Norman W. Garrick Mode Choice The introduction of congestion charging in London in 2003 is one example of a situation where mode choice modeling is needed.

Norman W. Garrick

Mode Choice

The introduction of congestion charging in London in 2003 is one example of a situation where mode choice modeling is needed.

The fundamental question is this case was “Will charging work to reduce congestion?”

The expectation was that people would switch modes – mode choice modes would be used to predict how many people would switch

Norman W. Garrick

Mode Choice

The introduction of the new AVE service between Madrid and Barcelona is another example of a situation were model choice modeling would be needed.

How many people would switch from plane or driving to train?

http://www.renfe.es/video.html

Norman W. Garrick

Factors affecting Mode Choice

In lecture 2 we talked about some of the factors affecting mode choice. These factors include

1. Type and purpose of trip 2. Ownership status 3. Cost (mostly out of pocket cost) 4. Door-to-door travel time 5. Convenience/service/comfort 6. Prestige 7. Availability 8. Accessibility of mode 9. Land use characteristics of start and end point

Obviously, not all of these factors can be effectively incorporated into a quantitative model of mode choice. We need to be cognizant of those factors that are important in influencing choice but are not fully accounted for in mode choice models.

Norman W. Garrick

Mode Split or Modal Choice Models

Mode Choice Models are used to try to predict travelers mode choice

Contemporary models are based on using UTILITY or DISUTILITY functions

These functions are meant to express the level of satisfaction (for utility functions) or dissatisfaction (for disutility functions) with a given mode

One the utility function is calculated for each mode, the probability that a given mode will be chosen can be calculated

Norman W. Garrick

Utility Function

A utility function takes the following form

uk = ak + a1 X1 + a2 X2 + ….. ar Xr + ε0

uk – utility function for mode k

ak – modal constant

Xr – variables measuring modal attributes such as cost or time of travel

ar – coefficient associated with each attribute

Where

ε0 – error term

Norman W. Garrick

Multinomial Logit Model

If utility function, uk, is assumed to be a Weibull Probability Distribution then the Multinomial Logit Model is used to calculate the probability that a traveler will chose a given mode

Multinomial Logit Model

p(k) = euk / Σ eux

Norman W. Garrick

Example

The mode available between Zone I and J are i) Passenger car (A), ii) Bus (B)

Find the market share for each mode given the following attribute table for the modes.

The utility function is

uk = ak – 0.025 X1 – 0.032 X2 – 0.015 X3– 0.002 X4

wherex1 – access plus egress time (min)x2 – waiting time (min)x3 – line haul time (min)x4 – out of pocket cost (cents)

Norman W. Garrick

Example … continue

The attribute table for each mode is given below

And aa = -0.10ab = +0.00

Therefore

u(A) = -0.625u(B) = -1.530

Probability of selecting PC, p(A) = e(-0.625) / [e(-0.625)+e(-1.530)] = 0.71

x1 x2 x3 x4

PC 5 0 20 100

Bus 10 15 40 50