A need-based approach to dynamic activity generation

Traffic congestion is still an everyday reoccurring problem. Models that predict traffic flows have been developed in the past and, in particular, considerable progress has been made in development and application of activity-based travel demand models over the last decade. However, several issues still need to be fundamentally addressed. Activity generation is a key factor in individual’s choices of trip frequency and trip purpose. Although several activity-based models made the transition to practice in recent years, modelling dynamic activity generation, especially the mechanisms underlying activity generation are not well incorporated in the current activity-based models. For example, current models assume that activities are independent, but to the extent that different activities fulfil the same underlying needs and act as partial substitutes, their interactions/dependencies should be taken into account. Hence, modelling dynamic activity generation is high on the research agenda in activity-based transport demand modelling. The concept of dynamic needs has been put forward as such a mechanism. The need-based activity generation approach, developed by Arentze and Timmermans (2009), is the first attempt to incorporate dynamic needs into an activity generation model. Within the need-based model, the utility of participating in an activity is defined in terms of its contribution to the satisfaction of dynamically changing needs. The model predicts the timing and duration of activities in a dynamic longitudinal framework, taking into consideration time budget constraints and needs at both household and person level. The mechanisms of the model allow for multi-day activity scheduling, within household interactions between individuals, and interactions between activities. The aim of this research project, therefore, has been to further develop and empirically test this dynamic need-based model for activity generation. Several components of the framework were studied. First, the need-based model was extended in order to consider influences of planned activities and events on the activity scheduling process. Second, the exact form of the utility functions of activity history and duration was obtained from a stated choice experiment. Third, the needs that underlie activity scheduling decisions were elicited and established using face-to-face interviews and a quantitative web-based questionnaire. Fourth, data was collected and analyzed with the purpose of estimating the parameters of the need-based model including the effects of events and planned activities and interactions between activities. The first objective was to develop and illustrate an extension of the need-based model of activity generation that takes into account possible influences of pre-planned activities and events. The theory considered both activity postponement when rational (i.e., a future day is more attractive in terms of an intrinsic preference and/or available time, such that it would be rational to postpone the activity) and future events (i.e., an event in the near future would be able to satisfy the same or similar needs, so that costs of conducting the activity can be saved). Simulations were conducted for six different activities and parameter values. The results showed that the model works well and that the influences of the parameters are consistent, logical and have clear interpretations. These findings offer further evidence of face and construct validity to the suggested modeling approach. In the future this extension of the model could be combined with previous work on event-driven activity generation. An experiment was conducted to estimate functions of several temporal factors on individuals’ propensity to schedule a given activity on a given day. The need-based theory on which the experimental design was based states that the probability of scheduling an activity is a complex and continuous function of how long ago the activity was lastly performed, the duration constraints for the activity and the amount of available time in the activity schedule of the day considered. Aurora, an existing model of activity scheduling, assumes S-shaped utility functions for the history as well as the duration functions, whereas most time-use studies assume monotonically decreasing marginal utilities. The stated choice experiment involved a range of flexible activities and a large sample of individuals, to measure the utility effects of a set of carefully chosen levels for the factors and test these specific assumptions. The results suggest that the amount of discretionary time on a day has no significant impact on the scheduling decisions, provided that enough time is available for the activity. The effects of other factors are as expected and show diminishing marginal utilities. Mixed evidence was found for an initial phase of increasing marginal returns, as assumed in an S-shaped function, for the duration and history component. Two related surveys were carried out to elicit and establish the needs that underlie activitytravel patterns of individuals. The first survey uses qualitative face-to-face interviews based on cognitive mapping techniques, to find out which needs and other factors are responsible for the discretionary activities individuals conduct in daily life. This resulted in nine needs to be included in additional research. In the second survey, subjects were asked to indicate to what extent they think the nine needs are influenced by 22 types of recreational, social, and sports activities. After looking at the similarities between the needs and their influences on performing activities, the set could be reduced to six independent needs, namely Social contact, Physical exercise, Relaxation, Fresh air/being outdoors, New experiences, and Entertainment. Many-to-many relationships between activities and needs support the hypothesis that substitution relationships may play a significant role in activity generation. This implies that current practice in activity-based modeling of focusing on activities may produce biased results when developing dynamic models of transport demand. In order to estimate the parameters of the need-based activity generation model, data had to be collected using a questionnaire that was designed specifically for this purpose. Therefore, a survey was carried out to collect activity data for a typical week and a specific day among an adequate sample of individuals. The diary data contained detailed information on activity history and future planning. The questionnaire included social, leisure and sports activities (as those activities are most likely to be substitutable). The survey was administered through the internet and approximately 300 respondents completed the questionnaire. Estimation of the need-based model involved a range of shopping, social, leisure and sports activities, as dependent variables, and socioeconomic, day preference, interaction, and activity anticipation variables, as explanatory variables. The results showed that several person, household, and dwelling attributes influence activity-episode timing decisions in a longitudinal time frame and, thus, the frequency and day choice of conducting the social, leisure and sports activities. Furthermore, interactions where found in the sense that several activities influence the need for other activities and some activities affect the utility of conducting another activity on the same day. Planned activities also showed some significant effects. In addition to the assumption that activities and events, planned in the near future, will affect the scheduling of activities that satisfy similar needs, other factors can explain in particular the negative estimates. For example, the need to conduct a preparation activity can be induced by a future event. Moreover, the planned activity can cause a lack of time on the future day, and consequently, in anticipation of this busy day, activities that have to be carried out, will be conducted on an earlier day. Overall, the research reported in this thesis led to a better understanding of the activity generation and scheduling process. In particular, the underlying needs, the need-growth function, interactions between activities, and the effects of events and planned activities were explored. The studies focused on leisure, sports, social and shopping activities, as those are more difficult to predict and, in addition, are more likely to be substitutable, compared to ‘fixed’ activities like paid work, courses, and going to school/university. The empirical studies show that it is possible to design specific surveys for the different components and use the collected data to test and develop the various elements of the needbased model.