The Effects of Age on the Driving Habits of the Elderly
The Effects of Age on the Driving Habits of the Elderly
Evidence from the 1990 National Personal Transportation Study
Final Report
October 1994
Prepared by
Xuehao Chu Center for Urban
Transportation Research
University of South Florida
4202 East Fowler Avenue, ENB 118
Tampa, Florida 33670- 5350
Prepared for
Office of University Research
and Education Research
and Special Programs Administration
U.S. Department of Transportation
Washington, D.C. 20590
Distributed in Cooperation with
Technology Sharing Program
Research and Special Programs
Administration U.S. Department of
Transportation Washington, D.C. 20590
DOT-T-95-12
TABLE OF CONTENTS
Table of Contents List of Tables iv Acknowledgments v Abstract vi
Chapter 1 : Introduction 1 Background 1 Issues and Hypotheses 2
Previous Studies 3 Approach and Organization of the Report 5
Chapter 2 : The 1990 Nationwide Personal Transportation Survey 5
Survey 5 Variables 6 Chapter 3 : The Effects of Age on How Much
the Elderly Drive 8
Number of Daily Vehicle Miles 8 Number of Daily Vehicle Trips 12
Distance of Daily Vehicle Trips 15 Chapter 4 The Effects of Age
on When the Elderly Drive 18 Driving at Night 18 Driving During
Peak Hours 21 Chapter 5 The Effects of Age on How the Elderly
Drive 24 Speed 24 Limited-Access Highways 27 Automobile Size 30
Number of Passengers Carried 34 Chapter 6 : Summary and Policy
Implications 37 Summary 37 Policy Implications 38 Endnotes 40
iii
LIST OF TABLES
Table 2.1 Definition of variables 7 Table 3.1 Average number of
daily vehicle miles by driver age group 8 Table 3.2 Tobit
analysis of daily vehicle miles 10 Table 3.3 Average number of
daily vehicle trips by driver age group 12 Table 3.4 Tobit
analysis of number of daily vehicle trips 14 Table 3.5 Average
distance of daily vehicle trips by driver age group 15 Table 3.6
Weighted regression of distance of daily vehicle trips 17 Table
4.1 Percent of miles driven at night by driver age group 18 Table
4.2 Logit analysis of driving at night 20 Table 4.3 Percent of
miles driven during peak hours by driver age group 21 Table 4.4
Logit analysis of driving during peak hours 23 Table 5.1 Average
speed on all roads by driver age group 24 Table 5.2 Average speed
on limited-access highways by driver age group 25 Table 5.3
Weighted regression of speed of daily vehicle trips 26 Table 5.4
Percent of miles driven on limited-access highways by driver age
group 28 Table 5.5 Logit analysis of driving on limited-access
highways 29 Table 5.6 Average size of automobiles by age group of
main drivers 31 Table 5.7 Weighted regression of automobile size
33 Table 5.8 Average occupancy of automobile trips by driver age
group 34 Table 5.9 Weighted regression of occupancy of automobile
trips 35
iv
ABSTRACT
The Effects of Age on the Driving Habits of the Elderly: Evidence
from the 1990 NPTS
This report examines the effects of age on the driving habits of
the elderly, using the 1990 Nationwide Personal Transportation
Survey (NPTS). Elderly is defined as persons 65 years or older.
Six aspects are considered: the amount of daily driving exposure,
driving by time of day, driving speed, driving by type of
roadways, vehicle size, and the number of passengers carried. The
scope of analysis is limited to the content of the 1991 NPTS and
those aspects of driving habits that are hypothesized to have
safety implications for the elderly. The scale of analysis is
limited to urban residents. Regression is used to isolate the
effects of being elderly while holding constant a set of
personal, household, and location characteristics of the drivers,
as well as a set of trip characteristics. Elderly drivers show an
increased effort of self-protection in their driving habits
relative to mid-aged drivers (persons between the ages of 25 and
64 years). Being elderly not only makes elderly drivers reduce
daily driving exposure, avoid driving at night, avoid driving
during peak hours, and avoid driving on limited-access highways,
but also make them drive at lower speeds, drive larger
automobiles, and carry fewer passengers. Despite their effort of
self-protection, however, the elderly still show a higher risk of
crash and injury per unit of exposure than the mid-aged. If
policies induce the elderly to further adjust their driving
habits to offset the external risks of their driving, their risk
of crash and injury would be reduced and society as a whole would
be better off. The elderly, however, are likely to be worse off
as a consequence of reduced mobility.The challenge to
policy-making is to balance these consequences of any policy
concerning the mobility and traffic safety of the elderly.
vi
ACKNOWLEDGMENTS
This project is made possible through a grant from the U.S.
Department of Transportation, University Research Institute
Program. Their support is gratefully acknowledged. Comments from
the following individuals are gratefully acknowledged: William L.
Ball, Michael R.Baltes, Patricia Henderson, Rosemary Mathias,
Steve Polzin, and Joel R. Rey.
v
Chapter 1 INTRODUCTION
The mobility and traffic safety of elderly drivers are of great
concern to the public.1 Much of this concern is due to the fast
growth in the number of elderly drivers and their driving. This
report examines the effects of age on the driving habits of the
elderly in the United States, as revealed in the 1990 Nationwide
Personal Transportation Survey (NPTS).2 Six aspects of driving
habits are considered that are hypothesized to have safety
implications for the elderly. A good understanding of the driving
habits of the elderly is essential not only to the provision of
public transportation to the elderly but also to the design of
policies that address the mobility and traffic safety of the
elderly.
BACKGROUND
Between 1985 and 1989, three national conferences were held to
discuss issues on the mobility and traffic safety of elderly
drivers.3 Initiated in 1986 by the Transportation Research Board
(TRB), the U.S. Congress requested in the Surface Transportation
Assistance Act of 1987 "a comprehensive study and investigation
of (1) problems which may inhibit the safety and mobility of
elderly drivers using the Nation's roads and (2) means of
addressing these problems.4 In 1987, Congress asked the U.S.
Department of Transportation to implement a pilot program of
highway safety improvements to enhance the mobility and traffic
safety of elderly drivers.5 In addition, elderly drivers
frequently make headlines in major magazines and newspapers
across the nation.6 The number of elderly drivers grew from 8.6
million in 1970 to 22.3 million In 1990, an increase of 148
percent, while the number of all drivers grew by 50 percent
during the same period. The number of elderly drivers as a
proportion of all drivers also increased from 8.0 percent in 1970
to 13.3 percent in 1990. 7 These increases reflect the growth in
the elderly population as well as in its licensure rate. The
elderly population grew from 20.0 million in 1970 to 31.1 million
in 1990, an increase of 56 percent, while the population of age
15 years or older grew by 34 percent during the same period.8 The
licensure rate of the elderly population increased from 45
percent in 1970 to 72 percent in 1990, while the licensure rate
of the population of age 15 years or older increased from 77
percent in 1970 to 86 percent in 1990. 9 The number of miles
driven by the elderly has grown more than the elderly population
and its licensure rate. The elderly drove 42.2 billion miles in
1969 and 153.7 billion miles in 1990, an increase of 264 percent.
The rate of growth for all drivers was 142 percent. The share of
miles driven by the elderly increased from 4.9 percent in 1969 to
7.1 percent in 1990.10 These trends are expected to continue. By
the year 2020, the elderly population is expected to reach 20
percent of all persons. The number of elderly drivers is likely
to exceed 20 percent of all drivers.11
1
ISSUES AND HYPOTHESES
This report considers six aspects of driving habits. These
aspects include the amount of daily driving exposure, driving by
time of day, driving speed, driving by type of roadways, vehicle
size, and the number of passengers carried. The scope of analysis
is limited to the content of the 1990 NPTS and to those aspects
of driving habits that are hypothesized to have safety
implications for the elderly. The scale of analysis is limited to
urban residents. In addition to age, other personal, household,
and location characteristics of the elderly also may influence
their drivingmhabits. Personal characteristics include
educational attainment and labor force participation. Household
characteristics include race, annual income, composition (size,
children), and vehicle ownership. Location characteristics
include the household location in an urban area (central city vs.
suburbs), the household location in the nation (the West vs.
other regions), the size of an urban area, and the population
density of an urban area. Many of these characteristics may
differ systematically between the elderly and others. Labor force
participation changes with aging. Household income may decline
with retirement from the labor force. Household composition may
change with aging. For example, the elderly are less likely to
live with young children than are younger persons. Vehicle
ownership may change with aging due to changes in household
composition and income. Household location may change with aging.
