Objectives: As response rates have been falling, we need to
analyze the places in a multi-stage survey where we have been consistently
losing respondents along the way. In the first stage of the traditional
household travel survey, the household is contacted and information about the
members of the household, such as age, sex, relationship, worker status, is
obtained. In the second stage of the survey we send a diary for each household
member to use to track his or her travel on the assigned travel day, and call
and collect the information about the travel. We often lose respondents at this
second stage, either as outright refusals, non-contacts (generally passive
refusals), and because some but not enough of the people have completed the
travel portion of the survey.
As part of this research proposal, the second-stage sample
of people within the household would be accomplished by randomly selecting one
or more adults in each household. The objective would include developing
selection criteria that would replicate the data collected from a fully
reporting household for the key travel demand variables, i.e. person trips and
vehicle trips per household by mode and purpose.
Some of the issues in developing a sample scheme for
selecting the person or persons in each household would include:
● How does household composition (life cycle) effect the
probability of selection? A person with the same age and sex living in either a
dual-earner household with no children or a dual-earner household with two
children have different amounts and types of trips.
● How does the number of vehicles per driver or worker
effect the probability of selection? A two worker household with two vehicles
or more will have different types and amounts of travel than a two worker
family with one car, or no vehicles at all. Important variables, such as auto
occupancy, transit use, and time of day of travel could be different in
households with different levels of vehicle ownership.
● How do the household composition and vehicle
availability interact? Do two earner households with no children and one
vehicle require different probability of selection than two earner households
with two cars, or with one car and two children?
● Are the land use, socioeconomic, and travel
relationships developed from these “synthesized” households statistically
significant or statistically different from those relationships derived from
more fully observed households?
One approach to be considered is one in which there is a
side by side field test with either the 50 percent rule of household completion
or a rule requiring all household members to be accounted for on one side, and
a sample person selection process using the scheme developed in stage one on the
The major questions from the field test would be:
● Are resources saved and how big is the savings?
● Does the lower burden per household increase response
(lower loss at the second stage?)
● Are the data improved (biases reduced, etc) by
targeting a specific person in each household rather than allowing whoever can
to complete (especially in larger households)?
● How does proxy reports affect the data? Are there more
or less proxy reports?
● How well does this method represent non-vehicle travel?
● Can this method fully represent the within household
dynamics researchers require?