ArticlePDF Available

Drivers' parking location choice under uncertain parking availability and search times: A stated preference experiment

Authors:

Abstract and Figures

To assess parking pricing policies and parking information and reservation systems, it is essential to understand how drivers choose their parking location. A key aspect is how dri-vers' behave towards uncertainties towards associated search times and finding a vacant parking spot. This study presents the results from a stated preference experiment on the choice behaviour of drivers, in light of these uncertainties. The attribute set was selected based on a literature review, and appended with the probabilities of finding a vacant parking spot upon arrival and after 8 min (and initially also after 4 min, but later dropped to reduce the survey complexity). Efficient Designs were used to create the survey design, where two rounds of pilot studies were conducted to estimate prior coefficients. Data was successfully collected from 397 respondents. Various random utility maximisation (RUM) choice models were estimated, including multinomial logit, nested logit, and mixed logit, as well as models accounting for panel effects. These model analyses show how drivers appear to accept spending time on searching for a vacant parking spot, where parking availability after 8 min ranks second most important factor in determining drivers' parking decisions, whilst parking availability upon arrival ranks fourth. Furthermore, the inclusion of heterogeneity in preferences and inter-driver differences is found to increase the predic-tive power of the parking location choice model. The study concludes with an outlook of how these insights into drivers' parking behaviour can be incorporated into traffic assignment models and used to support parking systems.
Content may be subject to copyright.
Drivers’ Parking Location Choice under Uncertain Parking Availability and
Search Times: A Stated Preference Experiment
Emmanouil Chaniotakisa,b, Adam J. Pelc
aCentre for Research and Technology Hellas – Hellenic Institute of Transport, 6th km Charilaou–Thermi Rd., 57001 Thermi,
Thessaloniki, Greece
bNational Technical University of Athens, School of Rural and Surveying Engineering, Zografou 15780, Greece
cDepartment of Transport and Planning, Faculty of Civil Engineering and Geosciences, Delft University of Technology,
Stevinweg 1, 2628 CN Delft, The Netherlands
Abstract
To assess parking pricing policies and parking information and reservation systems, it is essential to under-
stand how drivers choose their parking location. A key aspect is how drivers’ behave towards uncertainties
towards associated search times and finding a vacant parking spot. This study presents the results from
a stated preference experiment on the choice behaviour of drivers, in light of these uncertainties. The at-
tribute set was selected based on a literature review, and appended with the probabilities of finding a vacant
parking spot upon arrival and after 8 minutes (and initially also after 4 minutes, but later dropped to reduce
the survey complexity). Efficient Designs were used to create the survey design, where two rounds of pilot
studies were conducted to estimate prior coefficients. Data was successfully collected from 397 respondents.
Various random utility maximisation (RUM) choice models were estimated, including multinomial logit,
nested logit, and mixed logit, as well as models accounting for panel effects. These model analyses show how
drivers appear to accept spending time on searching for a vacant parking spot, where parking availability
after 8 minutes ranks second most important factor in determining drivers’ parking decisions, while parking
availability upon arrival ranks fourth. Furthermore, the inclusion of heterogeneity in preferences and inter-
driver differences is found to increase the predictive power of the parking location choice model. The study
concludes with an outlook of how these insights into drivers’ parking behaviour can be incorporated into
traffic assignment models and used to support parking systems.
Keywords: parking, parking uncertainty, choice behaviour, choice factors, stated preference
1. Introduction
One of the negative effects that can be associated to parking in urban areas is the presence of cruising
traffic. That is, drivers may need to drive around while searching for a vacant parking spot. This leads
to additional traffic on the urban network. Studies by Axhausen et al. (1994), Arnott & Inci (2006), and
Shoup (2006) found that at certain moments during the day up to 50 percent of the traffic is related to5
cruising for parking. This evidently is a contributor to problems regarding congestion levels, travel times,
emissions, and traffic safety (McCoy et al.,1990;Shoup,2006). These problems can be addressed by parking
policies (e.g., pricing policies and/or parking permits) as well as through the use of Parking Guidance and
Information (PGI) and Parking Reservation systems. Many studies in the literature have particularly focused
on designing efficient pricing policies (Gillen,1978;Anderson & de Palma,2004;Calthrop & Proost,2006;10
Mei et al.,2010;Ottosson et al.,2013;Millard-Ball et al.,2014) and permit schemes (Liu et al.,2014a,b;
Barata et al.,2011), also in light of drivers’ willingness to pay for parking (Hess & Polak,2004;Ibeas et al.,
Corresponding author
Email address: chaniotakis@certh.gr (Emmanouil Chaniotakis)
Preprint submitted to Transportation Research Part A July 23, 2015
2014). At the same time, others argue that what can be achieved via parking pricing and permit policies
is limited due to social and political constraints (Verhoef et al.,1995;Lam et al.,2006). In this regard,
Intelligent Transport Systems (ITS) aimed at informing and guiding drivers towards vacant parking spots15
and at allowing drivers to reserve parking spots have good prospects (Thompson & Bonsall,1997).
In accordance to the solutions proposed, researchers have developed traffic assignment and simulation
models to be used for their evaluation. To name but a few, Thompson & Richardson (1998) presented
a simulation parking search model in which individuals search for parking space and accept or reject the
vacant parking spots based on a disutility function as well as a function to represent the likelihood to20
select a particular turning movement. Lam et al. (2006) proposed a model for user equilibrium flows that
accounts for the departure time, parking location and parking duration. A Bureau of Public Roads based
(BPR) function is used as a cost function and the model includes multiple user classes and car parks.
In the same direction, Gallo et al. (2011) derived a simulation–based parking assignment model with a
hierarchical structure simulating parking location choices on a trip, a cruising and a walking layer, whereas25
Guo et al. (2013) presents an assignment model for strategic parking route choice decisions under uncertainty.
Leurent & Boujnah (2014) give a very detailed analysis of the network flows under the assumption of
drivers choosing their parking location and rerouting based on expected conditions in combination with the
stochastic realization of actual parking occupancies. Others have employed a different modelling approach
by specifically considering the searching behaviour of drivers. For example, Young & Taylor (1991) describes30
PARKSIM as a model that describes the parking search within a car park, and Benenson et al. (2008); Levy
et al. (2015) developed PARKAGENT; an agent-based model to explicitly simulate the spatial search process
and navigation behaviour of drivers.
Evidently, the (model–based) design and evaluation of such parking-oriented ITS and the derivation of
policies on parking require a valid understanding of drivers’ choice behaviour among alternative parking35
locations, as well as their willingness–to–pay, usage, and acceptance towards these systems. In this paper,
we focus on the former, here termed as parking location choice behaviour, which can improve the assignment
and simulation models developed for evaluation and has also clear implication for parking–related policies.
Earlier studies on parking location choice have aimed at identifying the determining factors for drivers’
parking decisions (Gillen,1978;Van Der Goot,1982;Polak et al.,1991;Brandley et al.,1993;Hunt &40
Teply,1993;Lambe,1996). Other studies have investigated parking location choice in combination with
related travel decisions, such as trip purpose (Van der Waerden & Oppwal,1995;Shiftan & Burd-Eden,
2001) and mode of transport (Hensher & King,2001;Hess,2001;Coppola,2004), as well as route choice
of cruising traffic (Guo et al.,2013;Leurent & Boujnah,2014). Typically, the influence of these factors is
studied through the estimation of discrete choice models, where these models are based on stated preference45
experiments. Attributes that are recurrently reported are parking costs, walking distance to destination,
access time, expected search time or waiting time at parking location, and to a lesser degree driver’s age, type
of parking facility, and parking duration. With respect to the design and evaluation of parking information,
guidance, and reservation systems, what is especially of interest is how drivers’ parking location choice
is affected by uncertainties in finding a vacant parking spot. In this regard, only the study by Van der50
Waerden (2012) explicitly accounts for the chance (probability) of finding a vacant parking spot. There,
parking availability probabilities are found to be significant for the choice set generation, where a higher
probability to find a vacant parking spot relates to a higher probability to choose that parking location. In
line, Kaplan & Bekhor (2011) presented a theoretical framework with a hierarchical structure where on-route
the preferred parking type determines the preferred parking location, which in turn determines the preferred55
route towards this destination.
The experiments reported in this paper were part of a larger study to investigate the viability of parking-
oriented ITS. Here the main target group of a (future) parking guidance and reservation system was chosen
to be drivers who make a trip for shopping purposes and park for a duration of 2-3 hours. Hence, this is
also the context for the remainder of this paper.60
This way, this paper first of all adds to the existing body of the literature (including the studies by
Kaplan & Bekhor (2011) and Van der Waerden (2012)) using stated preference experiments to understand
parking location choice behaviour in case of uncertain search times and parking occupancy levels. Second
of all, we analyse drivers’ parking location choice and show the importance of uncertain parking availability
2
as a determining choice factor, as well as the heterogeneity in preferences and inter-driver differences in this65
respect. Third of all, we estimate choice models that can be readily incorporated in both disaggregated
microscopic simulation models as well as aggregated macroscopic traffic assignment models. Finally, we dis-
cuss the quantified behavioural findings and conclusions within this paper with respect to their implications
on parking–related policies and transportation modelling.
