestrian crash data for the years 1993 - 1996 were obtained and analyzed. Florida data were selected because they included vehicle travel speeds for a large proportion of the crashes and because Florida has a relatively large population and, therefore, a large number of crashes. Statewide data are collected from police crash reports, which are filed whenever there is a personal injury or when alcohol use is suspected or an involved vehicle must be towed from the scene. There were 36,016 pedestrians involved in reported crashes for those years. In order to make the connection between vehicle speed and injury as directly as possible, only crashes involving one vehicle were retained for analysis. These included 32,651 pedestrians, nearly 91 percent of the total.
1. Data from NHTSA’s Fatality Analysis Reporting System (FARS) were examined for the years 1992 - 1996. The FARS database is an enumeration of all crashes on roads open to the public that result in the death of a vehicle occupant or a nonmotorist within 30 days of the crash. In those years, there were 27,934 pedestrian fatalities resulting from 27,450 crashes.

The purposes of these analyses were to relate pedestrian injury levels to striking vehicle speeds and to identify risk factors associated with higher-speed crashes. Two measures of striking vehicle speeds were used. The first, speeds estimated by the investigating officers, would be expected to be generally accurate, but they were present for relatively few crashes. The second, posted speed limits, were almost universally available, but they would likely be only general indications of actual speeds. These limitations allow valid general conclusions about the relationship of speed to injuries and crash conditions, but they prohibit deriving precise relationships.

While pre-crash vehicle speeds are very relevant to the conditions under which crashes occur, they are one step removed from the injury-causing event. Prior to the crash itself, the striking vehicle often, or usually, reduces its speed somewhat from its travel speed, either to perform a maneuver such as a turn or in an attempt to avoid or minimize the crash itself. Thus it was not possible to derive precise relationships between impact speeds and injury levels.

The primary method of analysis was based on crosstabulations, relating posted speed limits or, where available, vehicle travel speeds to injury severities. Because the likelihood of injury depends on the pedestrian’s age – most dramatically, older pedestrians are much more likely to be fatally injured – analyses were repeated for different pedestrian age groups.

It was also of interest to examine the relationship of other variables to the distribution of pedestrian injury severities. For the GES data, injury severity distributions were tabulated for pedestrian, crash, and vehicle characteristics. Also, general tabulations of the distribution of FARS fatalities across levels of pedestrian, crash, and vehicle characteristics were made.

The second major effort reported here was the review of literature related to vehicle speeds and crash results and to speed reduction and control strategies. There were three major steps involved.

First, searches of automated transportation reference databases available through the Transportation Research Board (TRB) were performed. The purpose was to identify relevant references – articles, books, and research papers – from around the world. Searches were made in three databases: TRIS (Transportation Research Information Services), with 310,000 records covering transportation-related publications from the U.S.; TRANSDOC, with 40,000 records on the social sciences of transportation published in European and associated countries; and IRRD (International Road Research Documentation), with 285,000 records covering all aspects of road research internationally. Two main searches were conducted: Articles on pedestrian fatalities in general, from 1992 on, and articles referencing pedestrian fatalities and vehicle rates of speed for all years. The first search yielded 271 possible "hits," and the second yielded 384. Other, more specialized, searches yielded another 20 or so references. Search results included article abstracts or summaries and names and addresses of publishers. The search results were combined with books, articles, and reports already in our possession.

Second, numerous articles were sought, from libraries, authors, and publishers. Contacts were initiated to acquire specific documents identified from the database search, and we also used the contacts to request other reports related to pedestrian safety and vehicle speed control. Sources contacted included, in the U.S., TRB, the Institute of Transportation Engineers (ITE), the Federal Highway Administration (FHWA), and researchers and traffic engineering practitioners. Foreign sources included individual authors and research organizations in Canada, Great Britain, France, Denmark, Austria, Finland, and South Africa. Additional countries represented in the research articles included Australia, Germany, The Netherlands, Greece, Norway, Sweden, Japan, Jordan, and Kuwait.

Third, discussions were held with researchers and practitioners in the U.S. and abroad. The discussions provided insights into research and applications programs, and in many cases the sources provided otherwise unpublished or interim materials.

The first part of this chapter reviews published studies relating vehicle speed to pedestrian injury severity. The second part presents analyses of three databases for the empirical relationship between speed limits and vehicle speeds and pedestrian injuries. The analyses also review other characteristics of the crashes and their relationship to vehicle speeds and pedestrian injury severities.

The idea that the faster a striking vehicle is traveling, the more damage is done to a struck pedestrian, is almost too obvious to require proof. Yet the relationship has been documented in a number of studies. Pasanen (1992) reviewed three studies relating collision speeds and pedestrian injury severity, finding their results quite consistent and that the probability of pedestrian death reached nearly 100% for speeds over 80 km/h (50 mph). Modeling the data from Ashton (1982), Pasanen estimated that about 5 percent of pedestrians would die when struck by a vehicle traveling 20 mph. The pedestrian fatality percentage would rise to about 40 percent for vehicles traveling 30 mph, about 80 percent for vehicles traveling 40 mph, and nearly 100 percent for speeds over 50 mph.

Numbers comparable to these are cited in a number of other references. For example, in the UK Department of Transport Traffic Advisory Leaflet 7/93 (TAU, 1993), figures quoted are, for 20 mph impact speeds: 5 percent death, 65 percent injured, and 30 percent uninjured; for 30 mph impact speeds: 45 percent death, 50 percent injured, and 5 percent uninjured; for 40 mph impact speeds: 85 percent death and 15 percent injured. The UK DoT values are illustrated in Figure 1.

In a separate effort, Pasanen (1993) took advantage of an intersection videotape surveillance system to examine ten pedestrian accidents in Helsinki. The speed of approaching vehicles ranged from 18 km/h (11 mph) to 62 km/h (39 mph), and Pasanen calculated the average probability of death as 0.16. Pasanen estimated that reducing approach speeds to a maximum of 40 km/h (25 mph), which would have slowed the vehicles in six of the crashes, would have eliminated two of the crashes entirely and reduced the estimated average probability of death to only 0.055. Pasanen also noted that all of the crashes occurred between pedestrians and free-moving vehicles, i.e., ones not part of and constrained by a queue of vehicles.

Anderson et al. (1997) projected the effects of lowered vehicle travel speeds in 176 fatal pedestrian crashes occurring in 60 km/h (37 mph) zones in Adelaide, Australia. In the actual crashes, impact speeds ranged from less than 10 km/h (6 mph) to about 100 km/h (62 mph), with a median of about 55 km/h (34 mph). Projected savings in fatalities ranged from 13 percent, assuming that all drivers obeyed the existing speed limit, to 48 percent, assuming all drivers traveled 10 km/h (6 mph) slower. Anderson et al. also developed curves relating probability of fatal injury with vehicle impact speed, based on data from the Interdisciplinary Working Group for Accident Mechanics (1986) and Walz et al. (1983). Anderson’s curve showed about 5 percent of pedestrians would die if struck by a vehicle at 25 km/h (16 mph), about 25 percent if struck at 40 km/h (25 mph), and about 80 percent if struck at 50 km/h (31 mph). Although the curve is a bit higher at lower speeds than Pasanen’s estimates, the general patterns are in quite good agreement.

Wazana et al. (1997) reviewed articles identifying risk factors for child pedestrian injuries. Two studies showed that higher speed limits were associated with higher risk of injury to child pedestrians. In New Zealand, Roberts et al. (1995) found an odds ratio of 3.22 for injuries on 40-49 km/h (25-30 mph) roads vs. roads with lower limits, and in the Seattle area Mueller et al. (1990) found odds ratios of 3.2 for child pedestrian injuries in 45-55 km/h (28-34 mph) zones and 6.0 for roads with limits above 64 km/h (40 mph) (vs. roads with speed limits below 40 km/h (25 mph)).

