An energy loss-based vehicular injury severity model

Recently published:

  • Ji, Ang and Levinson, D. (2020) An energy loss-based vehicular injury severity model. Accident Analysis and Prevention. 146 October 2020, 105730. [doi]

How crashes translate into physical injuries remains controversial. Previous studies recommended a predictor, Delta-V, to describe the crash consequences in terms of mass and impact speed of vehicles in crashes. This study adopts a new factor, energy loss-based vehicular injury severity (ELVIS), to explain the effects of the energy absorption of two vehicles in a collision. This calibrated variable, which is fitted with regression-based and machine learning models, is compared with the widely-used Delta-V predictor. A multivariate ordered logistic regression with multiple classes is then estimated. The results align with the observation that heavy vehicles are more likely to have inherent protection and rigid structures, especially in the side direction, and so suffer less impact.


21 Solutions to Road Deaths

So your civilization kills 1.35 million people per year in automobile-related crashes. This is a tragedy, but it is no “accident“. Road deaths are largely preventable, and the evidence is that some countries (the United States, Australia) have much higher rates of road death than others (Sweden). The strategies to reduce the carnage go under different names in different places, from Vision Zero to the Safe Systems Approach, but they are not really that distinct substantively.

Most crashes have multiple points of failure in the chain of causation. If only party A was paying more attention, it wouldn’t matter what party B was doing, and vice versa. If the road were designed differently, party A would have behaved differently.

Consider a vehicle-on-vehicle collision. There are at least six actors here: The driver, the other driver, the vehicle, the other vehicle, the road, the environment. Other vehicles (and their drivers) may also have a role to play, even if they are not directly involved in the crash.

The penalty for distraction or misjudgment should not be death. No one deserves to die for a momentary lapse of awareness. Driving is a hard skill, people are fallible, and in addition to the road deaths noted at the beginning of this post, there are hundreds of millions of severe crashes each year globally, many of which are injurious and life-changing, even if not life-ending.

The general approach is two-fold:

  • Prevent crashes.
  • Mitigate crash consequences: prevent death and reduce injury.

The good news is that the same things that prevent crashes often reduce the severity of crashes should they occur. Going slower gives drivers more time to react, but even if they fail to reach in time, (braking too late, or not at all), they still are less likely to be killed or injured, or kill or injure others, at a slow speed than a fast one.

The problem is that drivers often do not want to go slower, and the feeling is that the behaviour which is safer is also less efficient. This is true assuming no crash. Of course, being involved in a crash, or stuck in the congestion behind someone else’s is not efficient either.

The safety advocacy field has long used the 3 Es of Engineering, Enforcement, and Education to describe their strategies. This has been extended with other Es, including variously Emergency Services, Evaluation, Environment, Encouragement, and Everyone Else. Toole Design: has recently proposed a new set of 3Es: Equity, Ethics, Empathy. The Es are a useful mnemonic, but we need to dig deeper. This post includes 21 strategies that improve safety.  You might think of others. There is no one magic bullet for this complex problem, though some strategies are more effective than others depending on context.

Engineer for Safety

Design the system to be safer.

Foremost, this means design for slow speed. Wherever a pedestrian may come into conflict with a vehicle, the speed should be as low as possible, but start with 30 km/h (18 mph) as at that speed pedestrians are more likely to survive being hit by a car than die, while at higher speeds the reverse is true. If vehicles must go faster, the separation of pedestrians, bicyclists, and cars should be considered. There are numerous strategies to achieve low speeds on streets, including regulations (discussed below), enforcement of those regulations, and better, designing roads so that drivers only feel comfortable driving at low speeds.  There are numerous techniques to reduce the design speed of streets and roads, including traffic calming devices and shared space approaches.  Traffic calming aims to use street designs to reduce speed, by making it difficult or uncomfortable to travel faster. Solutions include narrowing the street, chicanes to alter direction of vehicles, and speed tables and wombats (raised pedestrian crossings, at intersections or midblock) to remind drivers they are in pedestrian areas. Shared spaces allow persons in and out of vehicles to use the same area and visually negotiate conflicts rather than requiring on an excess of signs, signals, and markings regulating behaviour. The Netherlands has probably done the most to improve the streetscape for the benefit of safety following the “stop the child murders” (Stop de kindermoord) campaign of the 1970s.

