Destination Choice Models

1
Destination Choice Models

• John Gliebe
• Parsons Brinckerhoff
• PB Consult, Inc.
• Portland, Oregon

2
Overview

• Trip distribution in 4-step models
• From gravity model to destination choice model
• Relationship to mode choice modeling
• Anatomy of a destination choice model
• Advantages of destination choice models
• Where destination choice models are used now

3
Why Destination Choice Models?

• More flexible extension of gravity models
• Utility-based interpretation
• Easier to market segment
• Get right workers to the right job locations
• Application is straightforward

4
Trip Generation

• Productions
• Rates estimated from expanded home interview
  surveys
• Usually cross-classification model of trips
  produced by households
• e.g. Size X Income X Autos
• Attractions
• Rates estimated from expanded home interview
  surveys
• e.g. Trip purpose X Attractor Type X Area Type
• Attractors (example)
• employees (retail, non-retail)
• students
• households

5
Trip Distribution in 4-step Model Systems the
Gravity Model
Production-constrained
Attraction-constrained
6
Friction Factors

• Derive a table from observed trip lengths
• or use an assumed impedance function

7
From Gravity Model to Destination Choice Model
8
Link to Mode Choice
Log sums Composite utility of all available
modes for i - j
9
Anatomy of a Destination Choice Model
Introduce market segments m
10
K-factors become interpretable
11
Advantages to Destination Choice Formulation

• Conceptually easier
• Simultaneous consideration of attractions and
  impedance
• Estimate with disaggregate data
• Transit service improvements reflected in log
  sums
• Easier to do market segmentation by auto
  ownership, and income level
• Consistent with Mode Choice logsums
• Interpretation of alternative-specific constants
  (replace k-factors) perceived impedance
• Get the rights jobs with the right income groups

12
Estimation and Calibration

• Estimate using disaggregate data (household
  interview survey)
• Sampling used to choose alternate zones
• Calibration--can use employment by zone (or
  district) to indicate proportion of HB work trips
  by district

13
Examples where applied

• Portland Metro (since 1980s)
• PB work
• Buffalo
• Thurston Co., WA
• Las Vegas
• Salt Lake City (ongoing)
• Los Angeles MTA (ongoing)
• Tour-based Models
• San Francisco CTA
• Columbus
• New York
• Oregon Statewide Model

14
Summary

• Destination Choice models offer a flexible
  alternative to Gravity model for trip
  distribution.
• Provide more informationmore sensitive to policy
  analysis
• Use readily available data
• Extendable to more sophisticated modeling tasks