Pitt et al. (1990) found a strong relationship between vehicle speed and pedestrian injury when they reexamined NHTSA’s Pedestrian Injury Causation Study (PICS) data, which was developed by investigative teams in five U.S. cities that examined police-reported pedestrian crashes between September 1977 and March 1980. Looking only at about 1,000 urban crashes with pedestrians younger than 20 years of age, Pitt examined cases based on fatality or Injury Severity Scores (ISS) of 16 or higher ("serious" injuries). Compared to vehicle travel speeds of 10-19 mph, the risk of serious injury (or death) was 2.1 for speeds of 20-29 mph, 7.2 for speeds of 30-39 mph, and 30.7 for speeds of 40 mph or more. A similar positive relationship was seen when injury severity was compared with posted speed limits, though it was weaker.

Harruff et al. (1997), by contrast, found no relationship between severity of injury and vehicle speed. However, they were looking only at fatalities in an urban area, and their indicators of speed were posted speed limit, which were not related with injury patterns, and roadway type. For the latter, pedestrians killed on thoroughfares (major roads) showed somewhat more serious injuries. Without a full range of injury levels, though, and with only indirect indicators of vehicle speed, the negative finding is not surprising and does not necessarily contradict the other findings.

In Denmark, general speed limits were introduced in 1974. They resulted in reductions of average vehicle speeds of 6 km/h (4 mph) and speed standard deviations by 3 km/h (1.9 mph), and pedestrian crashes (of all severities) dropped by 25 percent. In 1985, the urban speed limit was lowered from 60 km/h (37 mph) to 50 km/h (31 mph), and average speeds dropped by about 2-3 km/h (1.2-1.9 mph). At the same time, pedestrian fatalities dropped by 31 percent, serious injuries by 4 percent, and slight injuries by 9 percent. For all reported pedestrian crashes between 1986 and 1995, pedestrian injury severity distributions were plotted by speed limits. Results ranged from no fatalities (20 km/h (12 mph) or less) to 5 percent fatalities (50 km/h (31 mph)), 20 percent fatalities (80 km/h (50 mph)), and 35 percent fatalities (110 km/h (68 mph)). The probability of a fatality as a function of speed limit showed much less variability than the probabilities cited above as functions of actual vehicle speed. This suggests that, even for high vehicle travel speeds, enough speed can be reduced to bring most collisions into survivable-speed areas. It may also be that the impacts are often less than full head-on, which would reduce the severity of the impact and thus the injury. (All results cited in Jensen, 1998.)

Numerous additional European studies exist on the effects of "traffic calming" changes on crash reductions and pedestrian safety. They are reviewed in Chapter IV.

The General Estimates System (GES) database is a probability sample of police-reported crashes with all levels of severity (K (fatal), A (incapacitating), B (non-incapacitating), and C (minor) injuries as well as property damage only). Each crash in the GES database is weighted based on its probability of having been sampled, with these weights used to make national crash estimates. In the years 1994 through 1996, there were 5,921 pedestrian crashes in the database which involved a total of 6,171 pedestrians. Weighted, the crashes represent projected national figures of 273,440 crashes involving 283,828 pedestrians (across all three years). The projected totals include 9,546 fatalities, 64,076 A-level injuries, 88,700 B-level injuries, 96,162 C-level injuries, 14,935 uninjured, and 10,409 (3.7 percent) injuries of unknown severity.

GES results are reported without statistical significance testing. Although most of the results are based on crosstabulation distributions, the values in the cells are not raw frequencies but are projected frequencies based on a small number of actual cases and case-by-case weights. Tests of statistical significance require techniques such as those in the SUDAANÓ software programs (Shah et al., 1995), which were not available for this analysis.

The GES database includes the variable of vehicle speed, which provides the strongest evidence of the relationship between injury severity and speed. The results are shown in Table 1. Relatively few of the struck pedestrians were killed when the vehicle’s pre-crash travel speed was 35 mph or lower, but vehicles traveling at 36 - 45 mph killed about 16 percent of the pedestrians and vehicles traveling at 46 mph or above killed about 35 percent of the pedestrians.

Note that the speed value is for pre-crash motion, which will only be a rough estimate of the true impact speed. In most cases, the reported speed refers to the vehicle’s speed on approach to the crash. If the driver was able to attempt evasive or stopping maneuvers, the actual impact speed could have been considerably slower.

Although actual vehicle speeds are the most important values for relating injuries to speeds, they are missing for more than three-fourths (77 percent) of all the crashes in GES. (Travel speeds were more likely to have been recorded for fatality or A-injury crashes.) Additional analyses based on posted speed limits, which were known for nearly all reported crashes, were conducted. Posted speed limits are less accurate indicators of the travel speeds of vehicles in pedestrian crashes, but knowing them for nearly all crashes makes such analyses a useful complement to analyses based on estimated travel speeds.

Table 2 shows the distribution of injuries for pedestrians with known injury severity as a function of speed limit. The relationships are similar to those seen for actual travel speeds, but they show less drastic swings. For the lowest speed limits, just over one percent of struck pedestrians were killed. At speed limits of 50 mph or higher, 22.2 percent of the pedestrians struck were killed. In addition, the percentage of seriously injured pedestrians rose, from 15 percent (for speed limits up to 20 mph) to 31 percent and 26 percent (for speed limits of 40-45 mph and 50+ mph, respectively).

Table 3 examines the relationship between pedestrian injury and speed limits for crashes in which the vehicle was coded as going straight ahead. Overall, the proportion of fatally or seriously injured pedestrians was slightly higher than in Table 2. Also, the distribution of injury levels was somewhat more varied over speed limit. Fatalities increased from less than 2 percent for speed limits of 25 mph or less to 5 - 6 percent at speed limits of 30 - 35 mph, nearly 10 percent at 40-45 mph and to 24 percent at 50 mph or higher. The proportion of non-incapacitating injuries, minor injuries, and no injuries decreased steadily with increasing speed limits.

Related Crash Descriptors

Roads with different speed limits occur in different kinds of areas and attract different mixes of pedestrians and vehicles. To explore the way in which other crash descriptors might be related to pedestrian injury severity and speed limits, the distribution of different variables across speed limits was examined. The results are summarized in Table 4. Moderate differences are seen for most variables. (For example, males were struck somewhat more often on higher speed roads, and weekend crashes also occurred somewhat more often on higher speed roads.) Larger differences included:

• Young pedestrians (0 - 14 years old) were much more likely to be struck on roads with speed limits up to 25 mph, much less likely to be struck on roads with speed limits of 40 mph or more.
• High speed roads (50+ mph) were most often the site of late night (midnight and later) crashes; 40 - 45 mph roads saw more late evening crashes (8 p.m. – midnight) than ones at other times; and the lowest-speed roads were more likely to be the sites of midday crashes.
• Not-in-road crashes (ones on shoulders, medians, roadsides, etc.) were much more likely on roads with the highest speed limits.
• Smaller roadways (including two-lane halves of divided highways) more often had very low or very high speed limits.
• Intersections with traffic signals most often had moderate speed limits (25-45 mph); most of the high-speed-limit crashes were not at intersections.
• About twice the proportion of bad-weather crashes occurred on roads with speed limits of 50 mph or more, as compared to good-weather crashes.
• The most rural areas had more of the roads with highest and the lowest speed limits.
• Large striking vehicles, including buses and tractor-trailers, were more often found on the highest speed roads and less often on the roads with very low speed limits.
• Few turning vehicles were on roadways with speed limits of 50 mph or more, but a very large number of backing vehicles were on such roads.

In Florida in 1993 - 1996, 32,651 (91 percent) of the pedestrians in crashes were in single-vehicle crashes. These incidents were analyzed for the relationship between vehicle speed and pedestrian injury. For the pedestrians in single-vehicle crashes, 23,831 (74 percent) were struck by vehicles with estimated travel speeds and 21,864 (67 percent) were struck on roads with recorded speed limits. Of the pedestrians, 31,354 (96 percent) had known ages and 32,506 (99.6 percent) had known severity of injury (including no injury). The tables, figures, and text below are based on known data.

As noted earlier, younger pedestrians are generally more able to resist serious injury and death, while elderly pedestrians are much more susceptible to more serious consequences as crash victims. The distribution of fatality rates is shown in Table 6 and illustrated in Figure 2. Overall, pedestrians age 65 and older are more than 5 times as likely to die in crashes than pedestrians age 14 or less, and the likelihood of death increases steadily for ages in between. For vehicle travel speeds above 45 mph, pedestrians above age 65 die in about 5 of 8 crashes.

