Using panel data for modelling duration dynamics of outdoor leisure activities
Graphical abstract
Introduction
In transport-related research, the flexibility of outdoor leisure and recreational activities confers complexity to the demand models. The type of activity for which a journey is undertaken (such as sports or shopping) also determines the type of location (land use). In recent decades, the demand for recreational activities has increased because of increasing wealth, aging populations and changing lifestyles (Grigolon et al., 2013). It has therefore become important to analyse the expected demand for such locations, which is also associated with the duration of the undertaken activities and the transport needs of those locations. For example, the availability of parking facilities, operating times of public transport, proximity to leisure locations and connectivity by public transport can limit the duration and frequency of the leisure activities a person or family engages in.
Most non-work travel activity models focus on activity travel patterns for household interactions, recreational activities, and mode choice on holidays or weekends. Models for non-work journeys include a broad spectrum of trip purposes, such as family visits, shopping, as well as school- and church-related activities (Huang and Levinson, 2015). Bhat and Gossen (2004) looked at walking, jogging, riding a bicycle, and driving (touring) without having a specific destination as recreational activities. Bhat et al. (2004a) have examined how intra-household interactions influence weekday in-home and out-of-home activity generation; Habib and Hui (2015) have analysed the travel scheduling behaviour of older people for daily activities by using the framework of Bhat et al. (2004a). Arman et al. (2015) developed a model for household activity and vehicle allocation. Pozsgay and Bhat (2001) explored an attraction-end choice model of home-based urban recreational trips and Wang et al. (2015) investigated holiday travel behaviour characteristics in a trip chain. Overall, though, there is still little attention for non-work (leisure) activities.
The duration of outdoor leisure activities leads to dynamism in the transport market. Several factors influence the duration of leisure activities, such as a person's variety of activities and locations (intrapersonal variation). Studies based on longitudinal data have been conducted to analyse activity participation (Axhausen et al., 2002) and temporal and spatial variability of leisure activities (Schlich et al., 2004). These studies highlight the importance of analysing leisure traffic, and the relevance of modelling longitudinal data that covers both changing situations and sporadic behaviour.
Similarly, it is known that the mixture distributions can be discrete or continuous, and previous work show that any sample population may be decomposed into discrete segments that differ in their awareness of and proclivity towards certain mobility pattern (Vij et al., 2013). Multiple discrete continuous structures (MDC) have emerged as popular framework (Bhat, 2008), applied to goods consumption when continuous nature is involved, such as distance and time. Recently, MDC models have been extended to an integrated latent variable and choice model of time allocation (Enam et al., 2018), highlighting the importance of socioeconomic characteristics and individual's heterogeneity in activity durations. Also, some model studies treat activity type and duration as a continuous variable, e.g. multiple-discrete continuous extreme value models –MDCEV- (Calastri et al., 2017) rather than treating duration of activities as categorical manifestations. In most of the cases, this model structure assumes that individuals have a constant daily (24 h) budget of activity duration. An utility function for each (discrete) alternative is estimated and linked with the good consumption. And, a generic set of ‘consumption’ (e.g. duration) parameters is estimated.
However, individuals can have apriori preferences for a specific category of good consumption, in this case leisure activity duration (i.e. short or long duration leisure activity), depending on their preferences, activity patterns and available time budget during the day. So, longer durations do not always mean higher utilities. For example, a long duration leisure trip to an amusement park in the weekend versus a short duration lunch walk at a working day. Also, Calastri et al. (2017) indicated that consumer's preference structure might be such that she/he might not derive additional utility from additional consumption beyond a certain maximum level of consumption of a product. Exploring this required flexibility, some authors consider multiple constraints in time allocation models (Castro et al., 2012). Being flexibility a strong motivation in this paper to analyse duration of activities as categories instead of continuous variable, which allows the estimation of range-specific parameters, when an individual faces a single choice situation at the time.1 Previous research highlighted the advantages of discretization of alternatives in choice models. For example, Antonini et al. (2006) generated discrete categories of physical space, via the discretization of the set of walking alternatives. And, Vovsha and Bradley (2004) developed a hybrid discrete-duration model with temporal resolution (discretization) of 1 h. As described in Vovsha and Bradley (2004), this model has the advantages of a discrete choice structure (the flexibility and easy to estimate and apply) and the advantages of a duration model as the parsimonious structure with a few parameters that support any level of temporal resolution including continuous time.
Furthermore, the advantages of a repeated panel data are considered here to account for individual's heterogeneity as consumer's preference via (panel effects of) a mixed logit structure, see La Paix Puello et al. (2017). None of the previous modelling studies on discrete-continuous structures of activity durations have considered long term repeated panel data.
