2006 - Paez Et Al - Elderly Mobility Demographic and Spatial Analysis of Trip Making in The Hamilton CMA

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 33

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/240638471

Elderly Mobility: Demographic and Spatial Analysis of Trip Making in the


Hamilton CMA

Article in Urban Studies · January 2007

CITATIONS READS

130 182

5 authors, including:

Antonio Páez Darren M Scott


McMaster University McMaster University
234 PUBLICATIONS 9,371 CITATIONS 130 PUBLICATIONS 6,444 CITATIONS

SEE PROFILE SEE PROFILE

Dimitris Potoglou Bruce Newbold


Cardiff University McMaster University
84 PUBLICATIONS 4,392 CITATIONS 249 PUBLICATIONS 8,648 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Bruce Newbold on 20 December 2013.

The user has requested enhancement of the downloaded file.


Elderly Mobility:
Demographic and Spatial
Analysis of Trip Making in
the Hamilton CMA

A. Páez, D.M. Scott, D. Potoglou, P.S. Kanaroglou,


K.B. Newbold

014
January 2006
Elderly Mobility: Demographic and Spatial Analysis of Trip Making in the Hamilton CMA
A. Páez, D.M. Scott, D. Potoglou, P.S. Kanaroglou, K.B. Newbold

Abstract
Recent interest in the urban transportation challenges posed by the demographic
outlook of aging societies has prompted a growing body of scholarship on the
subject. The focus of this paper is on the topic of elderly trip generation and the
development of models to help formalize some important relationships between
trip making behavior, and personal, household, and contextual variables (e.g.
location). The case study is the Hamilton Metropolitan Area – an important
functional component of Greater Toronto, itself one of the regions in Canada
where the impact of aging is expected to be most strongly felt. Using data from
Toronto’s Transportation Tomorrow Survey and mixed ordered probit models,
we investigate the question of spatial and demographic variability in trip making
behavior. Our results support the proposition that trip making propensity
decreases with age. However, we also find that this behavior is not spatially
homogeneous, and in fact exhibits a large degree of variability – a finding that
highlights both the challenges of planning transportation for the elderly, and the
potential of spatial analytical approaches to improve transport modeling practice.

Keywords: Trip generation, demographic change, spatial analysis, mixed


ordered probit model, Transportation Tomorrow Survey

Short abstract: Interest in the transportation challenges posed by aging


societies has prompted a growing body of scholarship on the subject. In this paper
we focus on elderly trip generation and the development of models of trip making
behavior. The case study is the Hamilton Metropolitan Area, an important
functional component of Greater Toronto, itself one of the regions in Canada
where the impact of aging is expected to be most strongly felt. Using data from
Toronto’s travel survey and mixed ordered probit models, we investigate the
question of spatial and demographic variability in trip making behavior. Our
results indicate that trip making propensity decreases with age. However, we also
find that this behavior is not spatially homogeneous, and in fact exhibits a large
degree of variability. This finding highlights both the challenges of planning
transportation for the elderly, and the potential of spatial analytical approaches to
improve transport modeling practice.

1. Introduction
Canada, like most countries in the industrialized world, is an aging society that will face
significant increases in both the number and proportion of older people in the near future.
The median age of the Canadian population is now 35 years, and will reach 45 years
within the next three decades. It is estimated that by 2021 about 1 in 5 Canadians will be 65
years or older – a significant shift from the current situation. While the health care and
fiscal implications of this demographic trend are well recognized and have thus tended to
dominate the attention of the public, the transportation implications of aging societies are
also far reaching and have not gone unnoticed by the transportation research community.
At the same time that the aging situation has been recognized as one of the critical
challenges for transportation systems (Pisarski, 2003), research to date suggests that the
elderly in developed countries have become relatively more mobile, with a tendency to

Centre for Spatial Analysis - Working Paper Series


-1-
Elderly Mobility: Demographic and Spatial Analysis of Trip Making in the Hamilton CMA
A. Páez, D.M. Scott, D. Potoglou, P.S. Kanaroglou, K.B. Newbold

make more and more varied trips than previous elderly cohorts. Concurrently, they have
become more automobile dependent with less public transportation use (Rosenbloom,
2001).

Recent research suggests that these international trends in travel behavior are to some
extent shared by Canada (Newbold et al, 2005). In due course they are, therefore, expected
to have significant impacts on Canada’s transportation systems (e.g. traffic congestion),
urban environments (e.g. air pollution), and the ability of transportation systems and
service providers to meet the mobility needs of the elderly (a critical aspect of their quality
of life). Despite the clear economic and social importance of this array of issues, however,
our understanding of trip making by the elderly remains for the most part descriptive, and
tools to project their future transportation needs are sorely missing. The Urban
Transportation Modeling System (UTMS; Meyer and Miller, 2001), the most commonly
used tool for assessing urban travel in Canada, is still largely dominated by a focus on the
evaluation of work-related peak-period travel and ignores the impact of older drivers on
traffic congestion and air pollution.

Our objectives in this paper are twofold. First, we aim at contributing towards a better
understanding of these emerging demographic and travel trends. And secondly, we
demonstrate the use of spatial models and their potential for assessing travel demand (q.v.
Miller, 1999; Páez and Scott, 2004). In pursuit of these objectives, we concentrate on the
topic of trip generation analysis. A series of frequency-based trip generation models are
estimated for total trips, and two different trip purposes (i.e. work and non-work trips),
using ordered probit and mixed ordered probit models. Use of these models, and in
particular the mixed ordered probit, allows us to explore the hypothesis of geographical
and demographic variations in trip making behavior. In other words, in this paper we
address two research questions: are there age-related differences in trip generation rates,
and the factors affecting these rates? And, are these differences dependent on geographical
context? From an applied perspective, in addition, the models reported here are the first-
step in the UTMS, and are therefore a necessary element to project travel demand needs in
a region. Eventually, the models will be incorporated into a GIS-based decision-support
system, which will simulate travel demand, traffic congestion and emissions for both work
and non-work trips for elderly and adults over a day period (see Maoh et al, 2005). The
case study is the Census Metropolitan Area (CMA) of Hamilton, Canada. Data for analysis
are drawn from the Transportation Tomorrow Survey, a comprehensive trip diary-based
travel survey conducted every five years in the Greater Toronto Area, comprising the city
of Toronto and surrounding regions in south-western Ontario.

The remainder of the paper is organized as follows. Section 2 briefly reviews the literature
on elderly travel behavior. Section 3 gives an overview of trip generation analysis and
introduces the methods and related spatial analytical issues. Section 4 describes the study
area and data sources. The results of our analyses are presented and discussed in Section

Centre for Spatial Analysis - Working Paper Series


-2-
Elderly Mobility: Demographic and Spatial Analysis of Trip Making in the Hamilton CMA
A. Páez, D.M. Scott, D. Potoglou, P.S. Kanaroglou, K.B. Newbold

5. Finally, we summarize the main findings, and provide some concluding remarks in
Section 6, including directions for further research.

2. Elderly Travel Behavior


The travel behavior of the elderly, once an area of inquiry that took a back seat to the
development of advanced modeling techniques, or the study of gender, ethnicity, income,
and a host of other issues in travel behavior research, has seen a remarkable pace of
growth in the last decade, but most particularly in the last 5 years or so. This development
has been concurrent with, and perhaps at least in part encouraged by, an increased
awareness of the social aspects of transportation planning and research, including social
inclusion and well-being issues, to which the elderly appear to be particularly vulnerable
(e.g. Cvitkovich and Wister, 2001; Harrison and Ragland, 2003).

As part of a growing body of literature, previous research has studied various facets of the
travel behavior of the elderly. Some key papers have been relatively general in their
investigation of the overall mobility aspects of the aging situation. They include for
example Tacken’s (1998) exploration of elderly mobility in the Netherlands, Alsnih and
Hensher’s (2003) review of the evidence and reflections on the mobility and accessibility
expectations of the seniors, and Newbold and colleagues’ (2005) study of the travel
behavior of Canada’s older population. Consideration has been given as well to more
particular aspects of elderly transportation issues, as for example in Smith and Sylvestre’s
(2001) study of trip making behavior among the suburban elderly in Winnipeg. The
anticipated special mobility needs of the elderly have also led to an interest in the modal
choice of this population segment, as for example in Stern’s (1993) analysis of the
transportation choices of the elderly and disabled in rural Virginia, and Franklin and
Niemeier’s (1998) investigation of mode choice with respect to fixed-route transit and
paratransit services. Furthermore, the potentially serious repercussions of increased
automobile dependency on the one hand, and of driving cessation on the other, have
prompted researchers to pay attention as well to the sustainability dimensions of elderly
automobility (Rosenbloom, 2001), the spatial and racial aspects of elderly automobility
(Waldorf, 2003), and the access to private car by the elderly (Gardiner and Hill, 1996). As a
recognition that the knowledge generated by this body of research will be relevant only in
the measure that it influences planning and policy practice, other studies have
concentrated on the development of planning support tools, such as Hildebrand’s (2003)
system for the analysis of elderly travel demand from an activity perspective, or the
transport policy implications, as in Mercado and colleagues’ (2006) international
comparative survey of transport policy responses in aging societies.

