Safety Science 51 (2013) 165–177

Contents lists available at SciVerse ScienceDirect

Safety Science

journal homepage: www.elsevier .com/locate /ssc i

Selection of a relevant indicator – Road casualty risk based on final outcomes

Dragoslav Kukic a,⇑, Krsto Lipovac b, Dalibor Pešic b, Milan Vujanic b

a Road Traffic Safety Agency of the Republic of Serbia, Belgrade, Serbiab University of Belgrade, Faculty of Transport and Traffic Engineering, Department of Traffic Safety, Belgrade, Serbia

a r t i c l e i n f o

Article history:Received 17 January 2012Received in revised form 7 May 2012Accepted 12 June 2012Available online 27 July 2012

Keywords:Road safetyRatesRisksRisk mappingMunicipalitiesLinear correlation

0925-7535/$ - see front matter � 2012 Elsevier Ltd. Ahttp://dx.doi.org/10.1016/j.ssci.2012.06.016

⇑ Corresponding author. Address: Mihajlo Pupin BSerbia. Tel.: +381 113117928; cell: +381 648428028;

E-mail address: kukicdragoslav@gmail.com (D. Ku

a b s t r a c t

There are no dilemmas among the academics and experts whether it is important and necessary toanalyze the road casualty risk. The road casualty risk analysis is a very efficient way of filtering the mostdangerous sections, roads or specific territories. In previous analyses of road safety in Serbia, a value andtype of a specific risk according to the size of the observed area (state, region, district, municipality),section length or the importance of a road category, were not explicitly determined. Differences in valuesof the analyzed parameters could be expressed to such an extent that the acquired values of differences,among some of the units that are being observed, represent range divided into risk bands. These differ-ences are primarily the result of the severity of injuries and types of accidents used for calculating indi-vidual risk categories. In this paper, a model for selection of an ‘‘acceptable’’ risk in selectedmunicipalities in Serbia is presented. Here presented model will be used for future researches and finalassessments of the state of road safety, i.e. for the reliable risk mapping of the Serbian municipalities. Thepractical contribution of the risk analysis is in defining a reliable way of choosing acceptable final out-comes – rates for a defined unit of observation.

� 2012 Elsevier Ltd. All rights reserved.

1. Introduction

According to the data on the number and consequences of roadaccidents, population, registered motor vehicles, road networklength, AADT value (Average Annual Daily Traffic) or kilometrage,the values of so-called ‘‘public’’, ‘‘traffic’’, ‘‘collective’’ and ‘‘dynamic’’risks can be calculated. These are the most prominent relative indi-cators or final outcomes – rates of road safety in the scientific area ofRoad safety. They are most often used for risk mapping and also asthe most important elements for describing the level of road safetyon the particular territories, roads or road sections.

Researches in which road safety levels and also safety risksamong countries have been compared (Koornstra et al., 2002;Wegman et al., 2005 and Wegman et al., 2008), used several finaloutcomes, as well as several safety performance indicators (SPIs).Namely, the following final outcomes have been used:

� Distribution of fatalities per road transport mode (passengercars, commercial vehicles, motorcycles, bicycles, pedestrians,etc).

� Fatalities per road user’s age groups.� Fatalities per different road categories (highways, main urban

streets, rural roads, etc.).

ll rights reserved.

oulevard 2, 11000 Belgrade,fax.: +381 113117298.

kic).

Wegman and Oppe (2010) state that comparing road safetyamong countries is often conducted by using indicators – rates thattake into account the number of fatalities, i.e. fatalities per popula-tion, and these indicators represent the so called public risk. AlsoWegman and Oppe (2010) are of opinion that defining a risk whichis based on the public risk has a disadvantage, as the degree ofmotorization has not been taken into account, and suggest thatindicator obtained by fatalities and vehicle kilometers should beoften used as the indicator that gives better results of road safetylevels and risk assessments. This indicator is the so called ‘‘dynamictraffic risk’’ and it takes into account the mobility of the population.However, most countries do not register data about traveled kilo-meters. Therefore, indicator that is taking into account fatalitiesand the number of registered vehicles is used instead, and indicatoris so called ‘‘traffic risk’’.

Hermans et al. (2009) state that it is necessary to analyze avail-able data in order to make a review of road safety. For the purposesof ranking and comparing road safety among countries, data relat-ing to accidents and consequences (fatalities, serious injuries andslight injuries) per population could be used.

In road safety comparison of two countries (China and USA),Zhang et al. (2010) used the ‘‘traffic risk’’ (the number of fatalitiesrelative to the number of motorized vehicles and relative to thenumber of passenger cars), ‘‘public risk’’ (number of fatalities rela-tive to the population) and the relation of the number of fatalities tothe gross national product (hereinafter referred to as GNP). Also forthe purposes of defining and comparing road safety levels between

166 D. Kukic et al. / Safety Science 51 (2013) 165–177

countries, Zhang et al. (2010) have also determined and comparedthe trends of certain analyzed indicators.

In his doctoral dissertation, Al-Haji (2007) has thoroughly ana-lyzed the exposure (degree of motorization), ‘‘traffic’’ and ‘‘public’’risk (rates of final outcomes such as number of fatalities divided bythe number of vehicles or divided by the number of citizens), aswell as trends in the degree of motorization, traffic and public risksand the ability of assessing the state of road safety.

Applying absolute figures of fatalities or other road safetyoutcomes and dividing them by a chosen exposure is the simplestand most commonly used method for comparing road safetyperformance of countries (Ekler, 2010). Exler produced risk mapsfor Belgian municipalities based on Bayes relative risks. Focus is onthe final outcomes – fatalities per traffic and fatalities per traffictrend.

The Road Assessment Programme RAP (EuroRAP, usRAP andAusRAP) produces risk maps based on accident rates that com-bined effects of behavior, road and vehicle. RAP protocol focuseson fatal and serious accidents. RAP models are generally used innational targets and those that can have life-changing conse-quences. Consideration of fatal accidents alone would severely re-strict the average accident frequency per site and make resultsmore variable (Hill, 2010). RAP models are concretely intendedfor risk mapping of roads and road sections as observation units.

Nam and Song (2008) used Bayesian spatial modeling to esti-mate and map accident risk. This model analyses output indicators(the number of fatalities, serious injuries and slight injuries) andfinal outcomes -rates. The Model is based on account for spatial

Diagram 1. A model for selection of a re

dependence in modeling and corresponding statistical inference.The model is using a Global spatial autocorrelation. This is a globalmeasurement of spatial autocorrelation over the entire observa-tions over an area of interest, used for testing spatial autocorrela-tion to detect departures from spatial randomness. Global spatialautocorrelation is a measure of the overall clustering of the data.One of the statistics used to evaluate global spatial autocorrelationis Moran’s I (Anselin, 2005).

The reviews of the latest researches gave rise to the need fordetailed analyses of each and every indicator that is included ina comprehensive road safety assessment. It also became neces-sary to include more and more indicators, according to the qual-itative road safety assessment. In almost all recent researches,final outcomes have been used for the road safety assessmentat the researched territory. Final outcomes of the road safetyin Serbia are most often represented by a road casualty riskwhere the calculated risks are classified in five standardizedclasses that are labeled with marks ranging from 1 to 5. Thelowest mark (1) is associated with the highest levels of riskand reflects the most unfavorable value of the observed indica-tor, while the highest mark (5) indicates the lowest risk andthe most favorable value of the same indicator. The classificationof the levels of road casualty risks has been made according tothe EuroRAP model.

It is very important to select the indicator for the risk mapping inthe observed area (municipalities) because of the significant num-ber of various sorts of defined risks (public risk, traffic risk, etc.).In order to assess the road safety in the observed area, in the best

levant indicator – road casualty risk.