The pattern of differences varies by vehicle travel speeds. For speeds less than 20 mph, risk of fatality is about the same for all ages up to age 65, where the rate triples. For speeds of 21 - 30 mph, fatality rates are roughly the same to age 45, but the rate is about 2.5 times higher for pedestrians age 45 - 64 and doubles again for pedestrians age 65 and older. By speeds of 31 - 35 mph, the fatality rate of 25 - 44 year olds is nearly double the rates of younger pedestrians; by speeds of 36 - 45 mph the fatality rate for pedestrians 15 - 24 years old exceeds that of the youngest pedestrians, and the fatality rate for each succeeding age group is greater than that of the younger group.

Posted speed limits and pedestrian injuries

There was a strong relationship between the speeds of crash vehicles and the speed limits that were posted. Over 90 percent of striking vehicles were reported as traveling at their speed limit or slower, and half or more were reported as traveling 5 mph or more below their speed limit.

The relationships between posted speed limits and pedestrian injuries are shown beginning with Table 8. The percentage of pedestrians who were killed rose from less than 1 percent for speed limits of 20 mph or less to 28 percent for speed limits of 50 mph or more. Percentages of pedestrians receiving A injuries rose to a plateau of about 35 percent for speed limits of 35 mph or more. Pedestrians with B injuries remained at just over 40 percent for speed limits up to 30 mph and then dropped steadily with increasing speed limits. Pedestrians with minor (C) injuries or no injuries dropped steadily, from more than 40 percent at the lowest speed limits to just 12 percent at speed limits of 50 mph or more.

The Fatality Analysis Reporting System (FARS) is not suited to providing direct information on the relationship between speed and injury severity, since it looks only at fatalities. However, the FARS data can provide extensive information about the characteristics of the most serious pedestrian crashes. By relating these characteristics to vehicle speeds and speed limits, it is possible to identify the situations or conditions most strongly associated with pedestrian fatalities. This, in turn, can benefit countermeasure development by identifying factors that may be causal in the crashes and by limiting the countermeasure focus to the situations of greatest danger.

The distribution of pedestrian fatalities across travel speeds is shown for different pedestrian ages in Table 10. For all ages, the proportion of fatalities increases with speeds above 25 mph to about 45 mph. At speeds of 46+ mph, the proportion of fatalities for pedestrians between 15 and 44 years of age increases sharply, while it drops somewhat for younger and older pedestrians, possibly reflecting different exposure patterns on the highest-speed roadways.

• Males show relatively few fatalities at low travel speeds and speed limits, steadily increasing involvement with increasing speed/limits until 37% (38%) of pedestrian fatalities occurred with travel speeds 46 mph and above (speed limits of 50 mph and above). Females showed a much broader distribution across speeds and speed limits.
• Pedestrians of ages 15 - 44, as described above, showed sharply increasing involvement with increasing speeds and speed limits; the distribution was much broader for other age groups.
• Crash distributions with half or more occurring on sites with speed limits of 50 mph or more and travel speeds of 46 mph or more include: Pedestrians age 15-24; midnight - 5:59 a.m.; not in roadway; weather conditions of snow, ice, other unusual or unknown (not clear or rainy); rural locations; dark and not lighted; and two or more vehicles. For these conditions, very few of the fatalities occurred at locations with low speeds or speed limits.
• Crash distributions with 40 percent - 49 percent high-speed site occurrence include: Pedestrians age 25-44; pedestrian BACs of .10% or more; Saturday or Sunday; and pedestrians struck by trucks or buses. Non-intersection crashes just miss this category.
• Crash distributions which include fewer than 30 percent of the crashes at high-speed locations, and relatively more at low-speed sites, include: Female pedestrians and pedestrians age 14 or less or 65 or more; pedestrians of unknown (often, untested) BAC; daytime crashes, between 6 a.m. and 7:59 p.m.; crashes at intersections; crashes in the rain; urban areas; daylight and dark-but-lighted settings; and drivers age 75 and above.

Factors that seem relatively unrelated to the speed of striking vehicles or of roadways included roadway width and driver sex.

Taken together, the GES, Florida, and FARS crash data strongly support the relationship that higher speeds for vehicles striking pedestrians result in more serious consequences. The GES data and the FARS data were further analyzed to show some of the other characteristics of the crashes that were associated with more severe injuries and with higher vehicle speeds and higher speed limits. These factors will be relevant to the development of countermeasures to the consequences of high impact speeds that are presented in Chapter V.

Speed control is a very broad term which can cover any mechanism used to limit or reduce vehicle speeds. In the U.S., speed control has traditionally emphasized reduced speed limits and enforcement on continuous segments of roadway, and the installation of stop signs or traffic signals at intersections. Education, in the sense of informing the public of the dangers of excessive speed and the likely presence of police enforcement, has also been used. Increasingly in the U.S. and commonly in Europe, Australia, and Canada, roadways and intersections have seen engineering changes designed to encourage or require drivers to reduce their speeds. Engineering approaches are often given the general title of "traffic calming," and the title is also applied to plans that combine engineering changes with public information and education in order to affect whole neighborhoods, towns, or cities.

This chapter is divided into three sections. The first deals with speed management through means other than traffic engineering. The most common technique for this is using police enforcement along with posted speed limits, and this has received most of the research on effectiveness. One perspective makes speed enforcement a battle between police and speeders, and research on radar detectors, a prime weapon in the battle, is discussed. Finally, other speed management techniques are reviewed.

The second section reviews engineering-based approaches to speed management. While they often include aspects of speed management discussed in the first section of the chapter, the engineering-related approaches always include some physical change to the roadway or road environment intended to cause drivers to proceed more slowly. The final section of this chapter looks at the topic of traffic calming as it applies to wide area schemes for managing traffic and speeds.

The most straightforward approach to speed management is, if you want to slow people down, lower the speed limit. This tends to be only marginally effective. Reducing speed limits reduces speeds by, at best, about one-quarter of the speed limit reduction. However, several European studies which examined the broad implementation of lower urban speed limits showed that lower limits could be well accepted by all road users and that they could reduce speeds, to some extent, and reduce crashes and injuries.

As described in Chapter 3 (Jensen, 1998), in Denmark, general speed limits were introduced in 1974, and urban speed limits were reduced in 1985. In both cases, measured speeds came down (as did overall crashes and injuries and pedestrian crashes and injuries). When urban speed limits dropped 10 km/h (6 mph), corresponding speeds dropped 2 - 3 km/h (1.2-1.9 mph), consistent with ratios found elsewhere.

Johansson (1996) studied the long-term effects of reducing the speed limit from 110 km/h (68 mph) to 90 km/h (56 mph) on Swedish trafficways in June 1989. He analyzed monthly data from January 1982 through December 1991. Statistical models which included seasonal factors showed that fatalities, serious injuries, minor injuries, and vehicle-damage-only crashes all declined after the speed limit change; the reduction was statistically significant only for minor injuries and property damage crashes. He did not provide any actual speed data.

In France, the basic urban speed limit was decreased from 60 km/h (37 mph) to 50 km/h (31 mph) in November 1990. According to Page and Lassarre (1994), although there was a general public information campaign to introduce the change, the different cities and regions implemented the signage supporting the regulation gradually and on their own schedules. Enforcement was not increased for the new law. On main roads in small towns, between 1990 and 1992 speeds of cars decreased slightly during daytime (from 65 km/h (40 mph) to 60 - 63 km/h (37-39 mph)) but speeds were unchanged at night (about 74 km/h (46 mph)). For the entire country, injury crashes in urban areas decreased 14.5 percent from the two years before the speed reduction to the two years immediately after; in rural areas, where speed limits did not change, the decrease was 9.1 percent. Over the same period, fatalities decreased 12.3 percent in urban areas and 10.2 percent in rural areas. The effect was most prominent in urban areas of less than 5,000 inhabitants. The authors felt that the results, somewhat less than expected, might improve as areas better understand and implement the speed regulation (which also provides for 70 km/h (44 mph) arterials and 30 km/h (19 mph) neighborhood roads).

