Does the improved stability offered by motorcycle antilock brakes (ABS) make sliding crashes less common? In-depth analysis of fatal crashes involving motorcycles fitted with ABS

ABSTRACT Objective: This research investigated the following issue. Though several tests indicate that motorcycle ABS may increase motorcycle stability, thus reducing the risk of a sliding crash involving braking (i.e., the rider is separated from the motorcycle and slides along the road surface prior to collision), there is limited research showing to what extent sliding crashes are reduced by ABS in real-life conditions. Methods: The Swedish Transport Administration (STA) and the Norwegian Public Roads Administration (NPRA) carry out in-depth studies for all road fatalities. A total of 38 in-depth studies with ABS motorcycles were included: 22 in Sweden and 16 in Norway (2005–2014). These were compared with 98 cases in Sweden and 32 in Norway involving motorcycles of the same types but without ABS. The data sets were analyzed separately and also merged together. The difference between the proportions of sliding crashes regardless braking was analyzed; selective recruitment was handled with a sensitivity analysis. Induced exposure was used to calculate the reduction of all crashes and those involving braking. Results: Four ABS cases (11%) involved falling off the motorcycle prior to collision, and 35% of the non-ABS crashes were sliding (P =.004). The sensitivity analysis showed that the results were stable, with a relative difference of sliding crashes ranging between 65 and 78%. None of the 4 sliding crashes with ABS occurred during braking; that is, all ABS riders who braked prior to collision crashed in an upright position. In the 4 sliding cases with ABS, the riders lost control of their motorcycles: 2 while accelerating on asphalt with very poor friction, 1 while negotiating a curve with an excessive lean angle, and 1 by abruptly releasing the throttle in the middle of a curve. Although based on a limited number of cases, the distributions of sliding and upright collisions among crashes without braking were similar, thus suggesting that the crash posture would not be affected by ABS if no braking occurred. The calculations with induced exposure showed that upright crashes with braking were also reduced by ABS; all fatal crashes, regardless of braking, were reduced by 52%. Conclusions: Though this research was based on a limited material, it confirmed that sliding fatal crashes are significantly decreased by ABS. Considering that ABS will soon be mandatory in the European Union on all new motorcycles with engine displacement over 125cc, these findings should be taken into account in the future design and testing of motorcycle-friendly road barriers and integrated protection systems.


Introduction
Motorcycles may easily become unstable (Massaro et al. 2012;Pacejka 2002;Sharp 2001). A moving motorcycle is maintained stable by the forces and moments of forces produced by the wheels rolling on the road surface, including lateral forces acting on the tires. If one of these factors is compromised, the motorcycle could capsize, and the rider is most likely separated from it, falling to the ground (Highway Loss Data Institute 2009; Seiniger et al. 2012). In this case, limited actions can be taken by the rider, who may be even more exposed at high speeds where the crash protection provided by helmets and protective clothing has a limited potential (De Rome 2011). With regard to road barriers, previous research has also shown that sliding crashes (i.e., the rider is separated from the motorcycle and slides along the CONTACT Matteo Rizzi matteo.rizzi@folksam.se Folksam Research and Chalmers University of Technology, SE-, Stockholm, Sweden. Color versions of one or more of the figures in the article can be found online at www.tandfonline.com/gcpi. Associate Editor Alessandro Calvi oversaw the review of this article.
Supplemental data for this article can be accessed on the publisher's website.
road surface prior to collision) and being ejected into the barrier imply higher injury risks than crashing upright (Daniello et al. 2014;Rizzi et al. 2012). Antilock braking systems (ABS) have been introduced to address wheel locking and may be also complemented with combined brake systems (CBS), which make it easier for the rider to distribute the brake force between the back and front wheel more optimally (Green 2006). It has been suggested that, in a braking situation that would normally lock a wheel, ABS increases motorcycle stability, thus reducing the risk of a sliding crash. As early as in 1979, the Transport Research Laboratory performed braking maneuvers on a wet surface with a prototype version of motorcycle ABS, showing that falling off the motorcycle due to wheel locking was eliminated (Watson 1979). Though more recent tests seem to support these findings (Anderson et al. 2010;Gail et al. 2009;Green 2006;Kato et al. 1996;Vavryn et al. 2004), there is limited research showing to what extent sliding crashes are reduced by ABS in real-life conditions. A Swedish study (Olai 2011) based on interviews of 37 seriously injured riders with ABS showed that 5 (14%) fell off the motorcycle prior to collision, although it was reported that in none of these cases the riders had applied the brakes. However, this study did not include a control group including crashes with similar motorcycles without ABS, which made it difficult to draw general conclusions.
A further issue is that, at the present stage there is limited knowledge regarding how sliding crashes with ABS may happen and what safety countermeasures could be improved or added to prevent them. This aspect may be of particular relevance considering the legislation voted by the European Parliament to make ABS mandatory on all new motorcycles with engine displacement over 125 cc from 2016 (European Commission 2012).
The present article aimed at addressing these questions with the following objectives.
Objective 1: Investigate the distributions of sliding and upright fatal crashes involving motorcycles with and without ABS, regardless of whether the riders braked or not. Objective 2: Describe the main characteristics of sliding fatal crashes with ABS with regard to the road environment, the riders, the motorcycles and the crash dynamics. Objective 3: Calculate the reduction of fatal crashes involving braking with ABS, compared to similar motorcycles without ABS.

