MODELING HOUSEHOLD CAR OWNERSHIP LEVEL CHANGES IN AN INTEGRATED LAND-USE/TRANSPORT MODEL

Transcription

1 MODELING HOUSEHOLD CAR OWNERSHIP LEVEL CHANGES IN AN INTEGRATED LAND-USE/TRANSPORT MODEL Matthew Bediako Okrah, Corresponding Author Arcisstrasse, 0 Munich, Germany Tel: +---; matthew.okrah@tum.de Ana Tsui Moreno Arcisstrasse, 0 Munich, Germany Tel: +---; ana.moreno@tum.de Carlos Llorca Arcisstrasse, 0 Munich, Germany Tel: +---; carlos.llorca@tum.de Rolf Moeckel Arcisstrasse, 0 Munich, Germany Tel: +---; rolf.moeckel@tum.de Paper -00 //0 ACKNOWLEDGEMENT The research was completed with the support of the Technische Universität München Institute for Advanced Study, funded by the German Excellence Initiative and the European Union Seventh Framework Programme under grant agreement n.

2 Okrah, Moreno, Llorca, Moeckel INTRODUCTION Car ownership plays an important role in the interaction between land-use and transport. Both land-use design and the transport system heavily influence car ownership which in turn has substantial impacts on congestion and urban growth patterns (). Considering car ownership in an integrated land-use/transport model therefore facilitates the assessment of a wide range of impacts. To assist decision makers in impact assessments of car ownership, not only is it important to determine the number of cars owned by households. The ability to determine the changes to household car ownership over time is even more crucial. The acquisition or relinquishment of a car reflects a significant change in a household s mobility resources and can therefore cause a change in the mobility behavior of household members (). Whereas previous studies of car ownership focused on cross sectional analyses, the availability of large scale panel data sets has facilitated the examination of changes to household car ownership (). Panel studies have identified a number of factors that explain household car ownership level changes. A pseudo-panel analysis based on data from annual Family Expenditure Surveys in the UK showed that household car ownership increases as the household head reaches the age of 0 and thereafter decreases (). This mirrors the traditional family life cycle through which the household size increases by cohabitation and parenthood, and decreases by offspring leaving home ( ). The existing number of cars has also been found to be a strong predictor of the number of cars owned in the next period (, ). Using structural equation modeling to test a-priori hypotheses on the paths linking car availability, season ticket ownership and modal usage at three different time periods, car ownership was found to be highly stable over time (). Moreover, life events such as moving home, having children or changing jobs have been found to increase the likelihood of car ownership level changes as they have the potential to disrupt behavioral patterns and prompt people to reconsider and alter their behavior and resources (). This paper presents a microscopic car ownership model estimated with cross-sectional and panel data and implemented in an integrated land-use/transport model to model the changes to household car ownership level. METHODOLOGY Recognizing the interdependence of land-use and transport, and the role of car ownership in both systems, the changes to household car ownership level is modeled in an integrated landuse/transport modeling suite which simulates land use and travel behavior from 0 to 00 for the Munich Metropolitan area in Germany. The suite includes the Simple Integrated Land Use Orchestrator (SILO) module, the Microscopic Travel Demand Model Orchestrator (MITO) module and the MATSim module for traffic assignment. To determine the number of cars owned by each household in the start year (0), a cross-sectional model was estimated with data from the 00 German National Household Travel Survey (MiD). Considering the pros and cons of the three econometric methods suitable to estimate discrete values (), the logistic regression method was selected to model car ownership. A multinomial logistic regression formulation was therefore estimated to find the utilities of a household to own,, or at least cars relative to owning no car. Based on existing literature and other car ownership model applications, a number of variables ranging from household attributes to built environment variables were defined, with emphasis on variables that could be measured and/or easily obtained. The model was subsequently applied to the Munich Metropolitan Area to compute the number of cars owned by each household in the start year.