For example, the elderly may be more likely to live in the
suburbs and in the South. The elderly have more time available
for travel during the day. The elderly also may differ from
others in their activity patterns. The elderly may choose to
participate in activities that occur less frequently (e.g., once
a month instead of once a week). They may choose to participate
in activities that are closer to their homes. Or they may move
closer to activities in which they choose to participate. They
also may choose to participate in activities that occur during
the day or off-peak hours. However, the literature provides no
evidence of these hypothetical changes in the activity patterns
of the elderly. It is important to control for the
characteristics that differ systematically between the elderly
and others in order to isolate the effects of age on the driving
habits of the elderly. It is also important to control for these
characteristics in order to draw conclusions about the driving
habits of the future's elderly from the driving habits of today's
elderly because many of these characteristics may change in the
future for the elderly. For example, the future's elderly may
have higher vehicle ownership than today's elderly. The future's
elderly also may be more likely to live in the suburbs than
today's elderly. The elderly differ from others in two other
important characteristics that have not been discussed. First,
the majority of the elderly are not employed and will remain
unemployed for the rest of their lives. The elderly, therefore,
would lose less than younger persons in future labor earnings
from an injury. According to the foregone-labor-earnings approach
to measuring motor vehicle crash costs, elderly drivers are
likely to have lower costs of injuries than younger drivers.12
2
Second, cognitive and physical abilities generally decline with
aging.13 One consequence of this decline is that the driving
skills of the elderly are reduced. As a result, elderly drivers
are more likely to be involved in crashes than all drivers,
except those under the age of 25 years. 14 In the majority of
crashes in which elderly drivers were involved, they were at
fault for failing to yield the right-of-way, turning improperly,
ignoring traffic signals, or starting improperly into traffic.15
Another consequence of the decline in their physical abilities is
that the elderly are more likely to be injured than younger
persons in a crash. These two important characteristics of the
elderly may have two opposite effects on their driving habits. On
the one hand, elderly drivers may be more willing than younger
drivers to take risks because of their reduced costs of injuries.
On the other hand, elderly drivers may compensate for their
increased crash and injury risks. This behavior of risk
compensation can manifest itself in many ways. The elderly may
drive fewer miles to reduce exposure. They may feel less
comfortable with carrying passengers. They may find certain
driving conditions difficult, such as driving at night, during
peak hours, at high speeds, or on limited-access highways. They
also may feel vulnerable to the low crashworthiness of small
vehicles. While this study controls for many of the personal,
household, and location characteristics of the elderly discussed
earlier, it does not, however, control for the two important
characteristics just discussed. It is hypothesized that the
relative strengths of these two characteristics determine the
effects of age on the driving habits of the elderly.
PREVIOUS STUDIES
No known previous study exists that looks at the size of vehicles
that the elderly drive or the number of passengers they carry.
Previous studies on the amount of driving exposure, driving
speed, driving by time-of-day, and driving on limited-access
highways by the elderly have one drawback: they often fail to
control simultaneously for many factors that may influence the
driving habits of the elderly. This drawback has two
implications. On the one hand, any observed difference in the
driving habits between the elderly and others may be a mix of the
differences in age and other personal, household, and location
characteristics of the drivers that are not controlled for in
these studies. On the other hand, any difference observed in the
driving habits of today's elderly and others is unlikely to hold
true in the future because those personal, household, and
location characteristics of the drivers that are not controlled
for may change in the future. The evidence from previous studies
is mixed. Studies have found "no evidence that elderly drivers
who exhibit poor performance on driving simulators make any
compensating adjustment in the amount of driving exposure."16 One
reason given is that elderly drivers are unaware of the changes
in their cognitive and physical abilities and those driving
conditions that become more difficult as age advances. The other
reason given is that elderly drivers are unwilling to admit lack
of driving competence or to significantly reduce exposure.
Several U.S. studies, however, find that elderly drivers reduce
exposure more as they age and tend to avoid
3
high-risk conditions, such as driving at night and during peak
hours.17 A Canadian study concludes that "increased driver risk
due to medical conditions among elderly drivers was more than
offset by their adoption of new, less risky driving patterns."18
APPROACH AND ORGANIZATION OF THE REPORT
This study uses regression analysis to isolate the effects of age
on the driving habits of the elderly. Regression analysis
accomplishes this isolation by including variables measuring the
age as well as a set of other personal, household, and location
characteristics of the drivers. It is important to control for
factors that aging may affect. It is also important to control
for factors that aging does not affect, such as gender and race.
Under this regression framework, this study attempts to determine
whether or not age affects the driving habits of the elderly and,
if so, what the size and nature of the effects are. This report
is organized into six chapters. Chapter 1 is this introduction.
Chapter 2 describes the 1990 NPTS and the variables that are used
in this study. Chapter 3 examines the effects of age on how much
the elderly drive. The aspects examined include the number of
daily vehicle miles driven by individual drivers, the number of
daily vehicle trips taken by individual drivers, and the distance
of individual vehicle trips. Chapter 4 examines the effects of
age on when the elderly drive. The aspects examined include
driving at night and during peak hours. Chapter 5 examines the
effects of age on how the elderly drive. The aspects examined
include driving speed, driving on limited-access highways,
vehicle size, and the number of passengers carried. Chapter 6
summarizes the main results and discusses policy implications of
these results.
4
Chapter 2 THE 1990 NATIONWIDE PERSONAL TRANSPORTATION SURVEY
(NPTS)
This chapter describes the 1990 NPTS and defines the variables
that are used in this study. The 1990 NPTS compiles data on a
cross-section of personal travel in the United States for all
purposes and surface modes of transportation in 1990-1991.
SURVEY
The 1990 NPTS was conducted between March 1990 and March 1991
using random-digit dialing and computer-assisted telephone
interviewing. The sample was stratified by geography, quarter-of-
year,month-of-quarter, and day-of-week. A total of 73,579
telephone numbers was randomly selected to identify 26,172
households. Each of the identified households was contacted for
an interview. A total of 21,869 households participated. Each of
the participating households was assigned a 24-hour "travel day"
and a 14-day "travel period." For each participating household, a
household-level interview was conducted with an adult resident of
the household. This interview obtained information on the number
of household vehicles, household location, and household income.
In addition, a roster containing person data for each resident of
the household was completed.
A person-level interview was attempted for each resident of the
participating households who was five years or older. The
person-level interview was completed for 47,499 household
residents. Each resident older than 13 years was asked to report
all trips they had taken during the travel day, as well as trips
of 75 miles or longer taken during the travel period. A
"knowledgeable" household resident, older than 13 years, was
asked to report all trips taken by household residents between
the ages of 5 and 13 years. The 1990 NPTS data for this study are
contained in four files in the Statistical Analysis System (SAS)
format. The four files are the Household File, Person File,
Vehicle File, and Travel Day File. The Household File contains
household characteristics for 22,317 observations. The
information collected includes household race, household income,
household size, and household location, such as census region,
the location in an urbanized area, the size of an urbanized area,
and the population density of a zip-code area. Also included are
the sunrise and sunset times associated with the travel day. The
Person File contains the person-level attributes for 48,385
esidents of the participating households. The information
collected includes the age, educational attainment, driver's
license status, and labor force participation of each household
resident. Participating in the labor force means being employed
or actively looking for employment. The Person File also ontains
the number of vehicle miles and the number of vehicle trips taken
by each resident on the travel day.
5
The Vehicle File contains the attributes for 41,178 vehicles in
the participating households. The information collected includes
the model year, make, model, and main driver of each vehicle. The
Travel Day File contains the attributes of 149,546 trips taken by
residents of the participating households on the travel day. The
information collected includes the purpose, mode, occupancy,
length (both duration and distance), time-of-day, day-of-week,
and month-of-year of each trip. The survey also randomly selected
a private-vehicle trip for each resident of the participating
households (if any) to collect information on the various types
of roadways that were used on this trip. A total of 31,015 such
trips was sampled. The distance for each of these trips was
broken down by roadway classification. Weights were developed in
the 1990 NPTS to reflect the sample design and selection
probabilities, and survey non-response or non- coverage. The
Household and Vehicle Files have the same weight variable. The
Person and Travel Day Files have separate weight variables. A
weight variable was also developed for the randomly selected
private-vehicle trips.
VARIABLES The variables used in this study are defined in Table
2.1. They are organized into five groups: personal, household,
location, trip, and vehicle characteristics.
6
Click HERE for graphic.
Table 2.1 Definition of variables
7
Chapter 3 THE EFFECTS OF AGE ON HOW MUCH THE ELDERLY DRIVE
This chapter examines the effects of age on the amount of driving
exposure by the elderly. Three measures of driving exposure are
considered. These measures are the number of vehicle miles driven
by individual drivers on the travel day, the number of vehicle
trips taken by individual drivers on the travel day, and the
distance of individual vehicle trips on the travel day. Each of
these measures is first tabulated by driver age group and labor
force participation. Regression analysis is then used to isolate
the effects of age on each of these measures.
NUMBER OF DAILY VEHICLE MILES DRIVEN
TABULATION
Table 3.1 tabulates the average number of vehicle miles driven on
the travel day by driver age group and labor force participation.
On average, elderly persons in the labor force drive about l9
miles a day and those not in the labor force drive about 10 miles
a day. In comparison, mid-aged persons in the labor force drive
about 29 miles a day, and those not in the labor force drive
about 16 miles a day; and young persons in the labor force drive
about 27 miles a day, and those not in the labor force drive
about 3 miles a day.
Click HERE for graphic.
Table 3. 1 Average number of daily vehicle
miles driven by
Source: Computed from the Person File as the weighted average of
total vehicle miles driven by each responding driver on the
travel day.
REGRESSION Regression analysis is used to isolate the effects of
age on the number of vehicle miles driven by individual elderly
drivers on the travel day. Regression analysis isolates these
effects by including age and other personal, household, and
location characteristics of the elderly drivers as control
variables. The number of vehicle miles driven by individual
drivers is the dependent variable. The age and other
characteristics of individual drivers are the explanatory
variables.