Stop the child murders. Dutch National Archive
Stop child murders. Dutch National Archive

Use of roundabouts rather than traffic signals takes up more space, and increases the distance pedestrians have to walk, but by lowering speed at the intersection and deflecting cars, generally results in an overall safer situation than a signal or stop-controlled intersection. In addition to being safer, lower speeds also make driving less advantageous compared to other modes, and reduces demand for cars, which likely improves safety as well.

But it also means that designs for high-speed roads should be safer through better geometric design. Geometric design is usually about the design of horizontal and vertical curves, and the visibility at those curves determines the appropriate speed (or the design speed determines the curvature), which depends on driver perception and reaction times.  But there are other aspects as well.  Civil Engineers are responsible for designing roads and the traffic controls (signs, signals, and markings) that govern them. As an undergraduate student at Georgia Tech, my Transportation Engineering Professor, Paul Wright, was very much concerned with traffic safety. One of the problems he identified was that cars ran off the road at high speed and hit fixed objects, like light poles, or trees, which killed the driver or passenger, while the pole or tree remained standing. So he advocated ensuring the side of state highways were clear of fixed objects within 30 feet (~8 m) of the road edge. Light poles and signs would be redesigned to be breakaway, instead of being designed to outlast a car crash. Trees would be cut down. Bridges would have protected guardrails so that cars would be deflected back into the stream of traffic rather than crashing into a concrete pillar (as bridge pillars should not be breakaway for obvious reasons).  This apparently reduced deaths, though was unfortunately also applied in urban areas, encouraging higher rates of speed. Other similar strategies include cables in the medians of highways to reduce cross-over crashes (which are far worse as the speed of impact of a head-on crash is so high).

Different types of roads require different treatments. Limited access motorways (freeways) are generally safer at high speeds than city streets, and when faster (e.g. with higher freeway speed limits, to a point), attract traffic away from city streets and rural roads, which reduces overall statewide fatalities, though increases them on the relevant freeways, and probably induces increased demand overall.

Design road surfaces to reduce slip and increase friction. Pavement engineers consider not only the strength of pavements, but their surface condition.

Maintain Roads to improve traffic safety as well. This includes the general road condition, as well as snow and ice clearance. Toshihiro Yokoo, Mihai Marasteanu, and I found that good pavement quality is associated with lower crash rates in several conditions:

  • Snow – for fatal crashes,
  • Asphalt over concrete and sags for injury and property damage crashes,
  • Wet roads for injury crashes, and
  • Crests, and spring load restrictions for property damage crashes.

It should also be noted that very bad pavement quality also performs as a type of traffic calming, as people are uncomfortable traveling at higher speeds on bumpy roads.

Educate, Enforce, and Regulate Drivers

Regulate driver behavior by adopting severe rules about drunk driving (drink driving, driving under the influence, driving while intoxicated) and rules about maximum hours to ensure drivers are as alert (sober) as possible when making decisions. In a sense much of this is pre-crime, we regulate this behavior even if the driver has not actually driven badly or caused a crash because they are in a state where they might at a moment’s notice. But unlike walking, driving is a privilege, not a right. Other rules include things like prohibiting the use of mobile phones while driving, as distracted driving, unlike distracted walking, is a real problem. Some rules are downright counterproductive, like assuming pedestrians should only cross at marked crosswalks. The US is terribly inconsistent about which crosswalks are marked, leading to confusion and danger.