Taken as a body of research, the studies cited above make a persuasive case for the
relevance, and indeed the urgency, of achieving a better understanding of the travel needs
and behavior of the elderly, for a variety of compelling scholarly, planning, and social and
economic policy reasons. Much of the existing work, however, has remained conceptual or
descriptive in nature, an opinion voiced by Stern (1993), and more recently also by Kim

Centre for Spatial Analysis - Working Paper Series


-3-
Elderly Mobility: Demographic and Spatial Analysis of Trip Making in the Hamilton CMA
A. Páez, D.M. Scott, D. Potoglou, P.S. Kanaroglou, K.B. Newbold

(2003), who notes that few systematic studies of the mobility of the elderly have been
published. There is, it seems, a relative paucity of studies to formalize, by means of
multivariate analysis methods, some of the hypotheses suggested by the descriptive
studies of Tacken, Rosenbloom, Alsnih and Hensher, and others. This state of affairs, it
appears now, is being redressed with some of the research previously mentioned
(including Smith and Sylvestre, 2001), in addition to the work of, among others, Kim
(2003), Kim and Ulfarsson (2004), and Schmöcker et al. (2005a; 2005b). Some of these
papers are particularly relevant to the research reported here, as they document important
aspects of elderly trip making behavior, in particular a consistent finding that trip making
frequency decreases with age, but increases with license holding status, auto ownership,
and other variables that will become increasingly relevant as the pre-retirement
population segment enters old age to become the new and future elderly (q.v.
Rosenbloom, 2001; Newbold et al, 2005).

3. Trip Generation Analysis


3.1 Trip generation and the ordered probit model
As noted above, this paper is concerned with the trip generation of the elderly.
Conceptually, the study of trip generation can be thought of as trying to answer the
question of how many trips will each individual in a sample (or population) make. This
type of analysis is important for the insights that it can afford into the travel behavior of
individuals. It is also valuable as a tool for projecting the travel demand levels of a
city/region, and for assessing the need for transportation infrastructure and services.
Conventionally, trip generation analysis is the first step in the Urban Transportation
Modelling System widely used in practical transportation planning applications (for in-
depth discussions of this system see Ortúzar and Willumsen, 2001; Meyer and Miller, 2001;
Miller and Shaw, 2001). This step is followed in a sequential fashion by the analysis of trip
distribution (where are those trips going), modal split (how many trips are made by each
transportation mode) and assignment (route selection). Trip generation, being the initial
step in this modeling approach, provides the benchmark level of travel demand in a
system.

A number of methods have been developed over the years to study trip generation. These
include the use of simple or multiple linear regression analysis at the aggregate level
(zonal totals or rates), or even at the household level (e.g. Badoe and Chen, 2004). Models
of this type are used to identify variables that highly correlate with the number of trips
originating at the unit of analysis, based on variables that can be either of policy
significance (e.g. land uses), or of predictive interest (e.g. household structure, age). An
advantage of multiple regression analysis is its status as a mature technique whose
properties and characteristics are well understood (Washington, 2000). Linear regression
techniques, moreover, are relatively simple and widely known – a fact that explains their
continued use as state-of-the-practice tools in many applications. Their relative simplicity,
on the other hand, comes at a price, and the disadvantages of linear regression analysis for
trip generation have been noted before by Agyemang-Duah and Hall (1997), Boarnet and

Centre for Spatial Analysis - Working Paper Series


-4-
Elderly Mobility: Demographic and Spatial Analysis of Trip Making in the Hamilton CMA
A. Páez, D.M. Scott, D. Potoglou, P.S. Kanaroglou, K.B. Newbold

Sarmiento (1998), Schmöcker et al. (2005a), and others. Among its various limitations, it is
possible to note the lack of built-in upper and lower limits to the number of trips, a
situation that could potentially lead to unreasonable predictions as the model’s covariates
increase, or could result in negative number of trips when the covariate values are
relatively low. Conceptually, moreover, linear regression deals with the outcome variable
(number of trips) as a continuous variable. The assumption that the number of trips is
approximately continuous can be questioned – in particular when typical values for
number of trips are relatively low (most individuals make just a few, say two or three trips
in a day). Regression analysis using zonal averages or even household rates makes this
assumption more palatable, but on the other hand complicates the use of individual level
variables by assuming a typical (i.e. average) household, or the household as the trip
generating unit. Finally, the link between number of trips and covariates in a regression,
while it may be based on hypothetical ideas about the process of trip generation, lacks a
behavioral justification such as supported by the theory of random utility (e.g. Ben-Akiva
and Lerman, 1985).

Given the limitations of linear regression for trip generation analysis applications, it is not
surprising the many recent studies have tended to prefer alternative modeling approaches
that redress some of the above shortcomings, in particular the use of frequency-based
choice analysis at the individual (i.e. disaggregate) level (e.g. Boarnet and Sarmiento, 1998;
Schmöcker et al, 2005a; also see the analysis of relative desired commute by Redmond and
Mokhtarian, 2001). Frequency-based models include the ordered probit/logit models used
when the outcome variable is sequential, as is the case of number of trips or trip frequency
classes (the difference between probit and logit resides in the choice of distribution –
normal or extreme value respectively). As discussed by Train (2003), individuals facing
sequential decision processes (i.e. should I make one more trip?), can be thought of as
having some level of utility or opinion associated with the different possible alternatives.
In the particular case of trip making, an individual will derive utility from traveling as
follows (note that the thresholds are estimable from the data, and that there can be an
arbitrary number of trip making classes):
• If the utility is below a given threshold λ1 (i.e. if U < λ1 ), the individual does not
travel (i.e. number of trips is zero)
• If the utility is between λ1 and λ2 (i.e. if λ1 < U < λ2 ), the individual makes 1 trip

• If λ2 < U < λ3 , the individual makes 2 trips

• If U > λ3 , the individual makes 3 or more trips (see Figure 1)

The utility is in principle unobservable. However, the analyst can measure the outcome of
the decision making process (i.e. the number of trips made), and typically also other
variables that relate to the level of utility. The utility of individual i is decomposed into the
usual systematic and random components:

Centre for Spatial Analysis - Working Paper Series


-5-
Elderly Mobility: Demographic and Spatial Analysis of Trip Making in the Hamilton CMA
A. Páez, D.M. Scott, D. Potoglou, P.S. Kanaroglou, K.B. Newbold

Ui = Xi β + εi (1)

X i above is a vector ( 1× k ) of individual attributes, and β is a set of k coefficients that


map changes in X i to changes in the utility, and therefore the probability of making trips.
The term ε i is a random deviation from the systematic utility that captures
unobserved/unobservable factors, measurement/observational errors, and other
idiosyncratic variations (Ortúzar and Willumsen, 2001; p. 223). The utility in (1) can be
used to derive expressions for the individual’s trip making frequency probability. For
example, the probability of making 0 trips is:

Pr ( 0 ⋅ trips ) = Pr (U i < λ1 )
(2)
= Pr ( X i β + ε i < λ1 ) = Pr ( ε i < λ1 − X i β )

while the probability of making 1 trip is:

Pr (1 ⋅ trip ) = Pr ( λ1 < U i < λ2 )


= Pr ( λ1 < X i β + ε i < λ2 ) = Pr ( λ1 − X i β < ε i < λ2 − X i β ) (3)
= Pr ( ε i < λ2 − X i β ) − Pr ( ε i < λ1 − X i β )

The probabilities of observing other outcomes are derived in similar fashion. Assuming
that the random terms ε i follow the standard normal distribution, the ordered probit
model is obtained. Estimation of this model is discussed at length in Maddala (1983).

The ordered probit model, by treating the number of trips (or the trip frequency) as a set of
mutually exclusive and collectively exhaustive ordinal categorical variables, incorporates
built-in upper and lower limits. In addition, the model provides a behavioral framework
that directly links the number of trips to utility-based consumer and decision making
theory.

One final consideration is the use of spatial data, and the potential for incorporating
spatial effects into our transport modeling framework (see for example the discussion in
Miller, 1999; and the review of methods and applications in Páez and Scott, 2004). A major
analytical issue that should be considered when modeling spatial data is spatial
heterogeneity (e.g. Anselin and Griffith, 1988; McMillen, 1992). This effect is of particular
interest for the objective of the present paper, given our interest in learning whether trip
making behavior exhibits variation across space. In addition to the technical issues arising
from the use of spatially heterogeneous data (see Griffith and Layne, 1999; pp. 71-73),
renewed interest in the role of context and locality in spatial and transportation analysis
has led to a re-examination of spatial heterogeneity and techniques to incorporate this
effect (e.g. Bhat and Zhao, 2002).

Centre for Spatial Analysis - Working Paper Series


-6-
Elderly Mobility: Demographic and Spatial Analysis of Trip Making in the Hamilton CMA
A. Páez, D.M. Scott, D. Potoglou, P.S. Kanaroglou, K.B. Newbold

3.2 Spatial heterogeneity and the mixed ordered probit model


In statistical terms, spatial heterogeneity can be represented as systematic variation in the
mean of the process (spatial parametric drift), or alternatively as structural variation in the
definition of the variance. A mixed ordered probit is described in this section that belongs
to the latter class.

The mixed ordered probit model is part of a family of models alternatively known as
random coefficients, variance components, multilevel, or hierarchical models (for reviews,
see Jones, 1991; Duncan and Jones, 2000). The models of this family are characterized by
their ability to accommodate random variation of the coefficients, which makes them
suitable for exploring spatial variations in individual trip rates, and in particular the
relationships between these rates and the factors that influence them (e.g. location and
age). In our case, the term mixed is appropriate, in addition to its original meaning of
mixing distributions, because it allows for a mixture of variables at different levels of
geography. The ordered probit model can be generalized within a mixed framework, by
redefining the utility in equation (1), to incorporate fixed and random coefficients as
follows:

U iz = Yizα z + X i β + ε i (4)

The subscripts in the utility in (4) above denote decision maker i located at zone z. Vector
Yiz includes m variables at the individual or zonal level that are associated with a vector of
zone-specific coefficients α z (including a constant). Each of m coefficients in this vector
can be expanded in the following way:

α zm = α m + µ zm (5)

The two components of the coefficient are a region-wide (i.e. spatially constant) mean α m ,
and a zone-specific random term µ zm that quantifies the degree of dispersion or variation
around this mean. In what follows we assume that the random terms in each of the
coefficients are independent and normally distributed with constant standard deviation
σ m . It is worth noting that the use of other distributions (e.g. uniform, log-normal, etc.), or
the use of non-constant variance for these random terms, is also possible (for example, see
Bhat and Zhao, 2002). Other terms in equation (4) are as previously defined. In this way,
the model incorporates the typical fixed coefficients (constant throughout the population)
and coefficients that incorporate spatial variability (that is, they are specific to each zone).