3 Okrah, Moreno, Llorca, Moeckel 0 For the changes to the number of cars owned by each household in successive years, a transitional model was estimated with longitudinal data from the German Mobility Panel. Like the car ownership model, a multinomial logistic regression formulation was used to find the utilities of a household to add or relinquish a car relative to maintaining the existing number of cars. Drawing on the principal reasons of car ownership level changes () and the relationships between life events and car ownership level changes (), a number of variables related to changes in household characteristics were defined, with emphasis on variables that could be measured and/or easily obtained. The model was subsequently applied to two waves of the German Mobility Panel that had not been used in model estimation. FINDINGS Table shows the results of the multinomial logistic regression estimation used to determine the number of cars owned by each household in the start year while Figure shows the validation of the household car ownership model. TABLE Car Ownership Model Estimation Results Variable car cars + cars Intercept Driver license holders... Number of workers Household income (in Euro) Distance to transit (log of distance in meters) Area type (core and large cities) Area type (medium sized cities) 0... Area type (small cities) Area type (rural communities)...0 McFadden R = 0.

4 Okrah, Moreno, Llorca, Moeckel 0 FIGURE Validation of number of cars per municipality It can be seen from Table that the effect of the variables is highest on the first car ownership and tends to reduce with additional levels of car ownership. The number of driver license holders has the highest explained influence on household car ownership. Whereas all the included variables positively influence household car ownership, the negative intercept values indicate the existence of an unexplained resistance to car ownership. After applying the model to the Munich Metropolitan Area, the results from Figure shows a good match between modeled and observed number of cars especially for municipalities with fewer cars. For the car ownership level change, Table shows results of the multinomial logistic regression estimation while Figure compares the modeled and observed number of households in each car ownership level change category. TABLE Car Ownership Level Change Model Estimation Results Variable Add Relinquish Intercept Increase in household size.0 - Decrease in household size -.0 Increase in household income 0. - Decrease in household income - 0. Household member acquiring a license. - Change in residential location Number of cars in previous year McFadden R = 0.0

5 Okrah, Moreno, Llorca, Moeckel 0 0 FIGURE Validation of car ownership level change It can be seen from Table that the increase in household size has the highest influence on the acquisition of a car followed by the acquisition of a driver license by a household member. For the relinquishment of a car, it is the existing number of cars that has the highest influence followed by a decrease in household size. The existing number of cars in a household is a strong predictor of both addition and relinquishment of household cars confirming the presence of state dependence (). After applying the model to the unused waves of the German Mobility Panel data, the results from Figure shows a good match between modeled and observed changes for 0-, -, and to some extent - car households. The model however underestimates the share of - car households relinquishing a car. CONCLUSION This paper has presented a microscopic car ownership model estimated with cross-sectional and panel data for implementation in an integrated land-use/transport model to model the changes to household car ownership level. The study has demonstrated the value of panel data which has enabled the estimation of a multinomial logistic regression model to determine changes to household car ownership level. The results presented in the paper shows that there exists a high unexplained reluctance to car ownership in the study area. However, the number of workers and driver license holders in a household, household income, distance to transit and the area type of a household s residential location influence car ownership. Regarding household car ownership level change, the findings confirm the existence of state dependence where the change in car ownership level is strongly influenced by the existing number of cars. In addition, changes in household size, income, and residential location as well as the acquisition of a driver license by a household member influence household car ownership level change. While this study could not consider the effect of

6 Okrah, Moreno, Llorca, Moeckel shared mobility on car ownership for lack of data, it would be a valuable exercise to assess the impacts of such services on household car ownership level change as data becomes available. Modeling car ownership level changes in an integrated land-use/transport setup should enable the assessment of a wide range of impacts such as reduced parking costs, accident and emissions expected from the introduction of autonomous cars. Future research should focus on the implementation of autonomous car considerations in the modeling suite. Different scenarios of autonomous car penetration patterns and how they affect household car ownership vis-à-vis ownership by providers of shared mobility services should be explored.

7 Okrah, Moreno, Llorca, Moeckel 0 REFERENCES. Moeckel, R., and D. Yang. Car Ownership and the Impact of Built Environment, Demographics and Transport Systems. Journal of Civil Engineering and Architecture, 0. Clark, B., G. Lyons, and K. Chatterjee. Understanding the Process That Gives Rise to Household Car Ownership Level Changes. Journal of Transport Geography, 0. Dargay, J. M., and P. C. Vythoulkas. Estimation of a Dynamic Car Ownership Model: A Pseudo-Panel Approach. Journal of Transport Economics and Policy.. Volume, 0.. Clark, B., K. Chatterjee, and S. Melia. Changes in Level of Household Car Ownership : The Role of Life Events and Spatial Context. Transportation, 0, pp.. Simma, A., and K. Axhausen. Commitments and Modal Usage: Analysis of German and Dutch Panels. Transportation Research Record, 00.