Enforce the adopted rules and laws like the speed limit and other traffic laws that improve actual safety (that is, punish bad behavior). This is traditionally the responsibility of the police. Much of this can be automated with various types of `photocop‘ technologies, such as red-light running cameras and photo-radar speed enforcement. These generally improve safety. In contrast, ad hoc human-based enforcement often has racial bias, and has traditionally been used as a pretext to harass minorities. Still, if people believed there were serious consequences for violating road rules and killing or injuring other people with automobiles, they would probably be more careful.

Encourage and reinforce good behavior. While encouragement is not widely used, it is certainly possible to imagine earning points for good behavior, that might result in lower insurance premiums, or removing points from your driver’s license. However encouragement has its downsides:  imagine being pulled over by police, even if in the end they give you a citation of good behavior, the stress of the event, depending on your circumstances, especially if you had a record or points, may have taken several hours or days off your life)

Educate and License drivers, both new and continuing, on the state of the road rules, and their knowledge thereof. Driver education (and examination and licensure) is intended to produce better drivers. The US driver education system is not especially rigorous given the damage that can be caused, and there is no effective “continuing education” for drivers, requiring them to demonstrate continued improvement in driving skills, aside from occasional public service announcements and variable message signs. Practice improves skill, but experience creates relaxation and may induce over-confidence on the job (of driving). Licensure has been getting more rigorous in the US, and this has improved safety. Still, the only rigorous test are for the first time the license  is earned, and changes in road rules are never tested, much less memory of the existing road rules.


Equip vehicles with technologies that reduce injury to vehicle occupants in a crash, and equipment that reduces the likelihood of a crash in the first place.

Mechanical engineers have done better at the first of these, vehicle occupants are much less likely to die due to better-designed automobiles that now feature things like:

  • Seatbelts
  • Crash cages
  • Crumple zones
  • Shatterproof Glass
  • Airbags
  • Child safety seats (car seats)

Software and electrical engineers have helped develop systems such as

  • Rear cameras, for backing up,
  • Lane-departure warning
  • Automatic emergency braking systems
  • Adaptive cruise control
  • Auto-pilot, Supercruise, other advanced driver assistance systems

which improve safety, the latter ones especially on highways.

However, the increase in SUVs increases the severity of crash consequences for those outside the vehicle (in other cars, or on foot or on bike) both due to the mass of the vehicle, and the height, which collides with the soft body of the pedestrian nearer the vital organs than a lower sedan.

Materials engineers and mechanical engineers are involved in the design of tires, which is an important element in safety.

Equip humans with safety gear. If we can redesign vehicles, can we also redesign our unprotected road users, pedestrians and bicyclists?  Safety promotors have essentially proposed this. Some equipment that helps pedestrians and bicyclists be seen is that which increases visibility, including bicycle lights and increased reflectivity of the bicycle or the high-visibility clothes worn by road users. Or worse, asking pedestrians to carry flags to cross the street. By increasing visibility, the argument goes, the driver will be able to react and apply the brakes sooner, and either stop, or at least be slower, at the time of impact. The empirical evidence for safety is non-existent either way on reflective gear, though one imagines it helps, as it improves visibility in tests.

Vehicle equipment designs that reduce the impact of vehicles on bikes and people on foot should also be considered. The increasing height and mass of SUVs (and their increasing number) has been credited with the fast-rising number of pedestrian deaths in the US in recent years. The height makes it harder for drivers to see people (especially children and others who don’t play basketball professionally) just in front of the vehicle. The height also means the collision will occur on the upper rather than lower half of the body, where people keep most of their vital organs.