Derivation of the mixed ordered probit model follows the same logic described in
subsection 3.1 with respect to the conventional ordered probit model. A crucial difference
with the latter model is that each random element introduced in the specification of the
utility will increase by one the dimensionality of the integrals that need to be evaluated for
estimation of the coefficients. Despite the intuitive appeal of these models, an erstwhile

Centre for Spatial Analysis - Working Paper Series


-7-
Elderly Mobility: Demographic and Spatial Analysis of Trip Making in the Hamilton CMA
A. Páez, D.M. Scott, D. Potoglou, P.S. Kanaroglou, K.B. Newbold

constraint to their application was the complexity of the estimation procedures. Presently,
Hedeker and Gibbons (1994) have developed methods for estimating the mixed ordered
probit model that use numerical quadrature techniques. As an alternative to this approach,
Train (2003) provides an excellent discussion of simulation techniques, whereby random
numbers are generated to obtain a simulated log-likelihood function (the average of many
such simulations) that can be maximized to obtain estimates. Simulation is now routinely
used in the estimation of mixed multinomial logit models, and is also applicable to other
mixed models including the ordered probit and logit models (see Train, 2003; p. 186).

4. Case Study: Study Area and Data


Our case study is centered on the Hamilton Census Metropolitan Area (CMA) of Canada.
Hamilton is located on the west shore of Lake Ontario, approximately 70 km southwest
from the city of Toronto. The Metropolitan Area consists of three different administrative
units, namely the City of Hamilton, the City of Burlington, and the town of Grimsby. The
population of the area was 624,320 in 1996 and 662,401 in 2001. This metropolitan area is
thus the 4th largest urban centre in Ontario, after Toronto, Ottawa, and Mississauga (a
satellite of Toronto), and the 9th largest in Canada. Hamilton is an important functional
component of Greater Toronto, and is therefore part of one of the regions in Canada where
the impacts of demographic aging are expected to be most strongly felt in the next few
years (Hayward, 2001).

The population in the metropolitan area of Hamilton is by and large concentrated in the
City of Hamilton, which enumerates around 500,000 persons. Rapid growth has also been
observed in other parts of the city, in particular in association with suburban development
in Ancaster and Waterdown (southeast part of Flamborough), two areas where population
growth rates in the last few years have been rounding 30 and 20 percent respectively. With
regards to the demographics of the region, the Greater Toronto Area in general and
Hamilton in particular are among the fastest growing population centers in the country.
Provincial projections estimate that in the next 30 years the population of Ontario will
grow by about 4 million, and much of this growth is expected to be absorbed by cities in
Greater Toronto. The recent Places to Grow legislation (Bill 136 2005) of the Government of
Ontario, for example, identifies Hamilton as a provincial growth center.

In parallel to these population projections, the Canadian Census reveals a substantial


increase in the Hamilton metropolitan area of the elderly population aged 65 years and
over. A particular situation posed by this cohort is its characteristically dispersed
residential pattern, as illustrated by Figure 2. This pattern, it has been suggested, has
arisen over the years as a consequence of previous and still ongoing suburbanization
trends, and current attitudes towards aging in place (see Smith, 1998).

4.1 Data: The Transportation Tomorrow Survey


The data used in this study are drawn from the Transportation Tomorrow Survey (TTS).
This is a comprehensive travel survey conducted once every five years in Toronto and its

Centre for Spatial Analysis - Working Paper Series


-8-
Elderly Mobility: Demographic and Spatial Analysis of Trip Making in the Hamilton CMA
A. Páez, D.M. Scott, D. Potoglou, P.S. Kanaroglou, K.B. Newbold

surrounding regions that is moreover timed to coincide with the Canada Census. The first
such survey was conducted in 1986, and the data collection process and questionnaire
have been updated or substantially revised in successive years. Our analysis is based on
the 1996 survey, the third on the TTS series. For this particular survey, 115,000 households
in Greater Toronto were successfully interviewed by telephone, to give a 5% sample of the
households in the region. Details of the survey are documented in reports prepared by the
Joint Program in Transportation, the organization in charge of administering the survey
and of overseeing the maintenance and use of the data (e.g. Data Management Group,
1997). Other reports produced by this group also highlight notable developments detected
as part of each survey. For example, aggregate trip generation trends in the Greater
Toronto Area between 1986 and 1996 are described and discussed in a document prepared
by Miller et al. (1998).

For the purpose of our analyses, the subset of data corresponding to the Hamilton CMA
was extracted from the larger sample. This sub-sample includes detailed information in
four datasets, about the individuals and their household, as well as information about each
trip made by household members using private or public transportation during one
weekday. The sample contains approximately 10,500 households, 27,000 individuals and
51,300 trips recorded on the basis of traffic analysis zones (TAZ). In our analysis, we study
trip making behavior at the individual level, classifying trips as total trips (all day, all
purposes), and also as work trips (all day) and non-work trips (all day). Table 1 presents
the number of trips and trips per person in the sample. Clearly, in absolute numbers, the
65+ cohort makes fewer trips than any other cohort, although the difference is relatively
small compared to the young (less than 20) and the pre-retirement cohort (51-64). Very few
elderly trips however, are made for work purposes. Concerning non-work trips, on the
other hand, the difference in terms of trip generation rates is negligible for males but non-
trivial for the females of the 65+ age group, relative to other cohorts.

In addition to number of trips, a variable used to derive trip frequency classes, we also use
explanatory variables of three types in developing our models. Data from the
Transportation Tomorrow Survey provide information on: 1) the attributes of the
individual; and 2) the attributes of her/his household. We call these level-1 variables. In
addition, complementary information at the level of the zone (level-2 variables) was
extracted from the 1996 Canada Census, or as in the case of accessibility, was derived
based on the geography of the region (i.e. the traffic analysis zones).

Table 2 gives a cross-tabulation of age and gender by two relevant independent variables,
namely license holding status and vehicle ownership. It can be seen there that in absolute
terms there are fewer license holders in the pre-retirement and 65+ groups, compared to
other cohorts. More suggestive is the fact that licensing rates are relatively low for the 65+
cohort, in particular for females, of whom only about half of those sampled in this group
hold a license. As the table suggests, this situation is poised to change as the pre-
retirement group (with license holding rates of about 95% and 77% for males and females

Centre for Spatial Analysis - Working Paper Series


-9-
Elderly Mobility: Demographic and Spatial Analysis of Trip Making in the Hamilton CMA
A. Páez, D.M. Scott, D. Potoglou, P.S. Kanaroglou, K.B. Newbold

respectively) enter retirement. The difference between male and female rates moreover
becomes smaller with each successive younger cohort. This trend strongly suggests that in
the future, more people will enter their retirement years as licensed drivers than in the
past, and that this change is likely to be more significant for females (see Rosenbloom,
2001; pp. 378-380). Licensing rates give a base pool of potential drivers, but access to a car
is what enables these potential drivers to become automobile users. In terms of car
ownership, the rates shown in Table 3 indicate that access to a car is comparatively limited
among the members of the 65+ cohort (at about 87% and 71% for males and females), but
reaches almost saturation levels for every other cohort. The difference between other age
groups moreover is relatively small, suggesting a continuing trend towards vehicle
ownership. These statistics combine to suggest that in a matter of years there will be more
automobile-enabled licensed drivers in every cohort, a development which presumably
would have its most significant impacts in terms of trip making by the elderly in general,
and by old females in particular. Selection of variables for the analysis is discussed in
detail next.

4.2 Selection of Variables


A review of existing trip generation research helps to inform our choice of variables. Trip
generation has been studied in the past from a number of different topical and technical
perspectives, including the influence of land use on travel (e.g. Boarnet and Sarmiento,
1998), the spatial or temporal transferability of trip generation models (e.g. Agyemang-
Duah and Hall, 1997; Badoe and Steuart, 1997), and the relative proficiency of different
modeling approaches (e.g. Pas, 1978). More recently, the work of Schmöcker et al. (2005a)
has cast an explicit focus on the trip generation rates of the elderly and disabled, while
Smith and Sylvestre (2001) have explored the travel behavior determinants of the
suburban elderly. Our review of these researches reveals that there is substantial overlap
in the selection of explanatory variables between these papers, and also with our own
analysis. There are a few notable differences as well, but these seem to stem from the
particular focus of different papers, or from data availability issues.

Schmöcker et al. (2005a), in their study of trip making behavior, are interested in the effect
of aging and various forms of disability (physical and mental) in individual trip
frequencies in London, the argument being that the population affected by either or both
of these conditions (advanced age or impairments) may increase the demand for special
transportation services. Accordingly, their selection of explanatory variables includes a
detailed classification based not only on age cohorts, but also on the presence or absence of
different forms of disability. These data seem to be quite unique to the London Area
Travel Survey (LATS) used in their study, and variables of this nature are seldom reported
in other transportation studies (although they are sometimes obtained as part of purposive
surveys; see Smith and Sylvestre, 2001). Information on disabilities in contrast is not
collected as part of the Transportation Tomorrow Survey. A different paper by Boarnet
and Sarmiento (1998), on the other hand, aims to explore the links between land use and
trip making, with a particular mind to test claims that neighborhood design (e.g. Neo-

Centre for Spatial Analysis - Working Paper Series


- 10 -
Elderly Mobility: Demographic and Spatial Analysis of Trip Making in the Hamilton CMA
A. Páez, D.M. Scott, D. Potoglou, P.S. Kanaroglou, K.B. Newbold

traditional community design) may help to reduce the need for travel, and thus the
negative effects of congestion and pollution. Consequently, their selection of variables
prominently includes land use factors, such as street grid, population density, and retail
and service jobs per unit area. These differences in focus aside, there is an agreement that
modeling trip generation rates benefits from the use of personal variables (age, gender,
ethnicity, etc.) and socio-economic variables (income, employment status, educational
attainment, vehicle ownership, etc.) to explain the individual’s propensity to travel. Other
investigators also incorporate household structural variables either as number of people in
household, presence of children in household, or as categorical variables for different
household structures (see for example Agyemang-Duah and Hall, 1997).