1 PIARC 2008.ROAD SAFETY MANUAL, RECOMENDATIONS FROM THE ROAD WORLDASSOCIATION, Chapter 7 (Priority ranking).

2 Hill, 2010. EuroRAP202: Risk Mapping Manual, European Road AssessmentProgramme.

D. Kukic et al. / Safety Science 51 (2013) 165–177 167

possible and most realistic way, it will be necessary to select theroad safety indicator – risk that will assess the road safety in theobserved area in the best possible way.

2. Methodology

In this paper, a statistical analysis of the indicators – road casu-alty risks, has been made by calculating the value of linearcorrelation.

The first step in implementing a model for selecting a relevantroad casualty risk is to define an observation unit for the so calledentity (municipality, region or else). After having defined theobservation units, output indicators of road safety are being sin-gled out:

- Number of road accidents with casualties (includes the numberof road accidents with injured and the number of road accidentswith fatalities).

- Number of accidents with fatalities.- Number of fatalities.- Number of people with serious bodily injuries.- Number of people with slight bodily injuries.

In order to get final outcomes – rates, adequate parameters orexposure data are singled out and then compared to the outputindicators of road safety (values of road accidents and their conse-quences). The parameters are chosen depending on the select riskfor future researches. Thus, in order to calculate the value of publicrisk, it is necessary to get the information on the population; for acollective risk, we need data on the road network length, the lengthof a road or a road section; for calculating the dynamic traffic risk,the data on AADT (Average Annual Daily Traffic) and networklength will be necessary; and in order to calculate the traffic risk,information on the number of registered vehicles is needed. Thenumber of vehicle kilometers – the kilometrage – can be calculatedin the following way: AADT � L � 365 (veh. km/year) where L is thelength of the network (km). The number of registered motor vehi-cles within the territory of municipality can be used as a substitutePer and Al-Haji (2005), Rumar (1999), Smeed (1972), Sorensen(2002), Šliupas (2009). Transit traffic should also be added to thisvalue, i.e. by using fuel consumption data or AADT data dividedinto domicile and transit traffic, etc.

The dynamic traffic risk is still unavailable for the most munic-ipalities in Serbia, and therefore traffic risk that is calculated on thebasis of the number of registered motor vehicles within the terri-tory of municipality can be used as a substitute. For those countriesin which the motor vehicle kilometers are not available, the fatalityrate – defined as the number of fatalities per motor vehicle – willbe used instead (Wegman and Oppe, 2010).

The isolated final outcomes – rates are not the only ones, but arethe most common ones. A model for selection of a relevant risk rep-resents the calculation of five types of output indicators of roadsafety for each of the observed risks.

Input data for the observation of the public risk include: thenumber of inhabitants in a defined observation unit, the numberof road accidents and their consequences. All risks are calculateddepending on the type of road accidents and the severity of inju-ries. Those indicators are obtained on the basis of the weightednumber of casualties (1); on the basis of the absolute data onnumber of road accidents with casualties (2); number of fatalities(3); number of accidents with fatalities (4); and the number ofkilled and seriously injured (5). A model for selection of a relevantrisk as an acceptable indicator of the state of road safety is shownin Diagram 1.

Equations for the calculation of all risks which belong to thecategory of public risk, are presented in the text below.

The Eq. (1) shows the calculation of the Public risk based on theweighted number of casualties – PR WNC. This risk is based on thetype of consequences of road accidents weighted by correspondingcoefficients, depending on the level of injuries.

PRWNC ¼ Li � P1þ Si � P2þ F � P3Population

� 10;000 ð1Þ

where Li – number of slight injuries; Si – number of serious injuries;F – number of fatalities in road accidents. P1 = 1; P2 = 10; andP3 = 85. The values of coefficients that are associated with the cer-tain levels of injury, were taken from the Report of the Road WorldAssociation – PIARC, and they have been published in the documentRoad Safety Manual, Recomendations from the Road World Association(PIARC, 2008)1. The values of coefficients P1, P2 and P3 were takenfrom PIARC, because there is still no calculation of the accident costsin Serbia.

The second risk is the Public risk obtained on a basis of numberof road accidents with casualties – PR(RAc) – Eq. (2), and it is cal-culated as the ratio of the number of road accidents with casualties(number of road accidents with casualties – Num.RAc) and thenumber of inhabitants in a defined observation unit:

PRðRAcÞ ¼ Num:RAcPopulation

� 10;000 ð2Þ

The third risk is the public risk obtained on a basis of number offatalities in municipalities – PRf – Eq. (3). This risk is obtained asthe ratio of the total number of fatalities (number of fatalities –Num.F) and the number of inhabitants (population) in a definedobservation unit:

PRf ¼ Num:FPopulation

� 100;000 ð3Þ

The fourth risk in the category of public risks is the Public riskobtained on a basis of number of road accidents with fatalities –PR(RAf) – Eq. (4). The number of road accidents with fatalitiesand the number of fatalities are not the same number. These num-bers can vary by more than 10% (according to the number of roadaccidents and consequences in the Republic of Serbia from 2001 to2010 – source: Ministry of Interior of the Republic of Serbia). This iswhy it is important to monitor both of these indicators – the num-ber of road accidents with fatalities and the number of fatalities.

PRðRAf Þ ¼ Num:RAfPopulation

� 100;000 ð4Þ

where Num.RAf – the number of road accidents with fatalities; Pop-ulation – the number of inhabitants in a defined observation unit.

The fifth risk in the category of public risks is the Public risk ob-tained according to the number of fatalities and seriously injuredPRf + s – Eq. (5), obtained as the ratio of the sum of the number offatalities in road accidents (Num.F) and the number of serious inju-ries in road accidents (Num.S), and the number of inhabitants (Pop-ulation) in a defined observation unit. The researches of theEuropean Road Assessment Programme – EuroRAP2 give a specialimportance to this indicator, while the model itself is based on theanalysis of road accidents with fatalities and road accidents withseriously injuried (Hill, 2010).

PRf þ s ¼ Num:F þ Num:SPopulation

� 100;000 ð5Þ

Table 1Values of correlations of the Public risk in all municipalities in the Republic of Serbia (risk within the territory of the whole municipality).

Means Std. dev. PR WNC PRf PRf + s PR (RAf) PR (RAc) Mean value of the Public risks

Correlations (Spreadsheet1) marked correlations are significant at p < .05000 N = 161 (casewise deletion of missing data)PR WNC 117.5824 50.59794 1.0000 0.9506 0.8920 0.9247 0.7528 0.9407PRf 13.0582 7.24601 0.9506 1.0000 0.7278 0.9539 0.5445 0.7978PRf + s 78.5961 31.11583 0.8920 0.7278 1.0000 0.7287 0.8203 0.9402PR (RAf) 11.8421 6.21716 0.9247 0.9539 0.7287 1.0000 0.5712 0.8016PR (RAc) 196.4544 71.89993 0.7528 0.5445 0.8203 0.5712 1.0000 0.9270Mean value of the Public risk 83.5066 30.85363 0.9407 0.7978 0.9402 0.8016 0.9270 1.0000

3D Surface Plot (Public risks analysis 2006-2008.sta 6v*161c)Mean value of all public risks = Distance Weighted Least Squares

250 200 150 100 50 0

Fig. 1. A 3-D presentation of the risk values PR WNC, PR f + s and the Mean value ofall public risks.

168 D. Kukic et al. / Safety Science 51 (2013) 165–177

In order to get indicator that represents casualty risk in the bestway, the linear correlation between observed risk and Mean valueof all calculated risks in the respective categories has been tested.The main idea is to find indicator that has high linear correlationand because of that, the indicator could describe and assess roadsafety in a more realistic way. Also, idea is to develop approachfor selection of a relevant indicator – road casualty risk that willbe simple for use by others (decision makers, practitioners, etc.)that, by default, do not have special knowledge about statisticmethodology and calculation, multicriteria analysis, etc. Therefore,the last step of the presented model includes obtaining of just onerisk within the population of risks in an observation unit. It is therisk with the highest value of linear correlation in comparison tothe mean value of the observed risks.