In Graz, Austria, a city-wide 30 km/h (19 mph) limit on all residential streets (50 km/h (31 mph) on through "priority" streets) was implemented in September 1992 (Sammer, 1997; Pischinger et al., 1995). The change was implemented in response to increasing requests from citizen groups to participate in an area-by-area traffic calming scheme that was gradually including more areas over a ten-year period. The comprehensive areawide program included traffic regulation (signs, roadway markings), extensive and varied public information and awareness campaigns, and supervision (enforcement and speed display boards). About 75 percent of all roads became 30 km/h (19 mph). Injuries decreased from the year before the change to the year after. Minor injuries declined 12 percent, serious injuries dropped 24 percent, and all pedestrian injuries fell by 17 percent. Economic savings from the injury decreases were calculated to be about $6,000,000, a 26 percent drop. Mid-block average and 85th percentile speeds dropped immediately, then gradually recovered to a level slightly below pre-law speeds. Intersection speeds also dropped, by 2.5 km/h (1.5 mph) on average, and the proportion of extreme speeds dropped sharply. Drivers exceeding 50 km/h (31 mph) dropped from 7.3 percent "pre" to 3.0 percent "post." Surveys showed that approval of the reduced speed limits increased steadily after implementation, reaching 68 percent after 18 months for private car drivers, who were the least enthusiastic group throughout. Noise levels, measured on 30 km/h (19 mph) streets, decreased; overall air pollution did not change.

Moving to the U.S., Casey and Lund (1992) showed that the effects of increasing speed limits can extend beyond the roads on which speed limits are raised. In 1987, some California highways had their speed limits raised from 55 mph to 65 mph. Casey and Lund had studied speed adaptation – the tendency to drive faster on a medium-speed road after coming off of a high-speed road – while California’s maximum speed limit was 55 mph. They repeated the study after some roads, but not those at the test sites, had speed limits raised to 65. They found that: speeds on the still-55 mph freeways had increased, by 2 - 5 mph; speeds on adjacent roads had also increased, for speed-adapted drivers and non-adapted drivers; and the differential between adapted and non-adapted drivers remained. They concluded that raising speed limits in some locations may, through "some indirect process of speed generalization," also increase speeds elsewhere.

Parker (1997) described a study in which states and localities lowered and raised speed limits on short roadway segments. Sites included urban and rural community roads and rural roads. Speed limits were raised or lowered (only one change per site) by 5, 10, 15, or 20 mph. Actual speeds changed significantly, but only by as much as 1.5 mph. Crash rates did not change significantly, possibly because of the limited numbers of crashes overall but also possibly due to the very small change in mean speeds.

Police enforcement of speed limits has been a primary tool to reduce speeds, but it is a technique with long-recognized limitations. Driver expectations are key. According to general psychology guidelines, speed enforcement will have the greatest effects on drivers if it is: 1) believed likely to occur, 2) meaningfully costly to the offenders, 3) associated with driving in general rather than any specific time of day or roadways, and 4) not associated with any specific cues that signal the presence or absence of enforcement efforts.

For enforcement to have continuing effects, the first point is essential: drivers must have a continuing expectation that enforcement will occur. This is a basic weakness with nearly all real-world enforcement schemes. Bjørnskau and Elvik (1992) describe the enforcement system in game-theoretic terms, where driver behavior is influenced by the enforcement and the enforcement program is affected by driver behavior. Simply, higher enforcement will lead to reductions in driver violations, leading to reductions in arrest "benefits" to the enforcers, leading to a reduction in enforcement (as assets are diverted to now-higher priority problems), leading to decreased expectations of arrest, leading to increases in driver violations, leading to higher enforcement, ... etc. Reviewing a number of studies on enforcement, the authors make several points: 1) large increases in enforcement do reduce the violation rate and can also reduce crash rates; 2) road users modify their behavior according to the enforcement pattern as they understand it, i.e., less or no change outside the "danger zone" for enforcement and quick reversion to pre-enforcement behaviors once the enforcement is restored to normal; 3) enforcement agencies focus enforcement efforts on perceived problems; 4) the authors have no documentation of enforcement cutbacks because the program was judged successful, but such cutbacks are quite plausible; 5) stiffer penalties (above an attention-getting threshold) do not bring down violation rates; and 6) permanent surveillance can produce permanently low violation rates. A recommendation, though one not judged likely to be followed, is for enforcement to be allocated randomly according to a nearly permanent overall plan that ensures minimally effective results.

Specific studies are reviewed below.

One unique study examined the effects of a widely known, complete absence of enforcement, in Finland during a national police strike (Summala et al., 1980). During the two-week strike, mean speeds increased only slightly, but the percentage of speeding more than 10 km/h above the speed limit increased by 50 - 100 percent. This increased speed standard deviation by about 20 percent, likely increasing crash risks. Also, "suspicious-looking" cars parked beside the road, which ordinarily would have caused drivers to slow down, evoked no response during the strike.

Council (1970) reviewed earlier speed enforcement literature and studied the effects of stationary and moving marked police units on the speeds of oncoming vehicles, on two-lane roads with speed limits of 55 and 60 mph. He found that speeds were depressed by 5 - 6 mph alongside the stationary vehicle but that there was only minimal change alongside the moving vehicle. Speeds of vehicles 1.25 miles after passing the stationary marked car dropped slightly; speeds of vehicles 1.25 miles after passing the moving marked car were up slightly. No actual enforcement (stopping of motorists and issuing tickets) was done.

Dart and Hunter (1976) examined speeds at "treatment" points, two miles upstream, and two miles downstream on a two-lane rural roadway. Mean speeds alongside a partially concealed radar-equipped marked police car (with a visible "Speed Check" sign), a manned parked police car, and an "enforcement scene" of a police car with flashers activated parked behind an "arrestee’s" vehicle, all dropped by 5 - 6 mph compared to the same site with no treatment. Mean speeds alongside a visual speed indicator sign, "Your speed is ___" plus, for speeds over 55 mph, "Slow Down," were not significantly decreased compared to the no-treatment condition. In all three enforcement conditions, mean speeds had increased to recover about half the speed decrease by 1000 feet beyond the treatment, and by two miles downstream mean speeds in all conditions were comparable.

Edwards and Brackett (1978) noted that the effectiveness of enforcement depends on increasing the drivers’ belief that they may be apprehended if they speed, so that they will adjust their behavior and slow down. The goal is to make the "subjective probability" independent of time and place, so that drivers are encouraged to drive more slowly no matter when or where they are driving. They cited research that stationary marked-car enforcement is more effective than moving patrols or unmarked cars. Their study tested a two-phase approach: begin with intensive enforcement to effectively slow drivers down (two weeks), then continue with a schedule of minimum police presence that is still adequate to maintain the slower speeds (four weeks). They placed enforcement vehicles randomly along a 17-mile stretch. The strategy reduced average speeds and also extreme speeds, for up to 14 miles for the course of the study. The authors noted that vehicles with CB radio antennae traveled faster than vehicles without, although CB radio reports of enforcement activity did not have any effect on average vehicle speeds.

In The Netherlands, motorway speed enforcement routinely uses photo radar, with or without police present to stop some speeders, to cite all drivers exceeding a threshold speed level (de Waard and Rooijers, 1994). The authors tested several variations on 120 km/h speed limit roadways, based on the proportion of speeders who were apprehended, the presence of police to do visible roadside stops, and – for mail-out citations, whether mail notification arrived the next day or two weeks later. "Inconspicuous" radar sites and speed loops were about 5 km apart. When on-view stops were done, the stops were made in full view in the emergency lane between the radar and the speed loops. For all but the lowest enforcement level, speeds at the downstream loops were reduced; effects were greatest with the highest level of on-view enforcement, which was also the only condition to continue to show speed reductions in the four-week post-enforcement period. Speed variability (essentially, the prevalence of very fast speeders) also decreased during the treatments. (For the mail-out conditions, because speeds were reduced, the authors hypothesized that the enforcement camera’s flash must have been visible and that it caused the speed reductions.) Questionnaires were sent to apprehended speeders, speeders who had not been apprehended, and non-speeders, of whom 80 - 91 percent were male. Speeders reported regularly driving faster, more frequently thought the 120 km/h speed limit was too low, and rated speeding positively. There were no differences between apprehended and non-apprehended speeders. In a second study, de Waard and Rooijers were able to keep the level of speeding to 5 percent or less (about 40% below baseline levels) by adjusting their level of enforcement from week to week.

