Household activity-travel behavior : implementation of within-household interactions

R. Anggraini

Onderzoeksoutput: ScriptieDissertatie 1 (Onderzoek TU/e / Promotie TU/e)

871 Downloads (Pure)


Although the importance of households as a decision making unit has been recognized in seminal work in activity-based analysis of transport demand, most comprehensive models have relied on individual activity-travel patterns. The transformation of these models to household level models and the explicit consideration of resource allocation, task allocation and joint activity participation decisions is thus a challenge and research frontier in this field of study. To contribute to this expanding field, the aim of this PhD study is to develop such an activity-based model. More specifically, the slightly ad hoc treatment of household decisions in the ALBATROSS model is replaced with a systematic incorporation of household decisions. The new variant, based on the MON 2004 data, is compared in terms of goodness-of-fit and sensitivity with the previous version of the model. To this end, the thesis is organized as follows. Chapter 2 provides a review of past research efforts concerning the determinants of household decision making. We discuss how household decision making has been treated in comprehensive activity-based models of transport demand. This line of research started with analytical studies on household decision making taking into account car allocation and usage decisions. Further literatures addressed task and time allocation decisions. They found that household types, defined by the number of household heads and work status, strongly influence activity time allocation and trip chaining. The presence of children in the household has a positive effect on the duration of all out-of-home activities in household trip chaining, except for the duration of out-of-home discretionary activities of households having children under 5 years old. This suggests that the presence of children induces more chaining of trips and more time allocated to these trip chains. Households having more children of 16 years of age and over are more likely to spend time in trip chaining for out-of-home subsistence activities. Finally, they found that flexible work arrangements tend to be correlated with less trip chaining for the work trip. In addition to these studies, there is also a literature on joint activity participation. Several studies have examined the effect of household attributes on joint activity-travel behavior. They found that joint activities involving household heads are significantly affected by the presence of children. Couples without children living at home are more likely to pursue joint out-of-home non-work activities than couples with children. In households with children, most joint activities between adults are at home. In addition, the employment status of the household heads influences whether a joint activity originates from home or from an out-of-home contact point. In additional to analytical studies, existing comprehensive activity-based models are reviewed in this chapter in terms of their inclusion and treatment of household decisions. Comprehensive in this context means that the model allows predicting a combination of choice facets, at least compatible with those underlying traditional four-step models: i.e. activity generation, destination and transport mode choice. The discussion is restricted to fully operational models. Over the years, many activity-based models have been suggested in the literature, including constraints-based models, micro-simulation models, (nested logit) utility-maximizing models, suites of advanced statistical models and rule-based models. Most of these models either do not incorporate household decisions at all, or only in a limited way. Chapter 3 discusses the conceptual framework of this thesis for modeling household activity-travel behavior. Because the thesis is an attempt of elaborating the ALBATROSS model, we discuss this model in more detail, including its conceptual framework. Further, we explain the entire process underlying the ALBATROSS system and the inclusion of household decision making in the process, such as joint participation, activity allocation, car allocation for non-work tours, and some other choice facets. Household decision making is mostly applicable to non-work activities, but the problem of car allocation is highly relevant for work tours in car-deficient households. Further, we summarize the methodology that was used in this study: decision tree induction using a CHAID-based induction algorithm being the core method of ALBATROSS. The remaining chapters then present the results of the various derived decision tress for the sequential choice facets that together make up the ALBATROSS model. Chapter 4 describes the results for car allocation choice focusing on work activities. In this analysis work-tours as opposed to work trips are considered. The car allocation model focuses on car-deficient households (i.e., more drivers than cars present) and a joint decision between the two heads (mostly, a female and male). We also assume that both male-female are drivers and at least one of them has a work activity on the day considered. Furthermore, the model includes the option that none of the household heads uses the car, but some other means of transport. The results show that the propensity of men driving a car to the work place is higher than that of women, particularly, when women have no work activity or women’s work place is in the same zone as the home location. This finding is consistent with the common notion that women use a slow or public transport mode more often to travel to activity locations. Women tend to use the car when men have no work activities or men work at home. Chapter 5 reports the empirical derivation of a household decision model of activity choice taking into account joint participation and task allocation between household heads. These are considered household-level decisions given that they involve commitments of multiple persons, in particular the two-head households. Of the 10 activity categories concerned, 7 activity categories (non-work activities) are used in this study, i.e. bring/get, shopping to 1 store, shopping to multiple store, service-related, social, leisure, and touring. The first four activities are deemed task allocation activities and the rest are non-task activities (discretionary). Hence, two decision trees were derived from diary data. The activity participation model, given the large number of observations that could be derived from the data, included more than 300 condition action rules. The household task allocation model also involved an extensive set of decision rules, involving more than 90 condition-action rules. In both cases, the validity of the decision tree is satisfactory in the sense that the derived rules are readily interpretable and the overall goodness-of-fit of the model on a validation set is acceptable as well. Chapter 6 focuses on the joint participation of male-female heads in non-work activities and attempts to model the timing and duration decisions for these activities, using decision tree induction. Decision tree results indicated that there were 17 and 31condition-action rules derived for the duration model and start time model, respectively. The improvement in S-value (a measure of prediction accuracy) relative to a null model as well as an F-statistic indicates that there is a moderately strong association between condition variables at household, individual, activity and schedule level, on the one hand, and the decision, on the other. The S-value shows a more