Materials
Swedish and Norwegian in-depth studies of fatal motorcycle crashes from the period 2005-2014 were used. The Swedish Transport Administration (STA) and the Norwegian Public Roads Administration (NPRA) carry out in-depth studies for all road fatalities. Crash investigators at STA and NPRA systematically inspect the vehicles involved and record direction of impact, vehicular intrusion, seat belt and helmet use, airbag deployment, tire properties, etc. The crash site is also inspected to investigate road characteristics, collision objects, etc. Further information is provided by forensic examinations, witness statements from the police, and reports from the emergency services (STA 2005). Collision speeds are generally derived by vehicular deformation, and the initial driving speed is mostly based on eye-witness accounts, brake skids, etc. Precrash braking is also coded based on eye-witness accounts and brake and skid marks. The final results of each investigation are normally presented in a report. Because all fatal crashes are included in the sampling criterion, the material can be considered as fully representative for Swedish and Norwegian road fatalities. The data were collected case by case by a single analyst and cross-checked with the database containing the original data; particularly interesting cases were further discussed in a consensus group including 3 other road safety analysts. Crashes were analyzed according to the integrated safety chain illustrated in Figure 1. These were grouped depending on whether braking had occurred prior to collision, as well as the crash posture. As indicated by Bambach et al. (2010) and Daniello et al. (2014), there are 2 main crash postures in a motorcycle crash: upright and sliding. In the present article, sliding crashes were defined as when the rider was separated from the motorcycle prior to collision, thus including low-siding (for instance, when a motorcycle falls to the roadway on its side, on the inside of the corner; Bambach et al. 2010) and high-sliding (for instance, when a motorcycle flips over from the inside of the corner to contact the roadway on the outside side of the motorcycle; Bambach et al. 2010). Upright crashes were defined as when the rider was seated on the motorcycle during the crash (Bambach et al. 2010). Four combinations were labeled as follows.
• Crash type A: No braking prior to collision, rider crashed upright on the motorcycle. • Crash type B: No braking prior to collision, sliding crash; that is, rider slid on the ground into the collision partner. • Crash type C: Braking prior to collision, rider crashed upright on the motorcycle. • Crash type D: Braking prior to collision, sliding crash; that is, rider slid on the ground into the collision partner. There were a total of 22 ABS cases in Sweden and 16 in Norway involving helmeted riders of motorcycles with engine displacement >125cc. The types of motorcycles involved in these crashes were as follows (see also Table 1): • Touring • Sport touring • Standard • On/off-road • Scooter (STA database only, n = 1) • Suspersport (STA database only, n = 4) The highest vehicle age at the time of the crash was 14 years. However, due to the later fitment of ABS among supersport motorcycles, these were no older than 1 year at the time of the crash.
The same inclusion criteria were applied to all non-ABS cases; that is, only helmeted riders of motorcycles with engine displacement >125cc; same motorcycle types; and vehicle age at the time of the crash. This resulted in 98 relevant cases in Sweden and 32 cases in Norway (285 were excluded in Sweden and 206 in Norway, respectively). Table A1 (see online supplement) shows a detailed list of all make/models included in the analysis.
The databases were analyzed separately and also merged together, because the Swedish and Norwegian data sets were based on the same inclusion criteria, sampling period, and methodology (i.e., in-depth investigations of all road fatalities during the period 2005-2014).