In the UK, Holland and Conner (1996) implemented an intensive enforcement campaign on a busy commuter dual carriageway. The campaign was based on police speed check warning signs for three weeks and visible enforcement in the middle week. Speeds declined significantly during the intervention weeks, most when police were visible and active. One week afterward speeds were still somewhat depressed, and some residual effect was present six weeks later. Surveys were given to users of the road at service stations before and after the treatment. Young and male drivers (young males most) showed high "intentions to speed" before the treatment, as did all drivers who self-reported more frequent speeding. After the enforcement treatment, women under 25 had lower intentions to speed, but men under 25 showed even higher intentions to speed.

Vaa (1997) reported on a study of police enforcement on a 35-km segment of two-lane road with speed limits of 60 - 80 km/h. The police developed their own schedule of enforcement, emphasizing irregular timing and placement of enforcement locations and ultimately averaging about 40 hours/week for six weeks. Speeds were reduced by up to 5 km/h based on week and time of day, significant in all time periods. Percentages of speeding drivers also declined, with only morning rush hour speeds not significantly decreased. The effects lasted for up to eight weeks. A subsequent study was done to identify the minimum level of enforcement that would achieve comparable speed reductions (Vaa et al., 1995).

In 1992, the National Highway Traffic Safety Administration assembled its recommendations on speed enforcement into "Beyond the Limits: A Law Enforcement Guide to Speed Enforcement" (NHTSA, 1992). Much of the guide was devoted to police department activities necessary to implement a speed enforcement program, including department policy and commitment, training, staffing, and support. Major emphasis was also placed on: data evaluations to document the problem (speeding, speed-related crashes, etc.) and identify problem situations, designing an effective approach, strategic use of public information and education to increase knowledge and gain support, and program effectiveness evaluation. It was noted that "a comprehensive traffic program requires the joint efforts of law enforcement personnel, prosecutors, courts, driver licensing agencies, and public and private organizations."

In reviewing evidence on the contribution of speed to crashes on limited access roads, NHTSA (1992) cited TRB (1984) as concluding that, on limited access roads, the relationship between average speeds and crashes is not clear, but that vehicles traveling much faster or much slower than average are much more likely to be involved in crashes. The risk for vehicles traveling very slowly is problematic, because such vehicles may have problems or special circumstances, such as vehicle defects or special characteristics, driver limitations, or turning, merging, or stopping maneuvers – that make them particularly vulnerable. For very fast vehicles, though, excess speed and possible related risky maneuvers are the primary factors that distinguish them from vehicles traveling at average speeds.

The NHTSA guide did not offer suggestions on the kinds of enforcement strategies and patterns to obtain the most effective long term and wide area speed reductions, although it emphasized the importance of public information and education components. More than half the book was devoted to illustrations of active enforcement programs.

One topic within speed enforcement that has aroused considerable interest, if not emotion, is the use of radar detectors. Graham’s (1996) overview article reviewed the controversy between the enforcement community and the pro-detector community. She also described the kinds of police strategies and technology to render radar detectors ineffective, including "instant on" radars that measure speeds before drivers can react, narrow-beam radars, radars that use patterns or frequencies not picked up by most radar detectors, laser speed measurement devices, VASCAR (a vehicle average speed calculator and recorder, based on the time taken to travel a known distance), and drone radar that triggers radar detectors even though no enforcement is being conducted. Radar detectors are illegal in many countries, including France, where the driver’s vehicle can be seized, and in the District of Columbia and Virginia.

Drivers who use radar detectors drive faster and are less safe than drivers who don’t use them. Teed and Lund (1993) conducted speeding enforcement campaigns in Charleston County, South Carolina. When laser enforcement devices were used (compared to standard police radar), more speeding tickets were issued and speeders were four times as likely to have radar detectors. Cooper et al. (1992) compared the crash and speeding convictions of 174 drivers who had special insurance covering radar detectors with a sociodemographically matched comparison group. Those with detector insurance had significantly more crash claims and speeding convictions.

Teed et al. (1993) cited research showing sharp reductions in actual traffic speeds when hidden police radar was activated, indicating that users of radar detectors were among the fastest drivers on the roads. They examined the effects of radar on the travel speeds of drivers on Interstate 70 in Maryland, then a 55-mph-limit road. Drivers exceeding 65 mph were targeted by conventional police radar and their speeds were re-measured .05 miles later and again one, two, and five miles downstream. Forty-four percent of passenger vehicles were initially traveling more than 65 mph; this dropped to 32 percent immediately after they were hit by police radar, but the effect was essentially gone one mile later, where 42 percent were above 65 mph. Of 185 vehicles tracked at all sites, 81 showed brake lights and/or slowed by more than 5 mph just after being targeted by the police radar, likely indications that they were using radar detectors. They were going somewhat faster initially (than drivers who did not react, i.e., who probably didn’t have active radar detectors) but were slower immediately after, and the speed differences had vanished by 3 miles downstream. Fifty-eight percent of speeding tractor-trailers seemed to have radar detectors, as compared to 38 percent of light trucks and 32 percent of passenger cars.

Pezoldt and Brackett (1987) examined speeds along a thousand miles of 55-mph Texas highways. They found that trucks averaged 2 mph slower when radar was activated, and – largest for trucks and very fast passenger vehicles – the proportion of vehicles exceeding 70 mph dropped dramatically when radar was active. They noted, however, that their results did not show that users of radar detectors would drive faster than others if no one had radar detectors.

As early as 1987, Christoffel argued that radar detectors should be banned. He cited case law in Virginia, the District of Columbia, and Connecticut (where detectors were illegal until 1992) upholding the validity of those jurisdictions’ prohibitions. Fields and Hricko (1987) reviewed essentially the same court cases and came to the same conclusion. The one case in which courts ruled against enforcing a ban on radar detectors was cited in both articles and dismissed by the authors. Police in Michigan had attempted to charge a motorist under a Prohibition-era law against receiving police radio signals, but the courts eventually ruled that radar speed measurement devices were outside the scope of the law.

One constant in all radar (and laser) detectors so far has been that their only function has been to detect police speed enforcement efforts. Graham (1996) described the Safety Warning SystemÔ device, which can receive signals from special radar transmitters that could be placed at crash scenes, near construction zones, or at other hazards. The Safety Warning SystemÔ device would display one of up to 64 messages as to the nature of the hazard ahead. Naturally, the device would also respond to police speed radar, and the transmitted signals would cause standard radar detectors to trigger alarms. This device has not been tested in court.

A next step in enforcing speed limits is photo radar, which clocks vehicle speeds and photographs vehicles traveling above a variable threshold value, usually set at 5 - 10 mph above the speed limit. (Measuring vehicle speed can be done in any of a number of ways while being faithful to the basic premise of automatic speed enforcement.) The photos are used to identify the offending vehicle and mail speeding tickets to the registered owner. (If the owners were not driving, they are usually required to identify the drivers so that they can be cited.) The devices can yield a large number of citations with little on-site personnel expense. Traffic safety concerns about the devices are that, because the citation is delayed and remote from the actual violation, it may have little effect on speeding behavior. Legal concerns (Hoff, 1997) center on lack of officer discretion in deciding whether to cite, the delay between event and citation preventing the defendant from having an adequate chance to present a successful defense, and the relatively small consequences making it unreasonable to accept the expense of mounting a defense. Without directly relevant cases, Hoff concluded that, since photo radar is the combination of photographic evidence and radar speed measurement and since both are acceptable separately to courts, the combination is likely also acceptable.

Similar devices are used to detect, photograph, and cite drivers who run red lights. In this country, they are used in places like New York, Los Angeles, San Francisco, Paradise Valley, Arizona, Commerce City and Fort Collins, Colorado, Jackson, Michigan, and Fairfax City, Virginia. Installations typically include "Red Light Camera Ahead" and "Violators Photographed" signs, to maximize the changes in driver behaviors as well as provide fair warning.

