Everyday travel mode choice and its determinants: trip attributes and lifestyle

Everyday travel mode choice and its determinants: trip attributes and lifestyle Markéta Braun Kohlová marketa.braun.kohlova@czp.cuni.cz Faculty of Social Sciences and Charles University Environment Center Charles University in Prague Czech Republic

Topic: travel mode choice • Estimates of a discrete choice model • Latent variables capturing the differences in individual preferences of residence and access to downtown and amenities (traditional city-like versus suburban) • Based on the results of a qualitative survey on residential choices and its subsequent effects on accessibility, travel time and travel costs.

Presentation structure • Formulation of behavioral hypotheses on the effects of latent attitudinal characteristics on travel mode choice • Data description • Statistical model of travel mode choice • Latent variables constructs and the hypothesis test • Model estimates and result discussion • Key findings summary and outlines possible directions of future research

1.1 Behavioral hypotheses: effects of latent attitudinalchar.on travel mode choice A guiding idea: the incorporation of psychological or psychometric factors leads to a more behaviorally realistic representation of the choice process, and consequently, better explanatory power The choice model with latent variables allows inferring how preferences regarding place ofresidence and access impact travel mode choice

1.2 Qualitative survey Meaning of accessibility in decisions about residential relocation • the travel mode choice affected by the individual or family lifestyles comprising of housing and neighborhood preferences Typology: Great importance of a good access = Good PT connection The access does not matter because they „use car anyway“ Succeeded to meet the claim on good access Did not succeed to meet the claim on good access Expected travel time, travel cost and PT travel choice Car use as „reluctunt drivers or car passengers“

2. Data Data collection: • Spring 2008 • 7 Czech cities and their suburban areas (10 000 – 1 million) • 1723 adult individuals - analysis: 1438 • Quota sampling: gender, age, educational level, economic activity incl.retiree and residential locality • At least one trip during the previous working day and within the respective urban area • Professional agency • One-on-one interviews • In respondents household • Paper questionnaires Revealed preferences • The first trip and the following trips chain • Trips of different purposes (not only work trips) • Trips within the respective urban area (radius of public transport service or 25 kms) The total choice set (12 options): • Car as driver, car as passenger, motorcycle, tram, trolleybus, bus, subway, train, regional bus, bicycle, walk, other • The available options varied across localities • For this analysis six public transport modes were included into a common alternative (public transport)

3.1 Model formulation H0: people with different values of preference factors (not directly identifiable from the data) exhibit different travel behavior ►a sequential (two-stage) estimation method combining factor analysis and random utility choice model ► latent factors in the random utility choice model - multinomial logit (MNL) model

3.2 Choice probability: • multinomial logit (MNL) model:

3.3 The choice set: • The universal choice set : 1. car as driver (CAR) 2. car as passenger (PASS) 3. bicycle (BIKE) 4. walk (WALK) and 5. public transport (TRANS) Not all alternatives available for each individual The following rules were used, defined as follows: 1. Anybody without a driver’s license cannot drive alone. 2. Anybody in a household without a car cannot drive alone. 3. Anybody in a household without a bicycle cannot use bicycle.

3.4 Utility functions

4.1 Latent variables • Based on the qualitative survey: • desired characteristics of neighborhoods and • desired public transport accessibility Indicators: • five-point Likert-type scale I am willing to travel 15 minutes longer a trip to live in green, quiet locality and realize my hobbies. I like to live within walking distance to shops, restaurants and other amenities. I enjoy the hustle and bustle of the city. I consider very important the public transport access of my neighborhood from the city center. I consider very important the public transport access to frequently visited destinations of mine.

4.2 Rotated component Matrix (factor loadings)

4.3 Hypothesis testing Does the inclusion of latent variables capturing access and city life preferences into the travel mode choice model increase its explanatory power? Likelihood ratio test: (Ben-Akiva and Lerman, 1985) H0: βACCESS = βCITYLIFE = 0 -2 (LR – LU) χ2 with df = KU – KR - 2 (- 990.178 + 963.787) = 9.392 > 5.99 ►H0rejected (α = 0.05)

5.1 Travel mode choice: Multinomial logit (MNL) with latent variables

5.2 Multinomial logit model:

6. Summary: • Lifestyle preferences exist • They are determinants of travel mode choice • A more behaviorally realistic representation of the choice process and better explanatory power • Better understanding ofthe development of urban areas and the impacts that changes in urban form will have on modal split, environment and public health • Limitation of policies aiming at behavior modification by the large proportion of households who have strong preferences towards suburban and car oriented lifestyle Future extension: • Simultaneous estimate of latent variables scores or classes and choice model • To match the latent neighborhood and access preferenceswith the actual neighborhood characteristics

Acknowledgements This research has been supported by National Czech Foundation GAČR No. 403/08/1694, Application of the model of environmentally significant behavior in the Czech Republic and the by the Czech Ministry of Education Youth and Sports R&D project 2D06029 VaV/320/1/03 Analysis of distribution and social effects of sector policies funded This support is gratefully acknowledged.

Everyday travel mode choice and its determinants: trip attributes and lifestyle