Exploring the role of individual attitudes and perceptions in predicting the demand for cycling: a hybrid choice modelling approach

Maldonado-Hinarejos, Sivakumar, Polak, 2014, in Transportation

doi:10.1007/s11116-014-9551-4
Location London, England
Population General
Sample size 1985
Factor analysis type principal components, varimax rotation
Stepwise regression no
Removal of insignificant variables no
Reviewed by LCM

Abstract

Cycling is often promoted as a means of reducing urban congestion and improving health, social and environmental outcomes. However, the quantification of these potential benefits is not well established. This is due in part to practical difficulties in estimating cycling demand and a lack of sound methodologies to appraise cycling initiatives. In this paper we attempt to address this need by developing predictive models of cycle demand, relative to other transport modes, that capture not only the impacts of observed characteristics such as age and travel time but also the role of attitudes and perceptions. Using data from a stated preference survey, we estimate a hybrid choice model for cycle use that incorporates the role of attitudes towards cycling, perceptions of the image associated with cycling, and the stress arising from safety concerns. Model results indicate that the latent attitudes and perceptions explain an important part of the non-observable utility in a simple multinomial logit choice model. We also demonstrate policy analysis using the hybrid choice model, which allows comparisons of ‘hard’ policies such as the provision of parking facilities against ‘soft’ measures such as cycle promotion schemes. © 2014, Springer Science+Business Media New York.

Factors

Variable Pattern loading
Cycling is a safe way of getting about (Safety) 0.5
London is a city for cycling (Social norms) 0.6
Variable Pattern loading
Do you feel cyclists are law abiding? (Safety) 0.6
Do you feel cyclists are considerate? (Safety) 0.6
Do you feel cyclists are dangerous? (Safety) -0.6
Variable Pattern loading
Cycling is becoming much more popular (Other) 0.6
Cycling is a method of transport I would be happy to be seen using (Social norms) 0.7
Cycling is a convenient way of getting about (Convenience) 0.7
Cycling provides a good value for money (Cost) 0.6
Cycling is for everybody (Social norms) 0.6
Cycling is good for rush hour journeys in London (Other) 0.6
Cycling is a good way to get fit (Lifestyle) 0.6
Cycling makes a difference to improving the environment (Environmentalism) 0.7
Cycling is a family activity (Social norms) 0.6
It is important for children to learn to cycle (Social norms) 0.5
Do you feel cyclists are fit? (Other) 0.4
Do you feel cyclists are like you? (Social norms) 0.6
Variable Pattern loading
Cycling is an unpleasant experience (Other) 0.6
Cycling is stressful (Comfort) 0.7