The structure of this paper is as follows: We first introduce the stated preference experiments on drivers’70
parking location choice in Section 2. The estimated parking location choice models are then presented in
Section 3. The main behavioural findings from these models are discussed in Section 4. In Section 5, we
conclude with an outlook towards including drivers’ parking location choice behaviour in traffic assignment
models to support the evaluation of parking systems.
Note that throughout the paper, uncertainty concerning parking availability and search times refers to75
the situation in which the outcome is unknown, but the set of possible outcomes and their probabilities are
known. This terminology is consistent with the larger body of parking–related literature. However, note
that in many other lines of research this situation would typically be referred to as risk (while uncertainty
would be used to indicate that the set of possible outcomes and their probabilities are unknown).
2. Stated Preference Experiment80
In order to understand drivers’ parking location choice a series of stated preference experiments were
conducted in the form of internet–based questionnaires. In this section, we first describe the attribute set
and levels (in Section 2.2) where the set includes attributes accounting for the probability of acquiring a
vacant parking spot upon arrival, and after 8 minutes. The survey method is briefly illustrated (in Section
2.1), after which the experiment design is derived through two pilot studies (explained in Section 2.3). The85
collected data set of 397 successful respondents is discussed in terms of descriptive statistics (in Section 2.4).
This stated preference data is used in Section 3 to estimate various parking location choice models.
2.1. Questionnaire and Survey design
The survey conducted was internet–based using the LimeSurvey open-source platform and run on a
server at Delft University of Technology. The survey consisted of three parts.90
The first part pertained to questions on the personal characteristics of the respondent: gender, age,
education, income, and home postcode. Furthermore, two questions were asked regarding drivers licence
possession and frequency of private car trips to the city centre (“How often do you drive to the city centre
for shopping? ”). Only in case participants were with drivers licence and drove to the city centre at least
once per month, the survey continued.95
The second part pertained to questions on parking behaviour in an unfamiliar area. These questions
related to aspects of acceptable search times and on-route decisions to reroute towards another parking
location. The particular questions and formulation were refined during the pilot studies. The first question
of the second part was a closed question on their strategy users undertake when driving to a city centre they
are unfamiliar (“What do you usually do when you want to visit a city you are not familiar with? ”). The100
given answers were either search pre-trip for available parking locations, search upon arrival to destination or
search when arriving close to the destination. The second question referred to preference upon on-street or
off-street parking (“Which of the following options seems more attractive to you for 2-3 hours of parking?”),
while the third question referred to the reaction in case of not finding a vacant parking spot after 5 minutes
for their case of on-street or off-street parking (i.e. What do you do if you do not find a vacant parking105
spot at your selected parking lot after waiting for 5 minutes?”). Finally, the forth question referred to the
maximum search duration (or waiting time in case of off-street parking) that respondents were willing to
spend “What is the maximum duration of waiting before going to another parking facility?”).
The third part showed a series of choice situations where the respondent was every time asked to select
her/his preferred parking location from two alternatives. A screenshot of such a choice situation is shown110
in Figure 1. In total 12 choice situations were shown, based on the selected attributes (Section 2.2) where
the attribute levels were based on an efficient experiment design explained in Section 2.3.
Respondents had the option to complete the survey in either English or Dutch.
3
Figure 1: Screenshot of choice situation in survey
2.2. Attribute selection
Table 1gives an overview of the attributes that in earlier studies have been found to have a statistically115
significant influence on drivers’ parking location choice. As observed, parking location choice is a rather
complex decision process with many factors potentially playing a role. At the same time, the inclusion of a
large number of attributes creates large dimensions in a stated preference context that can result in biased
model estimation (Caussade et al.,2005;Hensher,2006).
For this reason, in our stated preference experiments we chose to limit the number of choice situations,120
attributes, and attribute levels, such that an efficient experiment design (Rose & Bliemer,2008) and rea-
sonable number of respondents would still yield satisfactory results. While evidently the primary attributes
as well as the attributes pertaining to probabilities in parking availability are included.
This attribute set consists of the attributes that are reported among earlier studies, and appended with the
probability of finding a parking spot upon arrival, and after 8 minutes. The uncertainty in finding a parking125
spot is included in this way, as it is believed to be intuitively understandable for respondents. Furthermore,
the influence of the probability of a parking spot upon arrival is interesting as it directly relates to the
benefit of parking information and guidance systems (where this probability would be 1). The probability
after 8 minutes is included to test how search times are appreciated, while 8 minutes is suggested to be
within the range of an acceptable search time before rerouting to a different parking area (found as average130
cruising time for on-street parking by Shoup (2006), and also suggested from our pilot studies, see Section
2.3.1).
For reasons of model estimation, we limit all attributes to 2 attribute levels. The initial values for the
4
Table 1: Parking choice model attributes reported in literature
Study Park
Cost
Walk
Dist.
Access
Time Search
Time
Dura-
tion TypeaAge
Illegal
fine
Pur-
pose Inc. PGIb
Occu-
pancy
cProb.d
1
2
3
4
5
6
7
8
9
10
11
aType: On-Street or Off-Street, bParking Guidance and Information Systems, cParked Cars to Capacity, dProbability
to find a vacant parking spot
where:
1. Gillen (1978) 7. Thompson & Richardson (1998)
2. Hunt (1988) 8. Dell’Orco et al. (2003)
3. Kanafani (1983) 9. Bonsall & Palmer (2004)
4. Axhausen & Polak (1991) 10. Ruisong et al. (2009)
5. Hunt & Teply (1993) 11. Van der Waerden (2012)
6. Lambe (1996)
pilot studies were chosen as lower and upper bound on a range that is deemed realistic. The final attribute
levels are determined after the two pilot studies.135
This way the final attribute level values were set as follows. The level values for parking price were set
at AC1.25 and AC2.50 per hour. Parking near shopping areas in the Netherlands is typically paid, where the
average hourly fee in a Dutch city centre is AC1.55 for on-street parking and AC1.52 for off-street parking (Van
Ommeren et al.,2012). Hourly parking fees in larger cities such as Rotterdam and Amsterdam tend to be
higher (respectively AC3.00 and AC5.00). In general, parking prices do not considerably differ among parking140
areas within the same vicinity of a shopping area, and free parking is not available within reasonable walking
distance. The level values for distance to destination were set at 100 meters and 700 meters. The upper
level is based on what is found as maximum reasonable walking distance in studies on catchment areas of
urban transit stops (Kittelson & Associates,2003, p.3-93). The level values for travel time to parking area
were set at 16 minutes and 24 minutes. These values are deemed realistic within the context of a car trip145
for shopping in a Dutch city centre. For the parking type, we distinguish on-street and off-street parking,
where the first relates to curb-side (on-street) parking and the latter relates to a (often monitored) parking
facility (off-street). This attribute was included as the way in which parking information and guidance
systems would operate may differ between these parking types. Finally, the probability values for finding a
vacant spot were set at 10, 40, and 70 percent, and at 40, 70, and 100 percent after 8 minutes. Infeasible150
combinations of level values where the probability would decrease after time were evidently excluded.
2.3. Experiment design
The design of the experiment followed a three phase procedure to fully benefit from using a well-
formulated efficient design (Rose & Bliemer,2009). The usefulness of such a multilevel approach is twofold.
Firstly, this approach allows the adaptation of the question’s formulation and focus, in order to make them155
realistic and understandable – especially with respect to the stated preference part of the survey. Secondly,
this approach allows the increase of the information to be obtained from the (final) experimental design –
due to a better model estimation–, against the conventional approach of using orthogonal designs (Louviere
et al.,2000;Antony,2003) and eliminating dominant alternatives, as it has been found that even with little
5
information on the estimates (priors) the experiment design would yield more information (Rose et al.,2008;160
Bliemer et al.,2009). The measure of information was evaluated on the minxXid det(Pθ) (D-optimality)
where Pθdenotes the covariance matrix.
A first round pilot study was designed based on prior coefficient estimates reported in literature. Then,
a second round pilot study was designed based on the model coefficient estimates from the first pilot study.
Finally, the estimated model from the second pilot was used to generate an efficient design for the final165
survey. These pilot studies are discussed first, before presenting the final design.
All reported experiment designs were generated using the Ngene software (ChoiceMetrics,2012) and all
reported choice models were estimated using the Biogeme software (Bierlaire,2006) and mLogit library for
R software (Croissant et al.,2012).
