You are on page 1of 25

Accepted Manuscript

Title: Can pedestrians and cyclists share the same space? The
case of a city with low cycling levels and experience

Authors: Andreas Nikiforiadis, Socrates Basbas

PII: S2210-6707(18)30926-0
DOI: https://doi.org/10.1016/j.scs.2019.101453
Article Number: 101453

Reference: SCS 101453

To appear in:

Received date: 19 May 2018


Revised date: 31 January 2019
Accepted date: 31 January 2019

Please cite this article as: Nikiforiadis A, Basbas S, Can pedestrians and cyclists share
the same space? The case of a city with low cycling levels and experience, Sustainable
Cities and Society (2019), https://doi.org/10.1016/j.scs.2019.101453

This is a PDF file of an unedited manuscript that has been accepted for publication.
As a service to our customers we are providing this early version of the manuscript.
The manuscript will undergo copyediting, typesetting, and review of the resulting proof
before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that
apply to the journal pertain.
Can pedestrians and cyclists share the same space? The case of a city with low
cycling levels and experience
Andreas Nikiforiadis, Socrates Basbas
Department of Transportation and Hydraulic Engineering, School of Rural and Surveying Engineering, Aristotle University of
Thessaloniki, Greece

E-mail addresses: anikiforiadis@topo.auth.gr (A. Nikiforiadis), smpasmpa@topo.auth.gr (S. Basbas)


Full postal address: 541 24 Thessaloniki, Greece

Highlights

T
 Social factors have great impact on pedestrians’ perceived LOS in shared use segments

IP
 The comfort and safety of pedestrians does not deteriorate significantly for bicycle volumes <0.17
bicycles/min

R
 Pedestrian unit flow rate determines to a significant extent pedestrians’ perceived LOS

SC
 Over-dimensioning of the bicycle lane has significant impact on pedestrians’ perceptions, regardless
of the bicycle volumes

U
N
Abstract
A
Transport policy objectives, as well as lack of space in the built environment make clear that the
M

coexistence of pedestrians and cyclists is an important issue. For that reason, a survey was conducted on two

sidewalks and one pedestrian street in the city of Thessaloniki, Greece where the cycling volumes and
ED

cycling experience are still low; however efforts are being made to promote active transport modes (e.g.

cycling, walking). In the analysis made in this paper, an ordinal regression model was developed that utilizes
PT

questionnaire survey data and field measurements. The aim of the ordinal regression model is to identify and
E

quantify the impact of geometric, functional and social factors on pedestrians' perception about the presence
CC

of bicycles and the infrastructure level of service (LOS). The results show a clear difference in the

perceptions according to gender and age. Females tend to perceive lower LOS and the same applies for the
A

younger people. Regarding the geometric and functional attributes, pedestrian unit flow rate has been found

to affect perceived LOS to a large extent. Another conclusion that emerges from the analysis is the

importance of a proper distribution of the available space between pedestrians and cyclists, as over-

dimensioning of the bicycle lane has a significant impact on the perceptions of pedestrians, regardless the

bicycle volumes.
Keywords: Pedestrians, Cyclists, Shared space, Level of service (LOS), Ordinal regression model

1. Introduction

For many decades, transport planning has been made from the perspective of motorized traffic, largely

ignoring users of non-motorized vehicles. There is now a shift in how to approach transport planning within

the urban environment. European Commission’s White Paper on Transport 2011, recognized the need to

T
promote walking and cycling in order to achieve a competitive and sustainable transport system (COM 144,

IP
2011). Moreover, European Commission suggested that local authorities of the Member States should

R
implement Sustainable Urban Mobility Plans (SUMPs). SUMP is defined as a “strategic plan designed to

SC
satisfy the mobility needs of people and businesses in cities and their surroundings for a better quality of

life. It builds on existing planning practices and takes due consideration of integration, participation, and

evaluation principles” (Wefering et al., 2013). U


N
This shift in transport policy is due to the considerable environmental, social and economic benefits of
A
non-motorized mobility. A number of surveys have demonstrated active transport (e.g. cycling, walking)
M

benefits in public and personal health, as well as congestion reduction, improvement in environmental
ED

conditions and economy (Litman, 2017; Dannenberg et al., 2017; Morettini et al., 2015; Oja et al., 2011;

Mulley et al., 2013; Blondiau et al., 2016). Moreover, it is very interesting that active modes found to have
PT

also a positive effect on travelers’ well-being (Friman et al., 2017; Singleton, 2018; Vaitsis et al., 2019).

On the other hand, promotion of active transport is prevented because of the relatively high number of
E

pedestrian and cyclist deaths and serious injuries. Thus, for safety reasons and taking into account the
CC

economic constraints that do not allow the implementation of exclusive traffic corridors for each category of

road users, in many cases bicycles are encouraged to use pedestrians' infrastructure (e.g. sidewalks,
A

pedestrian streets), in which fatality and serious injury risk is much lower (Grzebieta et al., 2011; Chong et

al., 2010). Considering the above, it is understood that the coexistence of pedestrians and cyclists is a matter

of great importance and extremely up to date. For this reason, it is necessary to properly design the

infrastructure in which pedestrians and cyclists coexist. This task initially requires the use of appropriate

methods for assessing the infrastructure.


To assess the pedestrian infrastructure, the level of service (LOS) concept has been used (Highway

Capacity Manual, 2010). The level of service measures the level of satisfaction that the traffic system

provides to the user. The most widespread methodology for determining the level of service is the one

proposed in the Highway Capacity Manual 2010. This methodology proposes six different levels of service,

where A describes the best and F the worst. Moreover, Highway Capacity Manual 2010 suggests a

methodology for the LOS calculation of pedestrian infrastructure based on the average pedestrian space and

the effect of external (motorized) traffic characteristics. It is easily understood that this methodology does

T
not take into account many additional parameters that potentially affect LOS.