The linear correlation is most often used for random variableanalysis. Despite the progress in establishing an efficient roadsafety system and reducing the number of road accidents like ran-dom events, it will be still impossible to estimate how many peoplewill be killed or injured in the road accidents. Proceeding from thisfact, assessment of the relation between variables, which are inthis case the risks obtained based on the number and conse-quences of road accidents, can be valid by using the coefficient oflinear correlation.

It is important to say that this research and the correlationstudy include all the data on the number of road accidents andtheir consequences, from all municipalities in Serbia, i.e. the dataon the total number of road accidents and their consequences from161 municipalities in Serbia, in the period 2006 to 2008. In thisperiod, there was a total of 202,529 road accidents, with 2783 per-sons killed, 15,293 seriously injured and 47,624 slightly injured –source: Ministry of Interior of the Republic of Serbia.

The advantage of the presented way of selecting just one risk fora particular unit of observation is the possibility of making a reli-able risk mapping on the observed territory by using available data(which are often insufficient). This model also makes possible tosingle out the reliable absolute indicator of road safety.

3. Results

Depending on the risk categories, the value of particular indica-tors describes the road safety situation in a different way. Largedispersion of results of the analyzed risks is important reason tochoice just one risk for future analysis and final assessment ofthe traffic safety situation. Therefore, the best way to extract oneor two risks that will be appropriate indicators of road safety, isto perform the selection of the parameters that are correlated themost with the Mean value of all calculated risks in the respectiverisk categories, and also mutually with all of the researched risks.

The values of the linear correlation between all the public risks,as well as between their mean values for all observation units –municipalities in Serbia – are given in Table 1. The risk whichcorrelates the most with the mean value of the public risks in pop-ulation is marked in blue, as the best describing risk based onincoming input for every municipality. Detailed information about

all the calculated risks for the municipalities in Serbia are given inthe Appendix A. All the assessed values of the calculated correla-tions are statistically significant (which was actually expected, con-sidering the origin of the variables whose correlations wereanalyzed), and they range between 0.54 and 0.95.

The greatest correspondence with all other risks in the popula-tion of the public risks was observed on PR WNC risk. The value ofthe linear correlation between this and all other calculated risksranges between 0.7528 in relation to the risk PR(RAc) and 0.9506in relation to the risk PRf. A minimum strength of correlation withregard to all other researched risks was recorded with the PR(RAc)risk, which is obtained on the basis of the number of road accidentswith casualties. The value of the strength of the linear correlationfor this risk ranges between 0.5445 in relation to the risk PRf and0.8203 in relation to the risk PR f + s.

The highest value of the linear correlation compared to theMean value of the public risks for all municipalities in Serbia,was obtained in the Public risk based on the weighted numberof casualties – PR WNC (0.9407) and the Public risk obtainedaccording to the number of fatalities and seriously injured PRf + s (0.9402). The difference in the strength of the linear correla-tion for these two parameters is only in the fourth decimal place,so if this detail is considered separately, it does not have a greaterimpact on the decision for selection of an indicator for the riskmapping. This especially makes sense if we accepted the logic

Fig. 2. Distribution of the final outcomes – risks with highest value of linear correlation, PR WNC and TR WNC for Serbian municipalities (2006–2008).

Table 2Values of correlations of the Traffic risk in all municipalities in the Republic of Serbia (risk within the territory of the whole municipality).

Means Std. dev. TR WNC TRf TRf + s TR (RAf) TR (RAc) Mean value of the Traffic risks

Correlations (Spreadsheet1) marked correlations are significant at p < .05000 N = 161 (casewise deletion of missing data)

TR WNC 57.1803 25.7265 1.0000 0.9592 0.8749 0.9261 0.7731 0.9481TRf 6.3702 3.7628 0.9592 1.0000 0.7240 0.9481 0.6003 0.8292TRf + s 38.0209 14.9781 0.8749 0.7240 1.0000 0.7164 0.8011 0.9225TR (RAf) 5.7775 3.2574 0.9261 0.9481 0.7164 1.0000 0.6247 0.8272TR (RAc) 95.4091 34.5146 0.7731 0.6003 0.8011 0.6247 1.0000 0.9291Mean value of the traffic risks 40.5516 15.2181 0.9481 0.8292 0.9225 0.8272 0.9291 1.0000

D. Kukic et al. / Safety Science 51 (2013) 165–177 169

of a reliable choice of risk in relation to the collection and pro-cessing of road accidents data, because the number of killedand seriously injured in comparison to the number of slightly in-jured represents more reliable data, which will also be needed forthe calculation of the risk PR WNC. However, the mutual compar-ison of the severity of the linear correlation between each risk,and significantly better results are on the side of the risk PRWNC. The values of the linear correlation between the risk PRf + s and other researched risks in the population of public risksrange between 0,7278 in relation to the risk PRf and 0,8920 inrelation to the risk PR WNC. Taking the above into account, therisk which coincides best with all the researched risks and alsowith the average value of all other risks in the population ofpublic risks, is the risk PR WNC.

Based on a three-dimensional presentation of the observedmeasures (Mean value of public risks, PR f + s and PR WNC), the sur-face of all obtained values has been designed (Fig. 1). On the y-axis,the results of the Mean value of public risks in all municipalities in

the Republic of Serbia are shown; the values of PR WNC risk areshown on the z-axis, while PR f + s risk is shown on the x-axes.(See Fig. 2).

By comparison with the farthest angles of the obtained values,the evenness of the influence of the observed measures on theMean value of public risks could be seen (Fig. 1).

The values of the linear correlation between all the trafficrisks, as well as among their mean values for all observation units– municipalities in Serbia – are given in Table 2. The highestvalue of the linear correlation compared to the Mean value ofthe traffic risks for all was obtained in the Traffic risk based onthe weighted number of casualties – TR WNC (0.9481). This riskcoincides best with all the researched risks and also with theaverage value of all other risks in the population of traffic risks.The second risk with highest value of the linear correlation com-pared to the Mean value of the traffic risks is Traffic risk obtainedon a basis of number of road accidents with casualties – TR(RAc)(0.9291).

Scatterplot (Public risks analysis 2006-2008.sta 6v*161c)PR WNC = -11,2398+1,5427*x; 0,95 Conf.Int.

Mean value of the Public risks: PR WNC: r = 0,9407; p = 00,0000

Cajetina

Velika PlanaAlibunar

Kanjiža

Sopot

Ub

0 20 40 60 80 100 120 140 160 180 200 220 240

Mean value of the Public risks

0

50

100

150

200

250

300

350

PR W

NC

Savski Venac

Fig. 3. Scatter plot diagram of differences of the PR WNC risk and the Mean value of public risks for the territory of the whole municipality.

Scatterplot (Public risks analysis 2006-2008.sta 6v*161c)Mean value of the Public risks:PRf; r = 0,7978; p = 00,0000

PRf = -2,5875+0,1874*x; 0,95 Conf.Int.

Cajetina

Savski Venac

Velika PlanaAlibunar

KanjižaPecinciSopot

Batocina

UbNovi Sad

Backi Petrovac

Vracar

0 20 40 60 80 100 120 140 160 180 200 220 240

Mean value of the Public risks

-5

0

5

10

15

20

25

30

35

40

45

PRf

Fig. 4. Scatter plot diagram of differences PRf risk and the Mean value of public risks within the territory of the whole municipality.