In our study we use personal and household variables. In particular, the papers reviewed
confirm that age is a significant factor to explain trip generation rates; moreover, they also
agree that trip making propensity tends to decrease with increasing age. This effect is
linear in the case of the models estimated by Boarnet and Sarmiento (1998), a team of
researchers that used age as a continuous variable (their results suggests that trip making
frequency decreases unbounded with increasing age). The models developed by
Schmocker et al. (2005a), on the other hand, indicate that while trip making propensity is
low for the elderly relative to other cohorts, the differences between various elderly
groups (65-69, 70-74, etc.) are relatively small (the difference of means between coefficients
for advanced age cohorts in their analysis tends not to be significant). Previous research
has additionally shown that different household structures also have an influence on trip
generation. Following similar lines to Schmöcker et al. (2005a), we include household
structural identifiers. This is in contrast to Boarnet and Sarmiento (1998) and Agyemang-
Duah and Hall (1997), who use number of people in household. The relative advantage of
using household structural identifiers is that these give explicit consideration to different
contextual situations within the household, not just the number of household members.

Vehicle ownership is shown to significantly increase trip making frequency by Agyemang-


Duah and Hall (1997) and Schmöcker et al. (2005a), while on the other hand Boarnet and
Sarmiento (1998) find in some of their models a counter-intuitive negative (but non-
significant) association between number of vehicles per person in household, and number
of trips. As part of our preliminary analysis, we initially used number of vehicles only to
obtain, as Boarnet and Sarmiento, non-significant coefficients. We therefore decided to
switch to a vehicle ownership indicator instead. Variables used in other studies but that
are not part of the TTS include indicators of ethnicity and educational attainment – both
found to be not significant in Boarnet and Sarmiento’s (1998) research.

In addition to disabilities, ethnicity and educational attainment, we faced other variable


selection constraints in working with the Transportation Tomorrow Survey. For example,
this survey does not ask a question about income. Intuition strongly suggests that income
effects could be a significant factor in trip making behavior, by providing the means to
enable more frequent trip making. The evidence concerning these effects however remains

Centre for Spatial Analysis - Working Paper Series


- 11 -
Elderly Mobility: Demographic and Spatial Analysis of Trip Making in the Hamilton CMA
A. Páez, D.M. Scott, D. Potoglou, P.S. Kanaroglou, K.B. Newbold

mixed. For example, Schmöcker et al. (2005a) find that, in general, higher income has a
positive and significant association with number of trips, while on the other hand Boarnet
and Sarmiento (1998) find that income is a positive but not significant predictor of number
of trips. Smith and Sylvestre (2001) equally find that income is not a significant predictor
of trip making for any purpose. Location is also important within the context of our
research questions (are there differences in trip making frequency that can be attributed to
location?). Schmöcker et al. (2005a), for example, use a crude but effective geography to
find that an indicator of residence in Inner/Outer London tends to return a significant
coefficient. This suggests the possibility that a more minute geographical definition could
yield equally significant and yet more informative results. Unfortunately for us, geo-codes
for the location of the households are collected and recorded as part of the survey, but not
reported out of consideration for the privacy of respondents. In compensation, a number
of aggregate-level geographies (traffic analysis zones, census tracts, planning district, etc.)
are available in the public data, and it is thus possible to locate households within zones.
The availability of this information enables the investigation of potential spatial patterns in
trip making behavior. Given the data constraints described in this section, it is worth
noting again that the use of the mixed ordered probit model described in Section 3 allows
us to mix variables at two different levels of geography (individual and zone). The zonal
variables can then be seen as contextual information that makes up to some extent for the
omissions of variables at the individual level (e.g. income and location). In addition, the
random components in the mixed model can help to account for between-place variations
(spatial heterogeneity), and between-place/between-people variations as well (personal
heterogeneity). The variables used in our analyses are listed in Table 3.

5. Results and Discussion


As previously noted, the models below are estimated using three different trip
classifications, namely total trips, work trips, and non-work trips. Work trips include the
first work trip of the day and every other subsequent work-related trip. Non-work trips,
on the other hand, comprise all other purposes, including shopping, recreation, social
visiting, etc. The independent variable in each model is the individual trip frequency class,
which was defined as follows (see tables 4, 5 and 6): 0 trips, 1-2 trips, 3-4 trips, 5-6 trips,
and 7 trips and over for all purposes; and 0 trips, 1 trip, 2 trips, and 3 trips and over for
work and non-work purposes. Furthermore, as shown in tables 4-6, we estimated three
models for each trip class: 1) A conventional ordered probit (Model 1 in each table) that
ignores potential spatial and demographic effects; 2) A spatial model (Model 2) that
incorporates a random component as part of the intercept, in order to capture between-
place variations; and 3) A spatial and demographic model (Model 3) with random
components for the intercept and the age cohort variables.

Our initial run of the models included the 7 accessibility variables indicated in table 3.
After some preliminary analysis, however, these variables were found by and large to be
non-significant jointly or individually. Their inclusion moreover led to long, cumbersome,
or even problematic estimation runs (e.g. lack of convergence), perhaps in part due to their

Centre for Spatial Analysis - Working Paper Series


- 12 -
Elderly Mobility: Demographic and Spatial Analysis of Trip Making in the Hamilton CMA
A. Páez, D.M. Scott, D. Potoglou, P.S. Kanaroglou, K.B. Newbold

particular correlation/colinearity structure. We therefore decided to exclude the


accessibility measures from further analysis and proceed with the rest of the variables
indicated in the table.

Final models for each trip class are reported below. There are some commonalities worth
noting at this point. First, since the coefficients indicate propensity to travel, in such a way
that positive values increase the probability of making more trips, and negative values
decrease the probability of making more trips, we note that all coefficients for which we
had reasonable a priori expectations are correctly signed. And secondly, there appears to
be but relatively small variation in the magnitude of the coefficients between different
models within a trip class. This suggests that the conventional models in our case study
may not be afflicted by large estimation bias. The significance tests on the performance of
the models, and the substantive findings, on the other hand, strongly suggest that the
conventional models, by ignoring between place and between people variations, overlook
important aspects of the problem, and thus give at best a limited perspective of the trip
making behavior of the elderly. Detailed discussion of results in each trip class follows in
the next three subsections.

5.1 Total Trips


The key points of these models are highlighted next.

MODEL SIGNIFICANCE.
Significance testing of the models was conducted based on the usual likelihood ratio tests,
which indicate that the spatial model (Model 2) is to be preferred to the conventional
ordered probit (Model 1). The test in this case returns a Chi-squared statistic (253.76, with
1 degree of freedom) that rejects the hypothesis of no spatial effects at the p=0.0000 level.
Similarly, the likelihood ratio test (4 degrees of freedom) also indicates that the spatial
model with demographic effects (Model 3) is to be preferred over the spatial model
(Model 2), at the p=0.0000 level. The best model in terms of information contents is
therefore the full specification that incorporates between-place and between-person
heterogeneity.

AGE COHORTS.
All coefficients associated with the age cohorts are significant, and suggest that while trip
making propensity is relatively low for the very young group (age <20), it increases with
age reaching a peak for the 34-50 group, and then decreases fairly steeply for the pre-
retirement, and 65+ groups. This finding is consistent across all models. However, we find
that although trip making propensity is relatively lower for the younger group (<20)
compared to the elderly, there is substantially more spatial variability in this behavior for
the 65+ cohort, as indicated by the standard deviation of the random component of the
coefficient (Model 3). It is worth in this regards recalling that the random coefficient is
composed of a systematic, system-wide component (in the tables the coefficient associated
with the variable), and a zone-specific random term whose standard deviation is reported

Centre for Spatial Analysis - Working Paper Series


- 13 -
Elderly Mobility: Demographic and Spatial Analysis of Trip Making in the Hamilton CMA
A. Páez, D.M. Scott, D. Potoglou, P.S. Kanaroglou, K.B. Newbold

(where applicable) below the systematic coefficient. The mean gives the average
relationship between trip making propensity and membership in an age cohort (e.g. the
trip making propensity of the average 51-64, 65+, etc., old individual), relative to the
reference group. The standard deviation, on the other hand, indicates dispersion around
this mean. As the results in the table show, this spatial variability increases with age and
reaches a maximum for the 65+ cohort, thus suggesting the extent to which elderly trip
making behavior is non-homogeneous across Hamilton. The distribution of the random
coefficient across the population of 65+ (in other words the coefficient plus dispersion)
indicates a negative relationship between membership in this cohort and number of trips
for individuals in about 66% of the zones. The complement of course means that the
relationship is actually positive for about 34% of zones, up from about 15% in the case of
the 51-64 cohort, about 29% for the 20-34 cohort, and about 2% for the younger than 20
group. Put differently, the elderly population in about one third of the zones have higher
trip making propensities, the largest share of any cohort (with the possible exception of the
reference group, for which dispersion is not estimated.)

OTHER PERSONAL ATTRIBUTES.


The results of the models suggest that females are more frequent trip makers than males.
Whereas Schmöcker et al. (2005a) appear surprised at finding no difference between males
and females in their London study, our result is in line with Boarnet and Sarmiento’s
(1998) analysis of non-work trips – where they find that females are significantly more
likely to undertake more trips. In our case, the positive coefficient is justified by the
statistics displayed in Table 2 which indicate that females do indeed travel more
frequently. It is tempting to hypothesize that females are more frequent trip makers for a
variety of reasons that relate to their dual role as members of the formal workforce and
housekeepers, including their involvement running errands, taking kids to school, picking
them up, etc. Interestingly, though, as the models for work and non-work trips in tables 5
and 6 below show, the propensity of females to undertake more trips is more pronounced
(the coefficient is larger) for work trips compared to non-work trips.

