j ourna l h o mepage: www.elsev ier .com/ locate /aap
nalysis of factors that increase motorcycle rider risk compared to car driver risk
ichael D. Keall a,∗, Stuart Newsteadb
Wellington School of Medicine and Health Sciences, Otago University, PO Box 7343, Wellington South, New ZealandMonash University Accident Research Centre, Melbourne, Australia
r t i c l e i n f o
rticle history:eceived 31 July 2010eceived in revised form 13 March 2011ccepted 4 July 2011
eywords:
a b s t r a c t
As in other parts of the Western world, there is concern in New Zealand about increasing popularity ofmotorcycles because of potential increases in road trauma. This study sought to identify important factorsassociated with increased risk for motorcyclists to inform potential policy approaches to reduce motor-cyclist injury, such as changes to motorcyclist licensing, training and education. Using data extractedfrom a register of all New Zealand licensed motor vehicles that were matched to crash data, statistical
njuryotorcycle
tatisticsrashworthinessisk
models were fitted to examine patterns of motorcycle risk in comparison with small cars. These showedgenerally elevated risks for motorcyclists compared to cars, but particularly elevated risks for motorcycleowners aged in their 20s or who lived in more urbanised settings. In crashes, motorcyclists have littleprotection from injury, putting the motorcyclist at high risk of injury. When comparing new motorcycleswith new cars, the odds of fatal or serious injury to a motorcycle rider involved in an injury crash werealmost eight times the odds for a car driver.
. Background
Motorcycles are relatively cheap to run and, particularly whenetrol prices are high, they can offer an attractive alternative meansf transport to some drivers who would otherwise use a car. Whathould be a major disincentive to motorcycle riding is the high riskf fatal and serious injury that motorcyclists experience, estimatedo be more than 20 times the risk of passenger vehicle drivers peristance travelled in Australia (Federal Office of Road Safety, 1997).ome 20 years ago, New Zealand motorcycle ownership was muchigher than it is currently, with disturbingly high motorcyclisteath rates: throughout the 1980s, there were at least 100 motor-yclist deaths annually (Ministry of Transport, 2009). A graduatedicensing scheme was introduced in 1987 to restrict the exposuref learner motorcyclists to certain riding situations thought to be ofigher risk (night riding; riding at speeds in excess of 70 km/h; rid-
ng with a passenger; riding motorcycles with capacity greater than50 cm3), which has been shown to be associated with a decline inospitalised injuries for young riders (Reeder et al., 1999).
Over the 1980s and 1990s, motorcycle ownership was graduallyeclining, accompanied by falling casualty rates (see Fig. 1). More
ecently, Fig. 1 shows that motorcycles have been growing in popu-arity and casualties have now started to increase correspondingly.he disturbing prospect of a return to historical patterns of high
motorcycle casualty rates provides motivation to study patterns ofmotorcycle ownership, usage and risk to develop effective policiesand programmes for motorcyclists. In particular, policies and prac-tices need greater emphasis on circumstances of elevated risk tomotorcyclists; there is already substantial focus on ways to reducecar occupant risk. This paper therefore focuses on the way that risksdiffer between motorcyclists and drivers of small cars. Small cars area likely alternative vehicle to motorcycles as they share some of thesame benefits of being relatively cheap to purchase and run.
2. Data and methods
New Zealand has very good data for studying risk by vehicletype. Each vehicle driven on public roads is legally required to belicensed and, excepting mopeds (defined below), is also requiredto be inspected periodically by a mechanic to certify that there areno significant safety-related problems with the vehicle. Vehiclesless than six years old are inspected annually and older vehiclesare inspected every six months. There are fines for drivers whoare caught driving a vehicle that is either unlicensed or does nothave its certificate of roadworthiness (the “Warrant of Fitness”).A driver can also be fined by police for driving a vehicle that ishas become un-roadworthy (e.g., with worn tyres) since its inspec-tion. At the time of these periodic vehicle inspections, odometerreadings are recorded by the inspecting mechanic. These data,
along with the results of the vehicle inspection, are entered on-lineand stored on the Motor Vehicle Register (MVR). Estimating vehi-cle distance driven is possible because two consecutive odometerreadings for any vehicle inspected at least twice can be extracted
Vehicle age was allowed to enter both models as a continuous
ig. 1. Number of new motorcycle/moped registrations and number of motorcyclistasualties by year in New Zealand (Ministry of Transport, 2009).
rom the database, along with the dates of the inspections. This pro-ides an estimate of distance driven over the period between thenspections. Such estimates can then be converted into estimatesf annual distance driven, providing a measure of exposure to risk.