Bicyclists are often asked or required to wear helmets in countries like Australia with a weak biking culture. Yet the safest countries for bicyclists, like the Netherlands, don’t require this. It is clear from the medical literature that if you are dropped on your head, a helmet helps. But that is not the whole story. The question is one of incidence, whether a helmet increases the likelihood of being dropped on your head, either because of more aggressive cars around helmeted, uniformed, and armored bicyclists or because of the false confidence that helmets provide bicyclists. By the same logic, vehicle passengers and drivers should wear helmets and goggles, which will reduce injury and death. We might want them wearing giant inflatable bags as well. This seems ridiculous, but we essentially ask this of bicyclists. The entire conception of bicycling as a race requiring special gear ensures we have othered (dehumanized) the bicyclist, which reduces driver empathy. Delbosc et al. have found that dehumanization is correlated with aggression.

Ergonomics, or Human Factors Engineering,  considers road user interaction with the vehicle and with the road environment. Out-of-vehicle, human factors engineers are responsible for standardising traffic signs, signals, and markings, and interfaces, which appear to be getting better, if too numerous. In-vehicle, human factors engineers interfaces like the dashboard, which appears to be getting worse, increasing the distractability of drivers.

Emergency Response

Emergency response saves lives if there is a crash, and this has been getting better, both in terms of reduced response time, due to almost instant notification of an incident from vehicle-based systems or from mobile phones, as well as faster dispatching of emergency response, as well as improvement in medical care technologies over the past century. Improvements in response time and medicine are responsible for about one-third of the reduction in fatality rates from crashes in the UK.


Fewer automobiles will result in fewer automobile-related deaths. No cars eliminate the risk entirely, but also is beyond the pale to consider, apparently.

Pedestrianisation, or reducing the number of roads where cars are permitted, naturally reduces car crashes on the pedestrianised sections. Similarly, better pedestrian networks and pedestrian priority will both increase pedestrian safety and reduce the number of drivers. A key example of this is the trip to and from school, which children used to do on their own

Improve public transport, walking, and bicycle networks, to reduce the number of people who feel the need to drive.

Change land use patterns to reduce driving. Reorganizing the location of activities so that origins and destinations are nearer means more trips can be by walking, biking, or public transport. Denser areas are generally safer on a per capita basis (that is the more people, the more collisions, but the rate of serious collisions per pedestrian or per vehicle drops), as there are safety-in-number effects, which we have corroborated for pedestrians and bicyclists. The relationships are complex.

Grid the street network. Marshall found that “Denser street networks with higher intersection counts per area are associated with fewer crashes across all severity levels.”

Economy is a factor, the better the economy, the more people driving, both to work and to non-work destinations. Our research has shown that as the gas prices rise (and one expects, as the economy deteriorates), the drivers who give up the road are typically worse than average, often younger, less responsible, and less able to hold down a job, while younger potential drivers defer licensure. It might be socially counter-productive to want high gas prices or a bad economy solely to reduce road deaths, and a bad economy may very well increase other deaths, though high gas prices may have additional beneficial effects from reduced pollution. Nevertheless, the correlation remains.

Emerging Technologies

Vehicle automation will certainly help in the long run, as computers are less likely to be vulnerable to the same imperfections that humans suffer when attempting to maintain concentration, and are programmable to behave in a more risk-averse way, as well as having faster reaction times and being more predictable and more likely to obey road rules. But we are at least two decades away from full automation, and many of the strategies here will help even with automated vehicles.


Like congestion and global warming, the road death toll can be significantly reduced, but there is little evidence that the United States, in particular, is collectively interested in solving it. While there are obviously advocates, they do not have the upper hand, otherwise deaths would not be rising in recent years off its 2014 lows.

Ezra Hauer argues convincingly that practice needs to engage better with research, and be more evidence-based. There are other strategies as well would result in safer systems that kill and injure fewer people each year.  Despite the high road toll, the issue seldom received the attention it deserves, overshadowed by more newsworthy types of death. This post also doesn’t cover the deaths due to air pollution, which is of a similar magnitude, and suggests another set of solutions.