IP
Many more surveys have focused on the issue of pedestrian infrastructure assessment. Mori and

R
Tsukaguchi (1987) proposed two alternative methods, one of which is based on pedestrians’ behaviour while

SC
the second one on pedestrians’ opinion. The results of this survey have shown that sidewalk width, lack of

obstacles and high green ratio are important parameters for the effective use of sidewalks. A completely

U
different methodology was followed by Jaskiewicz (2000), who used nine qualitative evaluation measures:
N
enclosure/definition, complexity of path network, building articulation, complexity of spaces,
A
overhangs/awnings/varied roof lines, buffer, shade trees, transparency and physical components/condition.
M

These measures were used in order to develop an overall pedestrian LOS. Qualitative criteria were also used
ED

by Rahaman et al. (2005) for the evaluation of the roadside environment of Dhaka city. These criteria were

safety, security, convenience/comfort, continuity, system coherence and attractiveness. The model
PT

developed by Jena (2014), also uses qualitative factors, such as comfort, safety, vendors’ attraction,

accessibility and environmental condition. Utilizing pedestrians’ perceptions Landis et al. (2001) developed
E

a mathematical model for LOS, using lateral separation elements between pedestrians and motor vehicle
CC

traffic, motor vehicle traffic volume, motor vehicle speed and motor vehicle mix as variables. The regression

model developed by Petritsch et al. (2006), also only includes quantitative factors. These factors are total
A

width of crossings at conflict locations and average 15-min volume on adjacent roadway. Similar

methodology was used in the study carried out by Bian et al. (2007). A model which combines qualitative

(comfort, problems caused by street vendor) and quantitative (pedestrian volume, number of pedestrian who

interact with vendors) variables was developed by Hidayat et al. (2011).


Moreover, several surveys reveal the impact of cycling on pedestrians’ perceptions. Muraleetharan et al.

(2003) with the aid of conjoint technique found out that bicycles can have a negative effect on pedestrians in

shared use segments. Jensen (2007) developed a cumulative logit model for pedestrian LOS, using citizens’

responses on specific video clips. The model he developed uses bicycle volume as one of the independent

variables. Furthermore, Tan et al. (2007) found out that bicycle volume along with vehicle volume,

pedestrian volume, driveway access frequency and distance between sidewalk and vehicle lane significantly

affect the pedestrian LOS. The same study also investigates the impact of users’ age, gender and walking

T
experience on pedestrian LOS. Kang et al. (2013) used a random parameters ordered probit model, in order

IP
to quantify the impact of specific factors on pedestrian perceptions. Through this study they found out that

R
bicycles have a strong negative impact on pedestrians’ perceived LOS. More particularly, bicycle flow rate,

SC
bicycle maximum speed and bicyclists riding against the pedestrian flow contribute to the degradation of

perceived LOS. Finally, the issue of pedestrian – cyclist interaction in shared use infrastructures is also

U
examined in surveys conducted by both Hummer et al. (2005) and Patten et al. (2006), but from the
N
bicyclist’s perspective. Table 1 summarizes the above methodologies, depending on whether they used
A
qualitative or quantitative factors.
M

Table 1
ED

Grouping of previous studies based on the use of qualitative and quantitative factors.

Qualitative factors Quantitative factors Mixed


Highway Capacity Manual, 2010 X
PT

Mori and Tsukaguchi, 1987 X


Jaskiewicz, 2000 X
Rahaman et al., 2005 X
E

Jena, 2014 X
Landis et al., 2001 X
CC

Petritsch et al., 2006 X


Bian et al., 2007 X
Hidayat et al., 2011 X
A

Muraleetharan et al., 2003 X


Jensen, 2007 X
Tan et al., 2007 X
Kang et al., 2013 X

The objective of this paper is to investigate pedestrians’ perceived LOS. For this purpose, an attempt has

been made so as a combined statistical model can be developed, quantifying the impact of quantitative and

qualitative factors, as well as pedestrians’ social characteristics on pedestrian LOS. Statistical models of this
type may help researchers understand pedestrians’ attitudes better and potentially contribute to a far better

design of new infrastructures or better management of existing infrastructures, taking into account the user’s

point of view.

2. Methodology and development of the analysis tool

To meet the above objective an ordinal regression model was developed based on a questionnaire survey

and field measurements. The motivation for the developing of such a model was provided by the significant

T
problems that occur in the city of Thessaloniki and in specific infrastructures due to the coexistence of

IP
pedestrians and cyclists. These problems stem from the existence of inadequate infrastructures in the city

R
and possibly from the pedestrians and cyclists’ limited familiarization with their coexistence. As a result, a

SC
great debate arises as to whether such coexistence is acceptable, while the complaints mainly from

pedestrians are quite numerous. Thus, the statistical analysis and the ordinal regression model could

U
comprise an important aid for transport and urban planners globally and especially in cities with
N
characteristics similar to those existing in Thessaloniki.
A
In order to obtain safe conclusions, the ordinal regression model was tested for its validity. It was decided
M

to use the specific type of model due to the ordinal nature of the dependent variable, which is the perceived
ED

pedestrian level of service. The key assumption in ordinal regression models is that the independent

variables have the same effect on the odds regardless of the threshold. This is called “assumption of
PT

proportional odds” and it can be examined through the test of parallel lines which compares the ordinal

model that has one set of coefficients for all thresholds with a model with a separate set of coefficients for
E

each threshold. Despite the fact that in most cases according to O’Connell (2006) this test results in the
CC

rejection of the “assumption of proportional odds”, the results from the model presented in this paper

demonstrate that the effects of any independent variable are consistent across the different thresholds and as
A

a result the use of this model is considered to be appropriate.

In order to develop the model, it was considered advisable to collect 300 questionnaires from

interviewees coming from three different segments, as well as to conduct field measurements concerning the

geometric and functional attributes of these specific segments. The three segments were chosen on the basis

of the fact that a bicycle lane is located on them, as well as because they have different characteristics,
which facilitate the investigation on the effect that these characteristics have on pedestrians’ perceived LOS.