170 D. Kukic et al. / Safety Science 51 (2013) 165–177

4. Discussion

In order to determine the differences between the values of re-searched risks in the best possible way and consequently the rea-son for selecting an indicator for the risk mapping, a comparisonbetween the risk with the highest value of the linear correlation

and the risk with the lowest value of the linear correlation (PRWNC and PRf) has been conducted. In order to best detect this dif-ference, a scatter diagram has been used because the type and levelof their connection can be observed in the best possible waythrough a graphical display of points determined by the values ofentities (observation units), at two variables simultaneously.

D. Kukic et al. / Safety Science 51 (2013) 165–177 171

Dispersion of the values obtained for the strength of the linearcorrelation is presented in Table 1 and indicates the possible differ-ences that occur in calculating certain types of risk. It could be ex-pressed to the extent in which the obtained value of the analyzedrisk for certain municipalities differs from the smallest to the big-gest defined risk band.

In other words, if we observe a municipality which, according tothe value of the public casualty risk, based on the overall numberof road accidents with casualties – PR (RAc), belongs to a munici-pality with a very low level of risk, the second indicator, such asthe Public risk of the weighted number of casualties – PR WNC,classifies this municipality into the class of municipalities with ahigh level of risk, compared to other municipalities in Serbia (forexample, the municipality of Vracar in the city of Belgrade).

In this case, the values of the observation unit are the Public riskof the weighted number of casualties – PR WNC (the highest valueof the linear correlation) and the Public fatality risk obtained on abasis of the number of fatalities in municipalities – PRf (the lowestvalue of the linear correlation).

The examination of the scatter plot diagram shows that the gen-eral trend of the points is highly linear, i.e. that it forms an elon-gated ellipse (Fig. 3). The strength of the correlation for theobserved entities – municipalities amounts to 0.9407 for thePRWNC and the Mean value of all public risks.

The value r = 0.9407 represents the value of coefficient of thelinear correlation (the third row in the diagram header), whilethe value p represents the probability of getting as big or a biggercoefficient of the linear correlation on the sample of a given size, ifthere is no linear connection in the population.3

P ¼ ðjtjP tcalculatejH0correctÞ ð6Þ

In this unique way, we can single out those municipalitieswhere one variable in the observed entity differs greatly fromthe increase or decrease of the value or other variable.

In other words, if we notice a considerable increase of the Publicrisk of weighted number of casualties – PR WNC, in relation to theMean value of public risks in some municipalities, these data couldmake us direct the research and analysis of the state of road safetyto the municipality with distinct nonlinearity of observed variables.

Fig. 4 shows the scatter plot for the risk with the lowest value oflinear correlation with regard to the Mean value of all public risks,it is the Public risk obtained according to the number of fatalities ofthe whole municipality – PRf. The coefficient value of the linearcorrelation in this example amounts to r = 0.7978. We can deter-mine that dispersion of points of the observed variables is muchhigher compared to the distribution of points for Public risk ob-tained on the weighted number of casualties – PR WNC.

On Fig. 4, likewise on Fig. 3, those municipalities with the big-gest deviation from the direction of correlation line have been sin-gled out. Apart from the fact that in some cases the deviation is farsmaller (municipality of Crna Trava), it is much bigger for themajority of municipalities which are singled out, compared tothe case we had with the risk PR WNC (Fig. 3).

With the risk PRf (Public fatality risk), there is a significant num-ber of other municipalities with much bigger deviation from thedirection of correlation line (which is expected if we consider thesmaller value of coefficient of linear correlation).

If we consider the fact that when choosing the PRf risk, a greaternumber of municipalities deviate from the direction of correlationline, and that this risk has the lowest value of linear correlationmutually with all other risks, and also with the Mean value of pub-lic risks, we can conclude that the PRf risk is not the best solutionfor the risk mapping of the observed unit.

3 Tenjovic, 2002. Statistics in Psychology – Manual; 2nd edition; Centre for appliedpsychology, the association of psychologist of Serbia, Belgrade.

5. Conclusion

In previous researches, the values and names of certain riskswere not explicitly assessed according to the calculated parame-ters, which would be acceptable for all further researches and riskmapping on the observed territory.4 In the most prominent papersin the field of road safety, the values of risk rates from estimations ofthe most serious consequences of road accidents (number of fatali-ties) in relation to the population (i.e. number of registered motorvehicles, number of vehicle-kilometers, AADT value, etc.), to takinginto consideration of all the consequences of a road accident,weighted by adequate coefficients, depending on their severity. Agreat number of different risks whose values could be calculatedfor every observation unit has been defined in this way.

The main difference between Model for selection of a road casu-alty risk, which is presented in this paper and the RAP model(EuroRAP, usRAP and AusRAP) is special analysis of the risk typeindicators (final outcomes) for risk mapping. Also, the RAP modelis intended for risk mapping on the roads and road sections, notfor the risk mapping for the territories of municipalities, countries,regions, etc. The RAP model does not analyze risk type indicators,but it analyzes (or more precisely calculates) two rates (1) crashrisk per kilometer traveled and (2) crash density, obtained fromthe data of road accidents with seriously injured and fatalities.

A Model-based risk map for roadway traffic crashes (Nam andSong, 2008) is using Global spatial autocorrelation to describethe risk on the observed territory. This model is analyzing all con-sequences of road accidents and in this sense it is similar to theModel for selection of a relevant risk that is presented in this paper.However, this model does not use statistical tools to compare therisks. The model is using statistical tools for the modeling a spatialdistribution of road accidents.

The Model for selection of a relevant risk is in its basis a model formonitoring the state of road safety based on final output which ispresented in rates. It is a fact that every single risk of the analyzedrisks describes the state of road safety by a certain quality level.However, due to the great dispersion of results of the analyzedparameters, we have great differences in the observed risks.

The best contribution of this Model reflects in the separation ofonly one risk which will be used for future researches and assess-ments of the state of road safety and the risk mapping. This Modelalso eliminates other risks which are not the best solution for areliable choice of relevant indicator for estimation of the real dan-ger. The use of this Model has shown that the most reliable finaloutput for a defined unit of observation – municipalities in Serbia,is the Public risk obtained on the basis of the weighted number ofcasualties – PR WNC. This risk could be used for future researchesand assessments of the state of road safety in a defined unit ofobservation.

By implementing the Model for selection of relevant risk, otherregularities of the observed entities have been noticed, that mightbe subjects to further researches. This primarily refers to the influ-ences of some occurrence or road user’s behavior on the level ofroad safety Elvik (2004). It is indisputable that there are great dif-ferences between the municipalities in terms of importance, fre-quency and quality of the state and local roads, road users’behavior, social attitudes with regard to the dangers of road trafficetc. In further researches, for the road safety assessment, someSPIs, suggested for i.e. by ETSC (2001), Hakkert and Gitelman(2007), Hermans et al. (2009), Gitelman et al. (2010), Wegmanand Oppe (2010) could be included Hakkert et al. (2007). SPIs thatcan be used for the future researches are:

4 The so-called risks, are not real risks, they are only rates, risk type indicators.Their value can be different from those between 0 and 1, and the exposure is only anestimation of the real danger.

172 D. Kukic et al. / Safety Science 51 (2013) 165–177

– % of drivers under its influence,– % of seatbelt using,– % of drivers who are speeding, etc.

In addition to the further researches we can do detailed analysisand classification of road network in Serbia, road safety assess-ments, fleet analysis, work of emergency services on the territoryof municipalities, etc.

Appendix A

All calculated Public risks for the municipalities in the Republic of Serbia

Num.