2.3.1. Pilot studies170
Two pilot studies were undertaken in order to derive prior coefficient estimates for the efficient design of
the final experiment. Both pilot studies were distributed among students and employees at Delft University
of Technology and at Dutch Organisation for Applied Research (TNO). Both pilot studies were accompanied
with an evaluation that was used to assess the complexity of the (internet-based) survey and for content
improvements.175
The first pilot study was created based on a combination of the available priors (0LPr) reported in the
studies by Van der Waerden (2012) and Axhausen & Polak (1991). In this design there were some dominant
alternatives in some scenarios and the information that could be acquired was not maximum (i.e., the design
was sub-optimal) mainly due to the fact that priors were harvested from two studies. However, this design
did accommodate more information than an orthogonal design (which yielded a higher number of dominant180
scenarios) and it was chosen to be implemented in the first round of the pilot study with a small sample.
After acquiring the answers from 11 respondents, a random utility maximisation (RUM) multinomial
logit (MNL) model was estimated. These RUM-MNL model coefficients are evidently not an accurate rep-
resentation of drivers’ parking location choice due to the setup of the survey design and small (unstratified)
sample size. However, these coefficients served well as priors to the second pilot (1LPr). Furthermore, a185
number of minor changes were made to the survey design, based on the evaluation feedback. The upper
level of the walking distance to destination was increased to 700 meters, because a number of respondents
indicated an indifference between 100 meters and 500 meters (initial value). The travel time to parking area
was increased (from initially 6 and 12 minutes to fianlly 16 and 24 minutes), as many respondents mentioned
that in case of initially chosen lower travel times, the bicycle would be the preferred mode of transport.190
The second pilot study was created based on the estimates from the first pilot study. The experiment
design was generated without dominant scenarios and had a D-error of 0.10. The survey was completed by
35 respondents, and again a RUM-MNL model was estimated. All coefficients had the correct sign, where
price, walking distance, and travel time are negative and parking type (where 1 indicates off-street parking)
and all vacancy probabilities are positive. Alternative specific constants were checked for, but found not195
statistically significant (i.e. there was no significant unexplained bias towards a specific choice situation
within the survey, which was expected as the alternatives were unlabelled). Heterogeneity in the preferences
was tested by estimating a RUM mixed-MNL model, but the variances in the coefficients were found not
statistically significant (i.e. there were no significant differences in the parking location choice preferences
of the respondents). The RUM-MNL estimates were therefore used in the final design.200
The survey evaluation feedback showed that the survey was experienced as time-consuming and rather
difficult to complete due to the high number of attributes. Many respondents indicated that they therefor
eliminated some attributes and made decisions based on those they considered as most important. This
corresponds with the findings by Hensher (2006) and Caussade et al. (2005) about stated preference survey
complexity where they caution for biased coefficient estimates. This issue was addressed by making the205
following adjustments
The attribute on the probability of finding a vacant parking spot after 4 minutes was dropped.
The lower level of the parking price was changed from AC1.50 to AC1.25 in order to increase the attribute
range.
6
The size of the design was set at 24 choice situations and divided into two blocks such that each210
respondent was shown 12 choice situations.
2.3.2. Final design
The attributes and levels included in the final design are summarised in Table 2. The final design for
the choice situations was based on the priors derived by the second round of the pilot study (2LPr). Again,
combinations of levels and model structures were tested using as a benchmark the D-error. As the priors215
of the model were fixed, the level values within the range of the level values used for the model calculation
resulted in different models. It was found that for the probability attributes, a design of 3 levels would have
better results in acquiring the most information (given the number of attributes and the number of choice
situations). The D-error of the final design is 0.01. The Fisher information matrix is shown in Table 3. The
final design is shown in Table 4.220
Table 2: Survey attributes and values in final design
Attributes Levels Level Values
Cost (C) 2 AC1.25 / AC2.5
Distance to destination (walking) (W) 2 100 meters / 700 meters
Travel time (T) 2 16 min / 24 min
Parking type (P) 2 On-Street / Off-Street
Probability upon arrival (P r0) 3 10% , 40%, 70%
Probability after 8 minutes (P r8) 3 40%, 70%, 100%
Table 3: Fisher Information Matrix of final design
Prior βCβWβTβPβP r0βP r8
βC8.082 -276.819 -8.516 0.401 0.881 0.732
βW-276.819 1862152 -2401.330 667.310 470.678 504.556
βT-8.516 -2401.330 331.049 2.840 4.392 3.123
βP0.401 667.311 2.840 5.173 -0.368 -0.148
βP r00.881 470.678 4.396 -0.368 1.232 -0.015
βP r80.733 504.556 3.123 -0.148 -0.015 1.554
2.4. Data set
Survey respondents were recruited via the Dutch company Respondenten Database who maintain a panel
of people who are paid to participate in this type of surveys. From 460 total respondents, 397 respondents
were found to have completed the survey successfully. This sample size is considered representative of the
car-driving population in the Netherlands (where the minimum sample size equals 384 respondents with225
a degree of accuracy of 0.05 according to the method by Krejcie & Morgan (1970)). Furthermore, the
distribution of gender, age, and highest level of education in the sample corresponds with the statistics
of the Dutch population (based on the Dutch Bureau of Statistics data for 2009 (CBS,2009)). Table 5
summarises the socio-demographic and personal characteristics of the respondents, while Table 6contains
the descriptive statistics for the Dutch population. Due to lack of information, the representativeness of the230
respondents across income levels was not checked.
In the second part of the survey, respondents were asked about their parking search strategy, their
preferred parking type, their behaviour after 5 minutes of searching (or waiting), and their maximum
searching (or waiting) time in case of being unfamiliar with the parking location.
Regarding parking search strategy, 145 respondents (36.6 %) indicated that they decide upon their235
parking area pre-trip, while 199 respondents (50.1 %) stated they start searching for parking once they
7
Table 4: Final experiment design
Choices Alternative 1 Alternative 2 Block
C1W1T1P1P r0
1P r8
1C2W2T2P2P r0
2P r8
2
1 1.25 100 24 1 0.1 0.4 2.5 700 16 0 0.7 1 1
2 1.25 100 24 0 0.1 1 2.5 700 16 1 0.4 0.4 1
3 2.5 100 16 1 0.1 1 1.25 700 24 0 0.7 0.7 2
4 2.5 100 24 0 0.7 0.7 1.25 700 16 1 0.1 1 1
5 1.25 700 24 1 0.1 1 2.5 100 16 0 0.7 0.7 2
6 1.25 700 24 1 0.4 0.4 2.5 100 16 0 0.1 1 2
7 1.25 700 24 0 0.7 1 2.5 100 16 1 0.1 0.4 2
8 2.5 700 24 1 0.7 1 1.25 100 16 0 0.1 0.4 1
9 1.25 700 16 0 0.1 1 2.5 100 24 1 0.4 0.4 1
10 1.25 700 16 0 0.7 0.7 2.5 100 24 1 0.1 1 1
11 2.5 700 16 1 0.4 0.4 1.25 100 24 0 0.1 1 1
12 1.25 700 16 1 0.1 1 2.5 100 24 0 0.7 0.7 1
13 2.5 100 16 0 0.1 1 1.25 700 24 1 0.4 0.4 1
14 1.25 700 24 0 0.1 1 2.5 100 16 1 0.4 0.4 1
15 2.5 700 16 0 0.7 1 1.25 100 24 1 0.1 0.4 2
16 2.5 100 24 0 0.1 1 1.25 700 16 1 0.4 0.4 2
17 1.25 100 16 0 0.1 0.4 2.5 700 24 1 0.7 1 1
18 2.5 700 24 0 0.7 1 1.25 100 16 1 0.1 0.4 2
19 1.25 100 24 0 0.7 0.7 2.5 700 16 1 0.1 1 1
20 1.25 100 16 0 0.4 0.4 2.5 700 24 1 0.1 1 2
21 1.25 700 16 0 0.1 0.4 2.5 100 24 1 0.7 1 2
22 2.5 100 16 0 0.1 0.4 1.25 700 24 1 0.7 1 2
23 1.25 100 24 1 0.1 1 2.5 700 16 0 0.4 0.4 2
24 1.25 700 24 0 0.4 0.4 2.5 100 16 1 0.1 1 2
Table 5: Socio-demographic characteristics of respondents
Respondents 397 (83.8% of total)
Mean age (standard deviation) 45.64 (14.9)
Age classes (class) 1.3% [18-19), 37.0% [20-40), 48.4%
[40-65), 13.4% [65-80)
Female respondents 215 (54.2% of completed)
Highest level of education 7.3% Primary school, 50.9% High
school , 34.5% Higher education ,
6.8% Masters degree , 0.5% Doctors
degree
Income classes (class) 17.9% [AC5000 - AC15000), 23.9%
[AC15000 - AC25000), 28.5% [AC25000
-AC35000), 15.9% [AC35000 -
AC45000), 5.0% [AC45000 - AC55000),
8.8% >AC55000
arrive at their destination, while 53 respondents (13.4 %) that they start searching for parking before
arriving at their destination.