The study was a population study of New Zealand small cars,otorcycles and mopeds licensed in the years 2005 and 2006.
he main data source for this study, the MVR, contains data onicensed vehicle type, age, make and model as well as owner infor-
ation, including age, gender and home address. Land Transportew Zealand provided a file of 101,126 motorcycles and mopedsnd 800,346 small cars that were licensed during either or bothf March 2005 and February 2006. This can be considered as theomplete population of licensed vehicles of this type over this timeeriod. If, for a given vehicle, owner data had changed betweenarch 2005 and February 2006, data from the latter date was used.
his meant that the owner data studied was primarily that of early006, with relatively few vehicles with owner data from 2005 (thatere not licensed in 2006). This was considered appropriate as theid-point of the period of the crash data used (see below) was early
006.Motorcycles and mopeds were identified according to their
lassification on the MVR, subject to the definition specified onhe Land Transport New Zealand website (Land Transport Newealand, 2006). Technical requirements for licensing are differ-nt for mopeds. Mopeds are not required to undergo the periodicehicle inspection required for motorcycles and cars. For vehicleegistration and licensing, a moped is defined as a two-wheeledehicle that has a power output of 2 KW or under and a maxi-um design speed of 50 km/h or under. A vehicle with a power
utput over 2 KW or a maximum design speed over 50 km/h is aotorcycle. It is illegal to register a motorcycle as a moped.About half of the motorcycles (54%) had missing distance driven
stimates. Of the cars analysed, only 11% had missing distanceriven. The difference in the availability of these data can bettributed to the way that periodic inspections are conducted inew Zealand and also to the reliability of motorcycle odometersnd the way that odometer data are recorded, such that recordsre evidently more accurately and completely obtained from thear odometers. This differential rate of missing distance data meanshat estimates of crash risk per distance driven may have systematiciases. For that reason, the main focus of this paper is on per-vehicleisk, although comparisons of crude risks per distance travelled arehown in Table 2.
The NZ Ministry of Transport provided a file of 24,676 crash-nvolved cars and 1767 motorcycles/mopeds (together with the
egree and number of injuries, driver age and gender) for the years005 and 2006. The injury crashes analysed involved an injury thatequired medical attention to at least one of the road users involved
is and Prevention 49 (2012) 23– 29
in the crash. This study relied on ownership data contained in theMVR to classify vehicles, so we only defined those vehicles as crash-involved if a match was made between the MVR and the crash databy the license plate number of the vehicle. 28% of crashed motor-cycles 9% of crashed cars were not able to be matched by platenumber to the MVR. This differential rate of matching therefore ledto underestimated crash rates for motorcycles relative to cars, asdiscussed below.
All variables, apart from crash involvement, were derived fromthe owner details in the MVR. Age groups were defined based onpreviously established groups that have been previously found tobe relatively homogeneous in terms of risk (according to prior NZresearch, viz. Keall and Frith, 2004): 15–19, 20–29, 30–59 and 60plus. The level of urbanisation of the owner’s address was defined infour levels according to the Local Authority code. These were: Auck-land; Local Authorities of other Main Urban Areas with populationover 100,000 (“Large Urban”); Local Authorities of smaller MainUrban Areas (“Other Urban”); all other Local Authorities (“Rural”).As the vicinity of the owner’s address can be expected to be the mainarea of driving exposure, these classifications provide a proxy fortypes of road, which are known to present different levels of risk. Forexample, urban speed limit roads present the highest risk of injurycrash involvement per distance driven (Keall and Frith, 2004).
Two models were fitted, one to estimate injury crash involve-ment risk for motorcycles compared to small cars (1), the other toestimate fatal or serious (hospitalised) injury risk for the drivers ofmotorcycles and cars given crash occurrence (2).