8
Model The first candidate model for this regression analysis
would be the standard linear regression model. This model can be
defined as follows:
Click HERE for graphic.
where yi is the dependent variable; i indicates an observation in
the data; b is a column vector of unknown parameters; xi is a
column vector of known values of the explanatory variables for
observation i; and u, is a disturbance term for observation, that
is independently and normally distributed across observations
with a zero mean and common variance. If the assumptions of this
model are not met, parameters estimated from the ordinary least
squares method may not have properties such as consistency or
efficiency. The current problem violates the assumption that the
disturbance term has a zero mean. About 40 percent of the
responding drivers reported no vehicle miles driven on the travel
day. This situation fits the Tobit model, which originally was
formulated to analyze survey data of consumer expenditures on
durable goods. Most households report zero expenditures on major
durable goods during any year. Among those households that report
any such expenditures, however, the amounts vary widely. The
Tobit model can be defined as follows:
Click HERE for graphic.
otherwise
where RHS refers to the right hand side and the other symbols are
as defined in the standard linear regression model in equation
(1). The ordinary least squares method in this situation leads to
inconsistent estimates of the unknown parameters. Consistent
estimates in the Tobit model can be obtained with the maximum
likelihood or Heckman two-stage method. The Heckman method is
easier to compute, but less efficient.1 Therefore, the maximum
likelihood method is used for this analysis.2
Results Many factors could affect the number of vehicle miles
driven on a given day by individual drivers. These factors
include the characteristics associated with the drivers as well
as the cost of driving. While the 1990 NPTS contains a set of
personal, household, and location characteristics of the drivers,
it does not, however, include information on the cost of driving.
As a result, the cost of driving is approximated by the statewide
average refiner/reseller sales price of motor gasoline plus state
gasoline tax in 1990.3 This cost of driving ignores any variation
in the refiner/reseller sales price of motor gasoline within a
state and in non-state local gasoline taxes. This cost of driving
also ignores other components of driving costs. This cost of
driving, in cents per gallon, will be referred to as gasoline
price. The results are shown in Table 3.2. The first column lists
the explanatory variables by category. The second column lists
the estimated coefficients, measuring the marginal effects
9
Click HERE for graphic.
Table 3.2 Tobit analysis of daily vehicle
miles driven
Source: Estimated from the Person File using the maximum
likelihood method with the SAS LIFEREG procedure. The dependent
variable is total number of vehicle miles driven on the travel
day by each responding driver. Whether a coefficient differs from
zero is labeled as follows: n significant at the 5 percent level
; insignificant at the 10 percent level ; others significant at
the 1 percent level.
10
of an explanatory variable on the dependent variable while
holding constant other explanatory variables. The last column
lists the corresponding chi- square (X2) statistics, indicating
the statistical significance of the explanatory variables. At the
bottom are the log likelihood at convergence, the number of
observations used in the estimation, and the proportion of
observations with zero miles. 4
Two issues are involved in the interpretation of the results.
First, the sign of a coefficient in a Tobit model measures the
direction of changes in the dependent variable from a change in
the corresponding explanatory variable. But to compute the
magnitude of these changes in the dependent variable is not
straightforward. The interpretation here focuses on the signs. 5
The second issue involved in the interpretation of the results
concerns dummy variables. Since the model includes a constant
term, the dummy variable coefficients are interpreted relative to
the omitted category. For example, the dummy variable for male
drivers is included, but the dummy variable for female drivers is
omitted. The omitted category becomes a benchmark. The dummy
variable coefficients for the remaining categories tell whether
or not each of the remaining categories differ from this
benchmark and, if so, by how much. There are two types of dummy
variables: those involving two categories and those involving
more than two categories. The two-category dummy variables
include gender, educational attainment, labor force
participation, Hispanic status, single status, location in an
urbanized area, month-of-year, and day-of-week. The
multi-category dummy variables include age, race, and census
region. The omitted category for age includes those persons
between the ages of 25 and 64 years; the remaining categories
include those persons age 24 years or younger and those persons
age 65 years or older. The omitted category for race includes
those persons who are neither White nor Black; the remaining
categories include White persons and Black persons. The omitted
category for census region is the West; the remaining categories
include the North East, North Central, and South regions. The
results indicate that the coefficient of the elderly dummy
variable is -3.8392 and differs from zero at the 5 percent level.
Thus, other things being equal, the elderly drive fewer miles
than the mid-aged. The other variables are organized into two
groups for interpretation. The first group includes those
variables whose coefficients differ from zero at up to the 10
percent level. The results indicate that, other things being
equal, persons in the labor force drive more miles than those not
in the labor force; males drive more miles than females; Whites
drive more miles than drivers who are neither White nor Black;
Blacks drive fewer miles than Whites; persons with higher
household incomes drive more miles; and persons from households
with more children under five years old drive more miles. In
addition, the young drive fewer miles than the midaged; persons
from households without vehicles drive fewer miles than those
with vehicles; persons living in areas with higher population
densities drive fewer miles; persons living in central cities
drive fewer miles than those living outside central cities; and
the number of daily vehicle miles driven by individual persons
decreases with an increase in gasoline price.
11
The second group includes those variables whose coefficients do
not differ from zero at the 1 0 percent level. The results
indicate that, other things being equal, Blacks drive the same
number of daily vehicle miles as those who are neither White nor
Black; Hispanics drive the same number of daily vehicle miles as
non-Hispanics; the size of an urbanized area does not affect the
number of daily vehicle miles driven by individual persons; and
census region does not make a difference in the number of daily
vehicle miles driven by individual persons.
NUMBER OF DAILY VEHICLE TRIPS
The number of vehicle miles driven combines the number and
distance of vehicle trips. The previous section has shown that
the elderly drive fewer miles than the mid-aged. Does this result
imply that the elderly take shorter trips as well as make fewer
vehicle trips than the midaged? The literature provides mixed
evidence.6 The number of vehicle trips taken on the travel day by
individual drivers and the distance of individual vehicle trips
are examined separately using both tabulation and regression
analysis.
TABULATION Table 3.3 tabulates the average number of vehicle
trips taken on the travel day by driver age group and labor force
participation. On average, elderly persons in the labor force
drive 2.56 vehicle trips per day and those not in the labor force
drive 1.64 vehicle trips per day. Midaged persons in the labor
force drive 2.99 vehicle trips per day and those not in the labor
force drive 2.22 vehicle trips per day. Young persons in the
labor force drive 2.92 vehicle trips per day and those not in the
labor force drive 0.35 vehicle trips per day.
Click HERE for graphic.
Table 3.3 Average number of daily vehicle
trips by driver age group
Source: Calculated from the Person File as the weighted average
of the number of vehicle trips driven by each responding driver
on the travel day.
REGRESSION This regression analysis is similar to that for the
number of vehicle miles driven by individual persons in the
previous section. The unit of observation is individual drivers.
The
12
same set of explanatory variables are used. As mentioned in the
previous section, about 40 percent of the responding drivers
reported no vehicle miles on the travel day. Thus, the Tobit
model in equation (2) is used along with the maximum likelihood
method for estimation. The results are shown in Table 3.4. The
results indicate that the coefficient of the elderly dummy
variable does not differ from zero at the 10 percent level. Thus,
other things being equal, the elderly drive just the same number
of vehicle trips per day as the mid-aged. The other explanatory
variables are organized into three groups for interpretation. The
first group includes those variables whose coefficients differ
from zero at up to the 10 percent level. The results indicate
that, other things being equal, persons in the labor force drive
more vehicle trips than those not in the labor force; persons
with more than a high school education drive more vehicle trips
than those with less education; Whites drive more vehicle trips
than those who are neither White not Black; Blacks drive fewer
vehicle trips than Whites; persons living with children under
five years old drive more vehicle trips than those not living
with children under five years old; and persons from
single-resident households drive more vehicle trips than those
from multi-resident households. In addition, the young drive
fewer vehicle trips than the mid-aged; persons from households
without vehicles drive fewer vehicle trips than those from
households with vehicles; people drive fewer vehicle trips on
weekend days than on weekdays; the number of daily vehicle trips
taken by individual drivers decreases with an increase in the
number of adults in a household; the number of daily vehicle
trips taken by individual drivers decreases with an increase in
the population density of a zip-code area; and the number of
daily vehicle trips taken by individual drivers decreases with an
increase in the size of an urbanized area. The second group
includes those variables whose statistical significance changes
in explaining the number of vehicle miles driven and vehicle
trips taken by individual drivers on the travel day. The results
in Tables 3.2 and 3.4 indicate that, other things being equal,
males drive more miles than females, but not more vehicle trips;
household income affects the number of miles driven, but not the
number of vehicle trips; gasoline price affects the number of
miles driven, but not the number of vehicle trips; and living in
central cities affects the number of miles driven, but not the
number of vehicle trips taken. In addition, the size of an
urbanized area has no effect on the number of miles driven, but
affects the number of vehicle trips taken by individual drivers.
The third group includes those variables whose coefficients that
do not differ from zero at the 10 percent level in explaining
both the number of vehicle miles driven and the number of vehicle
trips taken by individual drivers on the travel day. The results
in Tables 3.2 and 3.4 indicate that, other things being equal,
Blacks drive the same number of miles and take the same number of
vehicle trips as those who are neither White nor Black; Hispanics
drive the same number of miles and take the same number of
vehicle trips as non-Hispanics; and census region does not make a
difference in explaining the number of miles driven or the number
of vehicle trips taken by individual drivers.
