Temporal interdependencies in mobility decisions over the life course: A household-level analysis using dynamic Bayesian networks

Life trajectory analysis has been shown a powerful approach to understand the interdependencies between key life events, critical incidents and long-term mobility decisions such as residential move, job change and change in vehicle possession, which in turn constitute the context of daily activity-travel decisions. Because people in multi-earner households share resources, some of these long-term decisions affect them equally, while job change affects them differently because their job location likely differs. Current life course models in transportation research, however, have typically considered individuals' trajectories. To contribute to the further development of the relatively thin line of research in transportation studies, a dynamic Bayesian network approach is proposed to investigate the temporal interdependencies between life course events from a household perspective. Results show that the effects of child birth are much larger on residential and car ownership change than on job change for both household heads in dual-earner households. Moreover, the probability of residential and car ownership change increases when both spouses have relatively long commuting times. In case only the husband faces an excessive commuting time, households have a larger probability of moving house or purchasing an additional car. By contrast, in case only the wife faces an excessive commuting time, she is more likely to change job rather than the household taking particular actions to adjust to the problematic situation.     

The multiplicity of self-selection: What do travel attitudes influence first,residential location or work place?

Residential self-selection has been widely considered as an important issue in quantifying the impacts of the residential built environment on travel behavior and much empirical evidence regarding the nature and magnitude of the self-selection effect has been reported. Nevertheless, people may be based on travel attitudes/needs to self-select not only residential location, but also work place, car ownership, etc. In other words, the impacts of long-term decisions other than residential location choices (e.g., decisions on work place, car ownership, etc.) on travel behavior may also be biased by the self-selection effect. However, self-selection concerning these long-term decisions has not been explored much in the travel behavior literature. The role of residential self-selection would not be properly evaluated if self-selections concerning other long-term decisions were not considered because they are often related. This paper addresses this research gap in the travel behavior literature by exploring the multiplicity of travel-based self-selection. We jointly examine the possible self-selections concerning residential location, workplace, commuting distance and car ownership in an integrated framework, taking into consideration the interrelationships among these decisions. Data are derived from an activity-travel diary survey conducted in 2016 in Beijing, China. We classify the respondents into two groups based on the choice order of their current residential and work locations and conduct a comparative analysis using structural equation models. It is found that self-selection exists in all long-term choices examined in the study. The choices of residential location and work place are found to be mutually dependent. Consequently, both choices have indirect impacts on travel behavior through the other choice.     

Where are the electric vehicles? A spatial model for vehicle-choice count data

Electric vehicles (EVs) are predicted to increase in market share as auto manufacturers introduce more fuel efficient vehicles to meet stricter fuel economy mandates and fossil fuel costs remain unpredictable. Reflecting spatial autocorrelation while controlling for a variety of demographic and locational (e.g., built environment) attributes, the zone-level spatial count model in this paper offers valuable information for power providers and charging station location decisions. By anticipating over 745,000 personal-vehicle registrations across a sample of 1000 census block groups in the Philadelphia region, a trivariate Poisson-lognormal conditional autoregressive (CAR) model anticipates Prius hybrid EV, other EV, and conventional vehicle ownership levels. Initial results signal higher EV ownership rates in more central zones with higher household incomes, along with significant residual spatial autocorrelation, suggesting that spatially-correlated latent variables and/or peer (neighbor) effects on purchase decisions are present. Such data sets will become more comprehensive and informative as EV market shares rise. This work’s multivariate Poisson-lognormal CAR modeling approach offers a rigorous, behaviorally-defensible framework for spatial patterns in choice behavior.     

Interdependences between household residential and car ownership behavior: a life history analysis

Once a transportation system is built or a land-use policy is carried out, it influences people’s travel behavior and their lives for a long time period. It is therefore important for policy makers to understand people’s decisions on travel behavior and lives over a longer time period. However, little has been known about the interdependences between life domains, especially over the life course (i.e., biographical interdependences) in the context of residential and car ownership behavior. To fill this gap, this study aims to clarify households’ biographical interdependences relating to residential and car ownership biographies by explicitly incorporating the influence of household structure and employment/education biographies. Biography is defined based on a general concept of mobility that indicates a change occurring in a life domain. For this purpose, a Web-based life history survey was conducted in November 2010 and 1000 households living in major Japanese cities provided valid data. Aggregate analysis and exhaustive CHAID analysis were carried out, focusing on the occurrence times of mobilities in each biography. Results confirm obvious two-way cause–effect relationships over the life course between residential and car ownership biographies that are further influenced by household structure and employment/education biographies. Especially, not only short-term but also long-term state dependence and future expectations within and across life domains are clarified. Household structure and employment/education biographies are found to be more influential on residential biography than car ownership biography. Though residential biography is seen to be more influential on car ownership biography, the other two biographies also play an important role in explaining the car ownership mobility decision. All these findings suggest the necessity of developing intra-domain and inter-domain biographical interdependence models with flexible structures that capture the influences of state dependence and future expectations over different time scales in the life course in a unified framework.     

