Török Á.

A Novel Approach in Evaluating the Impact of Vehicle Age on Road Safety

ÁRPÁD TÖRÖK, Ph.D. Traffic Engineering

E-mail: [email protected] Original Scientific Paper
Budapest University of Technology and Economics Submitted: 16 Dec. 2019
Faculty of Transportation Engineering and Vehicle Accepted: 7 July 2020
Engineering (KJK)
Department of Automotive Technologies
Budapest, Hungary

A NOVEL APPROACH IN EVALUATING THE IMPACT

OF VEHICLE AGE ON ROAD SAFETY

ABSTRACT to be injured during accidents. The author has also

This study examines the correlation between road suggested reducing the number of older vehicles on
accident casualties and the age of the vehicle, assuming the roads, especially vehicles older than 15 years, or
that the age of vehicles and the improvements in their improving their safety aspects. In the current years,
safety designs are related. The study evaluates the impact the safety standards of the automotive sector can be
of the interrelationship between road segment character- characterized by a permanent development. These
istics and road accident type on vehicle age at the time safety improvements include the ability of the ve-
of the accident (AVC). To analyse the nested relationship
hicle to stop, the stability control of the vehicle, the
between these variables, a multinomial logistic regres-
passive safety systems (e.g. frontal and side airbags,
sion (MML) model has been developed. The result of the
analysis also duly finds that vehicle age has an emphatic roof crush strength), the ability of the vehicle to
role in the occurrence of accidents. perceive its environment (e.g. frontal and backward
sensors), and many other improvements related to
KEY WORDS driver assistance systems. In [6] it has been em-
vehicle age; road accident; multilevel model;
phasized that the remarkable safety improvements
hierarchical structure; of the recently purchasable cars can be expected to
have a significant influence on safety. In the United
States, improvements in safety-related vehicle tech-
1. INTRODUCTION AND LITERATURE
nologies are reported to reduce road accidents from
REVIEW
30 percent in the model year 2000 car to 25 percent
Every year, road accidents cause large econom- in the model year 2008 in case of new cars [7]. The
ic and social losses to societies and individuals [1]. report adds that the improvements of the 2008 ve-
Recently, identifying the causes of road accidents hicle fleet have been estimated to have saved about
has become a crucial objective for road safety ex- 2,000 lives during the investigated evaluation peri-
perts. Generally, road accidents are not randomly od.
located events, but they are rather related to various Most of these studies have been limited and they
circumstances like humans, vehicles, and other geo- included the use of simple models, which are rather
metric and environmental properties [2]. applicable to indicate the existence of some rela-
Identifying factors related to accidents can sig- tionship between the vehicle age and the accident
nificantly contribute to achieving safety objectives. data, then accurately describe the characteristics of
Relevant studies have reported a variety of different their relationship. Some of the researchers [8] have
factors contributed to road accidents [3, 4]. Meth- applied a logistic regression model to show the im-
ods that explain accident occurrence rarely consid- provements in vehicle safety related to vehicle age
er vehicle age dependency as an important factor. and model year. However, the field of association
The related studies have found a direct relationship analysis investigating the influence of the age of a
between the age of the vehicle and the increased vehicle at the time of the crash (AVC) on the type
likelihood or severity of the accident. Blows et al. of the accident and the surrounding environment
[5] have suggested that the increased proportion of is still under-researched. From a methodological
older vehicles increases the risk for their occupants viewpoint, most of the developed accident severity