13
Click HERE for graphic
Table 3.4 Tobit analysis of number of daily
vehicle trips
Source: Estimated from the Person File using the maximum
likelihood method with the SAS LIFEREG procedure. Whether a
coefficient differs from zero is marked as follows: n significant
at the 5 percent level; insignificant at the 10 percent level:
others significant at the 1 percent level.
14
DISTANCE OF DAILY VEHICLE TRIPS
TABULATION Table 3.5 tabulates the average distance of vehicle
trips taken on the travel day by driver age group and trip
purpose. For elderly drivers, the average distances are 6.55
miles for all trips, 8.30 miles for work trips, and 6.43 miles
for non-work trips. For mid-aged drivers, the average distances
are 9.25 miles for all trips, 11.54 miles for work trips, and
8.22 miles for nonwork trips. For young drivers, the average
distances are 8.91 miles for all trips, 9.98 miles for work
trips, and 8.54 miles for non-work trips. For all drivers, the
average distances are 8.98 miles for all trips, 11.23 miles for
work trips, and 8.10 miles for non-work trips.
Click HERE for graphic.
Table 3.5 Average distance of daily vehicle
trips by driver age group
Source: Calculated from the Travel Day File as the weighted
average of distances of individual vehicle trips on the travel
day in miles.
REGRESSION As with the models developed for the number of vehicle
miles driven and the number of vehicle trips taken by individual
drivers on the travel day, the purpose of this regression
analysis is to isolate the effects of age on the distance of
individual vehicle trips taken by elderly drivers on the travel
day.
Model The regression analysis in this section differs from those
in the previous sections in two important aspects. First, while a
large proportion of responding drivers reported no vehicle trips
on the travel day, the variable measuring the distance of vehicle
trips does not have this problem. Instead of the Tobit model in
(2), the standard linear regression model in (1) is used along
with the weighted least squares method for estimation. Second,
while the unit of observation in the previous sections is
individual drivers, the unit of observation in this section is
individual vehicle trips. As a result, an additional set of
explanatory variables measuring trip characteristics is also
included. These additional variables include time-of- day,
whether the driver carried any passengers, day-of-week,
month-of-year, and the purpose of a vehicle trip.
15
Results The results are shown in Table 3.6. The interpretation of
the standard linear model is straightforward. The coefficient of
an explanatory variable measures the expected change in the value
of the dependent variable from one unit change in the explanatory
variable, while holding other explanatory variables constant.
Another issue of interpretation is the set of dummy variables
that measures trip purposes. The 1990 NPTS classifies trip
purposes into ten categories. Four of these categories are
omitted from the model: trips for school or church, trips for
vacation, trips for pleasure driving, and trips for other
purposes. The remaining six categories are included in the model.
As a result, the coefficients of the dummy variables for these
remaining categories are interpreted relative to the omitted
categories. The results indicate that the coefficient of the
elderly dummy variable is -1.0471 and differs from zero at the
0.01 percent level. Thus, other things being equal, the elderly
drive about one mile shorter per trip than the mid-aged. The
other variables are organized into two groups for interpretation.
The first group includes those variables whose coefficients
differ from zero at up to the 10 percent level. The results
indicate that, other things being equal, male drivers take longer
trips than female drivers; drivers in the labor force take longer
trips than those not in the labor force; White drivers take
longer trips than those who are neither White nor Black; Blacks
take trips of shorter distances than those taken by Whites;
drivers with higher household incomes take longer trips; and
drivers living in larger urbanized areas take longer trips. In
addition, drivers living in central cities take shorter trips
than those living outside central cities; the distance of vehicle
trips decreases with an increase in gasoline price; drivers
living in areas with higher population densities take shorter
trips; trips for work- related purposes and for visiting friends
or relatives are longer than trips for those purposes that are
omitted from the model; and trips for other remaining purposes
are shorter than trips for those purposes that are omitted from
the model. The second group includes those variables whose
coefficients do not differ from zero at the 1 0 percent level.
The results indicate that, other things being equal, young
drivers take trips that are just as long as those taken by
mid-aged drivers; winter trips are just as long as nonwinter
trips; night trips are just as long as day trips; peak trips are
just as long as off-peak trips; Black drivers take trips that are
just as long as those taken by drivers who are neither White nor
Black; Hispanic drivers take trips that are just as long as those
taken by non-Hispanic drivers; and drivers in the North East or
South regions take trips that are just as long as trips taken by
those in the West.
16
Click HERE for graphic.
Table 3.6 Weighted regression of distance of
daily vehicle trips
Source:Estimated by Author from the Travel Day File using the
weighted least squares method. Whether a coefficient differs from
zero is labeled as follows: n significant at the 5 percent level,
u significant at the 10 percent level; insignificant at the 10
percent level: others significant at the 0.01 percent level.
17
Chapter 4 THE EFFECTS OF AGE ON WHEN THE ELDERLY DRIVE
This chapter examines the effects of age on driving at night or
during peak hours by the elderly. Night includes the hours after
sunset and before sunrise. Peak hours include 6:30-9:00 a.m. and
3:30-6:00 p.m. Whether a vehicle trip was taken at night or
during peak hours is determined by its start time. Driving at
night is examined first, followed by an examination of driving
during peak hours. For each analysis, the percent of vehicle
miles driven by time of day is first tabulated by driver age
group and trip purpose. Logit analysis is then used to isolate
the effects of age on the elderly's probability of driving at
night or during peak hours.
DRIVING AT NIGHT
TABULATION Table 4.1 tabulates the percent of vehicle miles
driven at night by driver age group and trip purpose. The elderly
drive about 18 percent of their miles at night for both work and
nonwork trips, while the mid-aged drive about 29 percent of their
miles at night for work trips and 23 percent for non-work trips.
The young drive about 29 percent of their miles at night for work
trips and 25 percent for non-work trips.
Click HERE for graphic.
Table 4. 1 Percent of miles driven at night
by driver age group
Source: Calculated from the Travel Day File. Each number
represents total miles driven by drivers of a given group at
night as a percentage of total miles driven by these drivers all
day.
REGRESSION The purpose of this regression analysis is to isolate
the effects of age on driving at night by the elderly, while
holding constant a set of the elderly's personal, household, and
location characteristics.
18
Model Similar to the regression analysis of the distance of
vehicle trips in the previous section, the unit of observation is
individual vehicle trips. This regression analysis, however,
differs from that for the distance of vehicle trips in four
aspects. First, the dependent variable here is binary, indicating
whether a vehicle trip on the travel day started at night. One
commonly used regression model for a binary choice problem is the
logit model, in which the probability of choosing to drive at
night has the logit form. If P is the probability of driving at
night, x is a column vector of the values of explanatory
variables, and b is a column vector of parameters, then:
Click HERE for graphic.
e b1x (3) P = 1 = e b1x
Second, speed may differ systematically by time of day. In
addition to a similar set of explanatory variables used in the
model for the distance of vehicle trips, speed is also included
in this analysis. Third, the ordinary least squares method does
not apply here. Instead, the maximum likelihood method is used
for estimation. Fourth, several variables are excluded because
convergence could not be reached when these variables are
included. These excluded variables are Black, Hispanic, and the
census regions. The reason that these particular variables are
chosen to be excluded is that they are thought to be less
important than others in the decision of driving by time of day.
Results The results are shown in Table 4.2. The coefficients in
this model are interpreted differently from those in a standard
linear or Tobit model. First, an increase in a variable with a
negative coefficient decreases the odds ratio of driving at
night. The odds ratio of driving at night is PI(1-P), where P is
the probability of driving at night. Second, the exponential
value of the coefficient of an explanatory variable determines
the percent change in the odds ratio of driving at night from one
unit change in that explanatory variable. For example, the dummy
variable for male drivers has a coefficient of 0.3070. Its effect
on the odds ratio of driving at night is 100*(e 0.3070 - 1) =36
percent. That is, males' odds ratio of driving at night is 36
percent higher than females' odds ratio of driving at night. The
results indicate that the coefficient of the elderly dummy
variable is -0.2183 and differs from zero at the 0.01 percent
level. Thus, other things being equal, the elderly are less
likely to drive at night than the mid-aged. In fact, the
elderly's odds ratio of driving at night is 20 percent lower than
the mid- aged's odds ratio of driving at night. The other
variables are organized into two groups for interpretation. The
first group has positive coefficients. The results indicate that,
other things being equal, the young are more likely to drive at
night than the mid-aged; males are more likely to drive at night
than females; persons in the labor force are more likely to drive
at night than those not in the labor force;
19
Click HERE for graphic.
Table 4.2 Logit analysis of driving at
night
Source: Estimated from the Travel Day File using the maximum
likelihood method with the SAS LOGISTIC procedure. Whether a
coefficient differs from zero is marked as follows: n significant
at the 1 percent level; others significant at the 0.01 percent
level.
20
persons living in central cities are more likely to drive at
night than those living outside central cities; the probability
of driving at night increases with an increase in household
income, the size of an urbanized area, and the population density
of a zip-code area; and trips for work-related purposes, visiting
friends or relatives, and other social or recreational purposes
are more likely to be taken at night than trips for those
purposes that are omitted from the model. The second group has
negative coefficients. The results indicate that, other things
being equal, persons with more than a high school education are
less likely to drive at night than those with less education;
Whites are less likely to drive at night than non-Whites; and
trips for shopping, other family or personal business, and
medical purposes are less likely to be taken at night than trips
for those purposes that are omitted from the model. Note that the
omitted category for race in this analysis is non- Whites.