Municipality PR WNC PRf PR f + s PR (RAf) PR (RAc) Mean value of Pablic risks

1

Savski Venac 282.63 25.10 215.66 25.10 579.54 225.60 2 Ub 155.12 15.57 101.75 15.57 440.24 145.65 3 Novi Sad 153.43 13.03 111.15 11.14 361.29 130.01 4 Lapovo 244.69 32.41 125.59 28.36 352.46 156.70 5 Lazarevac 231.64 26.21 152.68 24.50 343.53 155.71 6 Lajkovac 248.11 29.30 160.20 25.40 341.89 160.98 7 Ljig 185.48 15.95 164.06 15.95 328.12 141.91 8 Cajetina 284.75 36.26 174.90 36.26 319.94 170.42 9 In -dija 191.97 23.52 105.49 18.81 313.79 130.72

10

Zrenjanin 139.90 13.13 87.09 11.86 308.72 112.14 11 Backi Petrovac 131.24 9.08 122.61 9.08 304.25 115.25 12 Mionica 159.27 14.13 133.23 14.13 288.66 121.88 13 Smederevo 167.96 18.52 105.33 16.39 283.83 118.41 14 Temerin 174.83 18.86 124.96 16.50 282.94 123.62 15 Valjevo 125.36 11.71 86.47 11.02 277.32 102.38 16 Becej 122.15 11.39 78.89 11.39 276.51 100.06 17 Obrenovac 123.14 13.62 69.51 12.21 275.68 98.83 18 Cacak 146.92 15.09 101.08 13.67 275.33 110.42 19 Stara Pazova 146.75 16.77 90.27 15.29 273.77 108.57 20 Kraljevo 131.87 12.60 96.95 12.87 272.79 105.42 21 Ruma 187.48 23.33 102.21 20.55 272.75 121.27 22 Doljevac 150.98 17.04 85.20 13.63 272.65 107.90 23 Knic 173.81 20.64 111.47 18.58 268.35 118.57 24 Zabalj 219.05 29.08 121.15 24.23 266.54 132.01 25 Šabac 156.40 15.73 125.58 14.10 265.00 115.36 26 Bogatic 136.00 13.14 106.09 13.14 259.67 105.61 27 Vrnjacka Banja 117.39 10.07 93.11 13.84 255.42 97.97 28 Irig 176.28 21.63 108.15 16.22 254.14 115.28 29 Prijepolje 124.77 12.07 83.40 6.58 249.09 95.18 30 Kovin 176.26 21.74 105.07 18.11 247.27 113.69 31 Paracin 135.73 12.58 115.49 10.86 245.28 103.99 32 Cicevac 159.00 18.60 108.48 12.40 244.85 108.66 33 Pancevo 139.01 14.68 93.32 13.11 244.57 100.94 34 Cuprija 138.53 14.90 96.32 11.92 241.31 100.59 35 Sremska Mitrovica 133.33 13.97 90.03 11.25 240.97 97.91 36 Velika Plana 253.73 35.98 135.67 29.98 240.61 139.19 37 Svilajnac 141.77 16.99 83.62 16.99 240.42 99.96 38 Backa Palanka 104.92 10.39 63.97 9.84 238.93 85.61 39 Vladimirci 157.73 18.00 116.17 16.36 235.61 108.77 40 Pecinci 205.52 29.45 99.20 26.35 235.59 119.22 41 Odzaci 124.31 11.24 96.49 11.24 234.20 95.50 42 Stari Grad 84.62 6.00 74.42 6.00 233.45 80.90 43 Topola 149.72 15.82 104.12 15.82 231.96 103.48 44 Kikinda 98.11 8.95 71.14 6.96 230.34 83.10 45 Pozega 138.32 15.48 87.74 13.42 229.15 96.82 46 Senta 131.94 14.34 97.78 14.34 228.15 97.31 47 Veliko Gradište 144.57 16.14 98.42 16.14 227.50 100.55 48 Kovacica 168.88 21.51 95.61 21.51 227.08 106.92 49 Bela Crkva 114.73 11.46 80.20 11.46 225.86 88.74 50 Jagodina 105.46 8.93 89.81 7.05 225.69 87.39 51 Barajevo 173.02 20.29 109.57 16.23 224.56 108.73 52 Petrovac 98.52 7.73 86.93 8.69 224.08 85.19 53 Raca 135.30 12.86 110.61 12.86 223.78 99.08

D. Kukic et al. / Safety Science 51 (2013) 165–177 173

Appendix A (continued)

Num.

Municipality PR WNC PRf PR f + s PR (RAf) PR (RAc) Mean value of Pablic risks

54

Apatin 101.48 10.16 66.03 10.16 222.47 82.06 55 Šid 124.87 14.54 74.41 14.54 222.38 90.15 56 Srbobran 118.36 11.20 97.08 9.33 222.16 91.63 57 Mladenovac 135.90 15.88 79.38 14.61 219.72 93.10 58 Sombor 146.89 18.16 92.19 15.76 218.31 98.26 59 Brus 103.74 8.88 85.27 8.88 216.73 84.70 60 Alibunar 241.21 34.85 127.79 26.14 216.37 129.27 61 Batocina 181.94 27.28 73.65 24.55 215.49 104.58 62 Sremski Karlovci 154.99 18.86 98.05 11.31 214.96 99.63 63 Kruševac 108.47 8.88 95.66 8.63 214.16 87.16 64 Bajina Bašta 135.50 14.87 97.20 14.87 211.54 94.79 65 Backa Topola 160.46 20.05 106.33 15.69 210.05 102.51 66 Subotica 107.14 9.66 90.52 8.31 208.00 84.73 67 Vracar 60.52 1.71 66.80 1.71 206.67 67.48 68 Kula 122.43 12.41 98.58 11.03 199.92 88.87 69 Kosjeric 131.90 16.67 76.19 16.67 197.60 87.80 70 Zajecar 72.16 4.55 61.14 4.55 197.57 67.99 71 Gornji Milanovac 161.98 20.29 98.65 16.09 197.31 98.86 72 Nova Varoš 141.13 16.68 86.74 11.68 196.84 90.61 73 Razanj 121.38 14.66 67.43 14.66 196.44 82.92 74 Niš 76.99 6.39 54.95 5.59 195.73 67.93 75 Bela Palanka 176.16 25.50 71.85 23.18 194.70 98.28 76 Kuršumlija 98.11 10.80 53.99 10.80 194.37 73.61 77 Pozarevac 118.78 12.02 98.80 10.68 194.03 86.86 78 Despotovac 120.65 13.02 80.69 11.71 193.93 84.00 79 Novi Pazar 104.00 11.63 67.06 11.24 192.26 77.24 80 Palilula 117.55 12.61 84.88 11.33 192.00 83.68 81 Loznica 90.07 8.49 67.89 7.33 190.94 72.94 82 Kragujevac 85.68 7.39 66.55 7.02 186.57 70.64 83 Negotin 88.83 8.45 66.02 7.68 183.49 70.89 84 Beocin 127.65 16.58 74.60 16.58 182.35 83.55 85 Arilje 97.22 11.79 52.23 11.79 181.97 71.00 86 Koceljeva 125.35 12.79 110.86 12.79 181.21 88.60 87 Smederevska P. 102.24 11.31 70.22 9.52 180.32 74.72 88 Sokobanja 78.26 5.38 68.21 5.38 179.49 67.35 89 Surcin 127.32 13.78 95.62 13.78 179.18 85.94 90 Novi Becej 144.48 21.05 63.14 12.38 177.04 83.62 91 Bac 58.40 4.10 43.03 4.10 176.22 57.17 92 Aleksinac 112.50 14.43 51.95 13.85 172.01 72.95 93 Vozdovac 101.01 10.54 72.48 9.66 171.09 72.96 94 Leskovac 94.80 10.67 58.45 10.24 170.24 68.88 95 Ada 93.19 10.53 64.93 10.53 170.23 69.88 96 Opovo 103.79 12.10 66.57 12.10 169.45 72.80 97 Vršac 107.29 12.26 72.96 10.42 169.21 74.43 98 Merošina 101.49 11.25 60.76 9.00 168.78 70.26 99 Kanjiza 219.68 31.50 127.23 26.66 168.42 114.70

