Road trauma is a global public health problem that results in over 3000 deaths per day [1
]. Moving vehicles are the source of kinetic energy (EK) that causes injury in motor vehicle crashes (MVCs) and the phrase “speed kills” highlights the central role of vehicle speed in road trauma. Higher speed means more EK and more severe injury in event of a crash, regardless of the cause. Furthermore, higher travel speeds make the task of driving more difficult, because drivers must perceive, interpret, and respond to relevant stimuli at a faster rate. In complex driving environments, this may overwhelm a driver’s perceptual or cognitive capacity, resulting in failure to recognize or respond to hazards [2
]. Even when the driver perceives a hazard and responds appropriately, higher speed results in greater distance travelled by the vehicle during perception and reaction times [5
], and in exponentially greater braking distance [6
]. In addition, higher speeds make it more difficult to negotiate curves or manoeuver around road hazards, and faster vehicles are more difficult for other road users to avoid.
Most research shows that higher traffic speed results in higher rates of crashes and injuries. In 1982, Nilsson studied a series of speed limit changes on rural highways in Sweden. He found that, for a given highway, the rate of fatal crashes was proportionate to the fourth power of mean traffic speed, the rate of serious injury crashes to the third power, the rate of injury crashes to the second power, and the rate of property damage only crashes were proportionate to the first power of mean traffic speed [7
]. According to this model, a 5% increase in mean traffic speed (e.g., from 100 to 105 kph), would result in a 22% increase in fatal crashes. Several meta-analyses found that Nilsson’s “power model” is still a good fit for predicting the road trauma rate following speed limit changes on rural highways around the world [8
The relationship between speed and injury is complex [11
]. Roads can be designed for safe, high speed travel by controlling access points, using barriers to separate opposing traffic and to prevent drivers from running off the road, providing adequate lane width and gradual curves, and including crash mitigation features [12
]. Baruya analyzed speed and crash data from 139 European rural two lane highways and found a higher frequency of injury crashes in roads with more junctions (i.e., poorly controlled access) and/or narrower lanes, as well as roads with higher speed limits [16
]. The role of road design is reflected in the risk of fatality after a crash on different types of high speed roads. French researchers found that crashes on rural highways were more than twice as likely to be fatal as crashes on high speed motorways. Motorways have median barriers to prevent head-on collisions and controlled access to prevent side impact (T-bone) collisions [17
]. Crash mitigation features of the vehicle (e.g., airbags) or road (e.g., energy absorbing barriers) can also reduce injury severity following a collision. However, these measures are insufficient to prevent injuries in very high speed collisions or in certain types of crashes [18
]. The risk of fatality for properly restrained motor vehicle occupants increases steeply at speeds >50 kph in side impact collisions and at speeds >70 kph in frontal collisions [20
]. Conversely, side swipe or rear end collisions between vehicles moving at similar speeds might not cause injury even in higher speed collisions.
There are several approaches used to set highway speed limits. North American speed limits have traditionally been set according to the 85th percentile speed of vehicles in “free flowing traffic” [22
]. The rationale is that this speed is considered safe by most drivers and extensive enforcement will be not required, since most drivers “naturally” travel at or below this speed. Speed limits can also be set according to roadway features, such as road geometry (narrow, curving roads have lower limits than wide, straight roads) or roadside development (lower speed limits in areas with more “urban” roadside development and/or more driveways). Another approach is to choose “optimal speed limits” that minimize competing costs, such as travel time, crashes, noise, pollution, and road maintenance, as well as safety [9
]. Jurisdictions with the safest roads approach road safety from a “safe systems” perspective [21
] and set speed limits based on the type of crashes that could occur and the human body’s capacity to tolerate kinetic energy in different crash types. These limits are 30 km/h in areas where pedestrians or bicyclists may be struck by motor vehicles, 50 km/h in intersections where side impact vehicle–vehicle collisions may occur, 70 km/h on undivided highways, where head-on crashes may occur, and above 70 km/h on divided highways, where a median or guard rail provides protection from head-on crashes [6
In Canada, vocal motorist groups such as SENSE (Speed Education Not Speed Enforcement) in British Columbia (BC) [27
] and Stop100 in Ontario [28