License ownership relates positively with trip making frequency. It is interesting to note
that in fact this variable is a stronger predictor of number of trips than car ownership. We
use three other variables that relate to what has been termed “mobility tools” (Scott and
Axhausen, 2005): auto ownership, transit pass possession, and joint auto ownership and
transit pass possession. According to the coefficients, these variables correlate positively
with number of trips. Given the trends suggested by the statistics in table 3, in particular
those between cohorts, it is important to note that licensing and car ownership are, in
terms of the magnitude of their coefficients, two of the most important factors affecting
trip generation. Intriguingly, the coefficient for transit pass ownership is larger in
magnitude and significantly different compared to the coefficient for car ownership. This
result suggests that overall mobility may not necessarily be negatively affected by lack of
access to a car, presumably as long as transit remained accessible.

Centre for Spatial Analysis - Working Paper Series


- 14 -
Elderly Mobility: Demographic and Spatial Analysis of Trip Making in the Hamilton CMA
A. Páez, D.M. Scott, D. Potoglou, P.S. Kanaroglou, K.B. Newbold

Job related variables include full time employment, which returns a positive and
significant coefficient that is nonetheless relatively small in magnitude. Job type returns a
negative but not significant coefficient, thus indicating that the difference in trip making
frequency between blue collar and other workers is negligible. And free parking at
workplace, finally, is associated with a coefficient that is positive and significant. The
magnitude of the latter is comparable to the car ownership coefficient, suggesting some
potential for parking policies as traffic alleviation strategies.

HOUSEHOLD ATTRIBUTES.
We use as our reference class the household structure “Single” (households composed of
one adult alone). All other household classes have significantly lower trip making
propensities, with the exception of “Single parent with children”. Two possible
explanations for this result come to mind. First, it is possible that each individual in an
adult multi-person household makes fewer trips because of complementarities in trip
making (e.g. different individuals alternating for grocery shopping). But secondly, in
addition to this effect, it is also possible that an important factor could be the ease afforded
by individual decision making. This would be the case of single households, or a single
adult with children who do not partake of decision making, as opposed to household
structures where there could be a more complex (but unobservable) negotiation process
for task and vehicle allocation, scheduling agreements, and so forth. The idea that
household members influence each other as part of making decisions has been explored
(although not with regards to trip generation) from the perspective of intra-household
interactions, an area of research that currently is attracting substantial attention (e.g. Scott
and Kanaroglou, 2002).

ZONAL ATTRIBUTES.
The coefficient for median income is positive, suggesting that individuals in more affluent
zones tend to make more trips. However, while this result is significant in Model 1, its
importance vanishes as a consequence of incorporating random spatial variation. The
opposite occurs with the locational variables used to introduce a spatial trend. The
coordinates were meant to give an idea of whether there is a systematic spatial variation in
behavior. The coefficients for the spatial trend are positive, suggesting an eastwards and
northwards trend of increasing trip making propensity. In the ordinary (non-spatial
model) this spatial trend is not significant. However, in the spatial models it becomes
statistically significant for the east-west trend, if only marginally so. This results makes
intuitive sense, since moving east in the context of Hamilton, represents an increased
separation between place of residence and most major centers of activity in the city (e.g.
CBD, major commercial developments, in particular in the south Hamilton and Ancaster),
and also implies a movement away from the center of gravity of Greater Toronto.

It is interesting to contrast our finding that the zonal variables are non- or only marginally
significant, with the significant performance gains obtained from estimating spatial
models. This finding strongly suggests that there are spatial effects in operation that are at

Centre for Spatial Analysis - Working Paper Series


- 15 -
Elderly Mobility: Demographic and Spatial Analysis of Trip Making in the Hamilton CMA
A. Páez, D.M. Scott, D. Potoglou, P.S. Kanaroglou, K.B. Newbold

best only marginally captured by the contextual variables used in the models (i.e. income
and spatial trend). As a consequence, the heterogeneity measured in the models remains
for the time being unexplained, although the possibility remains open to explore the
potential influence of other contextual variables in future research.

5.2 Work Trips


MODEL SIGNIFICANCE.
As before, likelihood ratio tests were conducted to assess the performance of the three
models. Based on the results of these tests, we conclude that there are significant gains to
be had from using the spatial model (Model 1 with a randomly varying intercept), but
none from moving to the full model with random coefficients for the age cohorts (Model
3). This provides indication as before that trip making behavior exhibits a degree of spatial
variability with respect to work purposes, but variation within age cohorts is not
significant (i.e. the behavior is spatially homogeneous within age groups). This fits the
notion that non-discretionary travel with fixed origins and destinations, departure and
arrival times, etc., is the same for everyone within a given stage of their work lives, even if
intra-cohort differences in work-related trip making propensity exist (see below).

AGE COHORTS.
We find that advanced age correlates with decreased trip making frequency, and that the
difference between age cohorts is more pronounced for the case of work trips compared to
total trips. All coefficients are significant. On the other hand, we fail to find evidence of
spatial heterogeneity in this demographic effect.

OTHER PERSONAL ATTRIBUTES.


The results for work trips are qualitatively similar to total trips, with a few notable
differences. First, although the coefficients for car ownership and transit pass ownership
are still positive and significant, there is now a large difference in magnitude between
them, with transit pass ownership dominating the influence of mobility tools on trip
making frequency. License holding status remains a significant predictor of trip making.
The single most important factor is now employment status, followed closely by the
availability of free parking at the workplace. Job type, as before, does not make a
significant difference in terms of mobility.

HOUSEHOLD ATTRIBUTES.
The pattern of association between household type and individual trip making behavior
remains qualitatively unchanged compared to the signs of the coefficients displayed in
Table 4. The only difference is that it is not possible to distinguish between single
individuals and single parent with children, as the coefficient is not significant.

ZONAL ATTRIBUTES.

Centre for Spatial Analysis - Working Paper Series


- 16 -
Elderly Mobility: Demographic and Spatial Analysis of Trip Making in the Hamilton CMA
A. Páez, D.M. Scott, D. Potoglou, P.S. Kanaroglou, K.B. Newbold

Parallel to the case of total trips, the influence of these contextual variables is negligible, as
not even the marginal eastward trend detected before is significant in the models in table
5.

5.3 Non-work Trips


MODEL SIGNIFICANCE.
Significance testing based on the likelihood ratio statistics leads us to reject the hypothesis
that Models 1 or 2 are superior to the model that incorporates spatial and demographic
random coefficients (Model 3). All tests are significant at the p=0.0000 level.

AGE COHORTS.
All variables are significant, with exception of the age 20-34 cohort, which cannot be
statistically distinguished from the reference group. On average, trip making propensity
decreases rapidly for individuals in the pre-retirement and elderly cohorts. The degree of
spatial heterogeneity on the other hand is if anything more remarkable. Figure 3 below
shows the distribution of the random coefficients. Individuals in the young cohort (less
than 20) are less likely to make more trips relative to all other cohorts. The distribution
shows that when it comes to this population segment, there is a positive association with
number of trips only for about 17% of zones. Contrast this against the spread of the 20-34
cohort, and more particularly to the 65+ group. While trip making propensity by the
elderly is relatively low, there is a large degree of variability in this behavior. Even if the
random coefficient is positive for only 33% of zones, as opposed to 44% for the 20-34
group, the variation is considerably more pronounced – for example, the 20-34 age
coefficient is 0.2 or greater for the population of just about 1% of zones. The corresponding
figure for the 65+ cohort is 11%. Since the elderly population was about 86,500 people in
1996 (and given their spatial distribution), in absolute terms this means that between
20,500 and 34,600 persons aged 65 and older have a positive coefficient. The large spread
also means that between 4,000 and 14,500 individuals aged 65+ have a coefficient of 0.2 or
greater, and between 1,300 and 8,000 have a coefficient of 0.3 or greater. The significance of
this result is highlighted by the demographic trends. First, there will be more elderly
travelers both in absolute and relative terms. And secondly, the variability in behavior
tends to amplify the effect of licensing and auto ownership in terms of number of trips,
two trends that in turn will become more acute as more individuals enter retirement
holding a license and having access to a private vehicle. It could be said that, from a
systems perspective, trip making by those on the positive side of the distribution will be
offset by others on the negative side, of which in principle there are more. However, the
large variability on the negative side as well is also important because it suggests the
potential mobility challenges faced by a substantial segment of the elderly.

OTHER PERSONAL ATTRIBUTES.


All variables are significant, with the exception of job type. The results indicate that
females are significantly more inclined to travel more frequently, although the magnitude
of the effect is relatively small compared to work trips. Licensing again appears to be the

Centre for Spatial Analysis - Working Paper Series


- 17 -
Elderly Mobility: Demographic and Spatial Analysis of Trip Making in the Hamilton CMA
A. Páez, D.M. Scott, D. Potoglou, P.S. Kanaroglou, K.B. Newbold

single most influential factor in trip making behavior, followed by joint access to a vehicle
and transit pass ownership. For those individuals with access to only one mobility tool,
transit pass ownership continues to dominate over car ownership. Being employed full
time now correlates negatively with number of trips made, but the effect is relatively
small. Free parking at work has a positive if modest effect on number of trips, a fact that
could be attributed to the relative ease of making subsequent trips, even if not related to
work, once that a secure parking base is available.

HOUSEHOLD ATTRIBUTES.
The effect of different household structures remains qualitatively the same, except that it is
more pronounced for the case of non-work trips. That is, propensity to make non-work
trips is slightly lower for all adult multi-person households, relative to work trips, while
the opposite holds true for a single parent with children. The difference between the latter
household structure and single parents is only marginally significant.