Investigating household vehicle ownership,mode choice and trip sharing decisions using a combined revealed preference/stated preference Nested Logit model: case study in Bangkok Metropolitan Region

This study proposes a Nested Logit model to investigate household travel behaviour in respect to vehicle ownership, mode choice and trip sharing decisions. The model is analysed using revealed preference (RP) and stated preference (SP) data since a combined estimation of RP/SP data is an effective method of expressing complex travel behaviour and forecasting travel demand for new transport services. In the proposed model, the nesting structure has two levels. The upper level shows car ownership, motorcycle ownership, and no vehicle-ownership choices, and the lower level shows the mode choice combinations for two-traveller households. Trip sharing is considered as one of the mode-choice options in the model. The proposed model is analysed using data from the Bangkok Metropolitan Region. The analysis conducted informs that Central Business District (CBD) travel, long distance travel, household income, job status, age of travellers and presence of school children in households are key aspects in household travel decisions. Based on these aspects, households make important decisions on vehicle ownership, mode choice and trip sharing. In addition, this study reveals commuters’ hidden preferences for modes that are not in existence, in particular the Mass Rapid Transit System in the Bangkok Metropolitan Region due to be fully implemented in 2010.     

Determinants of car ownership among young households in the Netherlands: The role of urbanisation and demographic and economic characteristics

In the Netherlands, car ownership among young adults has slowly decreased in recent decades. The main causes of this trend are still unclear. Using a unique dataset in which vehicle registration data were combined with population and income register data for 2012/2013, this paper explores how car ownership among young Dutch households varies with household composition, urbanisation level (of household location), household income, employment status and ethnic background. Logistic regression analysis of this data revealed that urbanisation level and household composition are essential factors influencing car ownership. In addition, we found significant interaction effects between these two factors: the influence of urbanisation level on car ownership was much stronger for young couples than for young families or singles. Our results imply that increasing urbanisation and postponement of parenthood could reduce future car ownership among young adults in general. However, the increasing number of young families moving to more urbanised areas could increase future car ownership in cities.     

An examination of the factors that impact upon multiple vehicle ownership: The case of Dublin, Ireland

This paper examines the characteristics of households with multiple car ownership in Dublin, Ireland. Data from the 2006 Census of Ireland are analysed to ascertain the characteristics of these households. The analysis of multiple car ownership presented herein examines individual specific, transport availability, and household characteristics to provide an indication of the individuals most likely to have access to more than one vehicle. Understanding the characteristics of households with more than one car is important for many reasons, such as how policies for emissions reductions or pricing regimes might affect households. Ireland, like many countries, has recently launched a number of electric vehicle and car sharing schemes. Traditionally these schemes have been aimed at reducing multiple car ownership, therefore it is important to develop an understanding of the households that would most likely give up an extra car and use a car sharing scheme or an electric vehicle. Also from a sustainability point of view, greater levels of car ownership can result in unsustainable transport patterns.This paper examines the Census data using a multinomial logit regression model to determine the relationships between multiple car ownership levels and several household characteristics. The findings of the paper demonstrate that occupation, public transport availability and residential density all have an impact upon the decision to own more than one vehicle.     

Land-use and socio-economics as determinants of traffic emissions and individual exposure to air pollution

This paper presents a modeling framework developed for the City of Montreal, Canada, and is intended to quantify two indicators that can explain the spatial distribution of traffic-related air pollution at a metropolitan level. The indicators are estimated at the level of the traffic analysis zone (TAZ) and include: (1) the average level of emissions generated per individual and (2) the level of emissions occurring in a zone as a proxy for air pollution exposure. A regional traffic assignment model is extended with capabilities for emission modeling at an individual trip level while taking into account vehicle (type and age) and trip attributes (road type, speed, and volume). We observe that individuals who generate higher emissions from travel tend to reside in areas with lower exposure to traffic emissions while individuals associated with low levels of travel emissions (e.g. travel smaller distances, conduct less trips, and use alternative modes) reside in areas with high levels of traffic pollution. A regression analysis of the two indicators against a set of land-use and socio-economic variables shows that generated emissions per individual are positively associated with car ownership and larger vehicles, while being negatively associated with ownership of newer vehicles, and location in dense and walkable neighborhoods with high levels of commercial land-use. Meanwhile, exposure to emissions is positively associated with dense and walkable neighborhoods and negatively associated with car ownership and larger vehicles. These findings indicate major inequities in the generation of and exposure to traffic-related air pollution.     