] lobby for higher speed limits and reduced speed enforcement. Proponents of higher speed limits note that road trauma is decreasing [29
] despite increased traffic speed [31
]. Because modern vehicles are safer and handle better at high speed, they argue that older studies, such as those evaluating the US National Maximum Speed Limit (NMSL), no longer apply. Many proponents of higher speed limits suggest that increasing speed limits will reduce speed variance and therefore decrease dangerous encounters between vehicles, and ultimately improve road safety. The thinking is that drivers exceeding the speed limit are driving at their comfort level and will continue at that speed after the limits are increased, whereas slower drivers will drive faster if the limits are increased [27
]. Indeed, there is evidence that roads with higher speed variance have higher crash rates [19
]. However, instead of decreasing speed variance, the evidence suggests that higher speed limits either have no effect [37
], or increase variance [19
]. Another pro-speed argument is that, when speed limits are higher, fewer drivers exceed the speed limit, so police can divert enforcement efforts onto more dangerous roads [33
]. Some also suggest that increased freeway speed limits will divert traffic from unsafe secondary highways onto safer freeways. However, Grabowski (2007) analyzed US crash statistics from 1982–2002 and found that the repeal of the NMSL was associated with a 36% increase in fatalities on rural interstates with little support for any decrease in non-interstate driving or fatalities [42
On 2 July 2014, following a public consultation, the British Columbia (BC) Ministry of Transportation and Infrastructure (MoTI) increased speed limits on 1300 km of provincial highways (9% of BC’s paved highways). They increased the maximum speed in BC from 110 kph to 120 kph—the highest in Canada. MoTI stated that the goal was to improve road safety by reducing speed variance [43
]. This manuscript reports findings from a comprehensive evaluation of BC’s new speed limits. We studied adverse motor vehicle incidents (MVIs), including ambulance calls for road trauma, auto-insurance claims, and police-reported crashes occurring on the affected road segments. To account for potential spillover effects [44
], we also studied MVIs on nearby roads and across the province. We use an interrupted time series approach to account for pre-existing trends and we include gasoline sales in our model to account for changes in travel. Our evaluation will study the effects of BC’s speed limit increases on mean traffic speed and variance and will look for changes in MVI rates on secondary highways that were not directly affected by the speed limit changes.
The paper is organized into the following sections: Methods, Results, Discussion, Limitations, and Conclusions. The Methods section describes the data sources used in this evaluation and our analysis plan. In the Results section, we present our findings in detail. The Discussion section interprets our findings in the context of previous research. In the Limitations section, we briefly discuss how shortcomings of available data may have affected our findings. Finally, in the Conclusions, we succinctly summarize key findings and give recommendations.
Over the course of the study, there were annual averages of 265,187 crashes resulting in an insurance claim (2000–2016), 31,264 ambulance dispatches for road trauma (2004–2016), 41,733 police reported crashes (2000–2015), and 331 fatal crashes (2000–2015). Location data were missing for 29% of insurance claims, 7% of ambulance dispatches, 21% of police reported crashes, and 17% of fatal crashes (Table 1
According to our model, there were large and statistically significant increases in total insurance claims (43.0% increase; 95% CI = 16.0% to 76.4%, p
= 0.001)), injury claims (30.0%; 95% CI = 9.5% to 54.2%, p
= 0.003), and fatal crashes (118.05; 95% CI = 10.9% to 225.1%, p
= 0.031) on affected road segments. There was no significant change in ambulance dispatches on affected road segments (8.8%; 95% CI = −7.0% to 27.3%, p
= 0.291). For nearby road segments, our model estimated a 25.7% increase in insurance claims (95% CI = 16.1% to 36.1%, p
< 0.001). There were no other statistically significant changes on nearby segments, but it is worth noting that the point estimate for change in fatal crashes on nearby segments was −46.7% (95% CI = −115.6% to −22.2%, p
= 0.184). However, there were more fatal crashes on affected segments than on nearby segments and, in a model that included both affected and nearby segments, there was a 39.9% percent increase (95% CI = −18.5% to 98.2%, p
= 0.180) in fatal crashes. At the provincial level, our model indicated a slight increase in casualty claims (4.8%; 95% CI = 1.6% to 8.1%, p
= 0.003). Table 3
summarizes these results. Figure 2
and Figure 3
illustrate the change in fatal crashes on affected and nearby segments in the year before versus after the speed limit changes. Figure 4
shows time series plots.