The sample was set at 300 questionnaires since this number is considered to be satisfactory, for a target

population of one million, confidence level of 95% and sample error of 5.5% (Bartlett et al., 2001; Naing et

al., 2006). Moreover, it was considered desirable that the number of questionnaires would be equally

distributed among interviewees from all three segments. The field measurements refer to the following:

 Sidewalk/Pedestrian street total width, effective width and bicycle lane width,

 Pedestrian and bicycle volume,

T
 Pedestrian and bicycle speed.

IP
The survey was conducted for one hour every day of the week except for weekends (five hours of

R
measurements for every segment), in order to compute average values for a typical day in the morning.

SC
The ordinal regression model utilizes both measurements and questionnaires. In the last part of the

analysis, three sub-models have been developed, so as the differences in the results between the three

U
segments can be investigated. The reason why the three sub-models have been developed is to validate the
N
results of the main model and to further investigate whether the impact of some variables on the perceived
A

LOS is differentiated according to the characteristics of each segment.


M
ED

3. Description of the undertaken research

3.1. Study area


PT

The study area is the Municipality of Thessaloniki, located in northern Greece. The municipality is a

much smaller area than the metropolitan area and it comprises only one third of the total population of the
E

metropolitan area of Thessaloniki. It also occupies about 60 times smaller area. In Greece, there is a great
CC

need for a change in the way transport planning is considered and the change is sought through the
A

implementation of SUMPs, of which the design and implementation in Municipalities is highly encouraged

by the State. One of the Municipalities which is in the process of forming its own SUMP, is the Municipality

of Thessaloniki. The bicycle network currently in use in the Municipality of Thessaloniki has a total length

of 12 km (Kartsoviti, 2011), but it is anticipated that this number will increase significantly in the near future

through the implementation of the SUMP. A large part of the existing bicycle network uses sidewalks and
pedestrian streets, and a lot of the future network will also be covered through these infrastructures as this

solution is considered safer for cyclists under the prevailing conditions in the city.

More specifically, the three segments examined in the survey are:

 Ethnikis Amynis sidewalk: This sidewalk has extremely limited total width and the bicycle lane

occupies a large part of it. In addition, the bicycle lane is located along the fence of the Aristotle

University of Thessaloniki and goes past the entrance doors of the University, making the

infrastructure unsafe and uncomfortable, both for pedestrians and cyclists. Also, the existence of trees

T
at the edge of the sidewalk leaves a limited space for the pedestrians. It must be mentioned at this

IP
point that due to the existence of the University campus the majority of pedestrians are young people.

R
Pedestrian-cyclist segregation is achieved by the use of a yellow longitudinal line as well as a different

SC
color on the tiling.

 Aggelaki sidewalk: This sidewalk is characterized by a large width. However, there are planting zones

U
that result in a relatively narrow available space which becomes a lot narrower due to the land uses –
N
for instance cafes. The proximity of this sidewalk to the campus of the Aristotle University of
A

Thessaloniki and the International Exhibition Center of the city makes it an attraction pole for both
M

pedestrians and cyclists, who are also young in most cases. The segregation between pedestrians and
ED

cyclists is achieved by using a yellow longitudinal line as well as a different color on the tiling.

 Agias Sophias pedestrian street: The pedestrianization of this street completed in 2015 as an effort
PT

made by the city towards sustainable urban mobility. It is an important link in the city center, which

connects commercial streets and historical sites. This pedestrian street is widely used for shopping and
E

leisure by pedestrians and cyclists of all ages. The effective width of the pedestrian street is relatively
CC

wide, despite the existence of many cafes and fixed obstacles. These obstacles are scattered in the
A

pedestrian street, preventing in some cases the pedestrians to easily move. It is worthwhile to note that

the pedestrian-cyclist segregation, which is achieved by means of floor-mounted symbols and arrays

of lamps, is not distinct and as a result most pedestrians are not aware of the bicycle lane existence.

It is assumed that both sidewalks and the pedestrian street operate as shared space areas for both

pedestrians and cyclists and they are therefore treated in the same way. Fig. 1 presents the boundaries of the
Municipality of Thessaloniki, the 12 km bicycle network, as well as the location of the three segments

examined (a. Ethnikis Amynis sidewalk, b. Aggelaki sidewalk, c. Agias Sophias pedestrian street).

T
R IP
SC
U
N
Fig. 1. The bicycle network of the Municipality of Thessaloniki and the three segments examined.
A

Fig. 2 presents the characteristics of the three segments which were examined. Since the purpose of the
M

research is to identify quantitative and qualitative parameters that affect pedestrians’ perceived LOS, field
ED

measurements and questionnaire survey were conducted in those specific segments.


E PT
CC
A

Fig. 2. a) Ethnikis Amynis sidewalk, b) Aggelaki sidewalk, c) Agias Sophias pedestrian street.

3.2. Field measurements

The field measurements were conducted from July 18 2017 to August 4 2017, during morning peak hour,

by well-trained students. As it has already been said measurements were related to the geometric and

functional attributes of the three segments and the aim was to compute the average volumes and speeds

during a typical morning. Pedestrian average speed emerged from 274 counts in the case of Ethnikis Amynis
sidewalk, 176 counts in the case of Aggelaki sidewalk and 178 counts in the case of Agias Sophias

pedestrian street. The corresponding numbers for average bicycle speed were 28, 34 and 27. It should be

noticed that speeds were determined with the aid of a stopwatch. Table 2 presents the field measurements

results. Average pedestrian and bicycle speed broadly coincide with the values given in the literature

(Highway Capacity Manual, 2010; Botma, & Papendrecht, 1991; City of Copenhagen, 2013).

Table 2

Geometric and functional attributes of the three segments.