Shared autonomous vehicles and their potential impacts on household vehicle ownership: An exploratory empirical assessment

Emerging transportation technologies have the potential to significantly reshape the transportation systems and household vehicle ownership. Key among these transportation technologies are the autonomous vehicles, particularly when introduced in shared vehicle fleets. In this paper, we focus on the potential impact that fleets of shared autonomous vehicles might have on household vehicle ownership. To obtain initial insights into this issue, we asked a sample of university personnel and members of the American Automobile Association as to how likely they would consider relinquishing one of their household's personal vehicles if shared autonomous vehicles were available (thus reducing their household vehicle ownership level by one). For single-vehicle households, this would be relinquishing their only vehicle, and for multivehicle households (households owning two or more vehicles) this would be relinquishing just one of their vehicles. Possible responses to the question about relinquishing a household vehicle if shared autonomous vehicles are present are: extremely unlikely, unlikely, unsure, likely, and extremely likely. To determine the factors that influence this response, random parameters ordered probit models are estimated to account for the likelihood that considerable unobserved heterogeneity is likely to be present in the data. The findings show that a wide range of socioeconomic factors affects people's likelihood of vehicle relinquishment in the presence of shared autonomous vehicles. Key among these are gender effects, generational elements, commuting patterns, and respondents' vehicle crash history and experiences. While people's opinions of shared autonomous vehicles are evolving with the continual introduction of new autonomous vehicle technologies and shifting travel behavior, the results of this study provide important initial insights into the likely effects of shared autonomous vehicles on household vehicle ownership.     

A deeper investigation into the effect of the built environment on the use of ridehailing for non-work travel

Ridehailing has become a main-stream mobility option in many cities around the world. Many factors can influence an individual's decision to use ridehailing over other modes, and the growing need of policy makers to make built-environment and regulatory decisions related to ridehailing requires an increased understanding of these. This study develops a model that estimates how the built environment affects the decision to choose ridehailing for making non-work trips, while carefully accounting for a variety of confounding effects that could potentially bias the results (if ignored or improperly incorporated). These include: total number of trips, differences in supply between urban and non-urban areas, residential choice (e.g. urban versus non-urban areas), and household choice of whether to own a vehicle. We use individual-level data from a California travel survey that includes detailed attitude measurements to estimate an integrated choice and latent variable (ICLV) model to properly specify these effects. We include accessibility measures used elsewhere (e.g., Walkscore) plus measures developed for this study. Our analysis estimates the effect of these measures on ridehailing mode share, and how they differ between urban and non-urban areas. This analysis results in several major findings: we confirm that omission of latent preferences for residential location and vehicle ownership from the analysis can lead to biased results; previous studies may have overestimated the complementarity or substitution relationships between public transit and ridehailing by ignoring confounding effects; and even after accounting for other effects, individuals living in vibrant and walkable neighborhoods have a higher mode share for ridehailing (potentially using it instead of active modes).     

The built environment and vehicle ownership modeling: Evidence from 32 diverse regions in the U.S.

Vehicle ownership is an important determinant of the travel demand forecasting process. Vehicle ownership models are used by policy makers to identify factors that affect vehicle miles traveled, and therefore address problems related to energy consumption, air pollution, and traffic congestion. For the conventional travel demand forecasting, it logically follows land use forecasting, before trip generation, which is commonly treated as step one. The most critical limitation of the vehicle ownership models, especially in the conventional process, is that they are often related mainly to sociodemographic variables, not so much to built environmental variables. In this study, by pooling regional household travel survey data from 32 diverse regions (almost 92,000 households) of the U.S., and by controlling for socio-demographic and the built environmental variables, we estimated a vehicle ownership model that contributes to the understanding of vehicle ownership and improves the accuracy of travel demand forecasts. Two main findings of this research are: 1) The number of vehicles owned by a household increases with socio-demographic variables and decreases with almost all of the built environmental variables. For the urban planning and design practices, this finding suggests that car shedding occurs as built environments become more dense, mixed, connected, and transit-served. 2) We used both count regression and discrete choice models, and the results suggest that count regression models have better predictive accuracy. The model developed in this study can be directly used for travel demand modeling and forecasting by metropolitan planning organizations.     