Promet – Traffic&Transportation, Vol. 32, 2020, No. 6, 789-796 789

 Török Á. A Novel Approach in Evaluating the Impact of Vehicle Age on Road Safety

models have been limited to examining the effect of which are categorized in this research as follows:
the roadway and environmental feature on a specif- Type 1 (skidding, skidding and overturning, skid-
ic accident type. These models usually use a wide ding & hitting objects), Type 2 (hitting objects in
range of logistic regression models [9, 10], assum- the carriageway), and Type 3 (run-off-road).
ing a single relationship between all of the explan- In order to find the indirect effect of the envi-
atory variables and the target variable. In contrast, ronmental conditions on AVC the roads have been
Haghighi and others have explored the nested re- divided into homogeneous segments. The segmen-
lationship between individual crash characteristics tation process included factors related to traffic
and environmental and roadway features [11]. He volumes, speed limits, roadside hazards, and the
has applied a multilevel ordinal logistic regression presence of horizontal curves, as recommended by
to address the hierarchical structure of accident data the Highway Safety Manual specifications [20] in
and its impact on accident severity outcome. order to obtain homogeneous road segments with
This research applies a multilevel multinomial distinguished environmental characteristics. The
logistic regression (MML) model developing fur- segmentation process has resulted in 576 segments
ther the methodology of Haghighi [11], to investi- with an average length of 5.3 kilometres, and aver-
gate the multilevel relationship between accident age accident content that equals 12.3 accidents per
types (ex. skidding, overturning, etc.) and roadway segment per year.
geometric and environmental conditions (ex. road- The AVC has been divided into five age groups:
side hazard, horizontal curve, traffic volume, speed 0-2, 3-5, 6-8, 9-11, and >11 years that are repre-
limit, etc.) variables and its impact on AVC, assum- sented in the following sections by AVC1, AVC2,
ing that the age of the vehicle strongly affects the AVC3, AVC4, and AVC5, respectively. The analysis
level of in-built safety technology of the vehicle and of AVC attribute treats every vehicle individually as
the adapted improvements in its safety designs [12]. a single input object being involved at least by one
injury or fatal accident.
2. DATA
The used dataset has been obtained from the UK 3. METHODOLOGY
Department of Transport website and has included Vehicle age, as a target variable, is conditioned
two types of data. The first dataset ([13-15]) has on the fact that an accident has happened. The out-
included accident information (i.e. location, date, comes of the AVC have been divided equidistant-
time, number of vehicles involved, road character- ly into five age groups (i.e. AVC1, AVC2, AVC3,
istics, presence of junctions, weather conditions), AVC4, and AVC5). Accident frequencies between
crashed vehicle characteristics (i.e. type, model, these randomly divided groups may vary for the
age, engine capacity, type of accident), [16] and given time period. Multinomial logistic regression
description of casualties (i.e. severity, age, gender, is an appropriate analytical approach in such situa-
type). The second dataset ([17-19]) has included tions when there is an outcome with more than two
traffic volume characteristics and speed limits for events that are not ordered, or where the ordered na-
different road segments. Additional roadway char- ture is not so clear. The analysis of the data showed
acteristics (i.e. horizontal curves, roadside hazard, a hidden influence of a hierarchical relationship be-
lane characteristics) have been identified with the tween explanatory variables. A two-level approach
help of the Google Earth and ArcGIS software. has been used to examine the nested relationship.
Based on the available information, homogeneous The description of the two-level model is given in
road segments have been generated during the first the following.
step of the analysis.
This study has concentrated on examining the Level-one modelling
impact of the nested relationship between various Assuming the data have a multinomial distribu-
environmental characteristics (i.e. roadway geomet- tion, the total outcome categories can be denoted as
ric characteristics, traffic volume, and speed limits) C, and each individual category can be indexed by
and accident types and their impact on AVC as a c. The probability of being in c-th category P(Y=c)
target variable applying the data of the UK motor- is πc(c=1,2,..,C). The probability of the outcome of
way roads between the years 2014-2016. The data each of the other category πc is compared against the
includes three variables describing accident types, reference category πC using the expected cumulative

