Problem definition
Implementations of Utility Theory (UT) are commonly considered as the most useful concept
for the description and modelling of human choice behaviour and the prediction of its
outcomes. However, its descriptive-behavioural validity has been under discussion since the
1950s. Many experiments and surveys in several behavioural sciences have demonstrated
violations of its principles. Drawing on these findings, Prospect Theory (PT) was proposed
some decades ago as an alternative behavioural-economic model of choice behaviour. Though
researchers in mainstream economics and transport sciences are well aware that many
individuals and organizations violate UT’s principles, PT and alternative behavioural concepts
are only incidentally considered. Improvements to the structure and mathematical
formulations, particularly of stochastic elements of discrete choice models, followed by
calibration to empirical findings, are the dominant approach to coping with ‘inconsistent’
subjects.
The purpose of the current research is to examine whether the assumptions of PT, superadded
with insights from other social sciences, might be synthesized into a theory that offers a better
explanation of observed outcomes of human choice behaviour than UT. If such a theory can
be found, a further ambition is to investigate whether, after parameter estimation, it can be put
into operation for travellers’ choice prediction. As demonstrated hereafter both questions can
be answered positively. The present research that ended in this result evolved into a broad,
supra-disciplinary examination of human choice behaviour in general, including a large indepth
investigation of the real-life travel behaviour. It led to several methodological
challenges which are addressed in the introductory sections of Chapter 3, 4, 6 and 7. This
summary continues with an overview of the research activities as a frame of reference for the
findings that are listed in the following sections. It ends with the most prominent conclusions
and recommendations.
Overview of the research activities
In the first stage of this research (Chapter 2) the basic principles of human choice behaviour
are considered. Theories and empirical findings about human choice behaviour from
philosophy, behavioural economics, cognitive psychology and several other behavioural
disciplines are traced through the scientific literature. The analysis comprises a systemstheoretical
examination of the development of the scientific paradigm of choice-and-behaviour from about 1500 to date (Annex A). It results in the proposal of a generic Meta
Theory of Choice Behaviour. This offers a coordinating framework to facilitate comparison
and synthesis of concrete assumptions about choice behaviour.
The second stage of this research (Chapter 3…5) evaluates the assumptions of choice
behaviour in concrete contexts. First, the assumptions about human choice behaviour as
employed in microeconomics are fitted into the framework of the Meta Theory (Chapter 3).
These assumptions are compared with findings about choice behaviour reported from the
behavioural sciences (Chapter 4). Many findings are re-analysed, in some places quite
extensively (e.g. Annex B). An extension of PT with assumptions about the valuation of
attributes and the use of different decision rules yields a functional-descriptive concept of
choice behaviour that is able to describe most of the reviewed empirical findings to a larger
extent than the UT paradigm. These premises are further extended with a few assumptions
that draw on the cognitive consistency principle of Social Psychology and offer a weak
substitute for UT’s context-independent preference order. This enables predictions to be made
that apply to different contexts. The resulting ‘Extended Prospect Theory’ (EPT) has the
character of a generic choice paradigm rather than of a domain-specific or context-specific
model (Chapter 5).
The third stage of this research (Chapter 6) aims to assess the descriptive ability for travelrelated
choice behaviour of the five assumptions in EPT that distinguish it most from the
corresponding assumptions of UT. Application of discrete choice models that cover
assumptions that are at odds with UT appear uncommon in travel behaviour research. Most
widespread are several detached try-outs of PT. To examine the descriptive ability of the five
assumptions the travel behaviour literature is reviewed for information about observed actual
choices (Section 6.1). From 85 studies enough information on the choice context, research
design and observed choices is retrieved. These studies cover the whole range of domains and
contexts of travel behaviour, from strategic decision making to operational choice behaviour,
in experimental and real-life contexts, including choices from alternatives with certain,
probabilistic and uncertain outcomes (Section 6.2…6.4). After re-examination all but three
studies show a better descriptive performance for one or several assumptions in EPT
compared to the corresponding assumptions in UT (Section 6.5).
The fourth phase of this research (Chapter 7) investigates whether further improvement of the
descriptive and particularly the predictive performance of choice models can be attained when
the relevant assumptions of EPT are considered in connection with each other. The
development of car use by commuters in Singapore who, over a number of years have been
confronted with changes in road-pricing fares, offers sufficient information for such an
evaluation (Section 7.3, Annexes C, D and G). A discrete choice model is developed that is
suited to predicting the responses of car owners to the road-pricing measures according to the
UT and EPT paradigms (Section 7.4). The individuals’ value-of-travel-time-savings for both
implementations of the model are calibrated to the mode choices observed before road pricing
was introduced (Section 7.5). Next, the frequencies of different responses to the 1975
introduction of road pricing are predicted (Section 7.6, Annex F). Other predictions follow for
the responses to the sudden 1976 fare increase (Section 7.7) and to the long-term
developments in the travel context and socio-economic circumstances (Section 7.8). All
predictions are compared with the observed frequencies. Both for short-term and long-term
tactical travel choices the model that agreed with the EPT paradigm shows a better predictive
ability than the one that adhered to UT (Section 7.9).