T
Ethnikis Amynis Aggelaki Agias Sophias

IP
Type of infrastructure Sidewalk Sidewalk Pedestrian street
Total width [m] 4.8 14.5 19.5
Effective width [m] 1.55 4.45 12.10

R
Bicycle lane width [m] 2 1.5 2
Average pedestrian volume [ped/min] 2.15 5.24 19.12

SC
Average bicycle volume in bicycle lane [bic/min] 0.06 0.10 0.11
Average bicycle volume out of bicycle lane
0.01 0.02 0.06
[bic/min]
Average pedestrian speed [m/s]
Average bicycle speed [m/s]
1.35
4.30 U 1.26
4.41
1.31
4.69
N
A
3.3. Questionnaire survey
M

In order to investigate the impact of cycling in pedestrians’ perception of safety and comfort, as well as to

identify factors which affect their perception of LOS, a face to face Revealed Preference (RP) questionnaire
ED

survey was conducted at the shared use infrastructures by a well-trained postgraduate student. The survey

was done from July 18 2017 to August 4 2017, along with the field measurements (morning peak hour).
PT

Using the random sampling method, 300 valid questionnaires were completed (100 copies from every
E

segment) and were therefore analysed.


CC

A questionnaire was structured and modified slightly after the implementation of a pilot survey, which

was conducted on 13 July 2017. The 16 questionnaires which were completed during the pilot survey were
A

not included in the final sample. The final questionnaire consisted of three major parts. The first section was

relevant to the social and economic characteristics of respondents. The second part was relevant to

respondents’ mobility characteristics, as well as to their perception of the impact of cycling on their safety

and comfort in addition to other qualitative conditions of both the sidewalks and the pedestrian street. In that

part, the five point Likert scale was used (Likert, 1932). The third part, focused on how LOS was perceived.
The respondents were asked to give an overall A to F LOS evaluation, which usually applies to the three

segments. In order to facilitate the overall evaluation, appropriate cards were designed and displayed to the

respondents. Each card described a different situation regarding pedestrian flow, bicycle flow, surface

availability and conflicts. It is worth highlighting that respondents had to evaluate the usual situation of the

sidewalk or pedestrian street that they were using at that time.

4. Descriptive and inferential statistics

T
4.1. Descriptive statistics

IP
This section provides an overview of the outcomes derived from the descriptive statistical analysis.

R
Appendix A presents the coding of the variables, their levels and type of measurements, as well as the

SC
results of the descriptive statistics. It should be noted that the whole statistical analysis was carried out

through the SPSS v24.0 software.

U
The majority of the sample consisted of women (54.7%). Most of the respondents (41.7%) were between
N
25-39 years of age while a smaller group of pedestrians was observed aged aged more than 65 years old
A
(3%). This result is reasonable as it is observed in other similar surveys conducted in the city of Thessaloniki
M

(Basbas et al., 2013; Paschalidis et al., 2015). Additionally, most of the respondents (40.7%) usually use
ED

public transport for their trips while 20% of them usually perform their trips on foot and 5.3% of them by

bicycle.
PT

Furthermore, 8.7% of respondents experienced an accident with a cyclist while traveling as pedestrians,

and 28% of them barely avoided a similar accident. For those who had experienced an accident or had
E

scarcely avoided it, there was a question related to the liability of the accident which they were asked to rate.
CC

According to the respondents' replies, there is a slight tendency of putting the blame on cyclists. This

outcome coincides with the results of a survey conducted in Australia (Hatfield, & Prabhakharan, 2016).
A

Pedestrians were also asked to evaluate the impact of cycling on the sidewalks and the pedestrian street

related to their perception of both comfort and safety. Results showed that the negative impact felt by the

pedestrians on the current situation (e.g. low bicycle volume) is extremely limited. More particularly, 32.7%

of the respondents are absolutely satisfied and 36.7% of them are pretty satisfied with their perception of

comfort in sidewalks/pedestrian street due to bicycles. Respectively, 31.7 of the respondents are absolutely
satisfied and 35.3% of them are pretty satisfied by their perceived safety in sidewalks/pedestrian street due

to bicycles.

Moreover, respondents were asked about their preference for mixed infrastructure instead of the

segregated ones. The vast majority of pedestrians (91%) prefer the segregated type of infrastructure as they

feel safer and more comfortable when they use it. Pedestrians and cyclists preference for segregation is also

found in previous studies (Delaney, 2016; Paschalidis et al., 2017). Also, the results of a recent study

demonstrate that the more definite the segregation, the lesser the events experienced by pedestrians and

T
cyclists in shared use sidewalks and pedestrian streets (Basbas et al., 2019). Eventually, interviewees stated

IP
their opinion regarding the LOS of the sidewalks and the pedestrian street. It is observed that the majority of

R
the sample perceived LOS D (43.7%) or LOS C (23.3%).

SC
4.2. Inferential statistics

Before the development of the “perceived LOS” model, a correlation matrix was calculated in order to

U
reveal any possible relations among the variables. The correlation matrix, which contains only the most
N
important variables, is presented in Appendix B.
A
Regarding the examined quantitative factors, only bicycle lane width found to correlate with the
M

perceived LOS (r=.137, p<0.05). Considering the social characteristics of the pedestrians, both gender
ED

(r=.153, p<0.01) and age (r=-.173, p<0.01) are related to the perceived LOS. The results show that men tend

to perceive higher LOS, while younger pedestrians’ rating, as far as sidewalks and the pedestrian street are
PT

concerned, was lower.

Eventually, significant correlations were found between qualitative factors and perceived LOS. In
E

particular, the respondents who had evaluated comfort, safety, security and space sufficiency high, also
CC

perceived a high LOS. The effect of some of those qualitative parameters (comfort, safety, space sufficiency

due to obstacles) on pedestrians’ perceptions, has been also identified in former surveys (Hidayat et al.,
A

2011; Karndacharuk et al., 2016; Wicramasinghe, & Dissanayake, 2017).

5. Results

5.1. Model estimation


Considering the descriptive and inferential statistics results, an ordinal regression model was developed

in order to investigate the influential coefficients for the pedestrians’ perceived LOS on shared use

segments. An ordinal regression model is a differentiated version of a binary logistic regression model

which considers the ordinal coding of the dependent variable as well (see Norusis, 2005 and Gutiérrez et al.,

2016 for more details).