790 Promet – Traffic&Transportation, Vol. 32, 2020, No. 6, 789-796

 Török Á. A Novel Approach in Evaluating the Impact of Vehicle Age on Road Safety

log link function ηc. The general single-level model Table 1 – Two-level hierarchical structure between AVC and
to link the expected values of the outcome to the the explanatory variables
predicted values of ηc can be written as follows: Analysed object Evaluated attributes
Q
- Traffic volume
hc = log a r Cc k b 0^ c h +
r
/ X q b q^ c h (1) Roadway geomet- - Speed limit
q=1
Level-two rics and environ- - Horizontal curve
where: ηc is the log odd variable of being in a partic- mental features - Roadside hazard
ular category (c) versus the reference category C; β0 - etc.
is the intercept; Xq is the applied prediction variable - Crash type (shape)
Characteristics of
of predictor q; βq is the coefficient of predictor q. Level-one vehicle involved in
- Engine capacity
- Type of vehicle
The odd ratio can result from exponentiating the a single accident
- etc.
log odd coefficients (exp(ηc)). When the odd ratio
Vehicle age at the
is less than 1, the function value of the outcome Target
time of accident
Vehicle age at the time
becomes less than 0, while in case of an odd ratio variable of accident (AVC).
(AVC)
above 1, the function value becomes positive.
group j. Assuming there are no level-two predictors,
Level-two modelling (Developing multinomial the general form of the level-two model will be as
modelling) follows.
In the developed model, the multilevel model
treats the individual vehicle crashes as a member of b 0j^ c h = c 0j^ c h + u 0j^ c h (2)
a certain group. Each group corresponds to a specif- At level-two, more level-one intercepts (β0j(c))
ic road segment, including different environmental
can be modelled as a function of level-two variance
and roadway geometric characteristics. Since the in-
(uqj) and intercept (γ0) for different group j. The gen-
dividual crashes of a given group are likely to share
eral MML model is formed by combining both lev-
similar environmental characteristics, their causing
factors are assumed to be similar, more likely to els (Equations 1 and 2) in a single formula (Equation 3).
each other than to the causing factors of other indi- h ij^ c h = log b r ^ h l + b 0j^ c h
r ij^ c h
(3)
viduals from another group. Accordingly, the data ij C

of the analysed individual crashes cannot be consid- Intra-class correlation

ered independently of other crashes from the same
Intra-class correlation (ICC) is a method to anal-
group, whose aspect would violate the applicability
yse whether groups are significantly different from
of a single-model.
each other. In other words, it identifies the appli-
In the case of multilevel analysis, the relation-
ship between different groups of vehicles involved cability of the multilevel model. According to [22],
in accidents is modelled by identifying a hierarchi- even a small ICC value can have a substantial effect
cal data system that can describe latent relationships on the model. ICC is calculated based on the es-
of the clustered dataset. This means that the outcome timated values of level-two variance (u). Equation 4
of the model can represent a nested relationship be- presents the ICC that describes the proportion of
tween the lower-level properties (level-one) of indi- variance between groups.
vidual accidents and higher-level group properties u
ICC = u + 3c .29 (4)
(level-two) related to the environment features, as c

presented in Table 1. where the variance of the logistic distribution is giv-

Contrary to the fixed-effect of the slope and en as (π2)/3 or 3.29.
intercept of the single-model, the variation of the
multilevel model is random. The variation within a 4. RESULTS AND DISCUSSION
single cluster group j is measured by the intercept
(β0j). The value of the intercept is defined as the pre- A two-level continuous mixed regression (CMM)
dicted response either in the case of zero values of model has been developed to examine the safety ef-
the predictors or at their mean value [21]. fect of roadway environmental features on the age of
The MML model is used to predict the odds of the vehicle at the time of the crash (AVC). The model
belonging to an outcome category c versus the ref- has been applied several times in order to select the
erence category C in case of every individual i in best explanatory variables and eventually excluding