DRIVING DURING PEAK HOURS
TABULATION Table 4.3 tabulates the percent of vehicle miles
driven during peak hours by driver age group and trip purpose.
The elderly drive about 28 percent of their miles during peak
hours for non-work trips, 57 percent for work trips, and 30
percent for all trips. The mid-aged drive about 31 percent of
their miles during peak hours for non-work trips, 59 percent for
work trips, and 39 percent for all trips. The young drive about
38 percent of their miles during peak hours for nonwork trips, 50
percent for work trips, and 40 percent for all trips.
Click HERE for graphic.
Table 4.3 Percent of miles driven during
peak hours by driver age group
Source: Calculated from the Travel Day File. Each number
represents total miles driven by drivers of a given group during
peak hours as a percentage of total miles driven by these drivers
all day.
REGRESSION The regression analysis of driving during peak hours
is similar to that for driving at night. Again, the dependent
variable is binary, indicating whether a vehicle trip on the
travel day started during peak hours. The same set of explanatory
variables are included as in the regression
21
analysis for driving at night. The logit model is used along with
the maximum likelihood method for estimation. The results are
shown in Table 4.4. The results indicate that the coefficient of
the elderly dummy variable is -0.1251 and differs from zero at
the 1 percent level. Thus, other things being equal, the elderly
are less likely to drive during peak hours than the mid-aged. In
fact, the elderly's odds ratio of driving during peak hours is
about 12 percent lower than the odds ratio of driving during peak
hours by the mid-aged. This difference in the odds ratio of
driving during peak hours between the elderly and mid-aged is
smaller than that for the odds ratio of driving at night. This
change in the difference is consistent with that the elderly find
driving at night more problematic than driving during peak hours.
The other variables are organized into three groups for
interpretation. The first group includes those variables whose
coefficients are positive and differ from zero at the 10 percent
level. The results indicate that, other things being equal,
persons in the labor force are more likely to drive during peak
hours than those not in the labor force; persons with more than a
high school education are more likely to drive during peak hours
than those with less education; weekend trips are more likely to
be taken during peak hours than weekday trips; and work trips are
more likely to be taken during peak hours than trips for those
purposes that are omitted from the model. The second group
includes those variables whose coefficients are negative and
differ from zero at the 1 0 percent level. The results indicate
that, other things being equal, the young are less likely to
drive during peak hours than the mid-aged; males are less likely
to drive during peak hours than females; trips for shopping,
other family-or personal business, medical, visiting friends or
relatives, and other social or recreational purposes are less
likely to be taken during peak hours than trips for those
purposes that are omitted from the model. The last group includes
those variables whose coefficients do not differ from zero at the
10 percent level. The results indicate that, other things being
equal, Whites are just as likely as non-Whites to drive during
peak hours; household income or the size of an urbanized area
does not affect the probability of driving during peak hours;
persons living in central cities are just as likely as those
living outside central cities to drive during peak hours; and
winter trips are just as likely as non-winter trips to be taken
during peak hours.
22
Click HERE for graphic.
Table 4.4 Logit analysis of driving during
peak hours
Source: Estimated by from the Travel Day File using the maximum
likelihood method with the SAS LOGISTIC procedure. Whether a
coefficient differs from zero is labeled as follows: n
significant at the 10 percent level; insignificant at the 10
percent level; others significant at the 1 percent level.
23
Chapter 5 THE EFFECTS OF AGE ON HOW THE ELDERLY DRIVE
Chapters 3 and 4 have shown that age affects how much, as well as
when the elderly drive. This chapter examines the effects of age
on how the elderly drive. Four aspects are considered. These
include driving speed, driving on limited-access highways,
vehicle size, and the number of passengers carried.
SPEED
This section examines the effects of age on the driving speeds of
the elderly. Do the elderly drive at lower speeds than others? If
they do, do they drive on roads with lower speed limits? Or do
they drive slower than others on roads with the same speed
limits? The 1990 NPTS can be used to shed light on whether the
elderly drive slower than others on limited access highways. The
1990 NPTS does not, however, include the information necessary to
test whether the elderly drive on roads with lower speed limits
than others. In the following analysis, speed is first tabulated
by driver age group and trip purpose. Regression is then used to
isolate the effects of age on the driving speeds of the elderly.
This analysis is done separately for all roadways combined and
for limited-access highways.
TABULATION
Table 5.1 tabulates the average speed for vehicle trips using all
roads by driver age group and trip purpose. The elderly drive at
an average speed of 22 mph for all trips, 24 mph for work trips,
and 22 mph for non- work trips. The mid-aged drive at an average
speed of 29 mph for all trips, 31 mph for work trips, and 28 mph
for non-work trips. The young drive at an average speed of 32 mph
for all trips, 34 mph for work trips, and 31 mph for non-work
trips.
Click HERE for graphic.
Table 5.1 Average speed on all roads by
driver age group
Source: Calculated from the Travel Day File as the weighted
average of the speeds of individual vehicle trips. The speed of a
trip is measured as the ratio of its reported distance and
duration in miles per hour (mph).
24
Table 5.2 tabulates the average speed for vehicle trips using
limited-access highways by driver age group and trip purpose. As
expected, the average speeds for trips using limited access
highways are higher than those for trips using all roadways. On
average, the elderly drive at about 34 mph for all purposes, 36
mph for work trips, and 33 mph for non-work trips. The mid-aged
drive at about 39 mph for work trips, non-work trips, and all
purposes. The young drive at about 44 mph for all trips, 44 mph
for work trips, and 42 mph for non-work trips. All persons as a
group drive at about 39 mph for both work and non-work trips.
Click HERE for graphic.
Table 5.2 Average speed on limited-access
highways by driver age group
Source: Calculated from the sample of private-vehicle trips in
the Travel Day File as the weighted average of the speeds for
individual trips in this sample. The distance of each trip in
this sample is broken down by roadway classification.
REGRESSION This regression analysis is similar to that for the
distance of vehicle trips in Chapter 3. The unit of observation
is individual vehicle trips. The dependent variable is the speed
of individual vehicle trips, measured as the ratio of reported
distance and duration in miles per hour. The same set of
explanatory variables are included as in the analysis of the
distance of vehicle trips except gasoline price. The standard
linear regression model in equation (1) is used along with the
ordinary least squares method for estimation. The results are
presented in Table 5.3. The model for trips using limited-access
highways is shown in the second and third columns. The model for
trips using all roadways is shown in the last two columns. The
results indicate that the elderly drive at lower speeds than the
mid-aged for trips using all roads as well as for trips using
limited-access highways. The model for all roadways indicates
that, other things being equal, the elderly drive 3.9 mph slower
than the mid-aged for trips using all roadways. The model for
limited-access highways indicates that, other things being equal,
the elderly drive 3.7 mph slower than the mid-aged for trips
limited-access highways. The other variables are organized into
four groups for interpretation. Those in the first group have a
positive effect in both models. The results indicate that, other
things being equal, the young drive at higher speeds than the
mid-aged for both trips using all roadways and trips
25
Click HERE for graphic.
Table 5.3 Weighted regression of speed of
vehicle trips
Source Estimated from the Travel Day File using the weighted
least squares method. Whether a coefficient differs from zero is
labeled as follows: n significant at the 5 percent level; u
significant at the 10 percent level ; insignificant at the 10
percent level : others significant at the 1 percent level.2
using limited-access highways. Similarly, males drive at higher
speeds than females; persons with higher household incomes drive
at higher speeds; weekend trips have higher speeds than weekday
trips; and trips for medical and visiting friends or relatives
have higher speeds than trips for the purposes that are omitted
from the models. The variables in the second group have a
negative effect in both models. The results indicate that, other
things being equal, persons living in areas with higher
population densities drive at lower speeds for both trips using
all roadways and trips using limited-access highways. Similarly,
peak trips have lower speeds than off-peak trips. The variables
in the third group have a positive effect in the model for all
roadways, but have no effect in the model for limited-access
highways. The results indicate that, other things being equal,
persons with more than a high school education drive at higher
speeds than those with less education for all roadways, but at
similar speeds on limited-access highways. The size of an
urbanized area increases the speeds for trips using all roadways,
but has no effect for trips using limited-access highways. Since
limited-access highways generally have higher speeds than local
roadways, the positive relationship between the size of an
urbanized area and the speeds for trips using all roadways may
imply that trips in larger urbanized areas are more likely to use
limited-access highways. In fact, the analysis of driving on
limited-access highways in the next section confirms this
implication. Similarly, night trips have higher speeds than
day-time trips on all roadways, but have similar speeds on
limited-access highways; and work trips on all roadways have
higher speeds than trips for those purposes that are omitted from
the models, but have similar speeds on limited-access highways.
Also, carpool trips have higher speeds than single- occupant
trips on all roadways, but have similar speeds on limited-access
highways. It is reasonable that carpool trips have higher speeds
than single- occupant trips on all roadways because carpool trips
may be more likely to use limited-access highways. The variables
in the last group have a negative effect in the model for all
roadways, but have no effect in the model for limited-access
highways. The results indicate that, other things being equal,
persons living in central cities drive at lower speeds than those
living outside central cities for all roadways, but drive at
similar speeds on limited-access highways. Similarly, persons in
the North East or North Central regions drive at lower speeds
than those in the West on all roadways, but drive at similar
speeds on limited-access highways. Also shopping trips and trips
for other family or personal business have lower speeds than
trips for the omitted trip purposes on all roadways, but have
similar speeds on limited-access highways.