According to the objectives of the ordinal regression model, the dependent variable is the “ped_los”, that

is the perceived pedestrian level of service. Due to the fact that the distribution of responses was extremely

T
uneven (the vast majority answered LOS D or LOS C), the re-coding of this variable was chosen. Table 3

IP
describes the mode of re-coding the dependent variable.

R
Table 3

SC
Recoding of dependent variable.

Initial values Final values


LOS A
LOS B U
Very good
N
LOS C Good
LOS D Moderate
A
LOS E
Unacceptable
LOS F
M

Concerning the independent variables, the objective was to include quantitative factors, qualitative
ED

factors and social characteristics of respondents, utilizing the field measurements and the questionnaire

survey. Inferential statistics showed that the only quantitative factor which had correlation with “ped_los”
PT

was the bicycle lane width. However, it was found that the ratio of pedestrian volume to the effective width,
E

that is the pedestrian unit flow rate, had strong correlation with the perceived level of service (r=.150,
CC

p<0.01). In addition, due to the fact that SPSS software automatically uses the last class as a reference

category when applying ordinal regression, it was deemed necessary to re-code the nominal and ordinal
A

independent variables so that the reference category collects a relatively high number of responses. After

several “try and error” tests, the variables shown in Table 4 were finally included in the model.

Table 4

Model variables and reference categories.

Variable Reference category


Pedestrian unit flow rate -
Gender Female
Age Age group 18-24
Perceived space sufficiency due to bicycles Absolutely satisfied
Perceived space sufficiency due to obstacles Absolutely satisfied

In Appendix C both the parameter estimates and the overall fitting indices for the model are presented. In

the overall fitting indices table (see Appendix C), the Model Fitting Information indicates a statistical

significant improvement, as to whether the explanatory coefficients improve the model significantly,

compared to a baseline (intercept only) model that does not contain any independent variables. The

T
performed test compares the -2LL (Log Likelihood tests) of the baseline and the final model. The result of

IP
the p-value (sig.<.001) reveals a statistical reduction of the -2LL and thus a significant improvement.

R
Additionally, the Goodness-of-Fit tests applied suggest that the model is consistent to the data (p-

SC
values>.05) since the null hypothesis is that the fit is good. Additionally, the Nagelkerke R-Square suggests

that the final model can explain approximately 21.3% of the variance. Finally, the Test of Parallel Lines is

U
statistically insignificant (p-value>.05) and thus it can be concluded that the relation between the dependent
N
variable’s thresholds and the explanatory variables is the same for all the thresholds and therefore the key
A
assumption of the ordinal regression model has been met.
M

Finally, in order to measure the prediction power of the model, Table 5 presents the classification
ED

between the observed and the predicted cases. Summing the elements of the main diagonal (correct

predictions) and dividing by the total amount of predictions (300), the percentage of the correct predictions
PT

arises. The model manages to predict the observed responses in a percentage of 47.7%.

Table 5
E

Classification between observed and predicted cases.


CC

Predicted
Observed
Very good Good Moderate Unacceptable Total
A

Very good 6 2 31 0 39
Good 2 13 53 2 70
Moderate 3 8 110 10 131
Unacceptable 0 2 44 14 60
Total 11 25 238 26 300

5.2. Model interpretation


The interpretation of the ordinal regression models is based on the calculation of the odds ratios. It should

be noted that the odds ratios are calculated among the statistical significant intervals (confidence level 95%)

of response variables and their reference categories, as well as for those intervals which slightly overcome

the specific confidence level. The calculated odds ratios are presented in Table 6.

Table 6

Odds ratios results.

Variable Intervals Odds ratios


Pedestrian unit flow rate - 7.251

T
Male
Gender

IP
Female (reference category) 1.802
≥ 65
18-24 (reference category) 4.975

R
Age
55-64
18-24 (reference category) 2.188

SC
Not at all satisfied 5.906
Absolutely satisfied (reference category)
Perceived space sufficiency due to Somewhat satisfied 5.249
bicycles
U
Absolutely satisfied (reference category)
Neutral 2.893
N
Absolutely satisfied (reference category)
Not at all satisfied 4.932
A
Perceived space sufficiency due to Absolutely satisfied (reference category)
obstacles Somewhat satisfied 2.064
M

Absolutely satisfied (reference category)


ED

From the above table it could be stated that:

1. The only quantitative factor of the model, which is the pedestrian unit flow rate, has great impact on
PT

pedestrians’ perception about the LOS.

2. Females are more likely to perceive a lower LOS than males and the same applies to young people,
E

who have proved to be the toughest assessors of the LOS provided by the infrastructures.
CC

3. Concerning the qualitative factors of the model, both perceived space sufficiency due to bicycles and
A

perceived space sufficiency due to obstacles have significant impact on perceived LOS. More

specifically, pedestrians who have stated that space sufficiency is not restricted by bicycles or

obstacles are more likely to perceive higher LOS.

5.3. Comparison between the three segments

In this part of the analysis three ordinal regression sub-models have been developed, one for each

segment. The aim of the three sub-models is to identify possible differences in the impact of the independent
variables of the main model on the perceived pedestrian LOS by segment. It should be noted that from the

independent variables used in the main model, the pedestrian unit flow rate does not participate in the sub-

models as it is stable per segment and therefore cannot show any conclusions. The odds ratios have been

calculated only among the statistical significant intervals and their reference categories and they are

presented in Table 7. In the Table there is also a sketch for every segment, which depicts their total width,

the bicycle lane width and obstacles.

T
R IP
SC
U
N
A
M
ED
E PT
CC
A
Table 7

Odds ratios results for the three segments.