ZONAL ATTRIBUTES.
The significance of median income as a contextual variable vanishes again once spatial
heterogeneity is considered. The only significant variable in Model 3 is the east trend, at
the p=0.05 percent. The level of significance of the spatial trend is modestly enhanced in
Model 3 compared to Model 1.

6. Summary and Concluding Remarks


Canada’s population is aging, and this demographic trend is expected to have a significant
impact on urban transportation systems. The objective of this paper has been to conduct a
demographic and spatial analysis of trip generation in the Hamilton metropolitan area.
We approached this task by studying individual trip generation rates using variables at
two different levels: age cohorts and other personal attributes, and household structure at
the individual level; and median income and location contextual variables at the zonal
level. The use of mixed ordered probit models allowed us to mix variables at these two
different levels of geography.

The results of our models confirm previously published findings, in particular the
negative association between increasing age and trip making frequency. In addition,
however, we find that this behavior is not spatially homogeneous, in particular with
respect to non-work trips. Significant spatial variability (as indicated by model significance
testing) was detected in the case of work trips, but in addition to this effect, significant
spatial variability within age cohorts was also found in the case of non-work trips. Our
results suggest that a sizable segment of the 65+ cohort tends to engage in increased trip
making, relative to other cohorts, something that the mean coefficient reported in previous
studies tends to overlook. The magnitude of this heterogeneity effect is also substantially
larger for the elderly. While the overall impact in terms of number of trips system-wide is
not clear, the challenges suggested by our results are. Increased trip making by at least
parts of the elderly cohort would tend to reinforce (or be reinforced by) the effects of

Centre for Spatial Analysis - Working Paper Series


- 18 -
Elderly Mobility: Demographic and Spatial Analysis of Trip Making in the Hamilton CMA
A. Páez, D.M. Scott, D. Potoglou, P.S. Kanaroglou, K.B. Newbold

licensing and vehicle ownership, two trends that will become ever more relevant as more
people enter retirement both in possession of a license and with access to a vehicle. On the
other hand, our findings also suggest some potential for transit, as transit pass ownership
could offset lack of access to a vehicle (presumably as long as transit remained accessible)
without negatively affecting overall mobility. Currently, there is relatively little overlap
between transit pass and vehicle ownership. Parking policies, in particular at the place of
work, also appear as a promising strategy for traffic management efforts given the
strength of the relationship with number of trips.

Our experience in this research indicates that the use of spatial models can give more
detailed insights into trip generation. As noted in the introduction, the models reported
here will be incorporated as part of a GIS-based decision support system for evaluating
transportation strategies in the city of Hamilton. While much of the spatial variation
detected in our research of elderly trip making remains unexplained (i.e. as part of the
random component of the coefficients), this variability is made explicit by the mixed
ordered probit model, and can thus be incorporated as part of projections to obtain more
refined decision-making tools.

As for future research, we are currently investigating trip making from an intra-household
interactions perspective. It will be interesting to compare the performance of the models
reported in this paper, with respect to other modeling approaches, such as joint trip
making models that take into consideration said intra-household interactions, and also
with respect to models of trip rates at the level of the traffic analysis zone. Data from the
2001 Transportation Tomorrow Survey is as of yet unavailable to us. The 2006 Survey is
currently in progress. It would be interesting to use updated data in model estimation to
assess the temporal stability of the models, and also to investigate whether the
relationships formalized by our models are themselves evolving.

Centre for Spatial Analysis - Working Paper Series


- 19 -
Elderly Mobility: Demographic and Spatial Analysis of Trip Making in the Hamilton CMA
A. Páez, D.M. Scott, D. Potoglou, P.S. Kanaroglou, K.B. Newbold

Table 1. Number of trips and trip rates by age, gender, and purpose
Total Trips Work Trips Non-work Trips
Age Cohort All Male Female All Male Female All Male Female
>20 6806 3574 3232 187 98 89 6619 3476 3143
20-35 15151 7293 7858 3840 2148 1692 11311 5145 6166
36-50 16012 7551 8461 4298 2442 1856 11714 5109 6605
51-64 7423 3707 3716 1497 877 620 5926 2830 3096
65+ 5941 2930 3011 190 138 52 5751 2792 2959

Trips/Person Work Trips/Person Non-work Trips/Person


Age Cohort All Male Female All Male Female All Male Female
>20 0.95 0.96 0.94 0.03 0.03 0.03 0.92 0.93 0.91
20-35 2.36 2.36 2.35 0.60 0.69 0.51 1.76 1.66 1.85
36-50 2.58 2.50 2.66 0.69 0.81 0.58 1.89 1.69 2.07
51-64 2.12 2.18 2.06 0.43 0.52 0.34 1.69 1.67 1.72
65+ 1.58 1.77 1.43 0.05 0.08 0.02 1.53 1.69 1.40

Table 2. Individuals, licensing and car ownership by age and gender


Individuals in Sample Individuals in Sample (%)
Age Cohort All Male Female All Male Female
> 20 7173 3737 3436 26.5% 13.8% 12.7%
20 - 35 6433 3096 3337 23.8% 11.4% 12.3%
36 - 50 6202 3017 3185 22.9% 11.1% 11.8%
51 - 64 3499 1698 1801 12.9% 6.3% 6.7%
65+ 3764 1654 2110 13.9% 6.1% 7.8%

License Holders License Holders (%)


Age Cohort All Male Female All Male Female
> 20 634 352 282 8.8% 9.4% 8.2%
20 - 35 5777 2891 2886 89.8% 93.4% 86.5%
36 - 50 5726 2922 2804 92.3% 96.9% 88.0%
51 - 64 2993 1609 1384 85.5% 94.8% 76.8%
65+ 2447 1385 1062 65.0% 83.7% 50.3%

Car ownership Car ownership (%)


Age Cohort All Male Female All Male Female
> 20 6758 3548 3210 94.2% 94.9% 93.4%
20 - 35 5938 2907 3031 92.3% 93.9% 90.8%
36 - 50 5921 2906 3015 95.5% 96.3% 94.7%
51 - 64 3286 1619 1667 93.9% 95.3% 92.6%
65+ 2940 1444 1496 78.1% 87.3% 70.9%

Centre for Spatial Analysis - Working Paper Series


- 20 -
Elderly Mobility: Demographic and Spatial Analysis of Trip Making in the Hamilton CMA
A. Páez, D.M. Scott, D. Potoglou, P.S. Kanaroglou, K.B. Newbold

Table 3. Variable definitions


Dependent variable
Trip frequency class (e.g. 0 trips, 1-2 trips, … , 7+ trips)

Independent variables
Level-1 (Age and Socio-economic attributes)

Age cohorts Socio-economic variables


Age < 20 yes = 1; otherwise 0 Gender Female = 1; male = 0
Age 20-34 yes = 1; otherwise 0 License Holds license = 1; otherwise 0
Vehicle ownership 1 or more vehicles in household AND
Age 34-50 Reference
(only) owns no transit pass: yes = 1; otherwise 0
Transit pass possession Owns transit pass AND
Age 51-64 yes = 1; otherwise 0
(only) no vehicles in household: yes = 1; otherwise 0
1 or more vehicles in household AND
Age 65+ yes = 1; otherwise 0 Vehicle AND Transit pass
owns transit pass: yes = 1; otherwise 0
Employment status Full time = 1; other = 0
Job Type Blue collar = 1; other = 0
Free parking at workplace yes = 1; no = 0
Level-1 (Household structure variables)

Single Reference Couple with children yes = 1; ; other = 0


Couple yes = 1; other = 0 Single parent with children yes = 1; other = 0
Other yes = 1; other = 0
Level-2 (Zonal attributes)
Median income Accessibility variables:
X coordinate of zone 3 continuous variables Distance to major activity 7 continuous variables
centers (e.g. CBD)
Y coordinate of zone

Centre for Spatial Analysis - Working Paper Series


- 21 -
Elderly Mobility: Demographic and Spatial Analysis of Trip Making in the Hamilton CMA
A. Páez, D.M. Scott, D. Potoglou, P.S. Kanaroglou, K.B. Newbold

Table 4. Total Trips – Model Results


Model 1 Model 2 Model 3
Spatial
Ordered Probit Spatial Effects
and Demographic Effects
% in
sample Estimate p-value Estimate p-value Estimate p-value
Age Cohort (Level-1)
AGE<20 -0.2386 0.0000 -0.2358 0.0000 -0.2444 0.0000
26.5%
Std. deviation random term - - 0.1105 0.0001
AGE 20-34 -0.0497 0.0109 -0.0454 0.0121 -0.0489 0.0209
23.8%
Std. deviation random term - - 0.0896 0.0022
AGE 34-50 22.9% REFERENCE
AGE 51-64 -0.1118 0.0000 -0.1134 0.0000 -0.1151 0.0001
12.9%
Std. deviation random term - - 0.1134 0.0018
AGE 65+ -0.2013 0.0000 -0.1942 0.0000 -0.2065 0.0000
13.9%
Std. deviation random term - - 0.2436 0.0000
Personal attributes (Level-1)
FEMALE 51.2% 0.0767 0.0000 0.0739 0.0000 0.0740 0.0000
LICENSE 64.9% 0.7281 0.0000 0.7260 0.0000 0.7241 0.0000
VEHICLE ONLY 88.3% 0.3656 0.0000 0.3771 0.0000 0.3737 0.0000
TRANSIT PASS ONLY 1.9% 0.5362 0.0000 0.5497 0.0000 0.5495 0.0000
VEHICLE+TRANSIT PASS 3.5% 0.6347 0.0000 0.6398 0.0000 0.6420 0.0000
FULL TIME EMPL 35.1% 0.0968 0.0000 0.1059 0.0000 0.1087 0.0000
BLUE COLLAR 11.0% -0.0384 0.1285 -0.0287 0.2621 -0.0304 0.2386
FREE PARK@WORK 37.8% 0.2455 0.0000 0.2448 0.0000 0.2482 0.0000
Household attributes (Level-1)
SINGLE 9.1% REFERENCE
COUPLE 21.7% -0.1750 0.0000 -0.1780 0.0000 -0.1821 0.0000
COUPLE W/CHILD. 35.9% -0.2497 0.0000 -0.2437 0.0000 -0.2453 0.0000
SINGLE PAR. W/CHILD. 4.4% 0.1090 0.0097 0.1093 0.0075 0.1098 0.0070
OTHER 28.9% -0.2227 0.0000 -0.2243 0.0000 -0.2276 0.0000
Zonal attributes (Level-2)
MEDIAN INCOME - 0.0471 0.0032 0.0365 0.1737 0.0294 0.3167
X (Easting) - 0.0163 0.7393 0.1129 0.0485 0.1112 0.0724
Y (Northing) - 0.0423 0.3601 0.0429 0.5337 0.0153 0.8379
INTERCEPT -0.5317 0.0000 -0.5939 0.0000 -0.5547 0.0000
-
Std. deviation random term - 0.1088 0.0000 0.1105 0.0000
Thresholds (for identification threshold 1 = 0)
0 Trips 39.6% 0.0000 0.0000 0.0000
1-2 Trips 34.2% 1.0296 0.0000 1.0371 0.0000 1.0436 0.0000
3-4 Trips 17.2% 1.7991 0.0000 1.8094 0.0000 1.8197 0.0000
5-6 Trips 6.1% 2.3983 0.0000 2.4109 0.0000 2.4237 0.0000
7+ Trips 2.9%
Log-likelihood at convergence -32629.24 -32502.36 -32474.46
Level-1 observations = 27,071 (individuals)
Level-2 observations = 209 (zones)