Objective 1
Due to the limited number of ABS cases, the relative difference between the proportions of sliding crashes regardless braking was tested using Fisher's exact test (Agresti 1992). In order to verify the overall comparability of the ABS and non-ABS groups, initial checks were performed on the distributions of a number of factors such as rider age and gender, riding speed, use of protective gear, blood alcohol concentration, use of illegal drugs, etc. However, selective recruitment needed to be taken into account: the fitment of ABS is not randomly spread across the sampled population, which means that those who crashed an ABS-fitted motorcycle may have been more safety concerned than those who did not have ABS. This issue was handled with a sensitivity analysis of possible confounders: the proportions of sliding crashes were calculated for a number of different subgroups, by including only sober riders (blood alcohol concentration < 0.02% and no illegal drugs), licensed riders, cases without excessive speeding (less than 30 km/h over the speed limit), motorcycles without CBS, traction control (TC), or other than supersports.

Objective 2
The chain of events leading to sliding crashes with ABS was analyzed using the STA and NPRA crash reconstructions. Further investigations of the crash sites or vehicles could not be carried out because the most recent crash occurred in 2012. These crashes were later discussed in a consensus group.

Objective 3
The results in Objective 1 were further analyzed, because a reduction in the percentage of sliding crashes with ABS may also be due to an increase of the number of upright ones. In order to calculate which fatal crashes were actually prevented by ABS, induced exposure was used to calculate the reduction of fatal crashes involving braking (crash types C and D; see Figure 1). With this approach, the key point is to identify at least one crash type or situation in which motorcycle ABS can be reasonably assumed (or known) not to be effective. Then, the relation between motorcycles with and without ABS in that nonaffected situation would be considered as the true exposure relation (Evans 1998;Lie et al. 2006). It was assumed that all crashes without braking (regardless crash posture) would not be affected by ABS and therefore would represent the true exposure. If this assumption is correct, this means that the distributions of sliding and upright crashes with and without ABS should be similar in crashes without braking. Equations (1)-(3) below show the first steps for the calculations of the reduction in sliding crashes with braking, upright crashes with braking, and all crashes involving braking, respectively. In Eqs. (1)-(3), all crashes without braking were used as induced exposure: in other words, the basic assumption was that, if no braking occurred prior to collision, riders with ABS would have the same crash posture as riders without it.
R braking sliding = n braking sliding ABS n no braking ABS ÷ n braking sliding no ABS n no braking no ABS (1) R braking upright = n braking upright ABS n no braking ABS ÷ n braking upright no ABS n no braking no ABS (2) R braking = n braking ABS n no braking ABS ÷ n braking no ABS n no braking no ABS . ( The effectiveness of ABS in reducing fatal crashes involving braking was calculated as follows. The overall reduction of all fatal crashes with ABS (regardless of braking or crash posture) could be also calculated as follows, i.e. by multiplying the reduction of crashes involving braking (E braking ) with the overall share of these crashes.
n braking ABS +n braking no ABS n braking ABS +n braking no ABS +n no braking ABS +n no braking no ABS (7)

Objective 1
The results for the initial analysis of distributions across the ABS and non-ABS groups are presented for each data set separately and also merged together (see Table 1). In both countries, it was found that ABS riders were in general older and had a valid motorcycle driver's license for a longer time. In addition, impaired riding was less common among ABS riders and the use of full protective gear was more extensive. Though the fitment of CBS and TC was more frequent on ABS motorcycles, speed limit compliance and reported riding speeds were similar among Swedish cases. However, this was not the case in Norway, where ABS riders were reported to have better speed limit compliance (see Table 1).
It was found that the distributions of sliding and upright crashes (regardless of braking) were similar across Sweden and Norway; see Table 2. Among all 38 ABS cases, 4 (11%) involved falling off the motorcycle prior to collision, and 35% of the non-ABS crashes were sliding. This difference was statistically significant (P = .004). It was also found that in the majority of the ABS cases the rider did not brake: 15 out of 22 ABS cases in Sweden (68%) and 14 out of 16 cases in Norway (88%) did not involve braking prior to crash. Braking was coded as unknown in one Norwegian case. Table . Distributions of rider age and gender, motorcycle driver's license, motorcycle type, driving speed, speed limit compliance, use of protective gear, impaired riding, TC and CBS fitment across the ABS and non-ABS groups in Sweden, Norway, and merged.