The preferred parking type was off-street parking for 312 respondents (78.6 %). In case of a preference
for off-street parking and not being able to find a vacant parking spot within 5 minutes, the majority of240
respondents (189 respondents, 60.6 %) stated they would go to the next nearest off-street parking facility
while on-route searching for on-street parking, whereas 77 respondents (24.7 %) indicated they would go
directly to the next nearest off-street parking facility. Only 46 respondents (14.7 %) responded that they
8
Table 6: Socio-demographic characteristics of Dutch population (CBS,2009)
Percentage of female 50.5%
Age classes (class) 35.5% [20-40), 49.0% [40-65), 15.4%
[65-80)
Highest level of education 5.4% Primary school, 55.1% High
school , 32.2% Higher education ,
6.5% Masters degree , 0.6% Doctors
degree
would start searching for on-street parking. In case of a preference for on-street parking, the ma jority of
respondents (60 respondents, 70.6 %) indicated that after 5 minutes searching without finding a vacant spot245
they would continue searching for an on-street parking spot, while 25 respondents (29.4 %) stated that they
would in that case go to the nearest off-street parking facility.
The maximum searching or waiting time for a vacant parking spot (until moving to a different parking
area) is 8.4 minutes (std. dev. 4.7 minutes) for drivers who prefer off-street parking and 12.9 minutes (std.
dev. 8.4 minutes) for drivers who prefer on-street parking.250
The responses from the choice situations in the third part of the survey are used to estimate a number
of choice models presented in the next section.
3. Discrete Choice Model Estimation
In this section, a total of four choice models are estimated based on the stated preference experiments.
We start with the multinomial logit model under the assumption of random utility maximization. To test the255
correlation in parking type the nested MNL model under RUM assumption is estimated with parking type
as nests. To test heterogeneity in parking location choice preferences, the mixed MNL model under RUM
assumption is estimated. And finally, to account for correlation between the 12 answers by each respondent
a Panel Effect Mixed Logit is estimated. As mentioned earlier, choice models were estimated using the
Biogeme software (Bierlaire,2006) and they were validated using the mLogit Package in R (Croissant et al.,260
2012). The model results are presented in this section, after which the next section discusses what these
results imply for parking location choice behaviour.
All model estimations are presented in Table 7except for the Nested Logit model that did not yield re-
sults. Starting from the MNL model estimation, initially, alternative specific constants and socio-demographic
characteristics were also included, however the effects hereof were not statistically significant. Hence, Table265
7shows the results of the final model. Here all coefficients have the expected sign, where price, walking
distance and travel time are negative and parking type (where 1 indicates off-street parking) and parking
availability upon arrival and after 8 minutes are positive.
The Nested Logit model structure was selected to be evaluated as it was suggested by Hunt & Teply
(1993) that parking decisions are correlated within parking type, and hence can be modelled using the270
Nested MNL model where on-street parking and off-street parking belong to different nests. The diagnostics
of the estimated NMNL model showed that the nest structure on parking type does not give better model
results, while at the same time it is indicated that the region of trust is too small. Therefor the NMNL
model is rejected for further analyses.
For the RUM-MMNL model, the goodness of fit indicators and coefficient estimates are presented (Table275
7). Heterogeneity in taste was tested for all coefficients, however only the model where the parking availability
probability upon arrival was normally distributed yielded an increase in ¯ρ2. Hence Table 7shows the results
of the final model, where this coefficient is normally distributed and all other coefficients are constant. All
coefficients have the expected sign, and the RUM-MMNL model performs better than the RUM-MNL model
according to the log-likelihood and ¯ρ2.280
The model with panel data is estimated with the Panel Effect to be used for all coefficients and found
to be significant. Table 7shows the goodness of fit indicators and coefficient estimates for the Panel Mixed
9
Logit model. As can be observed from the log-likelihood and ¯ρ2, the model gives a much better fit than the
RUM-MNL model without panel data.
Table 7: Estimated Discrete Choice Models
Coefficient/
Constant
MNL Mixed Logit Panel Effect Mixed Logit
Estimate Robust
Asympt.
std.
err.
t-stat Estimate Robust
Asympt.
std.
err.
t-stat Estimate Robust
Asympt.
std.
err.
t-stat
βC-0.735 0.032 -22.85 -0.91 0.093 -9.83 -1.824 0.118 -15.43
σβC- - - - - - 1.826 0.133 13.69
βW-0.001 6E-05 -8.78 -0.001 0 -6.74 -0.001 0.000 -7.96
σβW- - - - - - -0.002 0.000 -10.16
βT-0.011 0.005 -2.44 -0.014 0.006 -2.52 -0.025 0.009 -2.63
σβT- - - - - - 0.070 0.017 3.99
βP0.119 0.035 3.43 0.161 0.047 3.44 0.248 0.074 3.35
σβP- - - - - - 1.676 0.137 12.26
βP r00.569 0.083 6.84 0.734 0.132 5.57 1.493 0.189 7.91
σβP r0- - - 1.43 0.151 9.51 -1.959 0.248 -7.89
βP r81.180 0.071 16.54 2.38 0.721 3.31 2.767 0.198 13.98
σβP r8- - - - - - 3.267 0.249 13.14
L(0) -
3293.84
-
3293.84
-
3293.84
L(ˆ
β) -
2912.33
-
2908.91
-
2305.70
¯ρ20.114 0.115 0.300
Iterations 11 23 32
Halton
draws
- 3000 3000
4. Findings on Parking Choice Behaviour285
The survey results and model estimates reported in the previous sections yield a number of findings
on parking location choice behaviour, in particular with regard to the (relative) importance of the tested
factors. These findings are discussed here.
Firstly, recall that in the second part of the survey, where respondents were asked when they choose
a parking location in case of an unfamiliar environment, the majority of respondents stated they start290
searching for parking once they approach (13 %) or arrive at (50 %) their destination, instead of deciding
upon their parking location pre-trip. This in part explains the prevalence of traffic searching for parking,
and the relevance of reducing cruising traffic by means of parking guidance and reservation systems. This
first finding on strategic parking location choice behaviour highlights the importance of adequate parking–
routing policies, as well as the potential benefits of parking guidance and reservation systems for parking.295
It also supports the need to include such strategic pre-trip and on-route parking location choice decisions in
urban traffic simulation models.
Secondly, from the various RUM logit–based models that were tested to describe drivers’ parking de-
cisions, the best fitting model is the mixed model where we account for the panel effects in the stated
preference data. Compared to the other models, the mixed model for panel data differs with respect to300
heterogeneous preferences related to inter-drivers differences in the appreciation of all attributes examined.
Thirdly, the importance of the tested factors in determining drivers’ parking decision is shown through
the coefficient estimates. Note that the coefficient estimates cannot be directly compared between models
10
(particularly with other models presented in literature) due to the fact that the coefficients are normalised
for identification, while the normalisation depends on the model type as well as the utility specification.305
Hence the importance of the factors, as measured by the choice coefficients, should be interpreted relatively.
One way of doing so is by considering the contribution of each factor to the total utility, as presented in Table
8, where the contribution is computed as the product of its coefficient estimate and its mean attribute level.
The factor that contributes most to the utility is the parking cost, followed by the probability of finding a
vacant parking spot after 8 minutes. Apart from the parking availability factors, the ranking between the310
contribution of the other factors is roughly in line with what is found in other studies reported in literature
(as for example presented by Van der Waerden (2012); Hunt (1988)) as well as with what was reported in the
pilot studies. The only difference is found in the contribution of travel time to parking location and walking
distance to destination. For these two attributes the walking distance was found to contribute more to the
utility compared to what was found in other studies (e.g., Van der Waerden (2012); Polak et al. (1991));315
something justified, given the larger walking distance and the shorter access times examined in these studies.
Table 8: Contributions in RUM-MMNL model for panel data
Coefficient Estimate Mean
attribute
level
Absolute
contribution
Relative
contribution
βC-0.821 1.875 1.539 44.7 %
βW-0.001 400 0.400 11.6 %
βT-0.013 20 0.260 7.6 %
βP0.132 0.5 0.066 1.9 %
βPr00.652 0.4 0.261 7.6 %
βPr81.310 0.7 0.917 26.6 %
The contributions of each factor in the other estimated models are presented in Table 9. The differences
among the models are minor, while the rankings of the factors are the same across all models.