Variable Intervals Odds ratios

T
Ethnikis Amynis Aggelaki Agias Sophias

IP
Male
Gender Female (reference
1.992* 2.653** 2.058*
category)

R
≥ 65
18-24 (reference category) 4.348* - -

SC
55-64
Age
18-24 (reference category) - 3.497* 5.682**
40-54
18-24 (reference category)
Not at all satisfied
-
10.814* U -
-
2.809*
-
N
Absolutely satisfied
(reference category)
A
Somewhat satisfied 6.006** 17.790** -
Perceived Absolutely satisfied
M

space (reference category)


sufficiency due Neutral - 3.667* 4.612**
to bicycles Absolutely satisfied
ED

(reference category)
Pretty satisfied - 2.183* -
Absolutely satisfied
(reference category)
PT

Not at all satisfied - - -


Perceived Absolutely satisfied
space (reference category)
E

sufficiency due Somewhat satisfied - - 9.610**


to obstacles Absolutely satisfied
CC

(reference category)
*
confidence level 90%
**
confidence level 95%
A

The results presented in the above table first of all validate the conclusion drawn from the main model

that females and younger people perceive lower LOS. Regarding the two variables expressing space

sufficiency due to bicycles and obstacles, differences exist between the three segments. It seems that the

presence of bicycles is more important in terms of the effect on the perceived LOS, in the first two segments

(Ethnikis Amynis and Aggelaki). This conclusion is partly in contrast to the values presented in Table 1,
where it is shown that the bicycle volume was increased in the Agias Sophias pedestrian street. So it turns

out that for such low bicycle volumes like the ones existing in Thessaloniki, the pedestrians’ perceived LOS

is not influenced by the number of bicycles, but it is influenced by the distribution of the space between

pedestrians and cyclists as well as by the geometry of the infrastructure. For example, although the bicycle

volume is extremely low on the Ethnikis Amynis sidewalk, the restrictions that pedestrians perceive are

strong due to the large width of the bicycle lane that leaves very little space available to pedestrians. The

presence of obstacles is of mayor importance in the case of Agias Sophias pedestrian street and this is

T
probably due to the fact that despite the large width of the infrastructure, there are many obstacles (benches,

IP
plantings, fountains, cafeteria tables and chairs) that are scattered and do not leave free corridors for

R
pedestrian movement.

SC
6. Conclusions

6.1. Development of the analysis tool U


N
Contrary to the majority of research efforts which develop statistical models based on geometric and
A
functional attributes of the pedestrian infrastructures, an attempt has been made through this study so that an
M

integrated model is shaped by taking into consideration both quantitative and qualitative parameters, as well
ED

the pedestrians’ social characteristics. For that reason an ordinal regression model was developed, which

quantifies the effect of specific parameters on pedestrians’ perceived LOS, utilizing questionnaire survey
PT

data and field measurements from three segments. The ordinal regression model was checked for its validity

by using the test of parallel lines, which examines the proportional odds assumption, as well as by
E

developing three sub-models (one for each segment) in order to ascertain whether the individual results
CC

match the results of the main model.

Despite the fact that the bicycle volumes are quite low in the city of Thessaloniki, descriptive statistics
A

showed that a relatively high percentage of pedestrians had been involved in an accident with a cyclist

(8.7%) or barely avoided one (28%). This is an indication of how important the development of the analysis

tool becomes, as this tool aims to assist transport and urban planners who are called to develop the

appropriate infrastructure for pedestrian and cyclist coexistence. This tool can be particularly useful in cities

with little bicycle familiarity, where efforts are being made to promote active modes of transport.
For the improvement of the tool it is advisable that data is collected from additional segments, which was

not feasible in the present study due to the lack of shared use infrastructures in the city of Thessaloniki.

Collecting data from more segments could facilitate the recognition of the impact of geometric and

functional attributes on pedestrians’ perceptions. Another important point is that this study examines the

issue by having the pedestrians in the core of the analysis. However, there is also one other important player

on the ground and it could be interesting and useful to have also cyclists responses to the research questions,

as they constitute an equal user of the shared use segments.

T
6.2. Social factors in comparison with similar research

IP
The results of the analysis tool have shown a clear difference in perceptions depending on age and

R
gender. These social factors are significantly influential regarding the LOS perception as shown by the main

SC
model, but also by the three sub-models. Moreover, this conclusion coincides with the results of a previous

study which was conducted in the same city, in two pedestrian streets but without a bicycle lane on them

U
(Lazou et al., 2015). Another study which was carried out in China, also came to the same conclusion, which
N
is attributed to the different psychological diathesis and physical force of the two genders, as well as to the
A
fact that older people walk slowly and usually for non-mandatory purposes (Tan et al., 2007).
M

However this does not seem to be always the case. A survey conducted on three sidewalks in Kuala
ED

Lumpur, Malaysia concludes that the LOS which males perceive does not differ significantly from what

females perceive (Arshad et al., 2016). Another study which was carried out in Delhi, India examines the
PT

satisfaction ratings for five different aspects of sidewalks and it was found out that some aspects like width,

surface and obstruction were rated higher by females; however, security was rated much higher by males
E

(Parida, & Parida, 2011). Thus, the role of gender in the assessment of pedestrian infrastructures is maybe
CC

related to the local context.

Concerning the impact of social factors on pedestrians’ perceptions, future research could examine how
A

the characteristics of the pedestrian trip such as purpose and duration in addition to the frequent use of

sidewalk/pedestrian street affect the perception of LOS by the young and the elderly, as well as the

perception by males and females.


6.3. Geometry as a factor for perceived LOS

Through the descriptive statistics the following conclusion emerges: pedestrians’ perceived nuisance due

to bicycles, in the current situation, is limited. It is understood that in the case of low bicycle volume (<0.17

bicycles/min), the pedestrians’ comfort and safety do not deteriorate significantly. The correlation analysis

revealed that the only quantitative factor with statistical significant impact on LOS is the bicycle lane width.