Studies of automated speed enforcement (ASE) systems were reviewed and results synthesized by Blackburn and Gilbert (1995). Programs from the United States, Australia, Canada, and Europe were studied, including many of the ones specifically referenced here. The authors concluded that the devices can reduce speeds, crashes, and injuries, but that the extent of benefits depend heavily on the details of the situations. They reviewed the history of the use of ASE systems in the U.S. and elsewhere, considered legal and technical requirements for using the systems, and reviewed problems in their operation. They recommended scientifically controlled tests of the effectiveness of the systems in reducing speeding and crashes, development of certification and training procedures for the use of the devices, and passage of state-level legislation permitting local jurisdictions to use ASE systems.

Oei (1997) reviewed three kinds of automatic speed management in The Netherlands, based on whether they were for specific locations (e.g., intersections), specific rural road segments, or provincial road networks. Speed limit signs that flash when approaching drivers are speeding reduced speeds significantly at an urban and a rural intersection, and crash savings of 24 - 65 percent were projected. On rural road segments, photo radar was used. The program began with an information campaign, and installations included a "Radar Check" sign, a speed limit sign, a variable "You Are Speeding" sign, and finally the photo radar. After a warning-only period, automatic enforcement was conducted for 3.5 months (with a brief interlude to repair vandalism damage, a problem cited frequently in reports of automatic traffic enforcement). Across four sites, average speeds and 85th percentile speeds dropped in the warning-only phase and dropped again during the enforcement phase. Speeders dropped from 38 percent in baseline to 28 percent and 11 percent in the test phases. Drivers were observed to brake when closing in on the devices and speed up after passing them. Injury (including fatal) and property damage only crashes dropped about 35 percent in the test periods when compared to control roads. In three provinces, 120 road segments were selected for photo radar enforcement. As police resources permitted, photo radar was operated from unmarked cars; downstream was a sign, "Your speed has been checked. Police"; and violators were cited by mail. Eighty-fifth percentile speeds dropped by 4 - 5 km/h, and the percentage of speeders dropped from 42 to 31 percent. Surveys were mailed to drivers in one province, and completed surveys were received from 76 percent. About three-quarters of them accept photo radar enforcement. Half said they comply with speed limits anyway, 70 percent said they would comply with monthly enforcement campaigns, and all said they would comply with weekly enforcement.

Norway has used photo radar for speed enforcement since 1988. Elvik (1997) examined crash reductions along 64 road sections with photo radar, finding an overall reduction of 20 percent in injury crashes and 12 percent in property damage only crashes (the former was statistically significant). After a number of installations had been made, warrants for their installation were developed (older units were not removed). Photo radar was appropriate if: crash rates were higher than for similar roads; there were at least 0.5 injury crashes per kilometer per year; and the mean travel speed was above the speed limit. On segments meeting the warrants, injury crashes dropped by 26 percent; on other segments, the decrease was only 5 percent. Traffic speeds were not directly measured, nor were other possibly relevant parameters and consequences, providing opportunities for subsequent research.

Photo radar installations have also been used in Kuwait (Ali et al., 1997). At eight test sites, drivers slowed significantly at the camera positions but were traveling at unreduced speeds at sites 1 km before and after the cameras. Some drivers used higher speeds downstream of some sites, perhaps in order to "make up" for the lost time passing the cameras. The authors note that even drivers "caught speeding" by photo radar almost never receive speeding citations in Kuwait, and that the absence of true enforcement makes photo radar essentially ineffective.

Portland, Oregon, is conducting a two-year demonstration project with photo radar in school zones and residential neighborhoods (Price and Hunter-Zaworski, 1998). The photo radar apparatus is mounted in a marked van and uses a reader board to advise drivers being ticketed for exceeding the speed limit. The city also posted signs at all entrances to the city advising of the use of photo radar. The authors summarize photo radar experiences in Australia, Canada, and various European countries. One U.S. study in National City, California, produced 14,000 speeding tickets in 20 months and reduced crashes from about 70 per month to less than 50 per month (Repard, 1993, cited in Price and Hunter-Zaworski, 1998). In Portland, five streets received photo radar enforcement, and three control streets were included in the study. Tested streets had 25, 30, or 35 mph speed limits. The test showed significant decreases in mean speeds and in percentages of drivers going more than 10 mph above the speed limits, on test streets as compared with control streets. The authors recommend continuing and expanding the program, citing several possible benefits that might come from a broader and better known program.

Freedman et al. (1990) conducted a telephone survey among residents of two communities (Paradise Valley, Arizona, and Pasadena, California) where photo radar is being used; they also surveyed residents of nearby communities. They found that between 75 and 96 percent of the respondents knew of the use of photo radar, 52 - 89 percent had seen it in use, 49 - 62 percent approved of its use, and – of those approving – 63 - 70 percent thought its use should be expanded. Almost half of the respondents who knew that photo radar was being used said that they were driving more slowly as a result.

More recently, Streff and Molnar (1995) surveyed drivers in Michigan in areas where there was a NHTSA-sponsored pilot test of automated speed enforcement devices (ASEDs). About 29 percent of drivers returned the mailed surveys, roughly equal percentages of all licensed drivers, drivers who received warning letters based on ASED detection, and drivers detected as speeding by the ASED but not warned. (Michigan law did not allow issuing tickets based on photo evidence.) The study did not change drivers’ speeds (not surprising to the authors because of the lack of "teeth" and minimal program publicity). Survey respondents favored use of ASEDs in select situations, particularly in school zones, in areas where traffic enforcement is dangerous for police, for heavy trucks, and in construction zones. The survey also showed opposition to ASED use on freeways, on bridges, and on "all roads." The observed speeders and persons who reported having multiple citations in the previous two years expressed greater opposition to the use of ASEDs than the general population.

Corbett (1995) conducted a similar survey in the UK after the experimental introduction of speed cameras. Overall, results indicated that speed cameras were effective in reducing the speeds adopted by some drivers, and there was some self-reported reduction in driving speeds in other areas. Drivers who reported driving fastest in general were most likely to report driving slowly as they passed the cameras but also did not reduce their speeds in other areas. In general, the surveyed drivers favored the speed cameras. The author recommended that placement of the cameras should be varied to increase the areas in which drivers slow down, while also noting that permanent placement at accident black spots may also meet objectives.

Hashimoto (1979) reported the use of police surveillance, without enforcement, on intersection problems in Japan. Posting one, two, or three uniformed police on the corners of a dangerous intersection, where they simply watched traffic in the intersection, led to reductions in "vulnerable behaviors." The behaviors were part of a model which described vulnerable behaviors as ones which placed drivers at risk of crashes, with subsequent circumstances, not under the control of the driver, key to whether a crash resulted. The author concluded that the vulnerable behaviors were valid indices of possible crash involvement and that they could be reduced by police presence.

Another technique used to reduce speeds is speed display boards, devices attached to trailers or police vehicles which display the speed of passing vehicles and may show a warning to speeders. Casey and Lund (1993) tested trailer-mounted speed boards labeled "police" and including the posted speed limit in Santa Barbara, California. When speed boards were deployed, speeds decreased by about 10 percent next to the boards and about 7 percent about one-half mile downstream, but effects rapidly disappeared when the boards were removed. Using varied deployment schedules, the authors found that speed reductions at the boards continued through two weeks of speed display board use but essentially vanished during the third week; throughout, downstream measurements showed little speed reduction. With intermittent enforcement, speed reductions continued through the third week at the speed display boards, but they were virtually gone at the downstream locations. When tested at school zones during hours when students would be present, speed boards produced drops of about 5 mph at three sites, 1 - 2 mph at two others with slower baseline speeds.

Webster (1995) reviewed vehicle-activated speed reminder signs in the UK, Europe, and the U.S. Most of the signs were "secret," i.e., with a blank face until activated by a speeding vehicle. Overall, the signs appeared to reduce vehicle speeds by a few mph, and some of the reduction was maintained downstream. In the UK where the signs were used at entrances to villages, speed reductions were sustained into the middle of the villages. Speed reductions appeared to be maintained over time. Crash reductions were not statistically significant, although all cited changes in crashes were in the proper direction. As a practical concern, the authors noted that vandalism had been a problem in The Netherlands and in the U.S. (rifle fire).