Promet – Traffic&Transportation, Vol. 32, 2020, No. 6, 789-796 791

 Török Á. A Novel Approach in Evaluating the Impact of Vehicle Age on Road Safety

insignificant variables. Tables 2 and 3 present the re- average accidents (in case of “off-road environ-
sulting random and fixed effect coefficients of the ment”, the effect of accident type on age is 0.7). If
model. the type of accident is hitting objects, then it is ex-
pected to include a newer vehicle (in case of “hit-
Table 2 – The random effect variance components of the CMM
model ting objects”, the effect of accident type on age is
-0.23). It needs to be mentioned that this is not sig-
Parameter Estimate Wald Z Sig. nificant at 95% confidence level (as the Sig. is 0.193
Residual 20.581 41.333 0.000 in Table 3). This finding is consistent with previous
studies. Boodlal and colleagues have suggested that
Intercept Variance
(Cluster ID)
0.086 1.025 0.031 crash occurrence and crash type can be partly ex-
plained by road geometric features, together with
The random effect coefficients represent the lev- vehicle characteristics [23]. Haghighi, Liu, Zhang,
el-two variance components of the intercepts. As & Porter have proved that geometric features can be
mentioned before, the inter-class correlation (ICC) applied to predict typical crash types [11].
describes the proportion of variance between groups. However, this outcome has to be interpreted very
In other words, it describes the characteristics of the carefully, since the risk of a vehicle characterized
nested relationship between the level-two explan- by a certain age group being involved in an acci-
atory variables (i.e. environmental conditions) and dent is also strongly influenced by the penetration of
the target variable (i.e. AVC). According to Table 2, the given age group compared to the whole vehicle
the residual parameter describes the variance due to fleet, and the traffic performance characterizing the
differences among individuals (i.e. AVC for each specific age group [24]. According to the penetra-
vehicle) within their respective units. In accordance tion-based comparison the oldest and the youngest
with the table, there is a significant variance with- age groups are significantly under-represented in
in the groups (Wald Z=41.33, p<0.001). Similarly, case of the investigated accident types [25, 26]. The
the intercept parameter indicates that the intercepts relative under-representation of newer cars in case
vary significantly across the different clusters (Wald of the analysed crash types seems to be rationally
Z=1.03, p=0.031). The Wald Z test provides a Z acceptable, since the new safety technologies can
statistic summarizing the ratio of the estimate to its strongly support the improvement of road safety
standard error. In other words, the developed CMM [27-29].
model can explain the hierarchical relation between Another interesting finding is the relationship
the hierarchical predictors and AVC. between the driver’s age and the vehicle age at the
The CMM has been built based on four predic- time of accident. The inverse relationship in Table 3
tors (i.e. driver age, accident type, vehicle type, (-0.11) indicates that older drivers are more likely to
and car-makers), all of which are statistically sig- be involved in new vehicles’ accidents while young
nificant. The estimated values of the predictors are drivers are more likely to be involved in old vehi-
presented in Table 3. cles’ accidents. It looks like the age compatibility
Table 3 provides the estimates of the fixed-effect between the car and its driver is more favourable to
coefficients. First, it can be seen that the intercept avoid accidents. With respect to vehicle type, newer
(β0=10.37) represents the average AVC when all trucks and buses are more likely to be included in
the predictor values are zero or at their reference accidents in comparison with cars. Regarding bus-
value. In the case of Accident type, two of which es, it has to be mentioned that it is not statistically
were found to be significantly related to the AVC. significant at a 95% confidence level (as Sig.=0.220
In the case of the first accident type (skidding, and in Table 3). This can be strongly influenced by the
skidding and overturning), accident risk is propor- commercial nature of these categories since com-
tional to AVC with positive estimated values (0.46 mercially operated vehicles are usually younger
and 0.35, respectively). In other words, accidents than the individually owned cars.
of older vehicles are more likely to occur due to Table 3 also estimated (with overall significant
skidding, and skidding and overturning. Similarly, value) the AVC for 44 different car makers but with
in case of the second accident type, if an accident different models. Among those, only 16 vehicles of
occurred in an off-road environment, the age of the same makers are statistically significant for the
vehicle is expected to be older than in the case of 95% confidence level. Figure 1 gives an indication of

792 Promet – Traffic&Transportation, Vol. 32, 2020, No. 6, 789-796

 Török Á. A Novel Approach in Evaluating the Impact of Vehicle Age on Road Safety

Table 3 – Estimated values and significance of the AVC multilevel model

Parameter Estimate Sig. Parameter Estimate Sig.

Intercept 10.37 0.000 MAN -1.21 0.403
Driver Age -0.11 0.016 MAZDA -1.87 0.079
Type1= Skidding and
0.46 0.018 MERCEDES -2.17 0.027
overturning
Accident type

Type1= Skidding 0.35 0.060 MERCEDES-BENZ -5.08 0.000

Type1= None (Ref.) 0 MG -0.80 0.586
Type2= Hitting objects -0.23 0.193 MINI -0.84 0.426
Type2= Off road 0.70 0.097 MITSUBISHI -1.39 0.212
Type2= None (Ref.) 0b NISSAN -2.56 0.010
Vehicle Type= Truck -2.52 0.000 PERODUA -1.82 0.585
Car type

Vehicle Type= Bus -1.24 0.220 PEUGEOT -1.30 0.158

Vehicle Type= Car (Ref) 0 PIAGGIO -0.66 0.676
AUDI -2.86 0.004 PORSCHE 0.50 0.768
BMW -2.88 0.002 Car-maker RENAULT -1.62 0.080
CITROEN -1.99 0.040 ROVER 0.73 0.571
DAEWOO 0.83 0.621 SAAB -1.97 0.188
DAF -1.82 0.129 SCANIA -1.72 0.167
FIAT -2.20 0.027 SEAT -3.15 0.004
FORD -1.78 0.049 SKODA -2.76 0.010
Car-maker