Sweden
Norway Sweden + Norway    Table 3 shows the results of the sensitivity analysis for sober riders, licensed riders, cases without excessive speeding, CBS fitment, TC fitment, and supersport motorcycles. Overall, it was found that results were stable. The relative difference in sliding crashes ranged between 65 and 78%, although the statistical power of some subgroups was reduced due to the limited number of cases (see Table 3).
Road barriers were involved in 8 crashes with ABS (21%), 3 of which were sliding. Among the non-ABS group, 17 crashes occurred against road barriers (13%); 6 of these were sliding.

Objective 2
In-depth analysis of the 4 sliding crashes with ABS showed that these did not occur during braking. All riders were reportedly rather experienced, were sober, and used full protective gear. The 2 Swedish cases occurred under very similar circumstances, in heavy rain on roads with high traffic flow. However, the asphalt friction at the crash site was found to be below the minimum requirement set by Swedish road authorities (the minimum coefficient of friction should be 0.5 in wet conditions; for further details see STA [2014]). While accelerating, both riders lost control as the back wheel spun off due to the very poor friction and were thrown off their motorcycles. One rider crashed into a light pole, and the other impacted a wire-rope barrier. None of these motorcycles were equipped with TC.
Both Norwegian cases occurred in a right curve under dry conditions. In one case, the leaning angle of the motorcycle was excessive, the front tire lost its lateral grip while negotiating the curve, and the rider fell off to the ground. He slid into a W-beam side barrier and impacted a post.  In the other case, the rider overtook a companion in a right curve without noticing an oncoming passenger car. He then abruptly released the throttle in the middle of the curve, the motorcycle weaved, and he was thrown off it. The rider landed on the top of a W-beam side barrier on the left side of the road.
All 4 sliding cases with ABS involved impacts with roadside objects. Among the non-ABS group, these were involved in 8 sliding crashes with braking (out of 39) and in 4 (out of 6) sliding crashes without braking, respectively.

Objective 3
Further comparison between the ABS and non-ABS groups showed that the distributions of sliding and upright collisions among crashes without braking were similar (see Table 4), thus suggesting that the crash posture would not be affected by ABS if no braking occurred. This finding was used in the calculations with induced exposure; the merged results showed that upright crashes involving braking were reduced with ABS by 79% (see Table 4). As mentioned in Objective 2, no sliding crashes involving braking occurred with ABS; that is, all ABS riders who braked prior to collision crashed in an upright position.
These findings are illustrated in Figure 2, where the chain of events leading to the fatal crashes with ABS included in this material is represented. The calculations performed according to Eq. (7) showed a 52% reduction in all fatal crashes with ABS (regardless of braking or crash posture).