Measures of Speeding from a GPS-based Travel Behavior Survey

Recently published


Objective: Lacking information about actual driving speed on most roads in the Minneapolis–St. Paul region, we determine car speeds using observations from a Global Positioning System (GPS)-based travel survey. Speed of travel determines the likelihood and consequences of collisions. We identify the road segments where speeding occurs. This article then analyzes the relationship between link length, traveler characteristics, and speeding using GPS data collected from 152 individuals over a 7-day period as part of the Minneapolis–St. Paul Travel Behavior Inventory.

Methods: To investigate the relationship, we employed an algorithm and process to accurately match the GPS data with geographic information system (GIS) databases. Comparing actual travel speed from GPS data with posted speed limits, we measure where and when speeding occurs and by whom. We posit that link length and demographics shape the decision to speed.

Results: Speeding is widespread under both high speed limits (e.g., 60 mph [97 km/h]) and low speed limits (less than 25 mph [40 km/h]); in contrast, speeding is less common at 30–35 mph (48–56 km/h). The results suggest that driving patterns depend on the road type. We also find that when there are many intersections, the average link speed (and speeding) drops. Long links are conducive to speeding. Younger drivers and more educated drivers also speed more, and speeding occurs more often in the evening.

Conclusions: Road design and link length (or its converse, frequency of intersections) affect the likelihood of speeding. Use of increasingly available GPS data allows more systematic empirical analysis of designs and topologies that are conducive to road safety.

Safety in Numbers and Safety in Congestion for Bicyclists and Motorists at Urban Intersections

Recent working paper:

The rate of number of crashes to traffic flow
The rate of number of crashes to traffic flow

This study assesses the estimated crashes per bicyclist and per vehicle as a function of bicyclist and vehicle traffic, and tests whether greater traffic reduces the per-car crash rate. We present a framework for comprehensive bicyclist risk assessment modeling, using estimated bicyclist flow per intersection, observed vehicle flow, and crash records. Using a two-part model of crashes, we reveal that both the annual average daily traffic and daily bicyclist traffic have a diminishing return to scale in crashes. This accentuates the positive role of safety in numbers. Increasing the number of vehicles and cyclists decelerates not only the probability of crashes, but the number of crashes as well. Measuring the elasticity of the variables, it is found that a 1% increase in the annual average daily motor vehicle traffic increases the probability of crashes by 0.14% and the number of crashes by 0.80%. However, a 1% increase in the average daily bicyclist traffic increases the probability of crashes by 0.09% and the number of crashes by 0.50%. The saturation point of the safety in numbers for bicyclists is notably less than for motor vehicles. Extracting the vertex point of the parabola functions examines that the number of crashes starts decreasing when daily vehicle and bicyclist traffic per intersection exceed 29,568 and 1,532, respectively.

Street wars 2035: can cyclists and driverless cars ever co-exist? | Guardian Cities

Laura Laker at Guardian Cities writes: Street wars 2035: can cyclists and driverless cars ever co-exist? | Guardian Cities

“Driverless cars appear unstoppable – except of course you can simply walk in front of one and force it to brake. Could this conundrum eventually mean a return to a dystopian world of segregated urban highways?”

I was interviewed, my quotes below …

A visualisation of data captured by an autonomous vehicle. Photograph: Elijah Nouvelage/Reuters
A visualisation of data captured by an autonomous vehicle. Photograph: Elijah Nouvelage/Reuters

Or how about prosecuting pedestrians or cyclists who get in the way of driverless cars? David Levinson, a professor at the School of Civil Engineering at the University of Sydney, is broadly supportive of AVs, but says: “It’s very big brother like, there’s a question of safety v freedom. How much risk to endanger yourself are we going to let you take?”

Thinking back to the kids stopping driverless cars on our imaginary future street, Levinson sees a future where blocking a driverless car could even be criminalised. “The car has a camera and the picture will be sent to the police department, and the police department will come and arrest you for annoying an autonomous vehicle.”