Speed data from the permanent vehicle count stations showed small increases in estimated travel speed and speed variance, on both affected and non-affected segments, in the year following the speed limit increases, compared to the year before. These changes were higher in spring and winter (Table 4
, Figure 5
Our evaluation has several limitations related to data quality and completeness. In particular, the findings related to auto-insurance claims should be interpreted with caution. In 2013, ICBC transitioned to a new system for reporting insurance claims. Although ICBC has made every effort to ensure that crash counts and location data generated from the new and old system are comparable, we recognize that claims data from 2014 onward may not be directly comparable to prior data. As a result, changes in the number of claims on affected segments in 2014 may be due to the data reporting system rather than the speed intervention. However, this explanation does not consider the fact that there was no change in claims at the provincial level. Note that this limitation only applies to auto-insurance claims and not to fatal crashes, nor to ambulance dispatches. Another limitation is the large percentage of crashes with missing location data. There were large improvements in location data for auto-insurance claims over time. Although our model included a percentage of crashes with missing location data, it is conceivable that the increase in insurance claims for crashes on affected segments is due to more mappable crashes, especially if there is a location bias in the missing data. Note that the percentage of fatal crashes with missing location data is fairly stable over time, so this problem is less likely to affect our conclusions for fatal crashes. We should also note that missing location data do not affect our conclusions for ambulance dispatches, since we restricted that analysis to events occurring after 2010, when all dispatch records included the GIS location that the ambulance was sent to. Gasoline sales data also have limitations. Sales are recorded in the month the sale occurred and do not necessarily reflect fuel used in that month. Furthermore, since some gasoline is used for other purposes, such as chain saws, lawn mowers, and mining equipment, sales volumes may not precisely measure fuel used for road transportation. However, most gasoline sales are for vehicles and trends in gasoline sales are likely to reflect temporal changes in vehicle kilometers travelled.
Vehicle speed data had many limitations. Speed data were aggregated into hourly speed bins instead of providing actual speed of individual vehicles. This was a particular problem for vehicles in the fastest speed bin and, as a result, we have little information on the speed of the fastest vehicles. Another major limitation was that the binning scheme was inconsistent with different binning schemes used at different count stations and over time on the same count station. Missing data were a limitation at many count stations. A final limitation is that there were relatively few count stations on affected road segments, providing a very limited picture of speed changes across the affected segments. Because of these limitations in speed data, we chose to report simple descriptive statistics pre versus post speed limit changes without any tests of significance.
Following the increase in rural highway speed limits in British Columbia, there was a marked deterioration in road safety on the affected roads. The number of fatal crashes more than doubled (118% increase) on roads with higher speed limits. Affected roads also had a 43% increase in total auto-insurance claims and a 30% increase in auto-insurance claims for injuries due to crashes.
Evidence of a spillover effect (i.e., where higher speed limits result in increased travel speed and more crashes on nearby roads) was mixed. In support of a spillover effect, we found a 26% increase in auto-insurance claims on nearby road segments. However, going against a spillover effect, we also found a reduction (although not statistically significant) in fatal crashes on nearby road segments. It should be noted that the absolute increase in fatal crashes on affected roads was larger than the decrease on nearby roads, and there was a net increase in the total number of fatal crashes that occurred on either affected roads or nearby roads.
The speed limit increases generated vigorous public debate, with pro speed advocates claiming, for example, that slower drivers were as dangerous as speeding drivers. There was concern that this “pro-speed rhetoric” would result in increased travel speed and more crashes across the province. Fortunately, there was only limited evidence of worsening road safety at a provincial level. We did find a 4.8% increase in auto-insurance claims for injuries due to crashes at a provincial level, but no significant worsening in other crash indicators.
Based on our findings, we recommend that British Columbia roll back the 2014 speed limit increases. Future speed limits should be set in accordance with the safe systems approach and not based on the 85th percentile of summer travel speed. Other jurisdictions, especially those with harsh winter climates or with highways that traverse mountainous terrain, should learn from this experience and resist pressure from pro speed advocates to raise speed limits without due consideration of road safety.