substantial improvement in the start-time model compared to the duration model. The results show that activity type has the most significant influence in both models. In addition, time availability for non-work activities during morning off-peak periods has a strong influence on start time decisions. The results also suggest that there is a substantial influence of duration decisions on start time decisions. Joint participation of household members in activities tends to lead to longer activity duration and earlier start times. Overall, modeling timing and duration of joint activity participation decisions at the household level proves to have some clear advantages. Chapter 7 discusses the development of the household location choice model taking into account the independent and joint activities, in particular non-work activities. In ALBATROSS, location choice is modeled for independent and joint activity participation of the household heads based on the concept of detour time. The detour time of a candidate location for an activity is defined as the extra travel time required to implement the activity in the context of the current activity schedule. There were two decision tree models for both independent and joint activity categories. The first model relates to the decision whether or not the activity is performed at the same location as the previous activity, whether the activity is done at the same location as the next activity, or whether it is conducted elsewhere. The second model relates to the last choice option in the first model and comprises 25 choice alternatives. It verifies the location in terms of a combination of size - distance classes. The size class depends on a particular activity type and the size of available facilities at the activity location. Size is classified into 5 categories based on employment in the relevant sector for the activity considered and distance is classified in terms of a detour travel time (by car) also into 5 categories. The tendency of conducting a particular activity at the same location as the previous activity is higher for independent activities than for joint activities. The same condition also applies to activities that are conducted at the same location as the next activity. These results imply that males and females are more likely to conduct multiple activities at one particular location independently than jointly. These results do make sense, since the activity-travel behavior of one person is different from the other person, even though male-female couples live in the same household. Chapter 8 is concerned with car allocation behavior for non-work activities. In this study, the assumption is similar to the assumption in Chapter 4 where tours are taken into account instead of trips. Travel for any activity episode or set of chained activity episodes that does not include a work activity is considered a non-work tour. The problem of modeling this allocation problem for non-work tours is more complex than for work tours because the decision at this stage depends considerably on the outcome of the previous stages in the scheduling process. Hence, the car can be allocated to male, female or none. Further, only overlapping non-work activities of the male’s and female that occur in the same time slot are taken into account. Overlapping tours are defined as a pair of tours conducted by respectively male and female of which the start and/or end times of each tour (simulating use of a car for the tour) defines a fully or partially overlapping episode. As a tour consists of a sequence of trips that starts and ends at a particular location (i.e., home), the primary activity in each tour needs to be determined. In order to identify the primary activity in a particular tour, we consider a hierarchical order of activity priority. In particular, 10 activity categories are considered in order of priority starting from work, business and other (mandatory)activities. A group of non-work activities is considered, such as escorting, shopping (daily and non-daily), service-related, social, leisure, and touring. Since business and other mandatory activities are not considered primary work activities, they are not dealt with in the first stage of the scheduling process and, hence, they are also considered as non-work activities in this model. The results show a satisfactory improvement in goodness-of-fit of the decision tree model compared to the null model. Gender seems to play an important role. A descriptive analysis indicates that men more often than women get the car for nonwork tours for which a car allocation decision needs to be made. Tour-level attributes are shown to influence the household car allocation decision for non-work tours. The decision to allocate the car is considerably influenced by the longest distance (travel time) from home to a particular location in a tour of men and women. The probability that the men and women get the car monotonically increases with increasing travel time. Socio-economic and situational factors have less influence on the car allocation decision. Overall, men have more influence on the car allocation decision for non-work tours, as indicated by the number of influential variables that relates to the males in the impact table. Chapter 9 discusses the results of the integrated model of ALBATROSS. In order to test the performance of the ALBATROSS system based on all decision tables for the assumed scheduling process, the validity and sensitivity of the integrated model were evaluated and compared with the performance of the old model. First, the validity of the model versions was compared by evaluating the extent to which the model is able to reproduce observed frequency distributions and mobility indicators in the MON dataset. In that sense, no major differences were expected. Instead, it was expected that the new model is able to reproduce the aggregated distributions as well as the existing model. A Second effort was to examine the sensitivity of the models by applying the models to a particular scenario of change in the Dutch population. It was expected that the new model was more sensitive to such scenarios. The scenario assumed an increase of 41 % in labor participation of women household heads (labor scenario) assuming the year 2000 as the base year. A fraction of 10% of the Dutch population in the year 2000 was generated using the synthesis module of ALBATROSS for the baseline and the labor scenario. As expected, in the context of validity test, the new model showed equal or slightly better goodness-of-fit for most choice facets, except for time of day and trip-chaining. The new model proved to be more sensitive to facets such as activity type, start time, trip-chaining, location, etc., in response to scenarios change. In particular, the new model predicted somewhat different responses that could be interpreted in terms of the better representation of opportunities and requirements related to task allocation and joint activity participation. In sum, by considering decisions of household heads in interaction, the system is able to predict with increased sensitivity activity-travel rescheduling processes of households in response to change.
Originele taal-2Engels
KwalificatieDoctor in de Filosofie
Toekennende instantie
  • Built Environment
  • Timmermans, Harry, Promotor
  • Arentze, Theo A., Co-Promotor
Datum van toekenning1 dec. 2009
Plaats van publicatieEindhoven
Gedrukte ISBN's978-90-6814-623-4
StatusGepubliceerd - 2009

Bibliografische nota



Duik in de onderzoeksthema's van 'Household activity-travel behavior : implementation of within-household interactions'. Samen vormen ze een unieke vingerafdruk.

Citeer dit