LIMITED-ACCESS HIGHWAYS
This section examines the effects of age on the elderly's choice
of driving on limited-access highways. It is unclear, at the
outset, how age may affect the elderly's use of limited-access
highways. Limited-access highways have the lowest fatal crashes
per mile driven.' But they are also likely to have higher injury
risks from crashes due to the high speeds. As
27
discussed in Chapter 1, however, driving on limited-access
highways is one of the commonly mentioned conditions that the
elderly find difficult. The percent of vehicle miles driven on
limited-access highways is first tabulated by driver age group
and trip purpose. Logit analysis is then used to isolate the
effects of age on the elderly's probability of driving on
limited-access highways.
TABULATION
Table 5.4 tabulates the percent of vehicle miles driven on
limited-access highways by driver age group and trip purpose. The
elderly drive 21 percent of their miles on limited-access
highways for work trips and 15 percent for non-work trips. The
mid-aged drive 28 percent of their miles on limited-access
highways for work trips and 26 percent for non-work trips. The
young drive 22 percent of their miles on limited-access highways
for work trips and 24 percent for non-work trips.
Click HERE for graphic.
Table 5.4 Percent of miles driven on
limited-access highways by driver age group
Source: Calculated from the Travel Day File. The 1990 NPTS
randomly selects a private-vehicle trip for each respondent (if
any), and breaks down its distance by roadway classification.
REGRESSION
This regression analysis is similar to that for driving at night
or during peak hours. The dependent variable is binary,
indicating whether a vehicle trip uses any mited-access highways.
The logit model is used along with the maximum likelihood method
for estimation. Two models are estimated in order to examine how
controlling for speed affects the elderly's choice of driving on
limited-access highways. The results are shown in Table 5.5.
Model 1 includes speed; Model 2 does not include speed. The
results in both models indicate that, other things being equal,
the elderly are less likely to drive on limited-access highways
than the mid- aged. The coefficients of the elderly dummy
variable are -0.5618 in Model 1 and -0.7364 in Model 2 and both
differ from zero at the 0.1 percent level. Thus, when speed is
not held constant (Model 2), the elderly's odds ratio is 52
percent lower than the mid-aged's odds ratio of driving on
limited-access highways. When speed is also held constant (Model
1), the elderly's odds ratio is 49 percent lower than the mid-
28
Click HERE for graphic.
Table 5.5 Logit analysis of driving on
limited-access highways
Source: Estimated from the sample of trips for which distances
are broken down by roadway classification. Whether a coefficient
differs from zero is labeled as follows: n significant at the 5
percent level . u significant at the 10 percent level u
insignificant at the 10 percent level: others significant at the
0.1 percent level.
29
aged's odds ratio of driving on limited-access highways. So, the
elderly's odds ratio of driving on limited-access highways
decreases slightly (from 52 to 49 percent) when speed is
controlled. This slight decrease seems to indicate that the
elderly avoid driving on limited-access highways mainly for
reasons other than high speeds. The other variables are organized
into three groups for interpretation. The first group includes
variables whose coefficients differ from zero at the 10 percent
level in both models. The results indicate that, other things
being equal, males are more likely to drive on limited-access
highways than females; persons with more than a high school
education are more likely to drive on limited-access highways
that those with less education; the probability of driving on
limited access highways increases with an increase in the size of
an urbanized area; limited-access highways are more likely to be
used for carpool trips than for non-carpool trips; limited-access
highways are more likely to be used for works trips than for
trips for purposes that are omitted from the models. In addition,
the probability of driving on limited-access highways decreases
with an increase in household income; persons in other census
regions are less likely to drive on limited-access highways than
those in the West; limited-access highways are less likely to be
used for peak trips and for off-peak trips; and limited- access
highways are more likely to be used for shopping and other family
or personal business than for trips for the purposes that are
omitted from the models. The second group includes those
variables that do not differ from zero at the 10 percent level in
either models. The results indicate that, other things being
equal, race makes no difference in the choice of driving on
limited-access highways; limited-access highways are more likely
to be used for night trips than for day trips; limited-access
highways are as likely to be used for weekend trips as for
weekday trips; limited-access highways are as likely to be used
for trips for medical, visiting friends or relatives, and other
social or recreational purposes as for trips for those purposes
that are omitted from the models. The last group includes
variables whose statistical significance changes between the two
models. The results indicate that, other things being equal,
greater population density increases the probability of driving
on limited-access highways when speed is held constant, but shows
no effect when speed is not held constant; living in central
cities increases the probability of driving on limited-access
highways when speed is not held constant, but shows no effect
when speed is also held constant; and persons in the labor force
are more likely than persons not in the labor force to drive on
limited-access highways when speed is not held constant, but are
as likely to drive on limited-access highways when speed is also
held constant.
AUTOMOBILE SIZE
This section examines the effects of age on the size of
automobiles that the elderly drive. Do the elderly drive larger
automobiles than others? The answer is not straightforward. As
discussed in the introduction, the increased injury risk and
reduced injury costs of the elderly may have two opposite effects
on the elderly's choice of automobile size. In addition, if one
30
assumes that the elderly value comfort or prestige more than
others, one may argue that the elderly may drive larger
automobiles for these reasons rather than for their
crashworthiness. The literature, however, provides no evidence
that the elderly value comfort or prestige more than others.
Also, the fact that elderly drivers take trips that are shorter
in distance, as shown in Chapter 3, suggests that the comfort of
an automobile is less important for the elderly than for others.
The 1990 NPTS associates each vehicle used on the travel day with
a main driver. This association allows one to link the
characteristics of the main drivers with the attributes of the
vehicles that they drive. The 1990 NPTS measures vehicle size
according to the National Accident Sampling System.1 The size of
an automobile is based on its wheelbase length and is coded on a
scale from one to six. For example, the size of a Ford Escort is
one and the size of a Toyota Camry is three. Only automobiles are
included in the analysis. Non-householdowned automobiles are
excluded because they cannot be related to household attributes
of the main drivers. The following analysis starts with a
tabulation of automobile size by age group of the main drivers
and labor force participation. Regression is then used to isolate
the effects of age on the size of automobiles that the elderly
drive.
TABULATION Table 5.6 tabulates the average size of automobiles by
age group of the main drivers and labor force participation. For
persons not in the labor force, the average sizes of the
automobiles they drive are 3.16 for the elderly, 2.85 for the
mid-aged, 2.52 for the young, and 2.88 for all. For those in the
labor force, the average sizes are 2.90 for the elderly, 2.61 for
the mid-aged, 2.35 for the young, and 2.58 for all.
Click HERE for graphic.
Table 5.6 Average size of automobiles by
age group of main drivers
Source: Calculated from the Vehicle and Person Files as the
weighted average of automobile sizes. The size of an automobile
is based on its wheelbase length, and is on a scale from one to
six.
31
REGRESSION The dependent variable is the size of an automobile
measured on a scale from one to six. Unlike the regression
analyses so far, where the unit of observation is either
individual drivers or vehicle trips, the unit of observation here
is individual automobiles. This analysis is similar, however, to
those for the distance and speed of vehicle trips in that the
standard linear regression model in equation (1) is used along
with the weighted least squares method for estimation. The
results are shown in Table 5.7. Two models are estimated. Model 1
includes a set of personal, household, and location
characteristics of the main drivers. In addition to these
characteristics, Model 2 also includes two vehicle attributes:
vehicle age and import status (whether a vehicle is
foreign-made). The results indicate that the coefficients of the
elderly dummy variable are 0.4039 in Model 1 and 0.2574 in Model
2, and both differ from zero at the 0.01 percent level. Thus,
other things being equal, the elderly drive larger automobiles
than the mid-aged. The other explanatory variables are organized
into three groups for interpretation. The first group includes
variables whose coefficients differ from zero at the 10 percent
level in both models. The results indicate that, other things
being equal, the young drive smaller automobiles than the mid-
aged; persons with more than a high school education drive
smaller automobiles than those with less education; persons in
the labor force drive smaller automobiles than those not in the
labor force; the size of an automobile increases with an increase
in household income, but decreases with an increase in the size
of an urbanized area; and persons in the South drive larger
automobiles than those in the West. The second group includes
variables whose coefficients do not differ from zero at the 1 0
percent level in either model. The results indicate that, other
things being equal, living in central cities does not affect the
size of an automobile one drives and persons in the South East
drive automobiles that are as large as those driven by persons in
the West. The third group includes variables whose statistical
significance changes between the two models. The results indicate
that, other things being equal, males are shown to drive larger
automobiles than females when vehicle age and import status are
not held constant (Model 1). But once vehicle age and import
status are held constant (Model 2), males drive automobiles that
are the same size as those driven by females. Similar changes in
statistical significance are also observed for Whites, Blacks,
household size, and persons living in the North Central region.
On the other hand, when vehicle age and import status are not
held constant (Model 1), Hispanics are shown to drive automobiles
that are the same size as those driven by nonHispanics. Once
vehicle age and import status are given (Model 2), however,
Hispanics are shown to drive smaller automobiles. Two
qualifications are in order. First, these models do not include
owning and operating costs as an explanatory variable, though
there is no reason to believe that including such a cost variable
would necessarily change the results. It is possible to estimate
these costs using other sources with the information on vehicle
make and model. However, estimating these costs would require
additional resources and is beyond the scope of this study.
