Stated Preference (SP) survey design

Motivation

• We don't have information of new/innovative transportation service.
• We need to estimate travel demand for new service (e.g. Flexible Transit) before introducing.

Example of choice experiment scenario (Frei et al., 2017)

Elements of SP survey

reference: https://www.b.kobe-u.ac.jp/~sanko/pub/Sanko2001_1.pdf

1. Available alternatives
   • Transit
   • Car
   • New service (On-demand transit feeder system)
2. Attributes
   • travel cost
   • in-vehicle travel time
   • walking time
   • waiting time
   • number of transfers
3. Attributes level
   • level should be selected within reasonable boundary (to reduce hypothetical bias).
   • Normally, vary from 2-5 level.
   • In the case of 5 level, -20%, -10%, 0(reference), +10%, +20%.

Example of attribute level (Frei et al., 2017)

Survey design

1. Orthogonal design (Use Excel or any programming language)
   • The basic way for survey design.
   • When there are $m$ attributes, each of which has $n$ levels, we need $n^m$ questionnaire.
   • When $m$ and $n$ get large, we need to recruit too many respondents.
2. D-efficiency design (choiceDes package in R)
   • reference : https://www.itl.nist.gov/div898/handbook/pri/section5/pri521.htm
   • Basic idea is to reduce the number of scenario while maintaining similar variance.

Discrete Choice Modeling

Applications

• Understand the preference for level of service attributes (e.g. travel time, number of transfers, etc) of transport mode.
• Understand how the preference would be heterogeneous depending on sociodemographics (e.g., age, income, etc).

Multinomial logit model