Extensive research is available on travel choice behaviour which occurs during evacuations in case of natural disasters. Due to the disadvantages of existing data collection techniques, more research is needed to better understand evacuation choice behaviour. The main objective of this thesis is twofold: (1) to develop, apply and assess a new experimental set-up to study evacuation choice behaviour and (2) to quantify the effect of herding on evacuation choice behaviour. The developed experimental set-up consists of the serious game Everscape and a questionnaire. In Everscape, participants are confronted with an earthquake and have to evacuate from a tsunami. The Everscape data consist per second of the exact location and viewing direction of each participant. The questionnaire collects information on characteristics of the participants (e.g. age, gender) and focusses on what participants did during the Everscape part of the experiment and why they did this. In total 14 experiments were conducted in which around 400 people participated. The data collected with these experiments were analysed and two main conclusions were drawn. On the one hand, the results support results from literature, meaning realistic evacuation behaviour is found with the experimental set-up. On the other hand, a first step is made towards validly quantifying the effect of herding behaviour on evacuation decisions with empirical data.@en

Insight into factors influencing the choices people make in case of an evacuation from a natural disaster can help governments and emergency management personnel to manage people in case of such a situation. One of the aspects that influences the choices that people make in such a situation is herding. Since herding has not been quantified, this paper focuses on quantifying the effect of herding on the decision to evacuate by using an experimental setup that is based on the serious game Everscape. Around 400 people participated in 13 experiments with this setup. Choice models were estimated with the data from these experiments by including observable characteristics of herding as an attribute into the utility function. It is concluded that an important step is made in quantifying herding. It is shown that the more people someone sees leaving, the more inclined this person is to leave. Seeing people leave has more impact than seeing people stay. When people have no information from official sources, they tend to use other people as a source of information. In case of a disaster, this might result in following people who make the situation even more dangerous (for themselves and possibly for others as well). The information provided by official sources is therefore essential in managing people in the best possible way in case of a natural disaster.@en