100

Aleksandrovac 100.49 10.21 78.26 6.81 166.73 72.50 101 Zitište 157.69 22.88 75.17 19.61 166.67 88.40 102 Cukarica 126.60 15.83 77.74 13.25 165.18 79.72 103 Bojnik 76.99 7.62 50.82 7.62 165.17 61.65 104 Secanj 109.30 14.25 54.96 12.21 164.87 71.12 105 Zitora -da 66.82 5.49 43.94 5.49 164.77 57.30 106 Vladicin Han 126.14 15.47 71.72 14.06 164.54 78.39 107 Coka 112.30 14.46 67.48 14.46 163.87 74.51 108 Raška 132.32 18.53 69.18 18.53 163.08 80.33 109 Nova Crnja 90.52 7.87 76.09 7.87 162.67 69.00 110 Knjazevac 81.15 8.97 50.22 8.97 162.31 62.32 111 Novi Beograd 75.84 6.73 58.62 6.73 161.33 61.85 112 Plandište 105.16 12.46 64.79 12.46 159.48 70.87 113 Mali I -doš 104.98 12.35 69.17 12.35 158.09 71.39

(continued on next page)

174 D. Kukic et al. / Safety Science 51 (2013) 165–177

Appendix A (continued)

Num.

Municipality PR WNC PRf PR f + s PR (RAf) PR (RAc) Mean value of Pablic risks

114

Pirot 79.32 8.36 50.16 5.75 156.24 59.97 115 Boljevac 125.77 16.83 67.30 10.52 155.64 75.21 116 Vrbas 142.20 20.36 76.33 13.09 154.85 81.36 117 Vranje 76.53 6.87 57.66 6.49 154.28 60.37 118 Trstenik 112.21 13.59 73.40 12.91 153.61 73.15 119 Aran -delovac 99.59 11.08 74.11 9.70 152.37 69.37 120 Uzice 77.37 7.63 52.60 6.83 151.77 59.24 121 Bujanovac 84.52 10.01 45.42 9.24 150.11 59.86 122 Rekovac 122.01 14.76 88.55 12.30 150.05 77.53 123 Gadzin Han 50.65 3.19 41.41 15.93 149.72 52.18 124 Grocka 125.44 15.46 81.71 13.25 149.29 77.03 125 Malo Crnice 140.76 16.84 105.87 14.44 149.19 85.42 126 Zvezdara 65.25 5.78 52.53 5.78 143.52 54.57 127 Prokuplje 62.68 6.19 38.49 6.19 142.95 51.30 128 Vlasotince 57.34 4.00 49.03 4.00 141.09 51.09 129 Blace 60.08 4.85 48.45 4.85 140.51 51.75 130 Svrljig 99.32 13.50 52.07 13.50 138.86 63.45 131 Ivanjica 95.17 11.29 62.07 9.40 137.30 63.05 132 Rakovica 77.88 7.41 61.28 6.40 136.36 57.87 133 Zemun 72.30 6.44 61.22 6.44 134.97 56.27 134 Lucani–Guca 66.22 5.42 50.11 5.42 134.07 52.25 135 Mali Zvornik 82.17 9.47 56.83 11.84 132.61 58.59 136 Varvarin 75.37 8.28 53.01 8.28 132.52 55.49 137 Kladovo 70.44 7.06 55.05 7.06 131.28 54.18 138 Golubac 62.21 6.73 40.35 10.09 131.14 50.10 139 Lebane 54.71 4.01 46.82 4.01 131.10 48.13 140 Novi Knezevac 85.81 10.28 56.52 7.71 131.02 58.27 141 Zabari 60.61 5.11 58.82 7.67 127.87 52.02 142 Preševo 81.18 10.51 38.20 8.60 127.02 53.10 143 Zagubica 71.74 6.75 58.47 6.75 123.68 53.48 144 Bor 58.70 5.37 43.00 4.18 123.02 46.86 145 Sopot 194.87 29.43 107.90 26.16 120.97 95.86 146 Kucevo 56.18 5.32 47.85 5.32 115.20 45.97 147 Krupanj 45.07 3.30 41.27 3.30 113.91 41.37 148 Titel 53.57 5.87 33.24 5.87 113.39 42.39 149 Osecina 49.55 4.40 37.44 4.40 107.92 40.74 150 Surdulica 72.71 10.52 27.04 10.52 106.65 45.49 151 Ljubovija 91.88 13.68 43.01 11.73 97.74 51.61 152 Majdanpek 66.94 7.03 53.44 7.03 97.03 46.30 153 Medve -da 25.71 0.00 34.08 0.00 96.03 31.16 154 Crna Trava 39.02 0.00 65.03 0.00 91.04 39.02 155 Sjenica 56.49 7.15 34.56 5.96 84.61 37.75 156 Bosilegrad 34.24 3.36 23.50 3.36 80.56 29.00 157 Dimitrovgrad 41.71 2.84 42.56 2.84 79.45 33.88 158 Priboj 25.41 1.62 17.80 1.62 78.50 24.99 159 Tutin 44.70 4.44 36.60 4.44 76.53 33.34 160 Babušnica 26.91 2.12 23.30 0.00 65.68 23.60 161 Trgovište 37.66 5.23 20.92 5.23 41.85 22.18

Appendix B. Appendix

All calculated Traffic risks for the municipalities in the Republic of Serbia

Num.

Municipality TR WNC TRf TRf + s TR (RAf) TR (RAc) Mean value of Traffic risks

1

Bela Palanka 147.11 21.37 59.78 19.36 161.23 81.77 2 Lapovo 141.62 18.67 73.27 16.36 205.31 91.04 3 Cajetina 139.62 17.72 86.43 17.72 156.66 83.63 4 Lajkovac 135.47 16.07 86.95 13.87 185.71 87.61

D. Kukic et al. / Safety Science 51 (2013) 165–177 175

Appendix B (continued)

Num.

Municipality TR WNC TRf TRf + s TR (RAf) TR (RAc) Mean value of Traffic risks

5

Velika Plana 128.57 18.26 68.58 15.19 121.67 70.45 6 Doljevac 127.62 13.69 76.28 12.34 240.40 94.07 7 Nova Varoš 104.02 12.25 64.09 8.64 146.23 67.05 8 Alibunar 103.61 15.01 55.03 11.02 91.71 55.28 9 Cicevac 99.77 11.70 67.82 7.77 152.90 67.99