The crash involvement risk model was estimated from data thatincluded all MVR records for motorcycles and mopeds as wellas small cars, typically cars with less than 1300 kg tare mass.Only statistically significant terms were included using backwardsvariable elimination in PROC LOGISTIC (SAS Institute, 1998). Thefactors included were: Intercept, agegrp (owner age group), town-class (urbanisation level of owner’s address), motorbike (whetherthe vehicle was a motorcycle or car), veh age (years since manu-facture, a continuous variable), owner gender, motorbike*agegrp,veh age*agegrp, motorbike*townclass. The “*” indicates an interac-tion term. The interaction terms can be interpreted as representingdifferences in the way that the vehicle and owner factors affectedmotorcycle risk compared to the way they affected small carrisk. For example, the estimated coefficients for the term age-grp*motorcycle measured how much greater the risk was formotorcycles owned by a given age group (e.g., young people) com-pared to small cars owned by that age group.
The data used in the model to estimate the odds of fatalor serious injury to the drivers or riders of the car or motor-cycle were passenger cars and motorcycles involved in crashesin 2005 and 2006 that were able to be matched to the MVRand were less than 60 years old. A variable was defined as1 if there was a fatal or serious injury to the driver/rider ofthe vehicle, defined as 0 otherwise. Only statistically significantterms were included using backwards variable elimination inPROC LOGISTIC (SAS Institute, 1998), starting with all first andsecond order terms: Intercept, agegrp (owner age group), town-class (urbanisation level of owner’s address), agegrp*townclass,motorbike (whether motorcycle or car), motorbike*agegrp, motor-bike*townclass, veh age (years since manufacture, a continuousvariable), veh age*agegrp, veh age*townclass, motorbike*veh age,owner gender, owner gender*agegrp, owner gender*townclass,motorbike*owner gender, veh age*owner gender. The final modelconsisted of the terms: Intercept, townclass, motorbike, veh age,motorbike*veh age.
variable modelled as linear in the logistic space as this was consid-ered to provide a proxy measurement for the crashworthiness ofcars. Previous work by Newstead and Watson (2005) has shown
M.D. Keall, S. Newstead / Accident Analysis and Prevention 49 (2012) 23– 29 25
Table 1Comparing 2005/2006 crashed motorcycle ownership details with driver details.
a Excludes 5 riders with missing information on sex.
hat the log of car crashworthiness can be expected to changepproximately linearly against the variable vehicle age.
. Results
Some initial data analyses were carried out before the mod-ls were fitted to compare details of crashed riders of motorcyclesith corresponding owner details. This was to show whether anal-
sis of owner data could be considered equivalent to analysisf driver/rider data. Table 1 shows that motorcycles owned byemales were nevertheless crashed by males in almost half the
njury crashes involving female-owned motorcycles. In contrast,
ale-owned motorcycles that crashed were being ridden by males4% of the time. The column for owner age “missing” includeshose motorcycles owned by companies, as well as other records
Fig. 2. Median distance ridden per year by mopeds (LHS of graph) and motorcy
for which owner age was not specified. Generally, Table 1 showsthat motorcycles owned by a given age group were also crashedby a member of that age group, particularly for those motorcyclesowned by young people (aged less than 30).
Rather than showing a comparison between owner and crasheddriver characteristics, the last section of Table 1 shows a com-parison between the owner address and the crash location. Thisis to show how well the owner address provides a proxy for theareas where the motorcycles is being ridden. As described above,owners’ addresses were divided into four approximately equaldivisions in terms of numbers of vehicles owned, in descending
order of urbanisation. Clearly, Auckland and Large Urban-ownedmotorcycles tended to be ridden (crashed) mainly in those areas.Even for urban-located motorcycles, significant percentages werecrashed in rural areas, which despite their relatively sparse resi-
cles by capacity of engine and age of vehicle. NZ data for 2005 and 2006.
2 nalysis and Prevention 49 (2012) 23– 29
dRa
ceatcmMpro
iadabgccwwriirmcFcoviafocTomecf
ccydmehaacrcTsp
mAm
otor
cycl
es
to
smal
l car
s:
nu
mbe
rs
lice
nse
d
in
NZ
2005
–200
6
by
age
grou
p, g
end
er
and
hom
e
loca
tion
;
also
mea
n
km
trav
elle
d
and
cras
h
rate
s
per
veh
icle
and
per
mil
lion
km
trav
elle
d.