Bloch (1998) compared photo radar, an unenforced speed display board, and speed display board with intermittent enforcement. The test was done in Riverside, California, on 25 mph speed limit residential collector roads with average daily traffic of 800 - 2400 vehicles in each direction. Speed data were collected for two baseline weeks, one week with the treatments in place, and one week after they were removed. Alongside the treatments, photo radar and the display boards reduced speeds by 4 - 5 mph below baseline speeds of 34 - 35 mph. Speeds had nearly returned to normal by 0.2 miles downstream for photo radar and the standalone speed display board, but they remained depressed for the condition with intermittent enforcement. Speed decreases had vanished by one week later, and in fact were absent during the treatment week during the hours when the treatments were not present.

Eagle and Winter (1980) tested speed warning signs in the UK: "Police: Speed Check Area." They found that speeds declined throughout a 12-week test period, more so when enforcement was added, but the effects disappeared when the signs were removed.

School zones are an area of concern in most communities. Studies by Aggarwal and Mortensen (1993) in California and Hawkins (1993) in Iowa tested school warning signs with flashers which were illuminated in periods when students would be going to or from school. Both studies showed significant speed declines when the flashers were operating, although the amount of decrease varied from site to site and was not significant at all sites. Hawkins found the effect persisted for twelve months with the signs in operation although the size of the speed reduction decreased by about 30 percent. In Philadelphia, Jordan (1998) examined child pedestrian crash patterns. Although only about 5 percent of the crashes occur in school zones during key time periods, the city is planning to select schools with greater crash problems to install school zone flashing warnings.

Advisory devices, when used in specific situations, can be effective in reducing speeds. Maroney and Dewar (1987) used transverse lines painted on the roadway at progressively diminishing distances. The objective was to produce an alerting response and, with the illusion of acceleration, an actual slowing of the vehicles to compensate. Tested on an exit to a freeway in Calgary, Alberta, Canada, the transverse lines reduced speeds initially but the effect began to disappear after three weeks. The lines also led to increased speed variance, as some drivers heeded the warning and others didn’t. Griffin and Reinhardt (1996) reviewed 10 studies, including Maroney and Dewar, and concluded that most studies showed the transverse painted lines could be effective in reducing speeds, more so at high speeds (such as 85th percentile) than for means. The conditions tested varied, along with the effects. Some studies showed no changes, some showed changes that persisted for long times, and others showed changes that dissipated rapidly. Griffin and Reinhardt suggested that the primary mechanism by which the stripes worked was as a warning device, not through psychophysical illusions. Griffin and Reinhardt also looked at patterns of converging chevrons on the pavement in Japan, and they concluded that the chevron patterns may reduce crashes by as much as 25 - 50 percent.

Retting and Farmer (1998) looked at pavement markings, "Slow" with a large arrow, on a rural road in Virginia just before a sharp curve. In their study, mean speeds dropped by up to 7 percent, compared to nearby untreated curves, and that the percent of drivers exceeding 40 mph on approach dropped significantly.

Speed control through reducing speed limits and providing a mix of enforcement and public information has proven to have modest but real effectiveness. The basic situation is that the roads and vehicles are such that a large percentage of drivers, with no other constraints, would travel faster than is desired by the authorities, and that the official efforts are to slow drivers down. The purpose of lower speed limits and public education messages, often in conjunction with enforcement, is to have more and more of the population believe that slower speeds are appropriate and reasonable. The purpose of enforcement is to increase the perceived negative consequences of driving fast and to draw more attention to the public information and education.

Changing speed limits has proved to have significant effects on average speeds. When speed limits are reduced, average speeds decrease by about one-quarter of the reduction in the speed limit. In these cases, the speed limit changes are to new levels deemed reasonable or appropriate by the authorities and thus might be characterized as "moderate" changes. It is not reasonable to assume that the relationship would continue if speed limits were changed more drastically – that, for example, a true 10 mph drop in speeds could be effected by a 40 mph drop in speed limits.

Enforcement is often used in conjunction with speed limits to control vehicle speeds, but its effects are also limited. Enforcement reduces speeds where and when the likelihood of apprehension is perceived to be high. However, enforcement is often an expensive, manpower-intensive operation, and the effects diminish rapidly away from the site and time of visible enforcement. Driver behavior suggests that many of them believe it is all right to drive quite a bit faster than the posted limits, and that the perceived high likelihood of enforcement is just a brief interruption to their normal driving patterns. The popularity of radar detectors, and the quick but transient response of radar detector users to perceived police radar, reinforces this conclusion. In the war of technologies, automated speed enforcement approaches such as photo radar allow enforcement to be more frequent, more broadly located, and more effective.

However, speed limits and enforcement are indirect means of controlling speeds. Direct approaches, like those which make it physically impossible, difficult, or unpleasant to travel faster than the authorities desire, are covered next.

This section reviews speed management approaches that include traffic engineering components. Particularly with respect to non-motorized participants in traffic, these engineering solutions are designed to make the traffic environment and the actions of motor vehicles safer and more pleasant.

Until recently, the goal of traffic engineering has seemed to be to increase the mobility of motor vehicles by providing for greater numbers of motor vehicles and allowing them to travel at higher speeds with fewer interruptions and delays. Under that approach, pedestrians and bicycles became second-class (or worse) roadway citizens. Their safety was achieved by separating them from motor vehicles. Separations could be temporal, through traffic signals, sometimes having pedestrian-only phases (and prohibiting pedestrian movement at all other times), or physical, through separate facilities like overpasses, underpasses, bike lanes, or totally separate bike paths. While these approaches achieved some safety benefits for pedestrians and bicycles, they also added inconvenience, delays, and often discomfort. This has discouraged many from walking and bicycling. If the activities are sufficiently inconvenient and unpleasant, and if desired destinations are too far away, then fewer people will walk or bicycle.

Another result of this approach has been an unacceptably high level of pedestrian and bicycle casualties. Many roadways are not well designed to accommodate pedestrians and bicyclists, but they use them anyway. In trying to do so, they often place themselves at risk because of impatience or ignorance or confusion about how they should negotiate the roadways. The result is crashes, injuries, and sometimes deaths.

Traffic engineering approaches have been developed to address the noxiousness of the traffic environment for non-motorists as well as the unacceptably high numbers of crashes, injuries, and deaths. This section of this chapter looks at individual techniques that have been used to control vehicle speeds and improve pedestrian and bicycle safety. It describes the techniques as well as studies of their effects, but it focuses primarily on their results for pedestrian safety. Although the interests of pedestrians and bicyclists overlap extensively, very little about bicycle safety and convenience will be covered, except where pedestrian solutions may have significant negative consequences for bicyclists.

"Road humps" can be successful in reducing both average and very high speeds, and the result is often a reduction in crashes, injuries, and deaths. Road humps are placed in roadways to cause vehicles to move up and down in a way that is uncomfortable if done too rapidly, thus encouraging drivers to slow down. Road humps have been used in foreign countries and in this country for decades, and a great deal of experience and research has been done to determine what shapes and spacings are appropriate to obtain what kinds of traffic control. (It should be emphasized that this does not refer to common American speed bumps, which are short, relatively high, and produce a jarring sensation if traversed at faster than walking speed.)

Speed humps are typically about 12 feet in cross-section, flat or rounded on top, and 3 - 4 inches high. Cars can cross them comfortably at 20 - 30 mph, depending on their exact shape. The British Department of Transport (DoT) (now the Department of Environment, Transport, and the Regions, or DETR), for example, has conducted extensive research and issued guidelines for hump design and placement, most recently in Traffic Advisory Leaflet 7/96, Highways (road humps) regulations 1996. That leaflet also contains recommended markings and signs for use in the UK, guidelines that provide useful input for U.S. applications. (See also Hodge, 1993, for a test of a variety of vehicles across speed humps; Webster, 1993a, for the danger of vehicles "grounding" on humps as a function of vehicle and hump dimensions; Webster, 1993b, for a study of hump implementations and effects on speeds, traffic flows, and crashes; Webster, 1994, comparing thermoplastic "thumps" and 50 mm high humps; Webster and Layfield 1996, reviewing the effects of 75 mm high humps at 72 sites; also DoT-TAU Traffic Advisory Leaflets 2/90, Speed control humps; 7/94, "Thumps," thermoplastic road humps; and 2/96, 75 mm high road humps).