10

Ljig 96.16 8.26 84.90 8.26 170.97 73.71 11 Pecinci 95.94 14.09 44.53 12.42 103.96 54.19 12 Knic 94.85 11.19 61.44 10.08 147.24 64.96 13 Temerin 93.69 10.30 66.36 9.04 146.01 65.08 14 Plandište 93.32 10.76 58.25 10.76 141.99 63.02 15 Vladimirci 93.31 10.70 68.44 9.70 138.62 64.15 16 Malo Crnice 93.13 11.95 64.83 9.96 91.47 54.27 17 Sopot 91.88 13.89 50.75 12.32 56.82 45.13 18 Razanj 90.35 10.91 50.26 10.91 146.22 61.73 19 Mionica 89.09 7.89 74.64 7.89 161.96 68.29 20 Smederevo 87.87 9.68 55.11 8.57 148.73 61.99 21 Stara Pazova 83.18 9.72 50.39 8.77 150.54 60.52 22 Vladicin Han 82.93 10.18 47.15 9.22 107.99 51.50 23 Ub 82.60 8.27 54.32 8.27 235.53 77.80 24 Lazarevac 82.21 9.29 54.35 8.71 121.71 55.25 25 Koceljeva 81.92 8.43 72.05 8.43 117.20 57.60 26 Novi Becej 80.81 12.06 33.84 6.68 91.65 45.01 27 In -dija 80.61 9.90 44.18 7.93 131.15 54.76 28 Boljevac 80.28 10.87 42.03 6.69 97.77 47.53 29 Zabalj 78.79 10.18 44.82 8.63 101.84 48.85 30 Irig 76.59 9.77 44.78 7.32 105.89 48.87 31 Batocina 75.53 11.33 30.56 10.18 89.29 43.38 32 Gornji Milanovac 75.42 9.45 45.92 7.48 91.64 45.98 33 Barajevo 75.23 8.86 47.53 7.06 97.12 47.16 34 Ruma 73.34 9.25 39.41 8.15 104.59 46.95 35 Rekovac 73.01 8.79 53.35 7.34 90.14 46.53 36 Bajina Bašta 72.47 8.00 51.75 8.00 112.43 50.53 37 Bogatic 72.16 6.97 56.33 6.97 137.75 56.03 38 Bojnik 71.13 7.08 46.53 7.08 152.18 56.80 39 Topola 70.77 7.48 49.22 7.48 109.63 48.92 40 Kovin 70.33 8.68 41.65 7.29 101.52 45.89 41 Kovacica 69.78 8.93 39.32 8.93 92.04 43.80 42 Zitište 68.69 9.79 33.68 8.63 76.80 39.52 43 Aleksinac 67.96 8.66 31.69 8.31 104.84 44.29 44 Preševo 66.98 8.68 31.44 7.10 104.49 43.74 45 Kula 66.94 6.81 53.97 5.98 108.05 48.35 46 Pozega 66.12 7.41 41.94 6.44 109.13 46.21 47 Kanjiza 65.90 9.42 37.89 8.08 52.58 34.77 48 Backa Topola 65.77 8.52 41.63 6.51 81.34 40.76 49 Raca 65.64 6.13 54.26 6.13 109.96 48.43 50 Svrljig 65.49 8.87 34.42 8.87 92.29 41.99 51 Vrbas 65.28 9.33 34.99 5.98 71.10 37.34 52 Prijepolje 65.14 6.29 43.65 3.44 130.16 49.74 53 Novi Pazar 65.00 7.22 42.12 6.99 121.31 48.53 54 Bujanovac 64.59 7.64 34.77 7.05 114.76 45.76 55 Šabac 64.09 6.46 51.39 5.80 108.48 47.24 56 Kuršumlija 63.75 7.01 35.11 7.01 126.54 47.88 57 Kosjeric 62.60 7.92 36.09 7.92 93.36 41.58 58 Merošina 62.51 6.92 37.45 5.55 104.01 43.29 59 Mladenovac 62.19 7.27 36.32 6.69 100.47 42.59 60 Ljubovija 61.25 9.08 28.87 7.79 65.96 34.59 61 Veliko Gradište 61.00 6.72 41.89 6.72 97.52 42.77 62 Odzaci 60.92 5.46 48.21 5.46 115.32 47.07 63 Raška 60.68 8.48 31.81 8.48 75.10 36.91 64 Paracin 60.35 5.60 51.28 4.84 108.94 46.20 65 Brus 59.90 5.13 49.35 5.13 124.99 48.90

(continued on next page)

176 D. Kukic et al. / Safety Science 51 (2013) 165–177

Appendix B (continued)

Num.

Municipality TR WNC TRf TRf + s TR (RAf) TR (RAc) Mean value of Traffic risks

66

Sombor 58.96 7.24 37.38 6.31 88.94 39.77 67 Kraljevo 58.43 5.59 42.96 5.71 120.76 46.69 68 Bela Crkva 58.41 6.05 39.23 6.05 113.66 44.68 69 Cacak 58.21 5.97 40.17 5.40 109.01 43.75 70 Secanj 58.07 7.92 28.14 6.39 80.21 36.15 71 Savski Venac 57.23 5.08 43.72 5.08 117.19 45.66 72 Sjenica 56.05 7.12 34.25 5.92 83.25 37.31 73 Obrenovac 55.27 6.10 31.25 5.47 124.13 44.44 74 Sremski Karlovci 55.18 6.54 35.89 3.92 79.81 36.27 75 Smederevska P. 54.99 6.09 37.73 5.14 96.93 40.18 76 Cuprija 54.11 5.81 37.64 4.65 94.30 39.30 77 Zrenjanin 53.42 5.04 33.15 4.52 116.90 42.61 78 Svilajnac 52.95 6.36 31.05 6.36 90.01 37.35 79 Valjevo 51.90 4.87 35.72 4.58 114.30 42.28 80 Šid 51.81 6.15 30.13 6.15 91.04 37.06 81 Pancevo 51.31 5.43 34.07 4.92 90.85 37.32 82 Grocka 51.03 6.31 33.15 5.39 60.37 31.25 83 Leskovac 50.96 5.74 31.38 5.50 91.48 37.01 84 Coka 50.95 6.31 31.72 6.31 80.14 35.09 85 Trstenik 50.91 6.18 33.16 5.88 69.66 33.16 86 Sremska Mitrovica 50.61 5.43 33.52 4.29 89.86 36.74 87 Mali I -doš 50.32 5.58 35.45 5.58 79.96 35.38 88 Zitora -da 50.21 4.15 32.99 4.15 123.73 43.04 89 Mali Zvornik 50.15 5.79 34.64 7.25 80.79 35.72 90 Zagubica 49.80 4.64 40.73 4.64 86.64 37.29 91 Beocin 49.69 6.36 29.65 6.36 72.17 32.85 92 Kruševac 48.99 4.00 43.24 3.89 96.87 39.40 93 Ivanjica 47.99 5.66 31.49 4.74 69.41 31.86 94 Jagodina 47.50 4.03 40.40 3.18 101.56 39.34 95 Kikinda 46.89 4.30 33.58 3.24 109.75 39.55 96 Becej 46.74 4.39 30.12 4.39 104.68 38.06 97 Pozarevac 46.53 4.70 38.76 4.18 76.01 34.04 98 Loznica 46.25 4.34 34.93 3.75 98.63 37.58 99 Despotovac 46.14 4.98 30.83 4.49 74.28 32.14

100

Aleksandrovac 46.11 4.66 36.06 3.11 76.78 33.34 101 Srbobran 45.78 4.39 37.22 3.53 84.39 35.06 102 Cukarica 45.19 5.64 27.82 4.73 58.99 28.47 103 Majdanpek 45.06 4.74 36.00 4.74 65.02 31.11 104 Sokobanja 45.03 3.17 38.91 3.17 101.90 38.44 105 Vrnjacka Banja 44.76 3.86 35.33 5.28 97.33 37.31 106 Knjazevac 44.34 4.90 27.44 4.90 88.71 34.06 107 Senta 43.75 4.66 32.81 4.66 78.06 32.79 108 Backi Petrovac 43.58 2.89 41.27 2.89 101.41 38.41 109 Arilje 43.25 5.29 23.02 5.29 80.34 31.44 110 Novi Sad 43.25 3.67 31.30 3.14 101.98 36.67 111 Opovo 43.20 4.92 28.84 4.92 70.48 30.47 112 Apatin 42.98 4.33 27.75 4.33 93.81 34.64 113 Backa Palanka 42.47 4.17 26.08 3.98 97.49 34.84 114 Aran -delovac 42.44 4.72 31.57 4.14 64.98 29.57 115 Blace 42.08 3.39 33.98 3.39 98.23 36.21 116 Palilula 41.95 4.51 30.29 4.05 68.48 29.86 117 Nova Crnja 41.59 2.72 41.72 2.72 84.97 34.75 118 Petrovac 41.52 3.27 36.52 3.66 94.47 35.89 119 Surdulica 40.25 5.82 14.98 5.82 59.10 25.19 120 Pirot 39.84 4.20 25.16 2.87 78.68 30.15 121 Tutin 37.91 3.78 31.08 3.78 64.24 28.16 122 Vranje 37.84 3.40 28.46 3.22 76.33 29.85 123 Negotin 37.78 3.58 28.11 3.26 78.25 30.20 124 Novi Knezevac 37.51 4.74 23.11 3.76 55.51 24.93 125 Lebane 37.47 2.77 31.83 2.77 89.37 32.84 126 Lucani–Guca 37.17 3.00 28.28 3.00 75.83 29.46

D. Kukic et al. / Safety Science 51 (2013) 165–177 177

Appendix B (continued)

Num.

