Abstract:
The research presented in this thesis was motivated by two factors. Firstly, little route choice research
has been undertaken in South Africa, especially in urban areas. This has resulted in significant gaps in
our understanding of commuter route choice preferences and associated willingness-to-pay measures
such as the value of travel time. Secondly, there are recognised limitations in the experimental
methodologies used for route preference data collection, i.e., field data collected using revealed
preference (RP) methods, and experimental data collected using stated preference (SP) methods. RP
methods have high external validity, but the analyst has limited experimental control. SP methods
have a high degree of analyst control of the experiment parameters, but the hypothetical nature of the
route alternatives provides lower levels of external validity. The research presented in this thesis
therefore had four objectives. Firstly, to provide a review of historical research and studies into mode
choice and route choice modeling in South Africa and highlight any gaps in our understanding of
commuter route choice preferences and the value of travel time. The findings confirmed that no route
choice research has been undertaken in South African urban areas for the last two decades, and large
gaps exist in our understanding of motorists’ route choice preferences these areas. The findings
confirmed the urgent need to undertake route choice research in South African urban settings,
especially in the light of the governments user pays policy for urban road and public transport
provision.
The second objective was to develop and demonstrate the proof-of-concept for an innovative,
smartphone-based application with the acronym RAPP-UP (Route Choice Application – University of
Pretoria), for collecting motorist route preference data in dense, congested urban road networks based
on real-time traffic conditions at the time of the trip. The author of this thesis designed and prepared
the specification for RAPP-UP, and an independent contractor was appointed to code the application
and make it available on the Google Play Store® for survey participants to download. RAPP-UP was
designed to achieve a better balance between external validity and analyst control. The third objective
was to use RAPP-UP to collect route preference data from a sample of commuters in Gauteng
Province, South Africa. The fourth objective was to estimate various types of discrete choice models to quantify different forms of route preference utility and estimate the associated willingness-to-pay
measures such as the commuter value of travel time.
RAPP-UP was designed for application in a self-validating survey context that included stated
preference (SP) and revealed preference (RP) components. A degree of analyst control was introduced
by allowing the analyst to factor the observed attribute levels before presentation to users in a predetermined
manner based on an unlabelled fractional factorial design. RAPP-UP’s innovation was its
ability to maximise external validity by generating two realistic alternative routes based on real-time
road network travel data between a user specified origin and destination, thereby anchoring the
experiment in a realistic and familiar setting. This innovation was enhanced by showing the route
alternatives on a detailed road map background to provide orientation for the trip origin and
destination locations, the routes themselves (highlighted on the road background), as well as the
utility attribute levels for each route in a choice set format. After trading-off the attribute levels for
each route, users were asked to choose their preferred route (SP component) and were then required to
drive their chosen route (self-validating RP component). The GPS function in the smartphone was
used to track the user to determine route adherence. An economic experiment was introduced by
deducting the toll cost of a chosen tolled route from a user survey account that was allocated to each
user at the commencement of the survey. The final survey account balance was paid to each user at
the end of the survey. As each trip is one observation, the use of RAPP-UP was required over several
days to obtain multiple observations from each user. RAPP-UP was designed to accommodate a
detailed form of utility expression that contained a disaggregated form of travel time that specified the
proportions of actual travel time (in minutes) in free-flow, slowed-down and stop-start travel
conditions. The trip petrol cost (in Rands), toll cost (in Rands) and the probability of on-time arrival at
the destination (in percent) were also included in the utility expression.
To illustrate proof-of-concept, a small sample of car commuters in the Gauteng Province of South
Africa was recruited to participate in a route choice survey using RAPP-UP. The road network in the
urban areas of Gauteng Province is dense and congested in the weekday peak periods, and the
motorways are tolled. The route preference data of the sample of commuters provided the basis for the
estimation of various forms of discrete route choice models. The models confirmed that the attribute
coefficients for each category of travel time were significant, thereby corroborating international
evidence. The congestion multipliers, i.e., the ratios of the travel time attribute coefficients for each
trip time category, were within the ranges determined in international studies. The petrol cost, toll cost
and probability of on-time arrival attribute coefficients were also significant. A toll road quality bonus
representing the unobserved factors of utility was introduced as a dummy utility attribute for routes
with tolled sections. The attribute coefficient had a negative sign, revealing that the survey
participants associated a disutility for routes with tolled sections for the unobserved factors of utility. All the objectives of the research were achieved. The research not only added to the body of literature
on the topic of route choice behaviour in urban areas, but also provided insights into the practicalities
of route choice data collection and model estimation.