Temporal dynamics in local vehicle ownership for Great Britain

This article explores the stability of local vehicle ownership rates in Great Britain using the technique of spatial Markov chain analysis. Non-spatial Markov chain processes describe the transition of neighbourhoods through levels of ownership with no regard to their neighbourhood context. In reality however, how a neighbourhood transitions to different levels of ownership could be influenced by its neighbourhood context. A spatial Markov chain accounts for this context by estimating transition properties that are conditioned on the surrounding neighbourhood. These spatial Markov chain properties are estimated using a long run census time series from 1971 to 2011 of household vehicle ownership rates in Great Britain. These processes show that there is different behaviour in how neighbourhoods transition between levels of ownership depending on the context of their surrounding neighbours. The general finding is that the spatial Markov process will lead to a greater homogeneity in levels of ownership in each locality, with neighbourhoods surrounded by relatively low ownership neighbourhoods taking longer than a non-spatial Markov process would suggest to transition to higher levels, whilst neighbourhoods of high ownership surrounded by high ownership neighbourhoods take longer to transition to lower levels. This work corroborates Tobler's first law of geography “Everything is related to everything else, but near things are more related than distant things” but also provides practical guidance. Firstly, in modelling ownership, spatial effects need to be tested and when present, accounted for in the model formulation. Secondly, in a policy context, the surrounding neighbourhood situation is important, with neighbourhoods having a tendency towards homogeneity of ownership levels. This allows for the effective planning of transport provision for local services. Thirdly, vehicle ownership is often used as a proxy for the social and aspirational nature of an area and these results suggest that these properties will persist for a prolonged period, possibly perpetuating and exacerbating differentials in society.     

Endogenous treatment of residential location choices in transport and land use models: Introducing the MetroScan framework

Within a random utility maximisation modelling framework, the paper develops a residential location choice model as part of an integrated transport and land use modelling system, called MetroScan – a quick scanning tool to evaluate transport and land use initiatives, including benefit-cost analysis and economic impact analysis. We describe how the developed model is integrated, as an empirically calibrated module, into the behaviourally richer transport and land use modelling system of MetroScan for practical application. A full application of MetroScan modelling system to Sydney West Metro link recently proposed by the New South Wales government is presented as a case study. The results demonstrate how the residential location choice model works with other inter-connected models, such as work and non-work location choices, dwelling tenure and dwelling type, and vehicle fleet size choice embedded in the modelling system, in simulating the impact of transport and housing development on household choices of residential location.     

Space-time dynamics: A modeling approach for commuting departure time on linked datasets

Commuters' departure time related decisions are important in time geography. Analytic tools have been proposed to capture the inherent choice determinants both in time and space. Although the dynamic aspects of the problem have been identified, most of the existing studies are based on static models. In this paper, a dynamic modeling framework is proposed to explore the relationship between commuters' departure time choices and the evolution of en route traffic. A data linkage method is developed to create an integrated dataset that enables the observation of commuters' reaction to changes in travel time and traffic conditions over time. A regional household travel survey is linked to travel information obtained from the Google Maps application program interface (API), creating a synthetic longitudinal dataset. Two decision rules are applied to model commuters' response to the evolution of traffic. The results indicate that travel time, distance to work location, flexibility in working schedule, expected arrival time, and commuters' sociodemographic influence departure time choices. It is also found that accounting for dynamics improves model fit and out-of-sample predictions. Both the dynamic model and the proposed data linkage method contribute to the understanding of human activities in space and time and can be used to enhance transportation demand analysis and urban policy studies.     

Effects of neighborhood types & socio-demographics on activity space

This paper examines the relationships between socio-demographic characteristics, travel time, the built environment and resulting average activity spaces for all activities and non-work activities separately using data from the 2012 Northeast Ohio Regional Travel Survey. Multiple regression models are developed to analyze these relationships at individual level. First K-means cluster analysis is conducted to create seven neighborhood types based on five built environment variables. These new neighborhood types are used as discrete explanatory variables to explain average activity spaces, while controlling for travel time, individual and household features, access to transit facilities and the job-population balance. The modeling results indicate that residential location characteristics have significant influences on activity spaces. People living in places away from suburban and rural areas and with a high mix of population and employment tend to have smaller activity spaces. Moreover, this study finds out that while the effects of some explanatory variables (such as age and gender) vary for all activities and non-work activities, socially disadvantaged people (such as the elderly and low income households) generally experience smaller activity spaces.     