Ove
rall
Ow
ner
gen
der
Ow
ner
age
Ow
ner
hom
e
add
ress
loca
tion
M
F
15–1
9
20–2
9
30–5
9
60+
Au
ckla
nd
Larg
e
urb
an
Oth
er
urb
an
Ru
ral
Car
800,
346
337,
127
374,
230
16,3
88
104,
148
361,
810
165,
558
227,
858
211,
550
173,
637
187,
301
M’c
ycle
101,
126
80,3
0312
,544
2362
9511
65,3
7283
5319
,295
24,9
6527
,239
29,6
27
cle
Car
100%
47%
53%
3%16
%56
%26
%28
%26
%22
%23
%M
’cyc
le
100%
86%
14%
3%
11%
76%
10%
19%
25%
27%
29%
Car
11,3
85
11,2
38
10,6
44
13,3
26
13,1
87
11,7
82
7705
12,0
09
10,6
79
10,5
54
12,1
96M
’cyc
le38
1337
9033
5248
71
4357
3770
2731
4231
3735
3664
3726
Car
7329
3103
3397
342
1167
3231
1179
2137
1869
1589
1734
M’c
ycle
d12
2393
318
165
191
713
4231
932
131
027
3
ap
erC
ar0.
92%
0.92
%0.
91%
2.09
%1.
12%
0.89
%
0.71
%
0.94
%
0.88
%
0.92
%
0.93
%M
’cyc
le
1.21
%
1.16
%
1.44
%
2.75
%
2.01
%
1.09
%
0.50
%
1.65
%
1.29
%
1.14
%
0.92
%
ap
erC
ar
0.80
0.82
0.85
1.57
0.85
0.76
0.92
0.78
0.83
0.87
0.76
M’c
ycle
3.17
3.07
4.30
5.65
4.61
2.89
1.84
3.91
3.44
3.11
2.47
te
per
veh
icle
, m’c
ycle
/car
b1.
3
1.3
1.6
1.3
1.8
1.2
0.7
1.8
1.5
1.2
1.0
tec
per
km, m
’cyc
le/c
ar
3.9
3.7
5.0
3.6
5.4
3.8
2.0
5.0
4.2
3.6
3.3
tes
for
two
year
s
(200
5
and
2006
);
not
e
that
this
un
der
esti
mat
es
actu
al
rate
s
as
abou
t
9%
of
cras
hed
cars
and
28%
of
cras
hed
mot
orcy
cles
cou
ld
not
be
mat
ched
.te
s
for
mot
orcy
cles
div
ided
by
that
for
cars
per
lice
nse
d
veh
icle
.te
s
for
mot
orcy
cles
div
ided
by
that
for
cars
per
km
dri
ven
or
rid
den
.t
thes
e
valu
es
for
n
wit
hin
own
er
cate
gori
es
are
dif
fere
nt
from
Tabl
e
1,
wh
ich
show
s
dri
ver
det
ails
6 M.D. Keall, S. Newstead / Accident A
ent population, contain considerable lengths of the road network.ural-owned motorcycles were often crashed in more urbanisedreas (40% in total).
The degree of exposure to risk is an important factor in theurrent analysis. Fig. 2 shows that median km driven per year gen-rally fell as the motorcycle or moped became older. This figurelso shows that the annual distance ridden generally increased withhe size of the engine. These associations mean that older, smallerapacity bikes are generally ridden less than newer bikes that areore powerful. Only about half the sample of motorcycles on theVR could provide a distance ridden estimate and only a very small
roportion of mopeds provided distance ridden: mopeds are notequired to undergo periodic vehicle inspection, which means thatdometer readings are not regularly recorded.