Strengths and limitations
The present article analyzed fatal motorcycle crashes to understand whether ABS improves stability in critical situations. Though this is supported by a number of tests (Anderson et al. 2010;Gail et al. 2009;Green 2006;Vavryn et al. 2004), to the authors' knowledge only one previous study based on serious crashes has investigated this issue (Olai 2011). The findings of the present article seem well in line with Olai (2011), showing similar distributions of sliding crashes among ABS riders: 14% in Olai (2011) and 11% in the present study. Further similarities were that the majority of ABS riders did not brake (at least 54% in Olai 2011) and that none of the sliding crashes with ABS occurred while braking. Though it should be kept in mind that the present study and Olai (2011) were based on different inclusion criteria (i.e., different injury outcomes), these results seem consistent with each other. Another important finding of the present article was that ABS reduced fatal crashes by 52%, which also seems to be in line with previous research from Sweden (Rizzi et al. 2016) and the United States (Teoh 2011), although the latter study indicated somewhat lower reductions in fatal crash rates (37%).
There are, however, a number of limitations and assumptions in the present study that need to be discussed. Theoretically, by only comparing the distributions of sliding and upright crashes, with and without ABS (Objective 1), it could be argued that upright crashes are increased; that is, a decrease in the percentage of sliding crashes could be explained by an increase in the number of upright ones. Though behavioral differences were shown not to affect the results to a large degree (see Table 3), this question could still be raised-Which fatal crashes are actually prevented by ABS and which ones are not? This issue was handled by Objective 3: though it may seem obvious to assume that crashes without braking would not be affected by a braking system, it was shown that this was actually the case (Table 4).
Those crashes were then used as induced exposure, showing significant reductions in sliding crashes.
However, it should be kept in mind that the material was limited. The present study used all in-depth reconstructions of fatal motorcycle crashes in Sweden and Norway. Though these materials are fully representative for Swedish and Norwegian fatalities, it is clear that the reductions presented in the present article do not necessarily mean that no crash had occurred at all. In other words, only fatal crashes were included, and it is possible that some of the "missing" crashes still resulted in injuries, although nonfatal ones.
Therefore, it is important that further research is carried out using materials based on a larger range of injury outcomes. It is most likely that in-depth studies would be needed, because most databases with mass data do not include the level of detail needed to classify crashes as in Figure 1 (i.e., sliding or upright, with or without braking, etc.). Other statistical methods could be used as well; for instance, regression analysis. In the present study, it was argued that the most suitable statistical analysis for Objective 1 was Fisher's exact test; that is, without assuming a normal distribution, due to the limited number of ABS cases. Though this approach by itself does not control for confounding factors, selective recruitment was handled with a sensitivity analysis of the results.
Another important issue is data quality: it is evident that a limited material would be more sensitive to miscoding in the crash reconstructions. Though precise routines have been set at STA and NPRA for data collection and crash reconstructions, the data were also cross-checked with the original database in order to ensure that these were reliable. However, it should be kept in mind that a detailed reconstruction of some motorcycle crashes may be a challenging task, due to their complexity.
Though it is rather straightforward to determine whether a collision is sliding or upright, based on the deformation of vehicles, marks on the asphalt, and the final position of the rider, this may not be the case for braking prior to collision. Skid marks are a clear indicator of braking without ABS, but these are prevented with ABS, which means that this information is mostly based on eye-witness accounts. However, in some cases the deformation of the thread of the brake lightbulb may suggest that braking did occur during an upright collision-for instance, in a headon crash-because a lit thread would be more flexible due to the heat and therefore be permanently bent due to the sudden deceleration. In the present article, comparison between the ABS and non-ABS groups showed that the distributions of sliding and upright collisions among crashes without braking were similar (Objective 3; see Table 4), thus suggesting that these particular data were reliable with regard to braking prior to collision.
A further limitation is that the initial driving speeds in the STA and NPRA reconstructions are sometimes based on eyewitness accounts. Though it could be argued that this issue would affect the ABS and non-ABS groups to a similar degree, some interesting differences were found between the Swedish and Norwegian data sets. For instance, it was found that the reported speed limit compliance among ABS riders in Sweden was almost identical to that among non-ABS riders. This was not the case in Norway, which could be explained by slightly different reconstruction techniques. On the other hand, it should be kept in mind that the Swedish and Norwegian road networks are quite different due to the nature of their landscapes. This could explain the poorer speed limit compliance among Swedish ABS riders. A further explanation may be given by the different distributions of motorcycle types between the ABS and non-ABS groups (see Table 1). In particular, the ABS group had a larger share of touring motorcycles, which have been shown to be associated with lower fatal crash rates and less risky beahviors than other types (Teoh and Campbell 2010). Though it was shown that these behavioral differences would not affect the overall results (see Table 3), further comparison between these materials is needed in order to clarify this issue.