Table 9: Relative contributions in estimated models
Coefficient MNL MMNL Panel Effect MMNL
βC44.30% 45.40% 44.7 %
βW12.90% 10.60% 11.6 %
βT7.10% 7.40% 7.6 %
βP1.90% 2.10% 1.9 %
βPr07.30% 7.80% 7.6 %
βPr826.50% 26.60% 26.6 %
Fourthly, we observe that the ratio between the coefficient estimate for the attribute pertaining to the
probability to park after 8 minutes and that of the probability to park upon arrival is approximately 2. This320
observation can be interpreted in two ways. Given that the parking probability after 8 minutes was tested
on a higher range than the parking probability upon arrival, one interpretation is that the appreciation
of uncertainty in parking availability is non-linear (regardless of the search time component). This would
imply that the mean marginal benefit of an increase in parking probability within the range of 40-100 %
is twice as high as the mean marginal benefit of an increase in parking probability within the range of325
10-70 %. An alternative interpretation is that drivers accept spending time for searching or waiting for a
parking spot to become available, and thus predominantly base their decision on the parking probability
after this search time (where the marginal benefit of an increase in parking probability does not depend
on the parking probability). This way, drivers’ assign the same amount of utility to x% of probability for
a vacant parking spot after 8 minutes as they do to 2x% of probability for a vacant parking spot upon330
arrival. These two interpretations are evidently not mutually exclusive. However, the latter interpretation
is supported by respondents’ responses in the second part of the survey where it was indicated that the
11
maximum searching or waiting time they deemed acceptable (until moving to a different parking area) is
on average 9.4 minutes. Furthermore, recall that it was found that this searching or waiting time appears
related to the type of parking spot that is preferred, where drivers with preference for off-street parking335
indicate a lower maximum search time (mean 8.4 minutes, std. dev. 4.7 minutes) than drivers who prefer
on-street parking (mean 12.9 minutes, std. dev. 8.4 minutes). The relevance of this finding for parking
related policies is, that it indicates that individuals will typically visit a parking location and subsequently
cruise for parking in cases when there is a high probability to find a vacant parking spot within an acceptable
amount of time (which appears to be as high as 8-13 minutes), for example in cases with high turnover rate.340
This clearly yields cruising traffic. From a modelling perceptive, it indicates that assignment models may
need to account for cruising for parking based on the probabilities of finding a vacant parking spot.
Fifthly, note that the responses towards uncertainty in parking availability and search times are evaluated
between the range of 10 to 70 % parking probability upon arrival and of 40 to 100 % parking probability
after 8 minutes. We believe this setup allows these results to be useful for preliminary testing of the345
potential benefits of parking guidance and reservation systems, where drivers would be given a parking
location alternative with very high probability upon arrival (up to 100 %). Nevertheless, the facts that we
found statistically significant differences in the way drivers appreciate the parking probability upon arrival,
together with that our previous finding may suggest non-linearity in the appreciation of parking probabilities,
underpins the relevance for further research into the behavioural responses towards such parking-oriented350
ITS.
5. Concluding Remarks
Being able to predict the decisions of drivers regarding their parking location is essential when developing
parking–related policies, evaluating the potential of parking–related ITS and hence also in traffic assignment
modelling. As the discussions in the earlier sections of this paper show, parking location choice behaviour355
affects the amount of traffic and distribution of traffic flows over the road network. Therefore the validity of
assignment models (particularly within the context of parking–related simulation studies) relies on a valid
model of drivers’ parking location choice behaviour. Similarly, a valid understanding of drivers’ parking
location choice factors and their importance is needed when designing, evaluating and implementing parking
policies and parking-oriented intelligent transport systems.360
In this paper we present the results from a survey conducted among a representative panel of 397 respon-
dents regarding parking location choice behaviour within the context of making a trip for shopping purposes
and parking for a duration of 2-3 hours. The survey consisted of various questions on respondents’ parking
location choices and a stated preference experiment. The results were analysed by means of estimating a
number of random utility maximisation choice models, where the mixed multinomial logit model account-365
ing for panel data was found to fit best. The fact that the estimated parameters are in accordance with
the body of literature on parking location choice suggests that the choice models and behavioural findings
in this study can be generalised to other settings, although this has not been explicitly tested for and a
meta-analysis would be required for a conclusive answer in this regard.
Main findings from the survey results and model estimates show:370
Searching for parking is prominent, where the majority of drivers will search for a parking spot as they
approach or arrive at their destination, and are willing to spend up to 8-13 minutes on searching for
a vacant spot.
(Higher) availability of parking after 8 minutes of search time is more important than (lower) parking
availability upon arrival.375
Uncertain parking availabilities rank second (for availability after 8 minutes) and fourth (for availability
upon arrival) most important factors in determining parking location decision, where parking costs is
ranked first, and walking distance to destination is ranked third.
Drivers differ in their appreciation of finding a vacant parking spot upon arrival at their parking
location.380
12
Furthermore, the fact that the best model fit was obtained when we accounted for panel effects in the
stated preference data implies that heterogeneity in driver characteristics also play a role in parking location
choices; however, this heterogeneity is uncorrelated with socio-demographic or socio-economic characteristics
that were tested for.
A note can be made that in this study we described drivers’ parking location choice under the assumption385
of utility maximisation. One may wish to investigate the suitability of this decision rule by considering
alternative models based on, for example, prospect maximisation or regret minimisation.
The estimated choice models allow predicting drivers’ parking decisions in terms of parking costs, walking
distance to destination, travel time, parking type, and parking probabilities. Hence this model structure
with these explanatory variables can be readily implemented in a traffic assignment model, where parking390
probabilities at a parking location can be computed based on the parking spot capacity, the arrival pattern
of drivers and their (mean) parking duration. Note that the estimated parking location choice models do not
include driver characteristics, and hence are suitable for both disaggregated microscopic simulation models
as well as aggregated macroscopic assignment models.
In this study we analysed drivers’ choice behaviour among alternative parking locations, in particular395
under uncertain parking availability and search times. As parking probabilities are shown to be important
determining factors in drivers’ parking decisions, this shows that parking guidance and reservation systems
could be useful in reducing the amount of traffic searching for parking. This warrants further research into
drivers’ acceptance and willingness-to-pay towards such systems.
Acknowledgements400
The authors wish to thank the Smart Mobility Group at TNO for financing and facilitating the survey,
and to thank Dr. Costantinos Antoniou, Iraklis Stamos, and Dr. Martijn van Noort for their helpful
comments on the research presented in this paper.
References
Anderson, S. P., & de Palma, A. (2004). The economics of pricing parking. Journal of Urban Economics,55 , 1 – 20. URL:405
http://www.sciencedirect.com/science/article/pii/S0094119003000779. doi:10.1016/j.jue.2003.06.004.
Antony, J. (2003). 2 - fundamentals of design of experiments. In Design of Experiments for Engineers and Scientists (pp. 6 – 16).
Oxford: Butterworth-Heinemann. URL: http://www.sciencedirect.com/science/article/pii/B978075064709050003X.
doi:http://dx.doi.org/10.1016/B978-075064709- 0/50003-X.
Arnott, R., & Inci, E. (2006). An integrated model of downtown parking and traffic congestion. Journal of Urban Economics,410
60 , 418 – 442. URL: http://www.sciencedirect.com/science/article/pii/S0094119006000386. doi:10.1016/j.jue.2006.
04.004.
Axhausen, K., & Polak, J. (1991). Choice of parking: Stated preference approach. Transportation,18 , 59–81. URL: http:
//dx.doi.org/10.1007/BF00150559. doi:10.1007/BF00150559.
Axhausen, K., Polak, J., Boltze, M., & Puzicha, J. (1994). Effectiveness of the parking guidance system in frankfurt/main.415
Traffic Engineering and Control,35 , 304–309.
Barata, E., Cruz, L., & Ferreira, J.-P. (2011). Parking at the {UC}campus: Problems and solutions. Cities,28 , 406
– 413. URL: http://www.sciencedirect.com/science/article/pii/S0264275111000412. doi:http://dx.doi.org/10.1016/
j.cities.2011.04.001.
Benenson, I., Martens, K., & Birfir, S. (2008). Parkagent: An agent-based model of parking in the city. Computers, Environ-420
ment and Urban Systems,32 , 431 – 439. URL: http://www.sciencedirect.com/science/article/pii/S0198971508000689.
doi:10.1016/j.compenvurbsys.2008.09.011. GeoComputation: Modeling with spatial agents.
Bierlaire, M. (2006). Biogeme: a free package for the estimation of discrete choice models. In Swiss Transport Research
Conference TRANSP-OR-CONF-2006-048.
Bliemer, M. C., Rose, J. M., & Hensher, D. A. (2009). Efficient stated choice experiments for estimating nested logit models.425
Transportation Research Part B: Methodological,43 , 19 – 35. URL: http://www.sciencedirect.com/science/article/
pii/S0191261508000635. doi:http://dx.doi.org/10.1016/j.trb.2008.05.008.