Models

Dependent variable Mode choice
Model type Multinomial logit
Sample size 15872.0
R2 0.456
Adjusted R2
Pseudo R2 (nan) nan
AIC nan
BIC nan
Log-likelihood at zero -11001.0
Log-likelihood at constants nan
Log-likelihood at convergence -5984.0
Car
Variable Coefficient p-value
Choice constant: car 1.61 0.0
Car time -0.02 0.0
Cycle
Variable Coefficient p-value
Choice constant: cycle 0.0 nan
No cycle lane 0.0 nan
Bus lane 0.0 1.0
Cycle lane -0.003 0.968
Segregated cycle lane 0.0 1.0
High volume traffic flow 0.0 nan
Low volume traffic flow -0.001 0.976
No parking facilities 0.0 nan
Simple parking facilities 0.191 0.001
Parking in a locked compound 0.364 0.0
Cycle time -0.032 0.0
Age 18-34 -1.01 0.0
Age 35-44 -0.627 0.0
Age 45-64 -0.171 0.003
Age 65+ 0.0 nan
Female 0.86 0.0
Male 0.0 nan
White British 0.39 0.0
Other ethnicity 0.0 nan
Outer residential area 0.142 0.006
Inner residential area 0.0 nan
PT
Variable Coefficient p-value
Choice constant: public transport 1.41 0.0
Public transport time -0.017 0.0
Walking
Variable Coefficient p-value
Choice constant: walking 1.43 0.0
Walking time -0.002 0.603
Dependent variable Mode choice
Model type Multinomial logit
Sample size 15872.0
R2 0.458
Adjusted R2
Pseudo R2 (nan) nan
AIC nan
BIC nan
Log-likelihood at zero -11001.0
Log-likelihood at constants nan
Log-likelihood at convergence -5965.0
Car
Variable Coefficient p-value
Choice constant: car 1.45 0.0
Car time -0.013 0.352
Random error (car time) 0.053 0.001
Cycle
Variable Coefficient p-value
Choice constant: cycle 0.0 nan
No cycle lane 0.0 nan
Bus lane 0.002 0.984
Cycle lane 0.006 0.936
Segregated cycle lane 0.001 0.992
High volume traffic flow 0.0 nan
Low volume traffic flow 0.001 0.984
No parking facilities 0.0 nan
Simple parking facilities 0.23 0.001
Parking in a locked compound 0.434 0.0
Cycle time -0.061 0.0
Age 18-34 -1.17 0.0
Age 35-44 -0.707 0.0
Age 45-64 -0.168 0.013
Age 65+ 0.0 nan
Female 0.96 0.0
Male 0.0 nan
White British 0.46 0.0
Other ethnicity 0.0 nan
Outer residential area 0.116 0.049
Inner residential area 0.0 nan
Random error (cycle time) 0.036 0.0
PT
Variable Coefficient p-value
Choice constant: public transport 1.19 0.0
Public transport time -0.022 0.0
Random error (PT time) 0.002 0.49
Walking
Variable Coefficient p-value
Choice constant: walking 1.28 0.0
Walking time -0.003 0.704
Random error (walking time) 0.012 0.194
Dependent variable Mode choice
Model type Multinomial logit
Sample size 15872.0
R2 0.506
Adjusted R2
Pseudo R2 (nan) nan
AIC nan
BIC nan
Log-likelihood at zero -11001.0
Log-likelihood at constants nan
Log-likelihood at convergence -5429.0
Car
Variable Coefficient p-value
Choice constant: car 1.74 0.0
Car time 0.001 0.944
Random error (car time) 0.065 0.0
Cycle
Variable Coefficient p-value
Choice constant: cycle 0.0 nan
No cycle lane 0.0 nan
Bus lane 0.0 1.0
Cycle lane 0.005 0.952
Segregated cycle lane 0.001 0.992
High volume traffic flow 0.0 nan
Low volume traffic flow 0.001 0.976
No parking facilities 0.0 nan
Simple parking facilities 0.23 0.001
Parking in a locked compound 0.436 0.0
Cycle time -0.043 0.0
Age 18-34 -0.869 0.0
Age 35-44 -0.593 0.0
Age 45-64 -0.233 0.001
Age 65+ 0.0 nan
Female 0.84 0.0
Male 0.0 nan
White British 0.22 0.0
Other ethnicity 0.0 nan
Outer residential area 0.164 0.009
Inner residential area 0.0 nan
Pro-bike -0.25 0.0
Context -0.85 0.0
Image -0.3 0.0
Stress 0.4 0.0
Random error (cycle time) 0.017 0.013
PT
Variable Coefficient p-value
Choice constant: public transport 1.66 0.0
Public transport time -0.016 0.0
Random error (PT time) 0.016 0.003
Walking
Variable Coefficient p-value
Choice constant: walking 1.16 0.0
Walking time 0.033 0.009
Random error (walking time) 0.039 0.0

The Attitudes and Travel Database is produced with support from the Center for Teaching Old Models New Tricks at Arizona State University, a University Transportation Center sponsored by the US Department of Transportation through Grant No. 69A3551747116.

sha256:a08d9e369743bf7e6d1c40d27347318209b40a7fb1543813fdcf31b898918815