Given these challenges, experts including Hickman and Levinson believe segregation and AV-only roads are inevitable. But wouldn’t that risk a return to the urban dystopia of the 1960s and 70s, when planners crisscrossed cities with elevated highways and erected barriers around roads with the aim of improving safety? The unintended consequences were fast, aggressive driving, and the splitting in two of countless communities.

“I think there will be some roads that will be transformed to higher speed roads,” says Levinson. “I’d be sceptical of someone who says we will not do any of that. But if you can move traffic away from the lower speed streets that pedestrians and cyclists want to use, that’s an improvement.”

Hickman believes “the case is overwhelming against AVs” but fears the powerful motor industry lobby means there is so much private and government money already at stake that the rise of driverless cars would be hard to stop.

Road network structure and speeding using GPS data

Recent working paper

Percentage of speeding across speed limit zone
Percentage of speeding across speed limit zone

This paper analyzes the relationship between road network structure and the percentage of speeding using GPS data collected from 152 individuals over a 7 day period. To investigate the relationship, we develop an algorithm and process to match the GPS data and GIS data accurately. Comparing actual travel speed from GPS data with posted speed limits we measure where and when speeding occurs, by whom. We posit that road network structure shapes the decision to speed. Our result shows that the percentage of speeding, which is calculated by travel distance, is large in high speed limit zones (e.g. 60 mph ) and low speed limit zone (less than 25 mph); in contrast, the percentage of speeding is much lower in the 30 – 50 mph zone. The results suggest driving pattern depends on  the road type. We also find that if there are many intersections in the road, average link speed (and speeding) drops. Long links are conducive to speeding.

Car Emissions vs. Car Crashes: Which One’s Deadlier? The answer actually surprised us.

Eric Jaffe at CityLab discusses my post from Monday:

The ever-thought-provoking David Levinson posed a question at his Transportationist blog earlier this week that’s worth a longer look: Are you more likely to die from being in a car crash or from breathing in car emissions? If your gut reaction is like mine, then you’ve already answered in favor of crashes. But when you really crunch the numbers, the question not only becomes tougher to answer, it raises important new questions of its own. …

You don’t have to live like a refugee |

Pedestrian refuge island, London, preceding the onslaught of the motorcar (via Flickr)
Pedestrian refuge island, London, preceding the onslaught of the motorcar (via Flickr)

Listen, it don’t really matter to me baby

You believe what you want to believe

You see you don’t have to live like a refugee (Don’t have to live like a refugee)

Yeah Somewhere, somehow, somebody

Must have kicked you around some

Tell me, why you wanna lay there

Revel in your abandon

It don’t make no difference to me baby

Everybody’s had to fight to be free

You see you don’t have to live like a refugee (Don’t have to live like a refugee)

Now baby you don’t have to live like a refugee (Don’t have to live like a refugee) No!

Tom Petty & The Heartbreakers – Refugee Lyrics | MetroLyrics

I think Tom Petty speaks not of the oppressed living in third world conditions, but rather his girlfriend. The lyrics however apply to the pedestrian trapped on refuge island between two stream of traffic (perhaps this post should be Islands in the Stream – nah)

The pedestrian refuge island allows the pedestrian to cross some of the lanes of a roadway without crossing all of the lanes of a roadway. If the lanes are going in two directions, this might decrease the travel time to cross the street, by increasing the likelihood of finding a safe gap in traffic (since you are more likely to find an acceptable gap in fewer lanes than more lanes) and reduce the number of objects the pedestrian is looking for.


The refuge island is presumably a safety improvement (the evidence is that all else equal, it is: See, e.g. Retting et al. (2003) for a review of this and other crash counter-measures). However pedestrians with pedestrian refuge islands may also be more aggressive and ignore traffic control devices since there is a refuge only a few lanes away. (I have observed this to happen almost daily), particularly when lights are timed with long cycles (e.g. resulting in waits > 50s). It also adds to the total crossing distance, and potentially time if there are no cars (remember most roads are empty most of the time). But these effects are smaller than the main  safety benefit.