Table 2 shows a comparison of motorcycles to small cars, accord-ng to: numbers licensed in NZ 2005–2006 by age group, gendernd home location. Also shown are comparisons of mean kmriven/ridden and crash rates. Note that owner age and genderre only available if the owner is an individual. Vehicles ownedy companies, partnerships etc. are excluded from the disaggre-ations by owner age and gender. Rows three and four show thatar ownership was relatively gender-balanced, with 60% of smallars licensed to male owners; contrastingly motorcycle ownershipas overwhelmingly (86%) male. Similar proportions of small carsere owned by younger people as were motorcycles (10% vs. 9%,
espectively). Motorcycles were somewhat more commonly ownedn rural areas (29% vs. 26% of cars), but much less commonly ownedn Auckland than were cars (19% vs. 28% for motorcycles and carsespectively). The last two rows of Table 2 show crash rates forotorcycles divided by crash rates of small cars per licensed vehi-
le (second-to-last row) and per distance driven/ridden (last row).or example, the car crash rate (injury crash involvement per vehi-le per two years) was 0.92% overall; that for motorcycles was 1.21%verall. Dividing 1.21 by 0.92 gives a relative crash rate per licensedehicle of 1.3, meaning that motorcycles had a 30% higher crashnvolvement rate per licensed vehicle than did small cars. This rel-tive involvement rate per vehicle appeared to be slightly higher foremale owners than males, and for owners aged under 20 than allther age groups shown; owners with the lowest rate of motorcy-le crashes relative to small car crashes was the 60 plus age group.he relative crash rate appeared to decrease with decreasing levelsf urbanisation, diminishing from 1.8 in Auckland to 1.5 in otherain urban centres, such as Wellington, Christchurch etc., to an
qual rate (1.0, showing that rural owned motorcycles and smallars had approximately equal per-vehicle crash involvement rates)or rural dwellers.
Mean distances travelled (Table 2) show that on average, a smallar was driven about three times the distance that a motorcy-le was ridden. Vehicles (both motorcycles and cars) owned byounger owners (aged under 25) tended to travel higher annualistances. Motorcycles owned in Auckland were generally riddenore than those owned elsewhere, with annual km ridden gen-
rally diminishing with decreasing degrees of urbanisation. Cars,owever, had highest mean distance driven when owned in ruralreas, with Auckland-owned cars having only the second-highestnnual distance. Although the crash rate per vehicle for motorcy-les was only about 30% higher than that for small cars, when crashates were adjusted for distances travelled, the rate for motorcy-les became almost four times that of small cars (bottom row ofable 2). These higher crash rates per distance travelled were con-istent across the owner classifications available, showing similaratterns to the crash rates per vehicle.
The last two rows of Table 2 provide crude (unadjusted) esti-ates of the crash rates of motorcycles compared to small cars.s described in the Methods section, a model was fitted to esti-ate the adjusted relative risks of crash involvement per vehicle Ta
ble
2C
omp
arin
g
m
n
lice
nse
d
%
per
veh
ity
pe
Mea
n
km
n
cras
hed
Cra
sh
rate
veh
icle
Cra
sh
rate
106
km
Rel
ativ
e
raR
elat
ive
ra
aC
rash
rab
Cra
sh
rac
Cra
sh
rad
Not
e
tha
M.D. Keall, S. Newstead / Accident Analysis and Prevention 49 (2012) 23– 29 27
Table 3Estimated parameters for the logistic model of crash risk per vehicle for motorcycles and small cars. The referent levels shown in brackets for the factors listed are: age (60 + ),sex (missing), motorbike (small car), townclass (rural).
Parameter Estimate Standard error Wald chi-square Pr > chi-square
a In the vast majority of cases where owner age or owner sex was missing, the ve
sing the data set of all licensed motorcycles and cars. This modelncluded interaction terms to show which owner characteristics
ere associated with statistically significant differences in crashates between motorcycles and small cars. Owner age and ownerome address location were the only such interaction terms foundo be statistically significant in the model (see Table 3).