One problem with speed humps is that different vehicles are affected in different ways. Trucks, buses, and most emergency vehicles bounce severely at speeds at which cars can travel comfortably. Kjemtrup (1988) reports on Danish studies into shapes of humps which could apply to different vehicle types, including the "K-hump" which includes two cross-sections, a standard speed hump in the center of the traffic lane and a hump with much longer cross-section at both edges. The K-hump intends for cars, which have relatively narrow tread measurements, to pass over the standard cross-section portion, and trucks and buses, with wide treads, to pass entirely on the extended-length hump sections. (The longer the hump cross-section, for any specific vehicle type, the faster the speed at which it can cross comfortably; their objective was to design the outer portion so that long-wheelbase buses and trucks could cross at the same speeds as cars could cross the center section.)

In the UK, the recommended solution to mixed car, bus, and truck traffic has become the "speed cushion." Speed cushions have the same cross section as standard speed humps, but they are at full height only in the center of the lane; toward the edges of the lane, speed cushions taper off until they are flush with the roadway. This requires cars, with narrow treads, to cross the cushions and be slowed by them, while allowing buses and trucks, with wider treads, to cross with their wheels at the tapered edges of the cushions and be much less affected. Research on their effectiveness, and recommendations for speed cushion dimensions, are presented in DoT-TAU Traffic Advisory Leaflets 4/94, Speed cushions, and 1/98, Speed cushion schemes.

Zaidel et al (1992) described road humps as proven speed control devices with three general obstacles to wider implementation: They are perceived as being obstructions and degradations to the smooth paved surface of roadways; their early designs (including rumble strips and speed bumps) were ineffective or hazardous; and there is concern that they may become overused. The authors explored community issues, noise and vibration, and the impact of road humps on pedestrians (positive), bicycles (neutral or positive), large buses and trucks (negative but solvable), and emergency vehicles (also negative but solvable). Public opinion is a major concern, according to the authors, because there are a number of relevant classes of people who may have differing opinions and perspectives about them and because some negative responses appear to be well-founded. They emphasize that obtaining public support requires a combination of proper education and consultation and proper project needs analysis, design, and implementation.

In the U.S., speed humps have been a relatively recent element of the speed management arsenal. Gonzalez (1993) describes their use in Bellevue, Washington, along with a number of other techniques. They have also been employed in Maryland (Walter, 1995) as part of a coordinated program of speed management. The Institute of Transportation Engineers (ITE) published Recommended Practice: Guidelines for the Design and Application of Speed Humps in 1993 (see ITE, 1993, for a summary). This document includes guidelines for developing community involvement and support as well as design specifics for speed humps and markings.

Other uses of vertical deflections include: Crosswalks, where the raised section has a flat surface and it is marked and intended as a pedestrian crosswalk; raised intersections, in which the entire center area of an intersection is raised (and flat) with appropriate markings, thus making the intersection stand out from others and requiring all traffic to slow; and gateways, transitions from standard roadways into traffic-calmed villages or neighborhoods with appropriate signage. These are discussed below.

Roads which are broad and straight encourage higher speeds. Making them narrower and less straight encourages lower speeds. Wallwork and Burden (1998) present graphic illustrations of numerous techniques used to accomplish this. The midblock techniques include roadway narrowings through buildouts and parking; medians to narrow the roadway and often redirect traffic; chicanes; and midblock barriers to create two short cul-de-sacs. In practice, these may be combined, and vertical deflections such as speed humps and raised crosswalks may also be mixed in. For intersections, techniques include roundabouts, sidewalk buildouts to shorten pedestrian paths and slow turning traffic, one-way entry or exit treatments to eliminate some possible traffic flows, various diverters to eliminate some possible traffic flows, and barriers across one or more legs to close them off, simplifying the intersection and creating cul-de-sacs. Particularly at intersections, specific treatments are often part of comprehensive neighborhood or wide area traffic management plans that involve significant traffic redirection.

Chicanes have been examined in test track and field implementations by TRL in the UK. Sayer and Parry (1994) tested chicanes constructed of interlocking plastic lane curbing on the TRL test track. They varied lane width (symmetric before and after the single deflection), stagger length (length from the beginning of the chicane to the end), free view width (offset between the near curb and the offside curb seen across the stagger), and visual restrictions (barriers to forward visibility installed at the beginning and end of the chicane). In general, narrow lanes, short stagger, negative free view width (wider offsets), and visual restrictions all reduce speeds through the chicane. The effects of chicanes depend heavily on the length and width of the vehicle passing through; large vehicles must track more precisely and turn more sharply, and thus go more slowly. Some chicanes are too tight and narrow for some large vehicles to pass at all, and "overrun areas," of contrasting paving and texture and not used by most vehicles, may be added to allow large vehicles to go through narrow chicanes. Chicanes may be single lane (shared, on a two-way roadway), two lane, or two lane with a center divider. Chicanes may be combined to present a complex path and generate additional slowing. (See also DoT-TAU Traffic Advisory Leaflet 9/94, Horizontal deflections.)

Davies et al. (1997) examined the safety aspects of road narrowings such as chicanes for bicyclists. Half of 62 local highway authorities indicated they did consider bicyclists in designing traffic calming schemes. Twenty-eight sites were examined. Ones safe for bicycles tended to have bypass pathways through the obstruction. Observations showed that motorists overtaking or passing bicyclists did not wait but passed them in the narrowings, often at reduced clearances, and motorists often intruded into bike lanes when going through the restrictions. Overall, bicycle-motor vehicle crashes decreased by about 35 percent, but this was not statistically significant and changes in bicycle flow rates were not monitored.

Sayer et al. (1998) studied 142 individual chicanes in 49 chicane schemes. Mean speeds through the chicanes were 23 mph, and 85th percentile speeds were 28 mph, both reflecting a 12 mph reduction from pre-chicane measurements. Speeds between chicanes, where more than one were installed, dropped by 7 - 8 mph from the "before" speeds for means and 85th percentiles. Traffic flows were reduced by about 15 percent at single-lane chicanes and 7 percent at two-lane chicanes (all on two-way roads). Several of the chicane sites had no injury accidents in the "before" period. For the 17 schemes with known crash data, there was a 54 percent decrease in crashes after the chicanes were installed. (See also DETR-TAU Traffic Advisory Leaflet 12/97, Chicane schemes.)

Traffic islands have also been used to reduce vehicle speeds. They can be used to narrow roadways, guide vehicles over speed cushions, and serve as refuges for pedestrians. They also are used in chicane schemes on wider roads to channel traffic through paths that lower speeds. In DoT-TAU Traffic Advisory Leaflet 7/95, a number of schemes are presented. Guidelines for accommodating pedestrians and bicyclists are also included.

One problem with chicanes (and roundabouts, below) is that when they are sized to evoke the desired behavior from cars they are too tight or small for trucks, buses, and most emergency equipment. One solution is "overrun areas," which are widenings or extensions that allow larger vehicles to pass through. Overrun areas are designed with slight rises and are paved of rough material like cobblestones so that car drivers will choose to stay on the smoothly paved roadways, not use the overrun areas, and be subject to the full slowing effects. (See, e.g., DoT-TAU Traffic Advisory Leaflet 12/93, Overrun areas.)

Roundabouts are essentially forms of intersections at which traffic entering the intersection area is deflected into a circular pattern and vehicles travel around the circle until they find their desired exit point. Key features of roundabouts are that: entering traffic always yields to traffic on the circle (which prevents gridlock in heavy traffic and provides a simple set of rules for drivers encountering unfamiliar roundabouts); traffic must be deflected from its original path in order to enter the circle (which enforces traffic slowing even in light traffic); there may be flared areas upstream (to add high capacity for holding vehicles waiting to enter the circle and allow several to enter at once when the way is clear); and there may be islands for each entry/exit pair (which direct the traffic flow and provide refuge for pedestrians).

(The other term is "traffic circle," which is reserved for designs