Bonsall, P., & Palmer, I. (2004). Modelling drivers car parking behaviour using data from a travel choice simulator. Transporta-
tion Research Part C: Emerging Technologies,12 , 321 – 347. URL: http://www.sciencedirect.com/science/article/pii/
S0968090X04000245. doi:10.1016/j.trc.2004.07.013.430
Brandley, M., Kroes, E., & Hinloopen, E. (1993). A joint model of mode/parking type choice with supply-constrained appli-
cation. In 21st Annual Summer PTRC Meeting on European Transport, Highways and Planning (pp. 61–73).
Calthrop, E., & Proost, S. (2006). Regulating on-street parking. Regional Science and Urban Economics,36 , 29 – 48. URL:
http://www.sciencedirect.com/science/article/pii/S0166046205000475. doi:10.1016/j.regsciurbeco.2005.04.002.
13
Caussade, S., de Dios Ortzar, J., Rizzi, L. I., & Hensher, D. A. (2005). Assessing the influence of design dimensions on435
stated choice experiment estimates. Transportation Research Part B: Methodological,39 , 621 – 640. URL: http://www.
sciencedirect.com/science/article/pii/S0191261504001055. doi:http://dx.doi.org/10.1016/j.trb.2004.07.006.
CBS (2009). Statistiek Netherlands. URL: http://www.cbs.nl.
ChoiceMetrics (2012). Ngene 1.1.1 User Manual & Reference Guide.
Coppola, P. (2004). A joint model of mode/parking choice with elastic parking demand. In M. Patriksson, & M. Labbe (Eds.),440
Transportation Planning (pp. 85–104). Springer US volume 64 of Applied Optimization. URL: http://dx.doi.org/10.
1007/0-306- 48220-7_6. doi:10.1007/0- 306-48220-7_6.
Croissant, Y. et al. (2012). Estimation of multinomial logit models in r: The mlogit packages. R package version 0.2-2. URL:
http://cran. r-project. org/web/packages/mlogit/vignettes/mlogit. pdf , .
Dell’Orco, M., Ottomanelli, M., & Sassanelli, D. (2003). Modelling uncertainty in parking choice behaviour. In 82nd Annual445
Meeting of the Transportation Research Board (pp. 1–20).
Gallo, M., D’Acierno, L., & Montella, B. (2011). A multilayer model to simulate cruising for parking in urban areas. Transport
Policy,18 , 735 – 744. URL: http://www.sciencedirect.com/science/article/pii/S0967070X11000217. doi:10.1016/j.
tranpol.2011.01.009.
Gillen, D. (1978). Parking policy, parking location decisions and the distribution of congestion. Transportation,7, 69–85.450
URL: http://dx.doi.org/10.1007/BF00148372. doi:10.1007/BF00148372.
Guo, L., Huang, S., Zhuang, J., & Sadek, A. (2013). Modeling parking behavior under uncertainty: A static game theoretic
versus a sequential neo-additive capacity modeling approach. Networks and Spatial Economics,13 , 327–350. URL: http:
//dx.doi.org/10.1007/s11067-012- 9183-1. doi:10.1007/s11067- 012-9183-1.
Hensher, D. A. (2006). How do respondents process stated choice experiments? attribute consideration under varying informa-455
tion load. Journal of Applied Econometrics,21 , 861–878. URL: http://dx.doi.org/10.1002/jae.877. doi:10.1002/jae.877.
Hensher, D. A., & King, J. (2001). Parking demand and responsiveness to supply, pricing and location in the sydney central
business district. Transportation Research Part A: Policy and Practice,35 , 177 – 196. URL: http://www.sciencedirect.
com/science/article/pii/S0965856499000543. doi:http://dx.doi.org/10.1016/S0965-8564(99)00054- 3.
Hess, D. (2001). Effect of Free Parking on Commuter Mode Choice: Evidence from Travel Diary Data. Transportation460
Research Record: Journal of the Transportation Research Board,1753 , 35–42. URL: http://dx.doi.org/10.3141/1753- 05.
doi:10.3141/1753-05.
Hess, S., & Polak, J. W. (2004). Mixed logit estimation of parking type choice. In Proc. of the 83rd Annual Meeting of the
Transportation Research Board (Washington D.C., USA, January, 2004).
Hunt, J., & Teply, S. (1993). A nested logit model of parking location choice. Transportation Research Part B: Method-465
ological,27 , 253 – 265. URL: http://www.sciencedirect.com/science/article/pii/0191261593900359. doi:10.1016/
0191-2615(93)90035- 9.
Hunt, J. D. (1988). Parking location choice: insights and representations based on observed behaviour and the hierarchical logit
modelling formulation. In Institute of Transportation Engineers (ITE), Annual Meeting, 58th, 1988, Vancouver, Canada.
Ibeas, A., dellOlio, L., Bordagaray, M., & de D. Ortzar, J. (2014). Modelling parking choices considering user heterogeneity.470
Transportation Research Part A: Policy and Practice,70 , 41 – 49. URL: http://www.sciencedirect.com/science/article/
pii/S0965856414002341. doi:http://dx.doi.org/10.1016/j.tra.2014.10.001.
Kanafani, A. (1983). Transportation demand analysis. Number τ. 1 in McGraw-Hill series in transportation. McGraw-Hill.
URL: http://books.google.gr/books?id=KS5PAAAAMAAJ.
Kaplan, S., & Bekhor, S. (2011). Exploring en-route parking type and parking-search route choice. In Proceedings of the 2nd475
International Choice Modelling Conference.
Kittelson & Associates (2003). Transit Capacity and Quality of Service Manual volume 100. Transportation Research Board.
Krejcie, R. V., & Morgan, D. W. (1970). Determining sample size for research activities. Educational and Psycholpgocal
Measurement, .
Lam, W. H., Li, Z.-C., Huang, H.-J., & Wong, S. (2006). Modeling time-dependent travel choice problems in road networks480
with multiple user classes and multiple parking facilities. Transportation Research Part B: Methodological,40 , 368 – 395.
URL: http://www.sciencedirect.com/science/article/pii/S0191261505000718. doi:10.1016/j.trb.2005.05.003.
Lambe, T. A. (1996). Driver choice of parking in the city. Socio-Economic Planning Sciences,30 , 207 – 219. URL: http://www.
sciencedirect.com/science/article/pii/0038012196000080. doi:http://dx.doi.org/10.1016/0038-0121(96)00008- 0.
Leurent, F., & Boujnah, H. (2014). A user equilibrium, traffic assignment model of network route and parking lot choice,485
with search circuits and cruising flows. Transportation Research Part C: Emerging Technologies,47, Part 1 , 28 – 46.
URL: http://www.sciencedirect.com/science/article/pii/S0968090X14002162. doi:http://dx.doi.org/10.1016/j.trc.
2014.07.014. Special Issue: Towards Efficient and Reliable Transportation Systems.
Levy, N., Render, M., & Benenson, I. (2015). Spatially explicit modeling of parking search as a tool for urban parking
facilities and policy assessment. Transport Policy,39 , 9 – 20. URL: http://www.sciencedirect.com/science/article/490
pii/S0967070X15000165. doi:http://dx.doi.org/10.1016/j.tranpol.2015.01.004.
Liu, W., Yang, H., & Yin, Y. (2014a). Expirable parking reservations for managing morning commute with parking space
constraints. Transportation Research Part C: Emerging Technologies,44 , 185 – 201. URL: http://www.sciencedirect.
com/science/article/pii/S0968090X14000916. doi:http://dx.doi.org/10.1016/j.trc.2014.04.002.
Liu, W., Yang, H., Yin, Y., & Zhang, F. (2014b). A novel permit scheme for managing parking competition and bottleneck495
congestion. Transportation Research Part C: Emerging Technologies,44 , 265 – 281. URL: http://www.sciencedirect.com/
science/article/pii/S0968090X14001028. doi:http://dx.doi.org/10.1016/j.trc.2014.04.005.
Louviere, J. J., Hensher, D. A., & Swait, J. D. (2000). Stated choice methods: analysis and applications. Cambridge University
Press.
14
McCoy, P., Ramanujam, M., Moussavi, M., & Ballard, J. (1990). Safety comparison of types of parking on urban streets in500
nebraska. Transportation Research Board, .
Mei, Z., Xiang, Y., Chen, J., & Wang, W. (2010). Optimizing model of curb parking pricing based on parking choice
behavior. Journal of Transportation Systems Engineering and Information Technology,10 , 99 – 104. URL: http://www.
sciencedirect.com/science/article/pii/S1570667209600271. doi:10.1016/S1570-6672(09)60027- 1.
Millard-Ball, A., Weinberger, R. R., & Hampshire, R. C. (2014). Is the curb 80impacts of san franciscos parking pricing505
experiment. Transportation Research Part A: Policy and Practice,63 , 76 – 92. URL: http://www.sciencedirect.com/
science/article/pii/S0965856414000470. doi:http://dx.doi.org/10.1016/j.tra.2014.02.016.