Table 3 shows all the terms included in the model with theirstimated parameters, standard errors and statistical significance.he levels of the variables used as reference levels are indicatedn the title for the table. Vehicle age is the only continuous vari-ble included in the model and its estimated coefficient indicateshe change in the log odds of crash involvement per year increasen vehicle age. Apart from vehicle age, the other variables refer toharacteristics of the owner. For the vast majority of vehicles forhom information on the age and sex of the owner was missing,
he vehicle was company owned,The estimates shown in Table 4 are for those variables estimated
o show a statistically significant difference in risk associated withotorcycles compared to small cars. They are the exponentiated
alues of the corresponding parameter estimates shown in Table 3.hese show that motorcycles had particularly high risk compared to
mall cars for young owners aged 20–29, with relative risk just overwice that for motorcycles owned by people aged 60 plus comparedo small cars owned by that age group. This result is particu-arly notable as young people already have a substantially elevated
able 4stimated odds ratios of crash involvement for motorcycles compared to cars –esults of logistic model showing statistically significant interactions with ownerharacteristics. 95% confidence intervals are in brackets.
risk when driving (Keall and Frith, 2004). The relatively low riskfor older motorcycle riders has been noted before in Australia(Australian Transport Safety Bureau, 2002). Compared to motor-cycles of owners aged 20–29, motorcycles owned by people agedless than 20 were estimated to have somewhat lower relative risk,although with overlapping confidence intervals. This can perhapsbe attributed to the effectiveness of the graduated licensing schemefor motorcycle riders as is discussed further below. The relative riskfor motorcycles compared to cars was also significantly elevatedin Auckland compared to the relative risk for rural-owned vehi-cles, with the relative risk gradually diminishing with diminishinglevels of urbanisation. The estimates from the model adjusted forother factors are therefore similar to the unadjusted patterns of riskshown in Table 2. Note that the apparent differences in relative riskby owner gender shown in Table 2 were not found to be statisticallysignificant once other factors were controlled for.
Studying differences in crash involvement risk for motorcyclesand small cars only provides one facet of safety differences: injuryseverity levels also vary considerably between crashes involvingthese different vehicles. Fig. 3 shows the output from the modelthat estimated the odds of a fatal or serious injury to the motorcy-cle rider or car driver, as explained by owner variables: owner age,gender, level of urbanisation of home address and vehicle charac-teristics: vehicle age, whether the vehicle was a car or a motorcycle.The statistically significant main effects and first order interactionsrepresented as odds ratios are shown in the figure. The first threebars show the estimated odds of fatal or serious injury to the driveror rider associated with the location of the owners’ address. Vehi-cles owned by Aucklanders had the lowest odds of fatal or seriousinjury given crash occurrence. Vehicles owned by rural people (thecomparison group for the model) had by far the highest odds of fatalor serious injury given crash occurrence. The odds increased withdecreasing levels of urbanisation, which is likely to be associatedwith the speed limits of the roads on which the vehicles are mainlyused.
The last three columns of the figure have a slightly more complexinterpretation as there was found to be a statistically significantinteraction between vehicle age and whether the vehicle wasa motorcycle or car. This finding can be interpreted as follows.
28 M.D. Keall, S. Newstead / Accident Analys
1.04 0.980.750.540.44
8.80
0.00
1.00
2.00
3.00
4.00
5.00
6.00
7.00
8.00
9.00
10.00
11.00
12.00
Aucklandvs rural
LargeUrban vs
rural
OtherUrban vs
rural
Motorcyclevs car Vehicle
age vs newMotorcycleage vs new
Fam
Cwttonoltoot
4
isbnTrdwhaffascsbe
cfpprett
ig. 3. Estimated odds of a fatal or serious injury to the driver/rider of motorcyclesnd cars, with 95% confidence intervals: by level of urbanisation of owner address;otorcycle compared to car; per year of vehicle age; per year of motorcycle age.
omparing a new car with a new motorcycle, the odds associatedith a fatal or serious injury were 8.8 times higher for a motorcycle
han for a car. However, there was also a small – although statis-ically significant – influence from car age, leading to higher riskf fatal and serious injury for drivers of older aged cars. There waso evidence of any effect of motorcycle age on the risk of a severer fatal injury given crash occurrence, the combined effect of theast two terms represented in Fig. 3. For a 10-year-old motorcycle,he odds of a fatal or serious injury to the rider fell to 6.9 times thedds for the driver of a car of the same vintage, and the comparativedds were estimated to continue to reduce with increasing age ofhe vehicles compared.