A plausible assumption for this conclusion is that pedestrians feel more restricted when there is an increase

in the bicycle lane width, as this situation enhances the feeling of the presence of bicycles in the

T
infrastructure. The importance of the correct distribution of the space between pedestrians and cyclists is

IP
also demonstrated by the results of the three sub-models; therefore, it is obvious that the over-dimensioning

R
of the bicycle lane may have extremely negative effects on the pedestrians’ perceptions, regardless of the

SC
bicycle volume. Transport planners and urban planners should be particularly careful when dimensioning

shared use sidewalks and pedestrian streets. Besides, in the case of too low bicycle volumes, they should not

U
exclude the possibility of designing mixed infrastructures instead of segregated.
N
The estimation of the main model has shown that pedestrian unit flow rate is a value that determines
A
pedestrians’ perceived LOS to a significant extent. Probably, that value is more appropriate for predicting
M

LOS than considering independently the pedestrian flow and the effective width. Moreover, perceived space
ED

sufficiency due to bicycles and due to obstacles, are two parameters which have been found to have great

importance for the LOS assessment in shared use sidewalks and pedestrian streets.
E PT

References
CC

Arshad, A.K., Bahari, N.I., Hashim, W., & Halim, A.G.A. (2016). Gender differences in pedestrian

perception and satisfaction on the walkability of Kuala Lumpur City Center. In MATEC Web of
A

Conferences, Vol. 47.

Bartlett, J.E., Kotrlik, J.W., & Higgins, C.C. (2001). Organizational research: Determining appropriate

sample size in survey research. Information Technology, Learning, and Performance Journal, 19(1), 43.
Basbas, S., Tetou, V., & Politis, I. (2013). Ordinal and binary logistic logit models for examination of

behavioral, infrastructure and perception factors influencing biking. WIT Transactions on the Built

Environment, 130.

Basbas, S., Nikiforiadis, A., Militsis, N., Grigoriadis, G., Theodoroglou, V., & Aifadopoulou, G. (2019).

Predicting cyclists and pedestrians hindrance on shared use sidewalks and pedestrian streets using log-

linear models. 98th Annual Meeting of the Transportation Research Board, Washington D.C., USA.

Bian, Y., Wang, W., & Lu, J. (2007). Pedestrian level of service for sidewalks. Journal of Southeast

T
University (Natural Science Edition), 37 (4), 695–699.

IP
Blondiau, T., van Zeebroeck, B., & Haubold, H. (2016). Economic Benefits of Increased Cycling.

R
Transportation Research Procedia, 14, 2306-2313.

SC
Botma, H., & Papendrecht, H. (1991). Traffic Operation of Bicycle Traffic. Transportation Research

Record, 1320, 65–72.

U
Chong, S., Poulos, R., Olivier, J., Watson, W.L., & Grzebieta, R. (2010). Relative injury severity among
N
vulnerable non-motorised road users: Comparative analysis of injury arising from bicycle–motor vehicle
A
and bicycle–pedestrian collisions. Accident Analysis & Prevention, 42, 290–296.
M

City of Copenhagen, international.kk.dk. (2013). Retrieved 25 November 2017.


ED

COM (2011). 144 White Paper: Roadmap to a Single European Transport Area – Towards a competitive and

resource efficient transport system. European Commission, Brussels.


PT

Dannenberg, A.L, Kraft, K., & Alvanides, S. (2017). Special Issue Editorial: Tools and practices for

understanding and promoting walking and walkability. Journal of Transport & Health, 5, 1-4.
E

Delaney, H., Parkhurst, G., & Melia, S. (2016). Walking and cycling on shared-use paths: the user
CC

perspective. Proceedings of the Institution of Civil Engineers. http://dx.doi.org/10.1680/jmuen.16.00033.

Friman, M., Gärling, T., Ettema, D., & Olsson, L. (2017). How does travel affect emotional well-being and
A

life satisfaction? Transportation Research Part A, 106, 170-180.

Grzebieta, R.H., McIntosh, A.M., & Chong, S. (2011). Pedestrian-Cyclist Collisions: Issues and Risk.

Australasian College of Road Safety Conference, Melbourne.


Gutiérrez, P.A., Pérez-Ortiz, M., Sánchez-Monedero, J., Fernández-Navarro, F., & Hervás-Martínez, C.

(2016). Ordinal Regression Methods: Survey and Experimental Study. IEEE Transactions on Knowledge

and Data Engineering, 28, 127-146, doi: 10.1109/TKDE.2015.2457911.

Hatfield, J., & Prabhakharan P. (2016). An investigation of behavior and attitudes relevant to the user safety

of pedestrian/cyclist shared paths. Transportation Research Part F: Traffic Psychology and Behaviour, 40,

35–47.

Hidayat, N., Choocharukul, K., & Kishi, K. (2011). Pedestrian Level of Service Model Incorporating

T
Pedestrian Perception for Sidewalk with Vendor Activities. Journal of the Eastern Asia Society for

IP
Transportation Studies, 9, 1012-1023.

R
Hummer, J., Rouphail, N., Hughes, R., Fain, S., Toole, J., Patten, R., Schneider, R., Monahan, J., & Do, A.

SC
(2005). User perceptions of the quality of service on shared paths. Transportation Research Record, 1939,

28–36.

U
Jaskiewicz, F. (2000). Pedestrian Level of Service based on trip quality. Transportation Research Circular,
N
Transportation Research Board, National Research Council, Washington D.C.
A
Jena, S. (2014). Perception based pedestrian level of service. MTech Thesis, Supervisor: Bhuyan P. K.
M

National Institute of Technology Rourkela, Department of Civil Engineering, India.


ED

Jensen, S. (2007). Pedestrian and bicyclist level of service on roadway segments. Transportation Research

Record, 2031, 43–51.


PT

Kang, L., Xiong, Y., & Mannering, F. L. (2013). Statistical analysis of pedestrian perceptions of sidewalk

level of service in the presence of bicycles. Transportation Research Part A: Policy and Practice, 53, 10–
E

21, doi: 10.1016/j.tra.2013.05.002.