Ottosson, D. B., Chen, C., Wang, T., & Lin, H. (2013). The sensitivity of on-street parking demand in response to price
changes: A case study in seattle, wa. Transport Policy,25 , 222 – 232. URL: http://www.sciencedirect.com/science/
article/pii/S0967070X12001886. doi:10.1016/j.tranpol.2012.11.013.510
Polak, J., Axhausen, K., & Errington, T. (1991). The application of clamp to the analysis of parking policy in birmingham
city centre. In PTRC Summer Annual Meeting, 18th, 1990, University of Sussex, United Kingdom.
Rose, J. M., & Bliemer, M. C. (2008). Stated preference experimental design strategies. Handbook of Transport Modelling,
Elsevier, Oxford, (pp. 151–180).
Rose, J. M., Bliemer, M. C., Hensher, D. A., & Collins, A. T. (2008). Designing efficient stated choice experiments in the515
presence of reference alternatives. Transportation Research Part B: Methodological,42 , 395–406.
Rose, J. M., & Bliemer, M. C. J. (2009). Constructing efficient stated choice experimental designs. Trans-
port Reviews,29 , 587–617. URL: http://dx.doi.org/10.1080/01441640902827623. doi:10.1080/01441640902827623.
arXiv:http://dx.doi.org/10.1080/01441640902827623.
Ruisong, Y., Meiping, Y., & Xiaoguang, Y. (2009). Study on driver’s parking location choice behavior considering drivers’520
information acquisition. In Intelligent Computation Technology and Automation, 2009. ICICTA’09. Second International
Conference on (pp. 764–770). IEEE volume 3.
Shiftan, Y., & Burd-Eden, R. (2001). Modeling Response to Parking Policy. Transportation Research Record: Journal of the
Transportation Research Board,1765 , 27–34. URL: http://dx.doi.org/10.3141/1765-05. doi:10.3141/1765- 05.
Shoup, D. C. (2006). Cruising for parking. Transport Policy,13 , 479 – 486. URL: http://www.sciencedirect.com/science/525
article/pii/S0967070X06000448. doi:10.1016/j.tranpol.2006.05.005. Parking.
Thompson, R. G., & Bonsall, P. (1997). Drivers’ response to parking guidance and information systems. Transport Reviews,
17 , 89–104. URL: http://www.tandfonline.com/doi/abs/10.1080/01441649708716974. doi:10.1080/01441649708716974.
arXiv:http://www.tandfonline.com/doi/pdf/10.1080/01441649708716974.
Thompson, R. G., & Richardson, A. J. (1998). A parking search model. Transportation Research Part A: Policy and530
Practice,32 , 159 – 170. URL: http://www.sciencedirect.com/science/article/pii/S0965856497000050. doi:10.1016/
S0965-8564(97)00005- 0.
Van Der Goot, D. (1982). A model to describe the choice of parking places. Transportation Research Part A: General,16 , 109
– 115. URL: http://www.sciencedirect.com/science/article/pii/0191260782900036. doi:http://dx.doi.org/10.1016/
0191-2607(82)90003- 6.535
Van Ommeren, J. N., Wentink, D., & Rietveld, P. (2012). Empirical evidence on cruising for parking. Transporta-
tion Research Part A: Policy and Practice,46 , 123 – 130. URL: http://www.sciencedirect.com/science/article/pii/
S0965856411001443. doi:10.1016/j.tra.2011.09.011.
Verhoef, E., Nijkamp, P., & Rietveld, P. (1995). The economics of regulatory parking policies: The (im)possibilities of parking
policies in traffic regulation. Transportation Research Part A: Policy and Practice,29 , 141 – 156. URL: http://www.540
sciencedirect.com/science/article/pii/0965856494E0014Z. doi:http://dx.doi.org/10.1016/0965-8564(94)E0014- Z.
Van der Waerden, P. (2012). Pamela, a parking analysis model for predicting effects in local areas. PhD dissertation Technische
Universiteit Eindhoven Department of Architecture, Building and Planning.
Van der Waerden, P., & Oppwal, H. (1995). Modeling the combined choice of parking lot and shopping destination. In 7th
World Conference on Transport Research July (pp. 16–21).545
Young, W., & Taylor, M. (1991). A parking model hierarchy. Transportation,18 , 37–58. URL: http://dx.doi.org/10.1007/
BF00150558. doi:10.1007/BF00150558.
15
... Some researchers have found other factors influencing parking decisions. It has been shown that a number of parking factors, such as parking time restrictions, walking distances to destinations, and the time it takes to find parking spaces substantially impact on parking decisions (Chaniotakis and Pel, 2015;Golias et al., 2002;Tsamboulas, 2001). Other studies have also found that socio-demographic factors such as monthly income, are significantly correlated with parking decisions (Brown, 1986;Sasaki, 1990). ...
... This indicates that motorcyclists may prefer these alternatives when egress time is a concern due to quicker access to final destinations compared to where they would park their motorcycles. These results are consistent with previous findings showing that walking distance from parking lots to final destinations sig- nificantly influences the decision of parking location choice (Chaniotakis and Pel, 2015;Golias et al., 2002;Tsamboulas, 2001) and public transport use (Zahabi et al., 2012). Car users also show a positive correlation with a shift to all alternative travel modes except a shift to taxis (i.e., without significant correlation). ...
Article
Full-text available
Although Jakarta has invested in various mass transport systems, these efforts have not successfully reduced private vehicle use. Due to this, this study aims to analyze the impact of implementing TransJakarta bus rapid transit corridor-based high parking tariffs on travel mode choice, including road- and rail-based public transport, ride-hailing, taxi, car, and motorcycle. Involving 478 private vehicle users and implementing a nested logit model, some variables, including respondents’ income, travel time, egress time, parking costs, parking distance, travel cost, and parking surcharge, are considered to understand to what extent these variables influence the use of proposed travel mode in the future. The nested logit model shows that not all variables significantly influence travel mode use, specifically related to rail-based public transport choice among motorcyclists. Meanwhile,parking distance insignificantly influences the choice of all travel modes except cars among car users. The results also indicate that increasing parking tariffs insignificantly influences the likelihood of both motorcyclists and car users shifting to public transport. Motorcyclists and car users tend to continue using motorcycles but change parking locations with higher tariffs. Additionally, some shifts towards ride-hailing services and TransJakarta Bus Rapid Transit are found, meaning that there is potential for these alternatives to play a significant role in reducing private vehicle use. Based on the model results, additional push-based policies, such as the odd-even license plate rule, are necessary to effectively support the transition from private vehicle use to public transport. Implementing these policies is expected to significantly contribute to reducing traffic congestion and promoting a sustainable and resilient urban environment.
... The study of drivers' behavior regarding uncertainties about search times for finding vacant spot has been carried out in Chaniotakis & Pel (2015), where stated preference experiments are applied to several discrete choice models. One of the findings of this research is that most of the drivers searching for a parking lot who make a trip for shopping purposes start the searching process when they approach or reach their destinations. ...
Preprint
Full-text available
This paper presents an optimization procedure to choose a parking facility according to different criteria: total travel time including transfers, parking fee and a factor depending on the risk of not having an available spot in the parking facility at the arrival time. An integer programming formulation is proposed to determine an optimal strategy of minimum cost considering the available information, different scenarios, and each user profile. To evaluate the performance, a computational experience has been carried out on Seville (Spain), where a historical city center restricts the traffic of private vehicles and encourages the use of parking facilities.
... In the first, LLM instances acted as a 72 survey participant, choosing between parking options based on attributes and 73 sociodemographic profiles, which we refer to as Personas. Previous research indicates 74 that drivers aim to minimize parking fees, walking distances, and search times 75 (Chaniotakis and Pel 2015 In experiment a1, we focused on parking prices. We set the price of parking spot A 277 at $5, while varying the price of spot 2 from $0 to $10 in increments of $1. ...