. Discussion
Novice motorcycle riders in NZ pass through a graduated licens-ng scheme (GDLS) with an initial “Learner” stage for at least the firstix months of licensing in which motorcycles over 250 cm3 cannote ridden, virtually no alcohol may be consumed before riding, lateight riding is restricted and speed is limited to below 70 km/h.he analysis presented above shows elevated risk for motorcycleiders aged in their 20s (and usually not subject to the graduatedriver licensing scheme). This risk was even elevated when theyere compared to cars with equally young owners, who alreadyave a well-established higher crash involvement risk (e.g., Keallnd Frith, 2003). Further research is required to establish a reasonor this, particularly whether the elevated risk is generated by dif-erent exposures (e.g., riding at night). More challenging exposuresre more likely to lead to errors in the control of any vehicle anduch errors are more likely to lead to injury when riding a motor-ycle than similar errors of control made in a car, which has greatertability due to having four rather than two wheels. As assertedy Williams et al. (2006), certain types of vehicles may imposelevated crash risks on young inexperienced drivers.
There are some limitations in the way the results of this studyan be interpreted. The choice of small cars as a comparison groupor motorcycles was designed to inform a question of interest toolicymakers. Motorcycle riding in many countries is increasing inopularity, leading to changes in road safety risks. But how is this
isk changing – in what circumstances and for which groups of rid-rs/drivers? The change in risk is ideally detected by looking at theravel modes being replaced by motorcycle riding on the assump-ion that a trip would still have occurred but by a different mode.
is and Prevention 49 (2012) 23– 29
Therefore the driving of small cars was chosen as the comparisonexposure to the riding of motorcycles, on the assumption that smallcars share some of the features that may lead people to ride motor-cycles, namely they are relatively cheap to purchase and to run. Butthere are other reasons why motorcycles are ridden, including theenjoyment of riding a two-wheeler. Also, motorcycles are ridden inmany developing countries in circumstances where driving smallcar is not a realistic alternative. In these circumstances the compar-ison of motorcycles with small cars – or with any other comparisongroup of vehicles, for that matter – is not relevant for investigatingchanges in risk.
A further limitation relates to the estimation of crash risk in thisstudy. As the matching rate of crashed motorcycles to the registerof licensed motor vehicles that provided the data for the analysisof crash risks was only 73% vs. the 91% matching rate for crashedcars, the relative risks for motorcycles compared to small cars willhave been underestimated by about 20%. The mechanism leadingto these differential matching rates is unclear, but related to theway that the licence plate of the vehicle is recorded by the policeofficer attending the crash. The second analysis, looking at the riskof serious or fatal injury given crash occurrence, will be less affectedby the matching rates as the analysis just focused on the crashedvehicles.
The higher risk found for motorcycles owned in more urbanisedareas is likely to be due to the higher levels of congestion in suchareas. Rural dwellers travel more on rural speed limit roads, asshown by the crash location of rural-owned motorcycles (Table 2).Rural-owned vehicles have higher risk of serious injury comparedto minor injury (see Fig. 3), likely to be due to travelling higherspeeds. At the other extreme of the urbanisation scale, Aucklanders– in New Zealand’s largest city – travel mostly on urban speed limitroads, encountering a high proportion of traffic stream conflicts atintersections etc., leading to a higher proportion of lower velocitycrashes. Motorcyclists are particularly vulnerable to even relativelylow velocity collisions as they have little protection compared tothat afforded to car occupants by the car chassis. The same crashwith little consequence for a car driver can be extremely harmful toa motorcycle rider for this reason. This lack of protection undoubt-edly leads to the finding of the almost 9-fold greater risk of fatal andserious injury for motorcycles compared to cars once a crash hasoccurred (Fig. 3). One important factor in the high rate of urbancrashes for motorcyclists is their visibility to other motorists. Alarge population case-control study in an urban setting (also in NewZealand) found that motorcycle riders wearing reflective or fluores-cent clothing, white or light coloured helmets and using daytimerunning lights could reduce serious injuries or deaths in crashes byup to one third (Wells et al., 2004).