Centre for Spatial Analysis - Working Paper Series


- 22 -
Elderly Mobility: Demographic and Spatial Analysis of Trip Making in the Hamilton CMA
A. Páez, D.M. Scott, D. Potoglou, P.S. Kanaroglou, K.B. Newbold

Table 5. Total Work Trips – Model Results

Model 1 Model 2 Model 3


Spatial
Ordered Probit Spatial Effects
and Demographic Effects
% in
sample Estimate p-value Estimate p-value Estimate p-value
Age Cohort (Level-1)
AGE<20 -0.6957 0.0000 -0.6997 0.0000 -0.6984 0.0000
26.5%
Std. deviation random term - - -
AGE 20-34 -0.0500 0.0254 -0.0509 0.0219 -0.0514 0.0233
23.8%
Std. deviation random term - - 0.0142 0.4728
AGE 34-50 22.9% REFERENCE
AGE 51-64 -0.1712 0.0000 -0.1719 0.0000 -0.1751 0.0000
12.9%
Std. deviation random term - - 0.1214 0.0132
AGE 65+ -0.8352 0.0000 -0.8358 0.0000 -0.8638 0.0000
13.9%
Std. deviation random term - - 0.1897 0.0667
Personal attributes (Level-1)
FEMALE 51.2% 0.0950 0.0000 0.0949 0.0001 0.0952 0.0001
LICENSE 64.9% 0.4558 0.0000 0.4527 0.0000 0.4547 0.0000
VEHICLE ONLY 88.3% 0.0777 0.1859 0.0893 0.1077 0.0891 0.1121
TRANSIT PASS ONLY 1.9% 0.3218 0.0012 0.3263 0.0018 0.3288 0.0017
VEHICLE+TRANSIT PASS 3.5% 0.3463 0.0000 0.3585 0.0000 0.3592 0.0000
FULL TIME EMPL 35.1% 1.1742 0.0000 1.1782 0.0000 1.1798 0.0000
BLUE COLLAR 11.0% 0.0123 0.6167 0.0169 0.5136 0.0169 0.5134
FREE PARK@WORK 37.8% 0.7975 0.0000 0.8065 0.0000 0.8070 0.0000
Household attributes (Level-1)
SINGLE 9.1% REFERENCE
COUPLE 21.7% -0.1154 0.0023 -0.1122 0.0133 -0.1124 0.0138
COUPLE W/CHILD. 35.9% -0.1931 0.0000 -0.1823 0.0001 -0.1817 0.0001
SINGLE PAR. W/CHILD. 4.4% 0.0141 0.8237 0.0299 0.6549 0.0293 0.6623
OTHER 28.9% -0.1045 0.0048 -0.0971 0.0298 -0.0971 0.0313
Zonal attributes (Level-2)
MEDIAN INCOME - 0.0401 0.0693 0.0311 0.2835 0.0276 0.3627
X (Easting) - 0.0316 0.6336 0.0689 0.3944 0.0777 0.3479
Y (Northing) - 0.0711 0.2609 0.0308 0.7389 0.0329 0.7248
INTERCEPT -1.7267 0.0000 -1.7350 0.0000 -1.7368 0.0000
-
Std. deviation random term - 0.1125 0.0000 0.1108 0.0000
Thresholds (for identification threshold 1 = 0)
0 Trips 68.6% 0.0000 0.0000 0.0000
1 Trips 28.0% 1.9647 0.0000 1.9762 0.0000 1.9788 0.0000
2 Trips 2.2% 2.4546 0.0000 2.4684 0.0000 2.4718 0.0000
3+ Trips 1.2%
Log-likelihood at convergence -12505.55 -12486.67 -12485.20
Level-1 observations = 27,071 (individuals)
Level-2 observations = 209 (zones)

Centre for Spatial Analysis - Working Paper Series


- 23 -
Elderly Mobility: Demographic and Spatial Analysis of Trip Making in the Hamilton CMA
A. Páez, D.M. Scott, D. Potoglou, P.S. Kanaroglou, K.B. Newbold

Table 6. Total Non-work Trips – Model Results


Model 1 Model 2 Model 3
Spatial
Ordered Probit Spatial Effects
and Demographic Effects
% in
sample Estimate p-value Estimate p-value Estimate p-value
Age Cohort (Level-1)
AGE<20 -0.1295 0.0000 -0.1324 0.0000 -0.1445 0.0001
26.5%
Std. deviation random term - - 0.1517 0.0000
AGE 20-34 -0.0131 0.5251 -0.0089 0.6466 -0.0118 0.5970
23.8%
Std. deviation random term - - 0.0907 0.0045
AGE 34-50 22.9% REFERENCE
AGE 51-64 -0.0858 0.0010 -0.0855 0.0013 -0.0849 0.0036
12.9%
Std. deviation random term - - 0.0864 0.0405
AGE 65+ -0.1142 0.0001 -0.1125 0.0000 -0.1261 0.0010
13.9%
Std. deviation random term - - 0.2605 0.0000
Personal attributes (Level-1)
FEMALE 51.2% 0.0717 0.0000 0.0683 0.0001 0.0689 0.0001
LICENSE 64.9% 0.7156 0.0000 0.7104 0.0000 0.7075 0.0000
VEHICLE ONLY 88.3% 0.3490 0.0000 0.3656 0.0000 0.3605 0.0000
TRANSIT PASS ONLY 1.9% 0.5470 0.0000 0.5663 0.0000 0.5666 0.0000
VEHICLE+TRANSIT PASS 3.5% 0.6402 0.0000 0.6523 0.0000 0.6525 0.0000
FULL TIME EMPL 35.1% -0.0810 0.0002 -0.0713 0.0006 -0.0688 0.0008
BLUE COLLAR 11.0% -0.0323 0.2226 -0.0276 0.3163 -0.0294 0.2896
FREE PARK@WORK 37.8% 0.1657 0.0000 0.1596 0.0000 0.1626 0.0000
Household attributes (Level-1)
SINGLE 9.1% REFERENCE
COUPLE 21.7% -0.1661 0.0000 -0.1683 0.0000 -0.1709 0.0000
COUPLE W/CHILD. 35.9% -0.2795 0.0000 -0.2724 0.0000 -0.2730 0.0000
SINGLE PAR. W/CHILD. 4.4% 0.0762 0.0713 0.0828 0.0655 0.0868 0.0508
OTHER 28.9% -0.2142 0.0000 -0.2169 0.0000 -0.2198 0.0000
Zonal attributes (Level-2)
MEDIAN INCOME - 0.0376 0.0198 0.0252 0.3666 0.0193 0.5220
X (Easting) - 0.0582 0.2403 0.1343 0.0274 0.1241 0.0475
Y (Northing) - 0.0406 0.3862 0.0267 0.7273 0.0103 0.8961
INTERCEPT -0.4882 0.0000 -0.5280 0.0000 -0.4894 0.0000
-
Std. deviation random term - 0.1128 0.0000 0.1047 0.0000
Thresholds (for identification threshold 1 = 0)
0 Trips 39.5% 0.0000 0.0000 0.0000
1 Trips 17.9% 0.4909 0.0000 0.4904 0.0000 0.4941 0.0000
2 Trips 20.6% 1.1060 0.0000 1.1119 0.0000 1.1197 0.0000
3+ Trips 22.0%
Log-likelihood at convergence -34177.87 -34033.29 -33997.62
Level-1 observations = 27,071 (individuals)
Level-2 observations = 209 (zones)

Centre for Spatial Analysis - Working Paper Series


- 24 -
Elderly Mobility: Demographic and Spatial Analysis of Trip Making in the Hamilton CMA
A. Páez, D.M. Scott, D. Potoglou, P.S. Kanaroglou, K.B. Newbold

Figure 1. Distribution of utility based on number of trips

Centre for Spatial Analysis - Working Paper Series


- 25 -
Elderly Mobility: Demographic and Spatial Analysis of Trip Making in the Hamilton CMA
A. Páez, D.M. Scott, D. Potoglou, P.S. Kanaroglou, K.B. Newbold

Figure 2. Elderly population distribution in the Hamilton CMA, 2001

Centre for Spatial Analysis - Working Paper Series


- 26 -
Elderly Mobility: Demographic and Spatial Analysis of Trip Making in the Hamilton CMA
A. Páez, D.M. Scott, D. Potoglou, P.S. Kanaroglou, K.B. Newbold

Figure 3. Random coefficients for age cohorts, Non-work Trips (Model 3, Table 6)

Centre for Spatial Analysis - Working Paper Series


- 27 -
Elderly Mobility: Demographic and Spatial Analysis of Trip Making in the Hamilton CMA
A. Páez, D.M. Scott, D. Potoglou, P.S. Kanaroglou, K.B. Newbold

References
Agyemang-Duah K, Hall FL, 1997, "Spatial transferability of an ordered response model
trip generation" Transportation Research Part A-Policy and Practice 31 (5) 389 - 402.