This paper presents an analysis of the duration of out-of-home leisure activities as discrete choice categories. The objective of our study was to investigate the factors that affect duration of outdoor leisure activities in a medium-term perspective (3 years). We explore shopping, sports/hobbies, tourism, walking, and other leisure activities and we took three years of data from The Netherlands Mobility Panel (MPN) to conduct the analysis on. The novelty in this paper comes from both the applied method as the characteristics of the data used. This paper develops an extended mixed logit modelling framework which uses discrete measurements of activity durations to disentangle range-specific effects of the estimated parameters, and a scale parameter of ‘zero-leisure days’ to incorporate temporal interdependencies between activities. Furthermore, we use data from the MPN, a unique panel data set in size (currently the largest ongoing mobility panel in the world) and scope (a very rich set of variables). This allows estimations of the effects of socioeconomic characteristics, trip characteristics and life events on outdoor leisure activities and leisure duration, and incorporation of interpersonal and intrapersonal variations in leisure activities and duration over time.
The remainder of this paper is organised as follows. Section 2 introduces the databases that were used for trip characteristics, socioeconomic characteristics, life events and other variables. Section 3 discusses the model setup (setup of alternatives, specification of variables, and conceptual model). Section 4 explores the statistics of the respondents and the variables collected from the database. Section 5 discusses the model results and Section 6 highlights the main findings of our study and provides recommendations for future work.
Section snippets
Literature on modelling duration of leisure activities
Activity duration has many effects on travel patterns for out-of-home activities. One effect is that the duration of the main activity can affect the propensity to undertake a secondary trip (Shiftan and Ben-Akiva, 2011). For instance, when the duration of a particular activity is long and is considered the primary/main activity (Katoshevski et al., 2015), there may not be any secondary trip. On the other hand, shorter periods of activity can be followed by one or more secondary trips towards
Description of The Netherlands mobility panel' (MPN)
The Netherlands Mobility Panel (in Dutch: MobiliteitsPanel Nederland, abbreviated as MPN) – a state-of-the-art household panel – aims to describe the dynamics in travel behaviour of individuals and households over time. The MPN is designed and implemented to understand social trends and their impacts on travel behaviour on an aggregated level, to identify and explain day-to-day variations in mobility, and look at the role of habits in travel behaviour (Hoogendoorn-Lanser et al., 2015). As an
Model structure
Discrete continuous model structures have been used to represent activity duration (Bhat, 2005) and joint models (see for example Srinivasan and Bhat (2006)) as well as mixed multinomial logit functions have been applied for e.g. out-of-home recreational episodes (Bhat and Lockwood, 2004). However, activity duration as categorical variable has received less attention, whereas this structure allows measuring the intrapersonal dynamics of activity duration via error components. The main advantage
Model estimation
Using the MPN data, we applied the discrete choice modelling approach to test the sensitivity of the attributes (travel oriented, socioeconomic) and life events for the duration of leisure activities. (See Section 3 for the definitions and descriptions of the categories of the alternatives.) We measured alternative-specific constants (ASCs) of the alternatives, assuming the highest duration (over 200 min) as a reference; this means that all ASCs are estimated except one. Different model
Conclusions
Understanding the dynamics of leisure activity duration is a very complex task; the diversity and flexibility of such activities are a challenge for modelling travel behaviour. In this study, a set of mixed logit scaled models was developed to explore the travel behaviour and dynamics of outdoor leisure and recreational activities. This paper provides an empirical analysis and discusses the dynamics and intrapersonal variation of duration of leisure activities; the analysis is based on discrete
Acknowledgements
This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors. We thank the KiM Netherlands Institute for Transport Policy Analysis for the fruitful collaboration in the development and running of the Netherlands Mobility Panel.
References (58)
- et al.
Mobile ICTs and physical mobility: review and research agenda
Transport. Res. Part A Pol. Pract.
(2012) - et al.
Discrete choice models of pedestrian walking behavior
Transp. Res. Part B Methodol.
(2006) - et al.
Information gain, novelty seeking and travel: a model of dynamic activity-travel behavior under conditions of uncertainty
Transport. Res. Part A Pol. Pract.
(2005) - et al.
On distinguishing between physically active and physically passive episodes and between travel and activity episodes: an analysis of weekend recreational participation in the San Francisco Bay area
Transport. Res. Part A Pol. Pract.
(2004) A multiple discrete–continuous extreme value model: formulation and application to discretionary time-use decisions
Transp. Res. Part B Methodol.
(2005)The multiple discrete-continuous extreme value (MDCEV) model: role of utility function parameters, identification considerations, and model extensions
Transp. Res. Part B Methodol.
(2008)- et al.
Intershopping duration: an analysis using multiweek data
Transp. Res. Part B Methodol.
(2004) - et al.
A mixed multinomial logit model analysis of weekend recreational episode type choice
Transp. Res. Part B Methodol.
(2004) - et al.
Does the social context help with understanding and predicting the choice of activity type and duration? An application of the Multiple Discrete-Continuous Nested Extreme Value model to activity diary data
Transport. Res. Part A Pol. Pract.
(2017) - et al.
Accommodating multiple constraints in the multiple discrete–continuous extreme value (MDCEV) choice model
Transp. Res. Part B Methodol.
(2012)