HONDA -0.70 0.468 SSANGYONG -0.77 0.818

HYUNDAI -3.38 0.002 SUBARU 0.27 0.878
ISUZU -5.95 0.004 SUZUKI -1.71 0.102
IVECO -0.20 0.876 TOYOTA -0.51 0.589
JAGUAR -1.70 0.172 TRIUMPH -3.41 0.010
KAWASAKI -1.99 0.112 VAUXHALL -2.14 0.017
KIA -3.05 0.012 VOLKSWAGEN -1.64 0.077
LAND -1.30 0.221 VOLVO -2.73 0.009
LEXUS -0.90 0.516 YAMAHA (Ref.) 0

0
-1
-2
Estimated value

-3
-4
-5
-6
-7
VAUXHALL
SEAT

FIAT
KIA

SKODA
BMW
ISUZU
MERCEDES-BENZ
TRIUMPH
HYUNDAI

AUDI

VOLVO
NISSAN

MERCEDES

CITROEN
FORD

Car make company

Figure 1 – The estimated values by car-makers (AVC model)

Promet – Traffic&Transportation, Vol. 32, 2020, No. 6, 789-796 793

 Török Á. A Novel Approach in Evaluating the Impact of Vehicle Age on Road Safety

the distribution of the model estimated value with target variable, considering it outstandingly relevant
respect to car-makers. The first to be noted from from the viewpoint of the accidents, since vehicle
Figure 1 is that all of the estimated values are neg- age actually represents the difference in the integrat-
ative, indicating an inverse relationship between ed safety systems in case of different vehicle gener-
the AVC and car-maker. In other words, newer cars ations. Accordingly, multilevel analysis in our case
have a higher probability of an accident in case all targets the investigation of the relationship between
the analysed car-makers. the different groups of crashed vehicles by identify-
In more detail, among the car-makers of Table 3, ing a hierarchical data structure that can contribute
the new models of Ford and Citroen have the least to using the advantages of the clustered dataset. This
predicted accident probability in comparison with multilevel approach is represented by a hierarchical
their old models (-1.78, -1.99, respectively). On the structure aimed to reveal the influence of group-level
contrary, the accident probability for newer models characteristics of data. A multinomial logistic regres-
from Isuzu and Mercedes-Benz were higher than sion model has been applied to compare the different
the older models that gave them higher negative categories of AVC outcomes using the cumulative
predicted values (-5.95, -5.08, respectively) in the odd log link function.
model. These probability values are influenced on Before the summary of the final results, it would
the one hand by the changes in the penetration and be good to introduce the limitation of the study brief-
on the other hand by the safety improvement efforts ly. It is important to emphasize that in most countries,
of the given producer. For example, if a producer’s a comprehensive statistical dataset is not available,
vehicles are more attractive in the present than in the which would be capable to describe the traffic per-
past, it causes an increased number of vehicles on formance of the vehicle types, categories, and age
the road from the certain producer, and thus, it can groups. It means that we cannot differentiate the cat-
increase accident probability. Besides if a producer egories based on the driven distances, which would
efficiently allocates higher resources on the safety be crucial to evaluate the accident risks and proba-
field, it can lead to safer cars and thus smaller recent bilities related to a specific category (e.g. age groups
accident probability values compared to the past. or vehicle types). In this case, if we characterize the
Due to the introduced factors and the relatively low effect of vehicle age on traffic safety without reliable
number of data used in the analysis for the different performance data, a relevant part of the methodology
car-makers the outcome has to be handled carefully. has to be based on assumptions and estimations. On
Furthermore, the variation in the odd ratio of the the other hand, we also need to mention that a data-
crash types between different AVC groups has proved base on the level of automation (especially focusing
the nested relationship between crash types and the on ADAS) related to specific vehicle categories and
environmental conditions related to different road age groups would strongly support the analysis of the
segments, which also support the applicability of the effect of automation on traffic safety.
multilevel approach. According to the results of the However, the model parameters of the current
analysis, older vehicles are more likely to be involved investigation have been estimated using road acci-
in traditional types of accidents. On the other hand,
dent data of the UK motorway network in the period
considering the penetration-based comparison, the
of 2014-2016. The first level of the model includes
oldest and the youngest age groups are significantly
parameters related to accident characteristics, while
under-represented in the case of the investigated ac-
the second level describes the environment and geo-
cident types. For instance, skidding type of accident
metric features of road segments where accidents oc-
has a higher probability of including casualties for
curred. The result has proven a significant influence
older vehicles compared to newer vehicles.
of the AVC by applying a hierarchical structure of
the considered explanatory variables. The values of
5. CONCLUSION the variance component have supported the assump-
In this paper, a multilevel multinomial logistic tion that the intercepts vary across the different AVC
regression has been applied to analyse the nested groups describing the correlation between the envi-
relationship between crash types, roadway geomet- ronmental features and different AVC groups.
ric, environmental conditions, and their impact on Finally, I would like to introduce the applicability
the change in the age of vehicles at the time of the of the results to other countries briefly. In this case,
accident (AVC). AVC has been investigated as the it seems to be reasonable to handle the issues of the