Article
Full-text available
Transport quality has long been recognized as an important factor in influencing travelers’ behavior, and transport terminal quality undoubtedly plays no small part. Indeed, transit promotion policies explicitly based on qualitative factors and high-standard architecture are increasingly being adopted in designing new bus terminals and stops. High architectural standards can be found in several bus terminals worldwide. Nevertheless, the literature in transport sector has yet to explore the impact of bus terminal “hedonic quality” on users’ behavior or use their willingness to pay (WTP) in cost–benefit analyses or other transport policy applications. Hedonic value is here intended as the aggregate of all elements related to travelers’ pleasure in spending time in a terminal where architectural beauty and the passenger services offered are arguably the most visible and representative attributes. Within this context, we propose a quantitative analysis of the perceived hedonic value of bus nodes in terms of users’ WTP (pure preference) for a high-quality bus terminal. To this end a discrete choice experiment based on a visual immersive experience was performed, and data were collected from 324 residents of Milan (Italy) traveling for tourism. Different binomial Mixed Logit models with panel data and random coefficients were estimated for the purpose. Monetary valuation of pure preference for using a high-quality bus terminal was estimated with a mean of about 25 % of the average trip cost observed for the case study considered. This pure preference means that the Italian tourist is willing to spend up to €4.35/trip more for a high-quality bus waiting space or travel up to 28.2 min/trip more, instead of using a traditional bus terminal for the same trip. Category-specific analyses show that females > 30 years old have a greater pure preference for a high-quality bus terminal (up to + 220 %) against males and young users. The employed also have a higher (+42 %) pure preference for the beauty of a bus terminal against others. The results of this paper should be compared with those from other case studies as they have a potential impact for transportation planning applications, underling the importance of also incorporating “hedonic quality” as explicit design variables for new/revamped transportation hub. At the same time, new challenges are posed for modeling user behavior and determining quality-related indicators and measures.
Article
Full-text available
This paper proposes two types of parking choice models, a static game theoretic model and a dynamic neo-additive capacity model, to capture the competition among drivers for limited desirable parking spaces. The static game assumes that drives make decisions simultaneously and with perfect knowledge about the characteristics of the parking system and the strategies of their fellow drivers in the system; the model thus captures only the rational aspect of parking choice behavior and pays no attention to modeling individual drivers’ psychological characteristics. The dynamic model, on the other hand, considers individual drivers’ psychological characteristics under uncertainty (i.e. optimistic and pessimistic attitudes) and thus captures the impacts of the irrational side of parking behavior in addition to the rational aspect. Following the formulation of the two models, they are both used to predict parking behavior as observed on a set of parking lots on the University at Buffalo north campus. Specifically for the dynamic model, the model is first calibrated based on real data collected from video recorded observations for a pair of parking lots, and then used to predict behavior on another pair. Validation results show higher predictive accuracy for the dynamic neo-additive capacity model compared to the static game theoretic model. This in turn suggests that the psychological characteristics of drivers play an important role in the parking lot choice decision process, and points to the potential for parking information systems to eliminate the unnecessary additional traffic generated by the parking search process.
Article
In order to improve the use efficiency of curb parking, a reasonable curb parking pricing is evaluated by considering individual parking choice behavior. The parking choice behavior is analyzed from micro-aspects, and the choice behavior utility function is established combining trip time, search time, waiting time, access time and parking fee. By the utility function, a probit-based parking choice behavior model is constructed. On the basis of these, the curb parking pricing model is deduced by considering the constrained conditions, and an incremental assignment algorithm of the model is also designed. Finally, the model is applied to the parking planning of Tongling city. It is pointed out that the average parking time of curb parking decreases 34%, and the average turnover rate increases 67% under the computed parking price system. The results show that the model can optimize the utilization of static traffic facilities.
Chapter
Publisher Summary Design of experiments refers to the process of planning, designing, and analyzing the experiment so that valid and objective conclusions can be drawn effectively and efficiently. To draw statistically sound conclusions from the experiment, it is necessary to integrate simple and powerful statistical methods into the experimental design methodology. The success of any industrially designed experiment depends on sound planning, appropriate choice of design, statistical analysis of data, and teamwork skills. In the context of design of experiment in manufacturing, one may come across two types of process variables or factors: qualitative and quantitative. For quantitative factors, one must decide on the range of settings and how they are to be measured and controlled during the experiment. The chapter illustrates three principles of experimental design, randomization, replication, and blocking, that can be utilized in industrial experiments to improve the efficiency of experimentation. These principles of experimental design are applied to reduce or even remove experimental bias. It is important to note that large experimental bias could result in wrong optimal settings, or in some cases it could mask the effect of the really significant factors.
Article
The engineering view of a measurable, supply-independent, demand for parking that can be expressed by “minimum parking codes” has been generally rejected during the last two decades and is gradually being replaced by “maximum provision” codes, limited parking development, and demand pricing. To assess new planning practices one has to estimate the drivers' reaction to proposed spatial–temporal parking limitations. The paper applies a high-resolution spatially explicit agent-based model termed “PARKAGENT” as a tool for this assessment. The model is used for evaluation of parking demand in the Diamond Exchange area in Ramat Gan, a city in the Tel Aviv metropolitan area, for estimating the effectiveness of planned parking facilities for different development scenarios in the area and assessing electronic signage system that directs drivers to vacant parking lots. The results strongly indicate the advantages of agent-based modeling over the current dominant engineering approach and show the potential benefits of using an intelligent parking guidance system.
Article
We examine car driver’s behaviour when choosing a parking place; the alternatives available are free on-street parking, paid on-street parking and parking in an underground multi-storey car park. A mixed logit model, allowing for correlation between random taste parameters and estimated using stated choice data, is used to infer values of time, both when looking for a parking space and for accessing the final destination. Apart from the cost of parking, we found that vehicle age was a key variable when choosing where to park in Spain. We also found that the perception of the parking charge was fairly heterogeneous, depending both on the drivers’ income levels and whether or not they were local residents. Our results can be generalised for personalised policy making related with parking demand management.
Article
The paper provides a novel network model of parking and route choice. Parking supply is represented by parking type, management strategy including the fare, capacity and occupancy rate of parking lot, and network location, in relation to access routes along the roadway network. Trip demand is segmented according to origin–destination pair, the disposal of private parking facilities and the individual preferences for parking quality of service. Each traveller is assumed to make a two stage choice of, first, network route on the basis of the expected cost of route and parking and, second, local diversion on the basis of a discrete choice model. Search circuits are explicitly considered on the basis of the success probability to get a slot at a given lot and of the transition probabilities between lots in case of failure. The basic endogenous model variables are the route flows, the lot success probabilities and the transition probabilities between lots. These give rise to the cost of a travel route up to a target lot and to the expected cost of search and park from that lot to the destination. Traffic equilibrium is defined in a static setting. It is characterized by a mixed problem of variational inequality and fixed point. Equilibrium is shown to exist under mild conditions and a Method of Successive Averages is put forward to solve for it. Lastly, a planning instance is given to illustrate the effects of insufficient parking capacity on travel costs and network flows.
Article
Morning commuters may have to depart from home earlier to secure a parking space when parking supply in the city center is insufficient. Recent studies show that parking reservations can reduce highway congestion and deadweight loss of parking competition simultaneously. This study develops a novel tradable parking permit scheme to realize or implement parking reservations when commuters are either homogeneous or heterogeneous in their values of time. It is found that an expirable parking permit scheme with an infinite number of steps, i.e., the ideal-scheme, is superior to a time-varying pricing scheme in the sense that designing a permit scheme does not require commuters’ value of time information and the performance of the scheme is robust to the variation of commuters’ value of time. Although it is impractical to implement the ideal-scheme with an infinite number of steps, the efficiency loss of a permit scheme with finite steps can be bounded in both cases of homogeneous and heterogeneous commuters. Moreover, considering the permit scheme may lead to an undesirable benefit distribution among commuters, we propose an equal cost-reduction distribution of parking permits where auto commuters with higher value of time will receive fewer permits.
Article
When total parking supply in an urban downtown area is insufficient, morning commuters would choose their departure times not only to trade off bottleneck congestion and schedule delays, but also to secure a parking space. Recent studies found that an appropriate combination of reserved and unreserved parking spaces can spread the departures of those morning commuters and hence reduce their total travel cost. To further mitigate both traffic congestion and social cost from competition for parking, this study considers a parking reservation scheme with expiration times, where commuters with a parking reservation have to arrive at parking spaces for the reservation before a predetermined expiration time. We first show that if all parking reservations have the same expiration time, it is socially preferable to set the reservations to be non-expirable, i.e., without expiration time. However, if differentiated expiration times are properly designed, the total travel cost can be further reduced as compared with the reservation scheme without expiration time, since the peak will be further smoothed out. We explore socially desirable equilibrium flow patterns under the parking reservation scheme with differentiated expiration times. Finally, efficiencies of the reservation schemes are examined.
Article
The city of San Francisco is undertaking a large-scale controlled parking pricing experiment. San Francisco has adopted a performance goal of 60–80% occupancy for its metered parking. The goal represents an heuristic performance measure intended to reduce double parking and cruising for parking, and improve the driver experience; it follows a wave of academic and policy literature that calls for adjusting on-street parking prices to achieve similar occupancy targets. In this paper, we evaluate the relationship between occupancy rules and metrics of direct policy interest, such as the probability of finding a parking space and the amount of cruising. We show how cruising and arrival rates can be simulated or estimated from hourly occupancy data. Further, we evaluate the impacts of the first two years of the San Francisco program, and conclude that rate changes have helped achieve the City’s occupancy goal and reduced cruising by 50%.