So far, so good – safety first and all. However, the existence of the refuge island makes it possible for the traffic engineer, and worse the driver, to even further subjugate the needs and rights of pedestrians. It creates an environment where the pedestrian must seek refuge from oncoming traffic (which implicitly has the right-of-way, rather one where the motor vehicles must yield to pedestrians who seek to cross.

Charlotte Complete Streets-Rozzelles Ferry Road  Charlotte completely revamped Rozzelles Ferry Road. The streetscape was enhance by the addition of street trees and planting strips, while pedestrian crossing opportunities--as indicated by the crosswalk and corresponding refuge median--were added along the length of the road to make walking a breeze. The bike lanes facilitate cycling on road that previously unsuitable for riding. Photo: (via Flickr, Creative Commons license)
Charlotte Complete Streets-Rozzelles Ferry Road
“Charlotte completely revamped Rozzelles Ferry Road. The streetscape was enhance by the addition of street trees and planting strips, while pedestrian crossing opportunities–as indicated by the crosswalk and corresponding refuge median–were added along the length of the road to make walking a breeze. The bike lanes facilitate cycling on road that previously unsuitable for riding.”
Photo: (via Flickr, Creative Commons license)

This is a problem of first best and second best. In a second best world, where pedestrians have no rights, this is the literal life-jacket being thrown to them so they don’t sink in the traffic stream. In a first best world, there would be no stream in which to sink. Life should not be a game of Frogger.

cross-posted at

Sioux City Safety

Sioux City has automatic red-light running detection cameras. These are doing their job. In the Sioux City Journal by Molly Montag article on the topic, the facts are all clear, unfortunately the headline takes the small negative instead of the large positive as the lede: “Sioux City data: Rear-end crashes increased at 5 red-light intersections”

"A truck passes a traffic safety camera at Lewis Boulevard and Outer Drive in Sioux City on April 8. State data show rear-end crashes increased at five of eight Sioux City intersections where red-light cameras were installed. ... photo by Tim Hynds, Sioux City Journal file"
“A truck passes a traffic safety camera at Lewis Boulevard and Outer Drive in Sioux City on April 8. State data show rear-end crashes increased at five of eight Sioux City intersections where red-light cameras were installed. … ” photo by Tim Hynds, Sioux City Journal file”

In Sioux City’s case, new police data obtained by the Journal show a 40 percent decrease in crashes from motorists running red lights at intersections with cameras and a 15 percent reduction in all crashes. Iowa Department of Transportation data also show a decline in accidents involving red-light running. “In general, this is a positive result, as rear-end crashes (though not desirable) are not as severe (or dangerous) as right-angle crashes that might otherwise occur,” David Levinson, professor of transportation engineering at the University of Minnesota, said in an email.

University of Missouri-St. Louis transportation studies professor Ray Mundy said rear-end crashes typically increase after systems are installed and drivers slam on the brakes when they see the camera flash. Such accidents usually decrease over time as people get used to driving through camera-controlled intersections. He echoed Levinson in saying more slower-speed rear-end crashes are a tradeoff for reducing higher-speed, T-bone crashes that happen when drivers run red lights.

The University of Minnesota’s Levinson agreed the decrease in red-light crashes showed the systems improved safety.

“That is the important takeaway,” he wrote.

I looked at the data. While in my analysis the total number of crashes did not change, it is clear that automated traffic enforcement reduced the number of “ran traffic signal-involved” crashes and increased the number of “rear-end-involved” crashes. The differences before and after installation by crash-type are statistically significant and meaningful in both cases. This is consistent with general results nationally about the effects of automated traffic enforcement. (And what you would expect if there were more sharp braking at intersections as people strive to avoid fines).

Cost to users is a transfer to the city (and should otherwise reduce some other taxes the city is collecting), and though there is some cost to administering the system, that is outweighed in general by the safety benefit.