Municipality TR WNC TRf TRf + s TR (RAf) TR (RAc) Mean value of Traffic risks

127

Zajecar 36.91 2.34 31.20 2.34 100.79 34.72 128 Subotica 36.91 3.30 31.23 2.87 72.33 29.33 129 Prokuplje 36.85 3.61 22.76 3.61 84.48 30.26 130 Varvarin 36.85 4.01 26.18 4.01 65.27 27.26 131 Vršac 36.29 4.12 24.75 3.52 57.96 25.33 132 Kragujevac 35.17 3.04 27.32 2.88 76.48 28.98 133 Osecina 34.66 3.19 25.67 3.19 73.17 27.98 134 Vozdovac 34.44 3.61 24.67 3.30 58.12 24.83 135 Trgovište 33.70 4.70 18.64 4.70 36.95 19.74 136 Bosilegrad 33.51 3.38 22.41 3.38 77.31 28.00 137 Crna Trava 33.04 0.00 54.64 0.00 78.20 33.18 138 Uzice 33.02 3.26 22.42 2.92 64.85 25.29 139 Ada 32.84 3.65 23.09 3.65 62.32 25.11 140 Golubac 32.68 3.57 21.03 5.41 67.91 26.12 141 Kladovo 31.84 3.19 24.91 3.19 59.48 24.52 142 Vlasotince 31.71 2.21 27.06 2.21 78.10 28.26 143 Krupanj 31.70 2.34 29.06 2.34 79.26 28.94 144 Gadzin Han 31.23 1.95 25.55 9.92 92.69 32.27 145 Rakovica 29.72 2.83 23.37 2.44 52.15 22.10 146 Niš 29.62 2.46 21.14 2.15 75.18 26.11 147 Zabari 29.04 2.45 28.17 3.68 61.47 24.96 148 Dimitrovgrad 28.48 1.93 28.98 1.93 54.36 23.14 149 Titel 28.31 3.03 18.25 3.03 61.26 22.78 150 Medve -da 27.42 0.00 36.44 0.00 101.78 33.13 151 Bor 27.39 2.50 20.16 1.95 57.35 21.87 152 Zemun 26.70 2.37 22.64 2.38 49.86 20.79 153 Kucevo 24.16 2.30 20.51 2.30 49.34 19.72 154 Priboj 23.09 1.47 16.18 1.47 71.35 22.71 155 Babušnica 22.52 1.82 19.18 0.00 54.25 19.55 156 Zvezdara 22.02 1.96 17.72 1.95 48.29 18.39 157 Bac 21.69 1.31 16.91 1.31 71.82 22.61 158 Novi Beograd 20.29 1.79 15.73 1.79 43.35 16.59 159 Stari Grad 18.91 1.35 16.60 1.35 52.09 18.06 160 Vracar 16.38 0.46 18.10 0.46 56.01 18.28 161 Surcin� (excluded) 0.00 0.00 0.00 0.00 0.00 0.00

References

Al-Haji, G., 2007. Road Safety Development Index (RSDI) Theory, Philosophy andPractice. Dissertation No. 1100, Department of Science and Technology,Linköping University, Sweden.

Anselin, L., 2005. Exploring spatial data with GeoDATM: a workbook. SpatialAnalysis Laboratory, 138.

Ekler, V., 2010. Measuring and understanding road safety performance at localterritorial level. Safety Science 48, 1197–1202.

Elvik, R., 2004. Dimensions of Road Safety Problems and Their Measurement.Institute of Transport Economics, Oslo, Norway.

European Transport Safety Council, 2001. Transport Safety Performance Indicators,Brussels.

Gitelman, V., Doveh, E., Hakkert, S., 2010. Designing a composite indicator for roadsafety. Safety Science.

Hakkert, A.S., Gitelman, V., 2007. Road Safety Performance Indicators: Manual.Deliverable D3.8 of the EU FP6 Project SafetyNet.

Hakkert, A.S, Gitelman, V., Vis, M.A., 2007. Road Safety Performance Indicators:Theory. Deliverable D3.6 of the EU FP6 project SafetyNet, 10.

Hermans, E., Brijs, T., Wets, G., Vanroof, K., 2009. Benchmarking road safety: lessonsto learn from a data envelopment analysis. Accident Analysis and Prevention 41,174–182.

Hill, J., 2010. EuroRAP202: Risk Mapping Manual. European Road AssessmentProgramme, Brussels.

Koornstra, M., Lynam, D., Nilsson, G., Noordzij, P., Pettersson, H.-E., Wegman, F.,Wouters, P., 2002. SUNflower: A Comparative Study of the Development of RoadSafety in Sweden, the United Kingdom, and Netherlands. SWOV Institute forRoad Safety Research, Leidschendam, The Netherlands.

Nam, C., Song, J., 2008. A Model – Based risk Map for Roadway Traffic Crashes.University of Arkansas.

Per, L., Al-Haji, G., 2005. Road Safety in Southeast Asia Factors Affecting MotorcycleSafety. Dept of Science and Technology, Linkoping University, Sweden.

PIARC, 2008. Road Safety Manual, Recomendations from the Road WorldAssociation (Priority Ranking) (Chapter 7).

Rumar, K., 1999. Transport Safety Visions, Targets and Strategies: Beyond 2000. FirstEuropean Safety Lecture, European Traffic Safety Council, Brussels.

Šliupas, T., 2009. The impact of road parameters and the surrounding area on trafficaccidents. Transport 24 (1), 42–47.

Smeed, R.J., 1972. The usefulness of formula in traffic engineering and road safety.Accident Analysis and Preview 4, 3–312.

Sorensen, M., 2002. Best Practice Guidelines on Black Spot Management and SafetyAnalysis of Road Networks. Brussels.

Tenjovic, L., 2002. Statistics in Psychology – Manual, second ed. Centre for AppliedPsychology. The Association of Psychologist of Serbia, Belgrade.

Wegman, F., Oppe, S., 2010. Benchmarking road safety performances of countries.Safety Science 48 (9), 1203–1211.

Wegman, F., Eksler, V., Hayes, S., Lynam, D., Morsink, P., Oppe, S. 2005.SUNflower+6: A Comparative Study of the Development of the Road Safety inthe SUNflower+6 Countries: Final Report. SWOV Institute for Road SafetyResearch, Leidschendam, The Netherlands.

Wegman, F., Commandeur, J., Doveh, E., Eksler, V., Gitelman, V., Hakkert, S., Lynam,D., Oppe, S., 2008. SUNflowerNext: Towards a Composite Road SafetyPerformance Index. SWOV Institute for Road Safety Research, Leidschendam,The Netherlands.

Zhang, W., Tsimoni, O., Sivak, M., Flannagan, M.J., 2010. Road safety in China:analysis of current challenges. Journal of Safety Research 41 (1), 25–30.