794 Promet – Traffic&Transportation, Vol. 32, 2020, No. 6, 789-796

 Török Á. A Novel Approach in Evaluating the Impact of Vehicle Age on Road Safety

methodology and the numeric results separately. Re- ABSZTRAKT

garding the methodology, we can conclude that the Jelen tanulmányban megvizsgálom és megahtáro-
applied model can easily be adapted to other coun- zom a közúti balesetek kockázata és a jármű kora közöt-
tries, to other national datasets. Accordingly, if a ti összefüggést, feltételezve, hogy a járművek kora és a
certain country has got a detailed road accident da- vezetést támogató rendszerek fejlettsége között össze-
tabase, it can perform the introduced evaluation and függés figyelhető meg. A cikkben figyelembe veszem az
can effectively investigate the relationship between útszakaszok jellemzői és a közúti balesetek típusa közöt-
ti összefüggések hatását a jármű életkorára a baleset
AVC and automotive manufacturer companies in
idején (AVC). Ezen változók beágyazott kapcsolatának
the case of a specific national vehicle fleet with the
elemzésére multinomális logisztikus regressziós (MML)
developed model framework. On the other hand, the modellt dolgoztam ki. Az elemzés eredménye azt is ig-
introduced numeric results can only be used to repre- azolja, hogy a jármű korának hangsúlyos szerepe van a
sent the safety situation in the UK since the penetra- balesetek előfordulásában.
tion of vehicle types, age groups, and manufacturers
can vary from country to country. KULCSSZAVAK
Furthermore, based on the introduced limitations jármű kora; közúti baleset; többszintű modell;
of the model, in the future, I would like to involve the hierarchikus struktúra;
factor of traffic performance in the research method-
ology. With this, the categories based on the driven REFERENCES
distances can be differentiated, which would be an
[1] Ghadi M, Török Á, Tánczos K. Study of the Econom-
outstandingly important contribution to the evalua- ic Cost of Road Accidents in Jordan. Period. Polytech.
tion of accident risks and probabilities related to a Transp. Eng. 2018;46(3): 129-134. Available from:
specific category (e.g. age groups or vehicle types). https://pp.bme.hu/tr/article/view/10392 [Accessed 18
Besides, I would also like to develop the database fo- June 2019].
[2] Huang Z. Extensions to the k-means algorithm for clus-
cusing on the level of automation (especially focus-
tering large data sets with categorical values. Data Min.
ing on ADAS) related to specific vehicle categories Knowl. Discov. 1998;2(3): 283-304. Available from:
and age groups to support the analysis of their effect doi:10.1023/A:1009769707641 [Accessed 18 Nov.
on traffic safety. 2019].
[3] Michalaki P, Quddus MA, Pitfield D, Huetson A. Explor-
ing the factors affecting motorway accident severity in
ACKNOWLEDGEMENT
England using the generalised ordered logistic regres-
The research reported in this paper and carried sion model. J. Safety Res. 2015;55: 89-97. Available
out at the Budapest University of Technology and from: https://www.sciencedirect.com/science/article/pii/
Economics was supported by the “TKP2020, In- S0022437515000833 [Accessed 3 Oct. 2019].
stitutional Excellence Program” of the National [4] Yannis G, Athanasios T, George P. Investigation of
road accident severity per vehicle type. Transportation
Research Development and Innovation Office in Research Procedia. 2017;25: 2081-2088. Available
the field of Artificial Intelligence (BME IE-MI- from: https://www.sciencedirect.com/science/article/pii/
FM TKP2020). The research reported in this paper S2352146517307081 [Accessed 5 Oct. 2019].
was supported by HAS (Hungarian Academy of [5] Blows S, Ivers RQ, Woodward M, Connor J, Ameratunga
Science) for providing the Janos BOLYAI Schol- S, Norton R. Vehicle year and the risk of car crash inju-
ry. Inj. Prev. 2003;9(4): 353-356. Available from: https://
arship. Moreover, the author is grateful for the sup- www.ncbi.nlm.nih.gov/pmc/articles/PMC1731030/ [Ac-
port of the New National Excellence Programme cessed 1 Sep. 2019].
Bolyai+ scholarship. [6] Høye A. Vehicle registration year, age, and weight –
Untangling the effects on crash risk. Accid. Anal. Prev.
Dr. TÖRÖK ÁRPÁD, Ph.D. 2019;123: 1-11. Available from: https://www.ncbi.nlm.
E-mail: [email protected] nih.gov/pubmed/30447490 [Accessed 6 Aug. 2019].
Budapesti Műszaki és Gazdaságtudományi Egyetem [7] Glassbrenner D. An Analysis of Recent Improvements
Gépjárműtechnológia Tanszék to Vehicle Safety. Ann. Emerg. Med. 2013;61(2): 222.
Available from: https://crashstats.nhtsa.dot.gov/Api/
A GÉPJÁRMŰVEK ÉLETKORÁNAK KÖZÚTI Public/ViewPublication/811572 [Accessed 7 Sep. 2019].
KÖZLEKEDÉSBIZTONSÁGRA GYAKOROLT [8] National Highway Traffic Safety Administration. How
vehicle age and model year relate to driver injury sever-
HATÁSAINAK ÉRTÉKELÉSÉRE ALKALAMS ity in fatal crashes; 2013. Available from: https://crash-
ÚJ MÓDSZERTANI KERETRENDSZER stats.nhtsa.dot.gov/Api/Public/ViewPublication/811825
BEMUTATÁSA [Accessed 17 Oct. 2019].

