A household-level activity pattern generation model with an application for Southern California

There is obviously some subjectivity in the activity purpose classification adopted here, though the overall consideration was to accommodate differences between the disaggregate activity purposes along such contextual dimensions as location of participation, physical intensity level, duration of participation, amount of structure in activity planning, and company type of participation (see Srinivasan and Bhat 2005). Of course, the classification was also based on the activity purpose taxonomy used in the 2000 Southern California Association of Governments (SCAG) survey that provided the sample for the current analysis. Note also that we retain a “work-related” purpose as a maintenance activity as opposed to a mandatory work activity, and predict the work-related time allocation of each individual in the household if the individual is employed. In this regard, we will refer to work-related activity as a “non-work” activity in the current paper. Further, since no work-related activity participation time of any individual was joint with other individuals in the household (based on the survey data), we do not allow jointness in work-related activity participation among household members.

On the other hand, the model developed in this paper is a household-level activity pattern generation model that determines time-use within a defined period (such as a weekday or an entire week) across all possible combinations of the members of a household (including individual members by themselves) and activity purposes. This includes the discrete choice of no participation in certain combinations and the continuous choice of time allocated to each combination in which there is participation.

See Apps and Rees (2007), Del Boca and Flinn (2012) and Kato and Matsumoto (2009), who discuss the many types of intrahousehold resource allocation models, including Becker’s (1965) unitary model and extensions, non-cooperative and cooperative bargaining models (including Nash bargaining models), Chiappori’s (1988) collective models that include altruism, and Samuelson’s generalized household welfare function (GHWF). The paper by Apps and Rees is of particular relevance here. These authors show how the GHWF approach is a very general formulation that can accommodate elements of conflict and cooperation in household decision-making through the appropriate specification of exogenous variables, as we have also discussed in the main text. The essential position of the GHWF formulation is that many different types of processes are likely to be at work in intrahousehold decision-making, and it is not necessary that the researcher should adopt one specific process as being the (only) basis for decision-making. Rather, there is value in “abstracting from the process by which an allocation or preference ordering is reached”, especially when we are still nowhere close to identifying the specific process (or combination of processes) at work (see Lundberg 2005; Del Boca and Flinn 2012), and adopting a general “reduced-form” function that nonetheless captures elements of several different processes at once.

Zhang et al. (2009) provide a good review of utility forms, and discusses the Nash utility form as a specific case of the iso-elastic utility function form.

In the SCAG survey sample used in the empirical estimation of the current paper, 23.4% of households did not have any non-work activity participation at all during the weekday.

This procedure may be viewed as a form of two-stage allocation, in which the household and its members can be thought of as optimally allocating total available non-work time between in-home and total OH time in the first stage, followed by the allocation of total OH time across the discrete alternatives.

While we use the simple MDCEV model in the current empirical context, it is possible to extend the MDCEV framework to accommodate more flexible error structures using generalized versions of the MDCEV models (please see Pinjari and Bhat 2010a; Bhat et al. 2006 for such applications in the time-use context).

Future studies would benefit from exploring alternate forms of accessibility as well as the consideration of transit and non-motorized mode network skims (in addition to the highway network skims used here). The transit and non-motorized mode skims were not considered in our study due to data-related quality limitations.