Alsnih R, Hensher DA, 2003, "The mobility and accessibility expectations of seniors in an
aging population" Transportation Research Part A-Policy and Practice 37 (10) 903 - 916.

Anselin L, Griffith DA, 1988, "Do Spatial Effects Really Matter in Regression-Analysis"
Papers of the Regional Science Association 65 11 - 34.

Badoe DA, Chen CC, 2004, "Unit of analysis in conventional trip generation modelling: an
investigation" Canadian Journal of Civil Engineering 31 (2) 272 - 280.

Badoe DA, Steuart GN, 1997, "Urban and travel changes in the Greater Toronto Area and
the transferability of trip-generation models" Transportation Planning and Technology 20
(4) 267 - 290.

Ben-Akiva M, Lerman SR, 1985, Discrete Choice Analysis: Theory and Applications to
Travel Demand (The MIT Press, Cambridge)

Bhat C, Zhao HM, 2002, "The spatial analysis of activity stop generation" Transportation
Research Part B-Methodological 36 (6) 557 - 575.

Boarnet MG, Sarmiento S, 1998, "Can land-use policy really affect travel behaviour? A
study of the link between non-work travel and land-use characteristics" Urban Studies 35
(7) 1155 - 1169.

Cvitkovich Y, Wister A, 2001, "The importance of transportation and prioritization of


environmental needs to sustain well-being among older adults" Environment and
Behavior 33 (6) 809 - 829.

Data Management Group, 1997, Transportation Tomorrow Survey 1996: Design and
Conduct of the Survey, Joint Program in Transportation, University of Toronto, Toronto
(http://www.jpint.utoronto.ca/PDF/conduct.pdf).

Duncan C, Jones K, 2000, "Using multilevel models to model heterogeneity: Potential and
pitfalls" Geographical Analysis 32 (4) 279 - 305.

Franklin JP, Niemeier DA, 1998, "Discrete choice elasticities for elderly and disabled
travelers between fixed-route transit and paratransit" Transit (1623) 31 - 36.

Centre for Spatial Analysis - Working Paper Series


- 28 -
Elderly Mobility: Demographic and Spatial Analysis of Trip Making in the Hamilton CMA
A. Páez, D.M. Scott, D. Potoglou, P.S. Kanaroglou, K.B. Newbold

Gardiner C, Hill R, 1996, "Analysis of access to cars from the 1991 UK census samples of
anonymised records: A case study of the elderly population of Sheffield" Urban Studies 33
(2) 269 - 281.

Griffith DA, Layne LJ, 1999, A Casebook for Spatial Statistical Data Analysis: A
Compilation of Analyses of Different Thematic Data Sets (Oxford University Press, New
York)

Harrison A, Ragland DR, 2003, "Consequences of driving reduction or cessation for older
adults" Transportation Research Record (1843) 96 - 104.

Hayward LM, 2001, "Local Planning for an Aging Population in Ontario: Two Case
Studies" http://socserv.socsci.mcmaster.ca/~sedap/p/sedap47.pdf

Hedeker D, Gibbons RD, 1994, "A Random-Effects Ordinal Regression-Model for


Multilevel Analysis" Biometrics 50 (4) 933 - 944.

Hildebrand ED, 2003, "Dimensions in elderly travel behaviour: A simplified activity-based


model using lifestyle clusters" Transportation 30 (3) 285 - 306.

Jones K, 1991, "Specifying and Estimating Multilevel Models for Geographical Research"
Transactions of the Institute of British Geographers 16 (2) 148 - 159.

Kim S, 2003, "Analysis of elderly mobility by structural equation modeling" Transportation


Research Record (1854) 81 - 89.

Kim S, Ulfarsson GF, 2004, "Travel mode choice of the elderly - Effects of personal,
household, neighborhood, and trip characteristics" Transportation Research Record (1894)
117 - 126.

Maddala GS, 1983, Limited-Dependent and Qualitative Variables in Econometrics


(Cambridge University Press, Cambridge)

Maoh H, Kanaroglou P, Scott DM, Newbold KB, Páez A, 2005, "A GIS-based decision
support tool to study the impact of elderly population on the transportation system in
Hamilton Census Metropolitan Area (CMA)" Proceedings of the 2005 Annual Meeting of
the Canadian Transportation Research Forum, Hamilton, ON, 11 - 5 August.

McMillen DP, 1992, "Probit with Spatial Autocorrelation" Journal of Regional Science 32 (3)
335 - 348.

Centre for Spatial Analysis - Working Paper Series


- 29 -
Elderly Mobility: Demographic and Spatial Analysis of Trip Making in the Hamilton CMA
A. Páez, D.M. Scott, D. Potoglou, P.S. Kanaroglou, K.B. Newbold

Mercado RG, Páez A, Newbold KB, Scott DM, Kanaroglou P, 2006, "Transport Policy in
Ageing Societies: An International Comparison and Implications for Canada" Paper
accepted for presentation in the 85th Transportation Research Board Annual Meeting,
Washington, D.C., 26 - 22 January.

Meyer MD, Miller EJ, 2001, Urban Transportation Planning, Second Edition (McGraw-Hill,
New York)

Miller, E. J., Dalton, P., and Briggs, R., 1998, GTA Trip Generation Rates 1986 -1996, Joint
Program in Transportation, University of Toronto, Toronto
(http://www.jpint.utoronto.ca/PDF/tripgen.pdf).

Miller HJ, 1999, "Potential contributions of spatial analysis to geographic information


systems for transportation (GIS-T)" Geographical Analysis 31 (4) 373 - 399.

Miller HJ, Shaw SL, 2001, Geographic Information Systems for Transportation: Principles
and Applications (Oxford University Press, Oxford)

Newbold KB, Scott DM, Spinney JEL, Kanaroglou P, Páez A, 2005, "Travel behavior within
Canada's older population: a cohort analysis" Journal of Transport Geograhy 13 (4) 340 -
351.

Ortúzar JD, Willumsen LG, 2001, Modelling Transport, Third Edition (Wiley, New York)

Páez A, Scott DM, 2004, "Spatial Statistics for Urban Analysis: A Review of Techniques
with Examples" GeoJournal 61 (4) 53 - 67.

Pas EI, 1978, "Empirical Comparison of Zonal, Household and Personal Models of Home-
Based Trip Generation" Traffic Engineering & Control 19 (2) 64 - 68.

Pisarski AE, 2003, "Prescriptions for research: reviewing the history of TRB's critical issues
in Transportation" TR News 226 30 - 35.

Redmond LS, Mokhtarian PL, 2001, "The positive utility of the commute: modeling ideal
commute time and relative desired commute amount" Transportation 28 (2) 179 - 205.

Rosenbloom S, 2001, "Sustainability and automobility among the elderly: An international


assessment" Transportation 28 (4) 375 - 408.

Centre for Spatial Analysis - Working Paper Series


- 30 -
Elderly Mobility: Demographic and Spatial Analysis of Trip Making in the Hamilton CMA
A. Páez, D.M. Scott, D. Potoglou, P.S. Kanaroglou, K.B. Newbold

Schmöcker JD, Quddus MA, Noland RB, Bell MGH, 2005a, "Estimating Trip Generation of
Elderly and Disabled People: An Analysis of London Data" 84th Transportation Research
Board Annual Meeting, Washington, D.C., 9 - 13 Januarya.

Schmöcker JD, Quddus MA, Noland RB, Bell MGH, 2005b, "Transport Mode Choice of the
Elderly and Disabled in London" 84th Transportation Research Board Annual Meeting,
Washington, D.C., 9 - 13 Januaryb.

Scott DM, Axhausen KW, 2005, "Household mobility tool ownership: Modeling
interactions between cars and season tickets" Transportation (forthcoming)

Scott DM, Kanaroglou PS, 2002, "An activity-episode generation model that captures
interactions between household heads: development and empirical analysis"
Transportation Research Part B-Methodological 36 (10) 875 - 896.

Smith GC, 1998, "Change in elderly residential segregation in Canadian metropolitan


areas, 1981-91" Canadian Journal on Aging-Revue Canadienne du Vieillissement 17 (1) 59 -
82.

Smith GC, Sylvestre GM, 2001, "Determinants of the travel behavior of the suburban
elderly" Growth and Change 32 (3) 395 - 412.

Stern S, 1993, "A Disaggregate Discrete-Choice Model of Transportation Demand by


Elderly and Disabled People in Rural Virginia" Transportation Research Part A-Policy and
Practice 27 (4) 315 - 327.

Tacken M, 1998, "Mobility of the elderly in time and space in the Netherlands: An analysis
of the Dutch National Travel Survey" Transportation 25 (4) 379 - 393.

Train K, 2003, Discrete Choice Methods with Simulation (Cambridge University Press,
Cambridge)

Waldorf B, 2003, "Automobile reliance among the elderly: Race and spatial context effects"
Growth and Change 34 (2) 175 - 201.

Washington S, 2000, "Iteratively specified tree-based regression: Theory and trip


generation example" Journal of Transportation Engineering-Asce 126 (6) 482 - 491.

Centre for Spatial Analysis - Working Paper Series


- 31 -
View publication stats

You might also like