Promet – Traffic&Transportation, Vol. 32, 2020, No. 6, 789-796 795

 Török Á. A Novel Approach in Evaluating the Impact of Vehicle Age on Road Safety

[9] Al-Ghamdi AS. Using logistic regression to estimate traffic-counts [Accessed 1 Aug. 2019].
the influence of accident factors on accident severi- [20] American Association of State Highway and Transporta-
ty. Accid. Anal. Prev. 2002;34(6): 729-741. Available tion Officials. Highway Safety Manual. 1st Edition; 2010.
from: https://www.sciencedirect.com/science/article/pii/ [21] Azen R, Walker CM. Categorical data analysis for the
S0001457501000732 [Accessed 3 Nov. 2019]. behavioral and social sciences; 2011.
[10] Yau KKW. Risk factors affecting the severity of single [22] Barcikowski RS. Statistical power with the group mean
vehicle traffic accidents in Hong Kong. Accid. Anal. as the unit of analysis. J. Educ. Behav. 1981;6(3):
Prev. 2004;36(3): 333-340. Available from: https://www. 267-285. Available from: https://www.jstor.org/sta-
ncbi.nlm.nih.gov/pubmed/15003577 [Accessed 3 Nov. ble/1164877?seq=1#metadata_info_tab_contents [Ac-
2019]. cessed 23 Nov. 2019].
[11] Haghighi N, Liu XC, Zhang G, Porter RJ. Impact of [23] Boodlal L, Donnell ET, Porter RJ, Garimella D, Le T,
roadway geometric features on crash severity on ru- Croshaw K, Himes C, Kulis P, Wood J. Factors Influenc-
ral two-lane highways. Accid. Anal. Prev. 2018;111: ing Operating Speeds and Safety on Rural and Suburban
34-42. Available from: https://www.ncbi.nlm.nih.gov/ Roads. Turner-Fairbank Highway Research Center. No.
pubmed/29169103 [Accessed 3 Nov. 2019]. FHWA-HRT-15-030, 2015.
[12] Greene D, Hossain A, Hofmann J, Helfand G, Beach [24] EurosStat. Available from: https://ec.europa.eu/eurostat/
R. Consumer willingness to pay for vehicle attributes: web/products-datasets/product?code=road_eqs_carage
What do we Know?. Transportation Research Part 2019.04.05. [Accessed 23 Nov. 2019].
A: Policy and Practice. 2018;118: 258-279. Available [25] Bair ST, Huang RJ, Wang KC. Can vehicle maintenance
from: https://www.sciencedirect.com/science/article/pii/ records predict automobile accidents?. Journal of Risk
S0965856417308546 [Accessed 3 Nov. 2019]. and Insurance. 2012;79(2): 567-584. Available from:
[13] Department for Transport. Road Safety Data 2014. https://onlinelibrary.wiley.com/doi/full/10.1111/j.1539-
[Online]. Available from: https://data.gov.uk/dataset/ 6975.2011.01433.x [Accessed 23 Nov. 2019].
cb7ae6f0-4be6-4935-9277-47e5ce24a11f/road-safe- [26] Knobloch K, Wagner S, Haasper C, Probst C, Krettek C,
ty-data [Accessed 1 Aug. 2019]. Otte D, Richter M. Sternal fractures occur most often in