Examining spatial variations in minimum residential parking requirements in Melbourne

Minimum off-street residential parking requirements are used in many cities as a way to accommodate parking demand associated with new residential development. In some cases, variations to these requirements are used in the form of reduced (or eliminated) minimums and/or maximum parking requirements to more actively manage parking demand. This paper assesses how such variations affecting new residential apartment development in Melbourne, known locally as parking overlays, compare against residential parking demand. Using household car ownership data as a proxy for off-street residential parking demand, a case-control analysis was undertaken to compare car ownership within and immediately outside areas affected by the parking overlays, while controlling for a range of built environment, public transport, demand management and socio-demographic variables. Key findings indicate that car ownership is generally lower in areas affected by parking overlays, yet this was either roughly the same or well below the actual parking requirement. Through regression modelling, the results highlighted the importance of public transport service quality, car parking requirements and demographics in influencing car ownership within and immediately outside the parking overlay areas. These results were used to develop a parking overlay index to identify other areas that could benefit from more flexible residential parking requirements. Despite parking overlays considered as a form of parking management, the results imply that, in Melbourne, they represent little more than a conventional supply-side approach to parking policy. The results indicate that residential off-street parking requirements could be reduced further in Melbourne, both within and outside of areas affected by parking overlays, to more actively manage parking demand.     

Modeling green vehicle adoption: An integrated approach for policy evaluation

This article employs an integrated discrete-continuous car ownership model to jointly forecast households’ future preferences on vehicle type, quantity and use, and to estimate greenhouse gas (GHG) emissions. The model system is estimated on a dataset collected from a web-based stated preference survey conducted in Maryland in 2014. The data contain vehicle purchase decisions and sociodemographic information of 456 households who were requested to state their future preferences over a 9-year period (2014–2022). In each time period, a respondent is faced to four alternatives that include the current vehicle, a new gasoline vehicle, a new hybrid electric vehicle, and a new battery electric vehicle. Intertemporal choices between conventional and “green” vehicles such as hybrid and electric cars capture dynamics in vehicle purchase decisions. Short run and medium-long run situations were predicted and compared based on the first 4-year data and the entire 9-year data of the dynamic panel. Vehicle GHG emissions were calculated correspondingly. We find the introduction of “green” vehicles makes a positive impact on car ownership and use, especially in a medium-long run. Two “green” taxation policies, gasoline tax and ownership tax, were proposed and their impact on vehicle use and emission reductions was evaluated. Results indicate that: (a) gasoline tax is a more effective way to reduce vehicle miles traveled and GHG emissions and (b) gasoline tax makes a higher impact on car use and emission reductions in the medium-long run, while ownership tax makes a higher impact in the short run.     

Estimating local car ownership models

Many studies in the transport demand literature have shown that income is an important factor in determining how many cars a household owns. When the models used to measure the strength of this relationship are estimated on cross-sectional data, they typically yield one overall value as the estimate. Local circumstances will, however, vary. This paper illustrates the use of the Geographically Weighted Regression technique to estimate the individual strength of this relationship for each of the United Kingdom electoral wards. Use of this type of model enables a wards’ income elasticity to be based on both the local estimate of the strength of this relationship and the current local level of car ownership. How the use of this local elasticity changes future forecasts of the size of the vehicle fleet is illustrated.     

Where no cars go: Free-floating carshare and inequality of access

Carsharing programs have demonstrated a potential to significantly shift incentives with regard to private vehicle ownership. The advent of free-floating vehicle fleets has enabled providers to offer ubiquitous vehicle access in designated urban areas. The ability of users to choose where to drop off vehicles presents the possibility that the density of available vehicles in particular areas will be insufficient to supply a reasonable level of service to local residents. The current paper will use exclusive data on vehicle location from a free-floating carshare service that operates in ten U.S. cities. Analysis will relate the availability of vehicles to census tract demographics. Results show vehicles cluster in tracts that are disproportionately populated by residents who are educated, young, employed, and white. Carshare systems have received significant in-kind incentives from government to operate. The mobility benefits of free-floating carshare systems appear to accrue disproportionately to advantaged populations.