CC

Karndacharuk, A., Wilson, D., & Dunn, R. (2016). Qualitative evaluation study of urban shared spaces in

New Zealand. Transportation Research Part D: Transport and Environment, 42, 119–134.
A

Kartsoviti, S. (2011). Cycling in combined urban transport: The case of Thessaloniki. MSc Thesis,

Supervisor: Pitsiava, M. Aristotle University of Thessaloniki, School of Civil Engineering, (in Greek with

English abstract).
Landis B.W., Vattikuti V.R., Ottenberg R.M., McLeod D.S., & Guttenplan M. (2001). Modelling the

roadside walking environment: A pedestrian level of service. Transportation Research Record 1773,

TRB, National Research Council, Washington, D.C.

Lazou, O., Sakellariou, A., Basbas, S., Paschalidis, E., & Politis, I. (2015). Assessment of LOS at pedestrian

streets and qualitative factors. A pedestrians’ perception approach. 7th International Congress on

Transportation Research, Athens, Greece.

Likert, R. (1932). A Technique for the Measurement of Attitudes. Archives of Psychology, 140, 1-55.

T
Litman, T.A. (2017). Evaluating active transport benefits and costs. Guide to Valuing Walking and Cycling

IP
Improvements and Encouragement Programs. Prepared for Victoria Transport Policy Institute.

R
Morettini, M., Storm, F., Sacchetti, M., Cappozzo, A., & Mazzà, C. (2015). Effects of walking on low-grade

SC
inflammation and their implications for Type 2 Diabetes. Prev Med Reports, 2, 538–547, doi:

10.1016/j.pmedr.2015.06.012.

U
Mōri, M., & Tsukaguchi, H. (1987). A new method for evaluation of level of service in pedestrian facilities.
N
Transportation Research Part A: General, 21(3), 223–234, doi: 10.1016/0191-2607(87)90016-1.
A
Mulley, C., Tyson, R., McCue, P., Rissel, C., & Munro, C. (2013). Valuing active travel: including the
M

health benefits of sustainable transport in transport appraisal frameworks. Research in Transportation


ED

Business & Management, 7, 27–34.

Muraleetharan, T., Adachi, T., Uchida, K., Hagiwara, T., & Kagaya, S. (2003). A Study on Evaluation of
PT

Pedestrian Level of Service along Sidewalks and at Intersections Using Conjoint Analysis. Annual

Meeting of Japanese Society of Civil Engineers (JSCE) Infrastructure Planning, Toyohashi, Japan.
E

Naing, L., Winn, T., & Rusli, B. N. (2006). Practical issues in calculating the sample size for prevalence
CC

studies. Archives of Orofacial Sciences, 1(1), 9–14.

Norusis, M. (2005). IBM SPSS Statistics 19 Advanced Statistical Procedures Companion. Prentice Hall.
A

O'Connell, A. (2006). Quantitative Applications in the Social Sciences: Logistic regression models for

ordinal response variables. Thousand Oaks, CA: SAGE Publications Ltd.

Oja, P., Titze, S., Bauman, A., de Geus, B., Krenn, P., Reger-Nash, B., & Kohlberger, T. (2011). Health

benefits of cycling: a systematic review. Scandinavian Journal of Medicine & Science in Sports, 21(4),

496-509.
Parida, P., & Parida, M. (2011). Appreciation of gender differences in development of qualitative level of

service for sidewalks. Women’s Issues in Transportation, 246.

Paschalidis, E., Basbas, S., Politis, I., & Prodromou, M. (2015). ‘‘Put the blame on. . .others!’’: The battle of

cyclists against pedestrians and car drivers at the urban environment. A cyclists’ perception study.

Transportation Research Part F: Traffic Psychology and Behaviour, 41, 243-260,

http://dx.doi.org/10.1016/j.trf.2015.07.021.

Paschalidis, E., Prodromou, M., Basbas, S., & Politis, I. (2017). Investigation of Cyclists' Attitudes and

T
Perceptions Towards other Road Users: Evidence from a Case Study in Thessaloniki, Greece.

IP
International Journal of Transportation, 5, 33-46.

R
Patten, R., Schneider, R., Toole, J., Hummer, J., & Rouphail, N. (2006). Shared-Use Path Level of Service

SC
Calculator – A User’s Guide. Final Report, FHWA-HRT-05-138.

Petritsch, T., Landis, B., McLeod, P., Huang, H., Challa, S., Skaggs, C., Guttenplan, M., & Vattikuti, V.

U
(2006). Pedestrian level of service model for urban arterial facilities with sidewalks. Transportation
N
Research Record 1982, 84–89.
A
Rahaman, K. R., Harata, N., & Ohmori, N. (2005). Evaluation of the Roadside Walkway Environment of
M

Dhaka City. Proceeding of the Eastern Asia Society for Transportation Studies, 5, 1751-1766.
ED

Singleton, P. (2018). Walking (and cycling) to well-being: Modal and other determinants of subjective well-

being during the commute. Travel Behaviour and Society.


PT

Tan, D., Wang, W., Lu, J., & Bian, Y. (2007). Research on Methods of Assessing Pedestrian Level of

Service for Sidewalk. Journal of Transportation Systems Engineering and Information Technology, 7(5),
E

74–79. doi: 10.1016/S1570-6672(07)60041-5.


CC

Transportation Research Board of the National Academies (TRB). (2010). Highway Capacity Manual

(HCM2010), Volume 3: Interrupted Flow, Chapter 23: Off-street pedestrian and bicycle facilities,
A

Washington, DC.

Vaitsis, P., Basbas, S., & Nikiforiadis, A. (2019). How Eudaimonic Aspect of Subjective Well-Being Affect

Transport Mode Choice? The Case of Thessaloniki, Greece. Social Sciences, 8(1):9.

https://doi.org/10.3390/socsci8010009
Wefering, F., Rupprecht, S., Bührmann, S., & Bohler-Baedeker, S. (2013). Guidelines. Developing and

implementing a sustainable urban mobility plan. Brussels.

Wicramasinghe, V., & Dissanayake, S. (2017). Evaluation of pedestrians’ sidewalk behavior in developing

countries. Transportation Research Procedia, 25C, 4072–4082.

T
R IP
SC
U
N
A
M
ED
E PT
CC
A

You might also like