Implications and future research
In order to understand the true implications of the present findings, these may need to be put in a broader context. Previous research has indicated that ABS may prevent crashes in the first place but may also lower the severity of the crashes that do occur (Rizzi et al. 2016). Though the possible reasons for these findings could not be fully investigated in Rizzi et al. (2016), the present article showed that ABS riders crashed more often in an upright position. In other words, some ABS riders would go from critical situations back to normal driving, whereas others would go further in the chain of events and actually crash. However, these riders would crash more often in an upright position, as shown in this article, which by itself would result in an overall lower injury outcome, even though leg injuries would not be addressed to the same extent (Rizzi et al. 2012). It has been suggested that integrated leg protectors would address the leg issue, due to the fact that the rider would more often be on the motorcycle during the crash. Though crash bars (usually made of loops of steel tubes projecting to the side of the motorcycle) have been proven not to have significant safety benefits in real-life crashes (Hurt et al. 1981), a previous study (Rizzi 2015) showed that impairing leg injuries were reduced by approximately 50% with ABS-equipped motorcycles fitted with a horizontally opposed flat-twin engine; that is, the cylinders are overhanging horizontally in front of the riders' legs (i.e., boxer-twin engine).
Considering that ABS will be mandatory in the European Union on all new motorcycles with engine displacement over 125 cc from 2016 (European Commission 2012), the combination of all of these findings may have significant implications for the future of motorcycle safety: an ABS-equipped (i.e., more stable) motorcycle with integrated protectors could give a much greater degree of protection during a crash but also create the conditions for other components of the road transport system to be more effective. In other words, this could mean that an integrated approach for motorcycle safety is actually possible, although further strategies to prevent injuries among motorcycle riders could need to be considered.
For instance, the role of the motorcycle itself could be further explored. It is likely that the development of electronic stability control for motorcycles would have significant implications from an integrated safety point of view (De Filippi et al. 2014), although the technical development of such systems may be particularly challenging (Seiniger et al. 2012). Other supporting systems could also address the portion of crashes that are not affected by ABS; that is, when the rider does not brake. Though autonomous emergency braking systems (AEB) on passenger cars have been proved to be effective in real-life crashes (Fildes et al. 2015), the development of similar technologies for motorcycles (MAEB, motorcycle autonomous emergency braking) is still ongoing with promising results (Savino et al. 2014).
In addition, motorcycle crashworthiness could be improved, at least with regard to some specific crash types (i.e., road barriers, turning cars at intersections). Today's motorcycles are not as compatible with road barriers as passenger cars are (Gabler 2007), but the development of integrated rider protectors would have great benefits considering that in future crashes the motorcycle will be likely upright through the entire chain of events as well as in the crash itself. In other words, the concept of "motorcycle-friendly barriers" could be brought to the next level of "barrier-friendly motorcycles. " Consumer testing may also be a powerful tool to encourage the development of more stable and crashworthy motorcycles.
It is evident, however, that the future road infrastructure will also play an important role for keeping motorcycles upright and for mitigating injuries during a crash. For instance, an interesting finding of the present research was that 2 (out of 4) sliding crashes with ABS were due to the very poor friction of the asphalt. Though other systems could have possibly prevented those crashes (i.e., TC), it could be argued that safety technologies on motorcycles will still need a certain level of friction to deliver their expected benefits. Although based on a very limited material, these results could suggest that proper maintenance of the road surface could be even more important for motorcycle safety in the future.
With regard to road barriers, it was found that 8 fatal crashes with ABS involved an impact with a road barrier (3 sliding crashes). Their design and testing have mainly focused on protecting riders who slide into it (Grzebieta et al. 2013). Most often, this is done by installing motorcyclist protective systems on a W-beam barrier. Though it is argued that motorcyclist protective systems do have positive effects in upright collisions as well (Nordqvist et al. 2015), the development of future road barriers will need to have much greater focus on upright crashes as well and on the possibility of interacting with protectors integrated on the motorcycles.
In conclusion, by analyzing the in-depth studies of fatal motorcycle crashes carried out by the STA and the NPRA during the period 2005-2014, it was found that • sliding fatal crashes were significantly reduced with ABS, regardless braking. • none of the 4 sliding fatal crashes with ABS involved braking; that is, all ABS riders who braked prior to collision crashed in an upright position. In these 4 cases, the riders lost control of their motorcycles: 2 while accelerating on asphalt with very poor friction, 1 while negotiating a curve with an excessive lean angle, and 1 by abruptly releasing the throttle in the middle of a curve. • upright crashes involving braking were also reduced with ABS. • all fatal crashes, regardless of braking, were reduced by 52% with ABS. • although based on a limited number of cases, the distributions of sliding and upright collisions among crashes without braking were similar. These results confirmed that ABS significantly improves stability during real-life critical situations, which should be taken into account in the future design and testing of motorcycle-friendly road barriers and integrated protection systems.