32
Click HERE for graphic.
Table 5.7 Weighted regression of automobile
size
Source: Estimated from the Vehicle and Person Files with the
weighted least squares method. Whether a coefficient differs from
zero is labeled as follows: n significant at the 1 percent level;
u significant at the 10 percent level; insignificant at the 10
percent level; others significant at the 0.01 percent level.
33
NUMBER OF PASSENGERS CARRIED
This section examines the effects of age on the number of
passengers that the elderly carry. Given that the elderly show
increased crash involvements per unit of exposure, one might
hypothesize that they feel less comfortable with carrying
passengers than younger persons. The following analysis first
tabulates the average automobile occupancy by driver age group
and trip purpose. Regression is then used to isolate the effects
of age on the number of passengers carried in each vehicle trip
on the travel day.
TABULATION Table 5.8 tabulates the average occupancy of
automobile trips by driver age group and trip purpose. The
elderly's average ccupancies are 1.39 for all purposes, 1.08 for
work trips, and 1.41 for non-work trips. The mid-aged's average
occupancies are 1.54 for all purposes, 1.14 for work trips, and
1.71 for non-work trips. The young's average occupancies are 1.44
for all purposes, 1.10 for work trips, and 1.56 for non-work
trips.
Click HERE for graphic.
Table 5.8 Average occupancy of automobile
trips by driver age group
Source: Calculated from the Travel Day File as the weighted
average of occupancies of individual automobile trips on the
travel day.
REGRESSION The dependent variable is the number of occupants in
an automobile trip on the travel day. This regression analysis is
similar to those for the distance and speed of vehicle trips in
two ways. First, the unit of observation is individual vehicle
trips. Second, the standard linear regression model in equation
(1) is used along with the weighted least squares method for
estimation. This analysis differs, however, from those for the
distance and speed of vehicle trips in that this analysis
includes additional variables that measure household composition
and vehicle ownership. The results are shown in Table 5.9. The
results indicate that the coefficient of the elderly dummy
variable is -0.0558 and differs from zero at the 1 percent level.
Thus, other things being equal, the elderly carry fewer
passengers than the mid-aged.
34
Click HERE for graphic.
Table e 5. 9 Weighted regression of
occupancy of automobile trips
Source: Estimated from the Travel Day File with the weighted
least squares method. Whether a coefficient differs from zero is
labeled as follows: n significant at the 1 percent level; u
significant at the 10 percent level : insignificant at the 10
percent level ; others significant at the 0.01 percent level.
35
The other variables are interpreted by category of
characteristics. Among the personal characteristics, the young
carry fewer passengers than the mid-aged and persons in the labor
force carry fewer passengers than those not in the labor force.
In addition, males carry just as many passengers as females.
Among the household characteristics, automobile occupancy
decreases with an increase in household income and vehicle
ownership; persons from household with more children between the
ages of 5 and 22 years carry more passengers; persons from
single-resident households carry fewer passengers than those from
multi-person households; and Blacks carry fewer passengers than
non-Blacks. Also, Whites carry as many passengers as those who
are neither White nor Black; and Hispanics carry as few
passengers as non-Hispanics. Among the location characteristics,
automobile occupancy increases with an increase in population
density, but decreases with an increase in the size of an
urbanized area; automobile occupancy is lower in the other census
regions than in the West. In addition, living in central cities
does not affect automobile occupancy. Gasoline price, as measured
in this analysis, has a positive but statistically insignificant
effect on automobile occupancy. Among the trip haracteristics,
night trips have higher occupancies than day trips; weekend trips
have higher occupancies than weekday trips; and long distance
trips have higher occupancies than short distance trips. In
addition, trips for other social or recreational purposes have
higher occupancies than trips for those purposes that are omitted
from the model; and trips for the other remaining purposes
included in the model (work-related, shopping, other
family/personal business, medical, and visiting
friends/relatives) have lower occupancies than trips for the
omitted purposes. The omitted purposes include trips for school
or church, trips for vacation, trips for pleasure driving, and
trips for other purposes.
36
Chapter 6 SUMMARY AND POLICY IMPLICATIONS
This report has examined the effects of age on six driving habits
of the elderly (persons age 65 years or older). This chapter
summarizes the main results and discusses the implications of
these results to policy-making in areas concerning the mobility
and traffic safety of the elderly.
SUMMARY
Elderly drivers show an increased effort of self-protection in
their driving habits relative to mid-aged drivers (persons
between the ages of 25 and 64 years). Elderly drivers not only
reduce daily driving exposure, avoid driving at night, avoid
driving during peak hours, and avoid driving on limited-access
highways, but also drive at lower speeds, drive larger
automobiles, and carry fewer passengers. The following summarizes
the results for each of the six driving habits examined.
* Daily Driving Exposure. The elderly reduce their daily driving
exposure by reducing not the frequency but the istance of vehicle
trips. The elderly drive fewer vehicle miles than the mid-aged.
They take as many vehicle trips as the mid-aged, but their
vehicle trips are shorter in distance than those taken by the
mid-aged.
* Driving By Time of Day. The elderly are less likely to drive at
night and during peak hours than the mid-aged. In addition, the
elderly are lesser likely to drive at night than to drive during
peak hours. This is consistent with the fact that the elderly
find driving at night more problematic than driving during peak
hours.
* Driving By Roadway Type. The elderly are less likely to drive
on limited-access-highways than the mid-aged. This avoidance
behavior by the elderly can be due to many characteristics of
limited-access-highways, such as high speeds. When speed is held
constant, however, the elderly still are found to be less likely
to drive on limited-access highways. In addition, the elderly's
likelihood of driving on limited-access-highways decreases only
slightly when speed is held constant. This slight decrease seems
to suggest that the elderly avoid driving on
limited-access-highways mainly due to characteristics of
limited-access-highways other than high speeds.
* Driving Speed. The elderly drive at lower speeds than the
mid-aged. They drive about 4 miles per hour (mph) slower than the
mid-aged for all trips. This is either because the elderly are
more likely to drive on roadways with lower speed limits or
because they drive slower on roadways with the same speed limits.
The evidence indicates that both
37
possibilities occur with the elderly. When only Vehicle trips
that use limited-access highways are considered, the elderly are
found to drive about 4 mph slower than the mid-aged. As indicated
earlier, the elderly also are less likely to drive on
Lmited-access-highways.
Automobile Size. The elderly drive larger automobiles than the
mid-aged. When the size of an automobile is measured by wheelbase
size on a scale from one to six, the average size of automobiles
driven by the elderly is 0.40 smaller then that by the mid-aged
when automobile age and import status are not held constant and
is 0.26 smaller when automobile age and import status are held
constant.
Number of Passengers Carried. The elderly carry fewer passengers
than the mid-aged. In fact, the elderly carry an average number
of passengers that is about 0.05 lower than the mid-aged.
These differences in the driving habits between the elderly and
mid-aged reflect the marginal effects of age difference between
the two groups. These differences do not reflect any effects of
the differences between the two groups in other personal,
household, location, and trip characteristics that are held
constant in this study.
POLICY IMPLICATIONS
Despite their increased effort of self-protection in their
driving habits, as summarized above, the elderly still show a
higher risk of crash and injury per unit of exposure than the
mid-aged.1 When the elderly adjust their driving habits, they
consider the risks they face, but not the external risks they
impose on others when they drive. If the elderly are forced to
adjust their driving habits further to offset the external risks
of their driving, their risk of crash and injury would be reduced
and society as a whole would be better off. Any further
adjustment in the elderly's driving habits, however, is likely to
make the elderly worse off due to reduced mobility. The challenge
to policy-making is to balance these consequences of any policy
concerning the mobility and traffic safety of the elderly. The
following discusses four existing policy options.
Removing Hazardous Elderly Drivers from Roadways.2 Removing
elderly drivers through the use of stricter licensing laws is
controversial. First, the removed drivers are forced to pay a
large price-loss of automobile mobility. Second, elderly drivers
have the lowest severe crash nvolvement per driver. If the
purpose is to reduce the maximum number of severe crashes per
removed driver, then removing younger drivers would be far more
effective than removing elderly Drivers. Third, the physical and
cognitive abilities vary widely among the elderly. Forth, such
removal has the appearance of discriminating against elderly
drivers. As a result, the higher the proportion of elderly
drivers that a state has, the harder to implement such an option.
The best example is Florida, where
38
the elderly population as a share of the total population is the
highest in the nation. Three attempts by Florida's legislature to
pass stricter licensing laws for elderly drivers have failed in
the past several years.3
Making Alternatives to Driving Available. 4 This option
accommodates the option of removing elderly drivers from
roadways. Alternatives to driving include walking, public
transit, specialized transportation, and the use of taxis. As
more elderly persons live in suburbs where the population density
is low, these alternatives become less feasible. Walking is
difficult for elderly persons in low density areas, and it is
extremely costly to expand public transit for the elderly in
these areas. Expanding specialized transportation to low density
areas is also expensive. Subsidizing the use of taxis may be more
expensive than specialized transportation.
Improving Vehicle and Roadway Design and Operation. 5 This option
attempts to accommodate the reduced physical and cognitive
abilities of elderly drivers. There is, however, strong evidence
that drivers become more risk-taking when the driving environment
becomes safer. 6 There is no reason to believe that elderly
drivers do not have such a behavior. This behavior would off-set
many of the intended benefits of improving vehicle and roadway
design and operation.