[14] Department for Transport. Road Safety Data 2015. old cars to seat-belted drivers without any airbag often
[Online]. Available from: https://data.gov.uk/dataset/ with concomitant spinal injuries: Clinical findings and
cb7ae6f0-4be6-4935-9277-47e5ce24a11f/road-safe- technical collision variables among 42,055 crash vic-
ty-data [Accessed 1 Aug. 2019]. tims. The Annals of Thoracic Surgery. 2006;82(2): 444-
[15] Department for Transport. Road Safety Data 2016. 450. https://www.sciencedirect.com/science/article/pii/
[Online]. Available from: https://data.gov.uk/dataset/ S0003497506005935 [Accessed 23 Nov. 2019].
cb7ae6f0-4be6-4935-9277-47e5ce24a11f/road-safe- [27] Szalay Z, Tettamanti T, Esztergár-Kiss D, Varga I, Bar-
ty-data [Accessed 1 Aug. 2019]. tolini C. Development of a test track for driverless cars:
[16] Zöldy M, Zsombók I. Modelling fuel consumption and Vehicle design, track configuration, and liability con-
refuelling of autonomous vehicles. MATEC Web of Con- siderations. Periodica Polytechnica Transportation En-
ferences. 2018;235: 37. EDP Sciences. Available from: gineering. 2018;46(1): 29-35. Available from: https://
https://www.matec-conferences.org/articles/matecconf/ pp.bme.hu/tr/article/view/10753 [Accessed 23 Nov.
abs/2018/94/matecconf_hort2018_00037/matecconf_ 2019].
hort2018_00037.html [Accessed 1 Aug. 2019]. [28] Hulse LM, Xie H, Galea ER. Perceptions of autono-
[17] Department for Transport-b. GB Road Traffic Counts mous vehicles: Relationships with road users, risk, gen-
2014. [Online]. Available from: https://data.gov.uk/data- der and age. Safety Science. 2018;102: 1-13. Available
set/208c0e7b-353f-4e2d-8b7a-1a7118467acc/gb-road- from: https://www.sciencedirect.com/science/article/pii/
traffic-counts [Accessed 1 Aug. 2019]. S0925753517306999 [Accessed 23 Nov. 2019].
[18] Department for Transport-b. GB Road Traffic Counts [29] Zöldy M. Legal Barriers of Utilization of Autonomous
2015. [Online]. Available from: https://data.gov.uk/data- Vehicles as Part of Green Mobility. In: Burnete N, Varga
set/208c0e7b-353f-4e2d-8b7a-1a7118467acc/gb-road- B. (eds) Proceedings of the 4th International Congress of
traffic-counts [Accessed 1 Aug. 2019]. Automotive and Transport Engineering (AMMA 2018).
[19] Department for Transport-b. GB Road Traffic Counts Proceedings in Automotive Engineering. Springer, Cham;
2016. [Online]. Available from: https://data.gov.uk/data- 2019. Available from: https://doi.org/10.1007/978-3-
set/208c0e7b-353f-4e2d-8b7a-1a7118467acc/gb-road- 319-94409-8_29 [Accessed 28 Oct. 2019].

796 Promet – Traffic&Transportation, Vol. 32, 2020, No. 6, 789-796