The finding from the model whose estimated coefficients arein Fig. 3 that the fatal and serious injury risk increased for carswith increasing years since manufacture is consistent with previ-ous studies of crashworthiness (Newstead et al., 2004), reflectingimproved occupant protection for cars with improved engineering,technology and higher standards of regulation. There is no par-ticular reason why improved technology for motorcycles shouldresult in lower rates of fatal and serious injuries compared tominor injuries, however, as the motorcycle cannot provide thewide range of secondary safety features available in cars. There arealso currently no regulations or manufacturer initiatives focusedon motorcycle design improvements to achieve injury reductions.The principal aspect of the physical environment that can reducemotorcyclist trauma is, of course, the road environment. Consid-eration of road design is outside the scope of the present study,
but with increasing motorcycle popularity, blackspot analysis ofthe road network needs to have greater focus on the safety needsof motorcyclists. Specific road features that have been identifiedas contributing particularly to motorcycle crashes include: poor
nalysi
rmac
5
mrsctme
iciif
A
RA
M.D. Keall, S. Newstead / Accident A
oad maintenance (potholes, slippery sealants, fuel spills and looseaterial such as gravel); roadside features such as pavement ridges
nd signage; inadequate warning of potential hazards for motor-ycles (U.S. Department of Transportation, 2000).
. Conclusion
Statistical models fitted to compare crash involvement risk ofotorcycles compared to the risk of small cars showed elevated
isks for owners aged in their 20s or who lived in more urbanisedettings. In crashes, the odds of fatal or serious injury to a motor-ycle rider involved in an injury crash were more than eight timeshe odds for a car driver. It is this finding, rather than our estimated
odest increase in per-vehicle crash risk, which should motivatefforts to reduce crash rates for motorcyclists.
There will be an inevitable increase in road trauma withncreases in motorcycle riding, as indicated in this study by thelear association between increases in motorcyclist casualties andncreases in numbers of newly licensed motorcycles. This loom-ng epidemic of motorcycle injury is unlikely to diminish withoutocused efforts on the part of policy makers.
cknowledgements
We thank Members and Observers of the Used Car Safetyatings Project Steering Committee for their ongoing support ofustralasian road safety studies.
s and Prevention 49 (2012) 23– 29 29
References
Australian Transport Safety Bureau, 2002. Motorcycle Safety, Monograph 12.Federal Office of Road Safety, 1997. Vehicle Type and the Risk of Travelling on the
Road, Monograph 17.Keall, M.D., Frith, W.J., 2003. An evaluation of young drivers’ risk of crash involve-
ment with respect to driving environment and trip characteristics. In: RoadSafety Research, Policing and Education Conference, Sydney, Australia.
Keall, M.D., Frith, W.J., 2004. Older driver crash rates in relation to type and quantityof travel. Traffic Injury Prevention 5, 26–36.
Land Transport New Zealand, 2006. Factsheet 43: Mopeds: Road Rules and Equip-ment, 2007.
Ministry of Transport, 2009. Motor Vehicle Crashes in New Zealand 2008. Ministryof Transport, Wellington.
Newstead, S., Cameron, M., Watson, L., 2004. Vehicle Crashworthiness and Aggres-sivity Ratings and Crashworthiness by Year of Vehicle Manufacture: Victoriaand NSW Crashes During 1987–2002, Queensland, Western Australia and NewZealand Crashes During 1991–2002. Report No. 222. Monash University Acci-dent Research Centre.
Newstead, S., Watson, L., 2005. Trends in Crashworthiness of the New ZealandVehicle Fleet by Year of Manufacture: 1964 to 2002, Report No. 238. MonashUniversity Accident Research Centre.
Reeder, A.I., Alsop, J.C., Langley, J.D., Wagenaar, A.C., 1999. An evaluation ofthe general effect of the New Zealand graduated driver licensing system onmotorcycle traffic crash hospitalisations. Accident Analysis & Prevention 31,651–661.
SAS Institute, 1998. SAS/STAT Software: Changes and Enhancements ThroughRelease 8.02. SAS Institute Inc., Cary, NC, USA.
U.S. Department of Transportation, 2000. National Agenda for Motorcycle Safety,DOT-HS-809-156, Washington, DC.
Wells, S., Mullin, B., Norton, R., Langley, J., Connor, J., Jackson, R., Lay-Yee, R., 2004.Motorcycle rider conspicuity and crash related injury: case-control study. BMJ328, 857.
Williams, A.F., Leaf, W.A., Simons-Morton, B.G., Hartos, J.L., 2006. Vehicles driven byteenagers in their first year of licensure. Traffic Injury Prevention 7, 23–30.