Re-Educating Elderly Drivers. 7 Re-educating elderly drivers
would be an appropriate policy if elderly drivers were not fully
aware of their reduced cognitive and physical abilities and the
consequences to their traffic safety.
As the number of elderly drivers continues to grow, the welfare
of the society as a whole becomes increasingly dependent upon the
mobility and traffic safety of elderly drivers. While this study
has implications to policy-making, policy recommendation is
beyond the scope of this report. Future research needs to examine
the impacts of existing policies, as well as to develop new
policy options that would better balance the effects on the
elderly and society as a whole.
39
ENDNOTES CHAPTER 1
1. Elderly is defined as age 65 years or older. This is the most
commonly used definition in the literature on the mobility and
Safety of elderly persons.
2. Federal Highway Administration, 1990 Nationwide Personal
Transportation Survey.- User's Guide for the Public Use Tapes,
Advance Copy (Washington, 1991).
3. Summary of Findings and Recommendations: Highway Mobility and
Safety of Older Drivers and Pedestrians (Washington, D.C.:
Highway Users Federation for Safety and Mobility, 1985);
Transportation Research Board, "Executive Summary," in
Transportation in an Aging Society. Improving the Mobility and
Safety of Older Persons, Vol. 1, Committee Report and
Recommendations (Washington, D.C.: National Research Council,
1988); and Conference on Research and Development Needed to
Improve Safety and Mobility of Older Drivers (Washington, D.C.:
National Highway Traffic Safety Administration, 1990?).
4. The TRB effort and Congressional request resulted in a
two-volume report by TRB, Transportation in an Aging Society.-
Improving Mobility and Safety for Older Persons, Special Report
218, Vol. 1: Committee Report and Recommendations, Vol. 2:
Technical Papers (Washington, D.C.: National Research Council,
1988).
5. The result is a report BY THE U.S. Department of
Transportation, Older Driver Pilot Program: Report of the
Secretary of Transportation to the United States Congress
(Washington, D.C.: Federal Highway Administration, 1990).
6. For example, Max Israelite, "Take Away My License: I Would
Rather Stop Driving Too Soon Than Too Late (Elderly Automobile
Drivers)" in Newsweek (May 9, 1994): 1 1; Joan E. Rigdon, "Car
Trouble: Older Drivers Pose Growing Risk On Roads As Their
Numbers Rise; They Crash More Than Many, Yet Taking Away Wheels
Leads To Isolation, Anger; A Man Runs Over His Wife" in Wall
Street Journal (October 29, 1993): Al; Lisa J. Moore, "Drive on
Miss Daisy (older automobile drivers)" in U.S. News & World
Report (June 22, 1992): 8384; Alan L. Otten, "Older Drivers
Appear Safer But More Frail (National Institute On Aging Study
Reveals Older Drivers More Likely To Die In Auto Accidents Than
Younger Drivers)" in Wall Street Journal (June 1, 1992): Bl;
"Safety And The Older Driver: When Difficult Issues Collide
(Federal And State Authorities Struggle To Identify Aged Drivers
Who Pose A Hazard While Not Discriminating Against Those Who Do
Not)" in New York Times (May 4, 1992): Al; Sandy Rovner, "Driving
Difficulties Increase With Age" in Washington Post
40
(October 30, 1990): WH 1 6; and James Camey, "Can A Driver Be Too
Old? Fender Benders And Fatalities Raise Fears Over Elderly
Motorists" in Times (January 16, 1989): 28.
7. U.S. Federal Highway Administration (FHWA), Highway
Statistics, 1990 (Washington, D.C.: FHWA, 1991), Table DL-20; and
FHWA, Highway Statistics, Summary to 1985 (Washington, D.C.:
FHWA, 1987), Table DL-220.
8. U.S. Bureau of the Census, Statistical Abstract of the United
States, 1992 (Washington, D.C.: U.S. Government Printing Office,
1992), Table 14.
9. FHWA, Highway Statistics, 1990, Table DL-20; and FHWA, Highway
Statistics, Summary to 1985, Table DL-220.
10. Ruth H. Asin, Characteristics of 1977 Licensed Drivers and
Their Travel.- Report 1, 1977 NPTS (Washington, D.C.: FHWA,
1980), Table 16; and Ezio C. Cerrelli, Crash Data and Rates ffor
Age- Sex Groups of Drivers, 1990 (Washington, D.C.: National
Center for Statistics & Analysis, 1992), Table C.
11. The elderly population is expected to reach 20 percent of all
persons by the year 2020, according to Census Bureau, Projections
of the Population by Age, Sex, and Race for the United States,
1983-2080 (Washington, D.C.: Government Printing Office, 1984),
No. 952, Series P-25, cited by TRB, Transportation in -an Aging
Society, Vol. 1: 22. In 1990, the elderly population was 12.5
percent of all persons, while the number of elderly drivers was
13.3 percent of all drivers.
12. Finn Jorgensen and John Polak, "The Effect of Personal
Characteristics on Drivers' Speed Selection," Journal of
Transport Economics and Policy, 27 (September 1993): 237-252.
13. TRB, Transportation in an Aging Society, Vol. 1: 61, 72.
14. Ibid.: 39-40.
15. J. Peter Rothe, The Safety of Elderly Drivers: Yesterday's
Young in Today's Traffic (New Brunswick:Transaction Publishers,
1990), p. 64.
16. S.J. Flint, K.W. Smith, and D.G. Rossi, "An Evaluation of
Mature Driver Performance," paper presented at the 14th
International Forum on Traffic Records Systems, San Diego (1988),
cited by J. Peter Rothe, The Safety of Elderly Drivers, 127.
41
17. P.A. Brainn, Safety and Mobility Issues in Licensing and
Education of Older Drivers (Washington, D.C.: NHTSA, U.S.
Department of Transportation, 1980), cited by Sandra Rosenbloom,
"The Mobility Needs of the Elderly," in Transportation in an
Aging Society., improving Mobility and Safety for Older Persons,
Special Report 218, Vol. 2, Technical Papers (Washington, D.C.:
National Research Council, 1988), 40.
18. R. Risser and C. Chaloupka, "Elderly Drivers: Risks and Their
Causes," in Proceedings of the Second International Conference on
Road Safety, ed. by J.A. Rolhengafter and R.A. de Bruin (Assen,
Netherlands: Van Gorcum, 1987), cited by Sandra Rosenbloom, "The
Mobility Needs of the Elderly," 40.
CHAPTER 3
1. G.S. Maddala, Limited-Dependent and Qualitative Variables in
Econometrics, Econometric Society Monographs, No. 3 (Cambridge,
Mass.: Cambridge University Press, 1983): 149- 165.
2. SAS/STAT User's Guide, Version 6, Fourth Edition (Cary, NC:
SAS Institute Inc., 1989): 1005-6.
3. Bureau of Census, Statistical Abstract, 1991 (Washington,
D.C.: U.S. Department of Commerce, 1992), No. 762.
Refiner/Reseller Sales Price of Motor Gasoline, by Grade and
State: 1989 to 1991; and No. 998. State Gasoline Tax Rates, 1990
and 1991, and Motor Fuel Tax Receipts, 1990.
4. The SAS procedure used for estimation, LIFEREG, does not
report the log likelihood at zero (i.e., when all explanatory
variables are excluded).
5. For more on the interpretation of Tobit models, see John F.
McDonald and Robert A. Moffift, 'The Uses of Tobit Analysis," The
Review of Economics and Statistics 62 (1980): 318-321.
6. Rosenbloom, "The Mobility Needs of the Elderly," Vol. 2:
33-34.
42
CHAPTER 5
1 FHWA, User Guide to the 1990 Nationwide Personal Transportation
Survey, Appendix J: National Accident Sampling System Vehicle
Make and Model Coding Dictionary (Washington, D.C.: Department of
Transportation, 1991).
2. A more appropriate tool would be grouped data regression or
ordered probit regression (William H. Green, Econometric
Analysis, New York: MacMillian Publishing Company, 1990).
3. Kenneth Train, Qualitative Choice Analysis: Theory,
Econometrics, and an Application to Automobile Demand, MIT Press
Series in Transportation Studies, Marvin L. Manheim, ed.
(Cambridge, Mass.: M.I.T. Press, 1986): 143-144.
CHAPTER 6
1. See Chapter 1.
2. TRB, Transportation in an Aging Society, Vol. 1: 76-103.
3. A.D. Burch, "Bill Targets Old, Young For Added Driving Tests"
in The Orlando Sentinel (March 3, 1994): C-1.
4. TRB, Transportation in an Aging Society, Vol. 1: 76-103.
5. Ibid.
6. Sam Peltzman, "The Effects of Automobile Safety Regulation,"
Journal of Political Economy, 83 (June 1975): 677-725.
7. TRB, Transportation in an Aging Society, Vol. 1: 76-103.
43
NOTICE
This document is disseminated under the sponsorship of the U.S.
Department of Transportation in the interest of information
exchange. The United States Government assumes no liability for
its contents or use thereof.
The United States Government does not endorse manufacturers or
products. Trade names appear in the document only because they
are essential to the content of the report.
This report is being distributed through the U.S. Department of
Transportation's Technology Sharing Program.
DOT-T-95-12
