A Multiattribute Analysis of Goals For Intelligent Transportation System Planning

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 15

TrcmspnRes.-C. Vol. 4. No. 2. pp.

97-l I I, 1996
Copyright 0 1996Elsevier Science Ltd
Printed in Great Britain. All rights reserved
Pergamon
0968490X/96 $15.00 + .OO

A MULTIATTRIBUTE ANALYSIS OF GOALS FOR


INTELLIGENT TRANSPORTATION SYSTEM PLANNING

JONATHAN LEVINE and STEVEN E. UNDERWOOD


The University of Michigan, Ann Arbor, MI 48109, U.S.A.

(Received 8 May 1995; in revised form 26 January 1996)

Abstract-Transportation planning in general, and planning for Intelligent Transportation Systems


(ITS) in particular, are notable both for muhiple goals and for multiple constituencies. In response to
this policy environment, multicriteria decision analysis has often been utilized to evaluate alternative
transportation investments. This approach is extended here to assess stakeholder valuation of broad
goals of an ITS planning process, the FAST-TRAC operational field test in Oakland County, a
suburban region of metropolitan Detroit, Michigan, U.S.A. Representatives of stakeholder groups,
ranging from emergency response firm employees to city managers to environmental groups, were
interviewed. Using a modified Analytical Hierarchy Process, implicit preference weights for transporta-
tion planning goals were derived, and inter- and intragroup comparisons made. Overall, collision
reduction emerged as a dominant goal accounting for nearly 35% of the overall valuation of all goals.
In contrast, travel time reduction and energy/environmental impacts each accounted for about 20% of
the total valuation. Stakeholder group affiliation appeared to affect transportation system preferences
most strongly with regard to environmental preferences and reduction in commercial travel time; with
regard to other goals, individual interests seemed to dominate those of the ostensible stakeholder
group. In an environment such as that of ITS, in which policy goals are diverse and potentially
conflicting, the methodologies presented here can aid in policy and system design by gauging the rela-
tive preferences of strongly interested individuals and groups. While the specific findings presented
here are not generalizable to other regions, they underscore the relative importance of a range of
ITS goals apart from simple reductions in travel times. Copyright 0 1996 Elsevier Science Ltd

I. INTRODUCTION

The nationwide Intelligent Transportation System (ITS) program is notable for its broad
range of hoped-for benefits. ITS in its various forms is designed to increase the capacity
and speed of transportation facilities, enhance environmental quality, save energy, increase
safety, boost productivity and more (IVHS America, 1992). Indeed, there is little doubt
that virtually all approaches to urban transportation planning - whether supply or demand
based - could benefit from enhanced communications and computing capability. By
offering policy makers more options, ITS may well increase the range of goals that are
achievable - may “stretch the envelope” - in transportation planning (Shladover,
1993). At the same time, it should be understood that a number of goals and outcomes
of any transportation planning or policy option may interfere with one another, potentially
necessitating significant tradeoffs.
It is inevitable that different interest groups will focus on different aspects of ITS benefits,
and even that what is perceived to be a benefit by one group will be seen as a net cost by
others. For example, in the case of real-time route guidance or adaptive traffic controls,
transportation engineers, emergency services or commercial trucking may anticipate the
benefits of travel time savings, while environmentalists may be concerned about increased
auto emissions from induced travel (Downs, 1992; Gordon, 1992). Similarly, neighborhood
groups may be concerned about the routing of diverted traffic and potentially deleterious
effects on their immediate surroundings. Given this diverse policy environment, a search for
a unitary overall societal valuation of a particular ITS deployment in the style of traditional
benefit-cost analysis may be futile. Worse, by reducing varied impacts to commensurate
terms and by masking variation in preferences, such analysis may limit the role of public
debate regarding the directions of publicly financed involvement in ITS.
97
98 Jonathan Levine and Steven E. Underwood

At the same time, large scale ITS demonstration projects are under way in various parts
of the country, and evaluators are called upon not only to analyze technical and system
outcomes, but also to illuminate some notion of the systems’ societal value. Under these
circumstances it is critical for policy makers to understand the often competing preferences
of the various groups sharing intense interests in transportation planning options in general,
and in ITS policy in particular.
This study investigates the preferences of members of identified stakeholder groups in an
ITS evaluation process in Oakland County, Michigan, U.S.A., a growing suburban region
northwest of Detroit, through a multiattribute analysis of transportation planning goals.
Underpinning much evaluation research of this sort is the notion that programs have both
multiple goals and multiple constituencies, and that evaluations should be structured
accordingly (Edwards and Newman, 1982; Giuliano, 1985; Chen and Underwood, 1988).
Thus, this study does not seek a single optimal design of an ITS installation; rather, the
analysis of the varying preferences of stakeholder groups can be used in the search for
compromise or even win-win design solutions and in the identification of possible ranges
of reactions to ITS policies and proposals by the various groups.
The analysis is intended as a tool, in advance of large scale deployment, both to facilitate
identification of groups’ concerns and priorities and to aid in designing systems that will be
responsive to these concerns. The assessment of stakeholder preferences regarding system
attributes and goals is designed to be combined with technical system outcomes to gauge pref-
erences for alternative system designs by the various stakeholder groups. System design is thus
seen not as an engineering problem alone, but as a process in which technologies are devel-
oped and deployed in response to identified, if sometimes competing, needs and preferences.

2. PREFERENCE MEASUREMENT IN MULTIATTRIBUTE DECISION ANALYSIS

The planning, marketing and operations research literatures abound with approaches to
multiattribute analysis in decision making. Methodologies such as Multiattribute Utility
Theory (MAUT) (Keeney and Raiffa, 1976) Analytic Hierarchy Process (Saaty, 1980;
Saaty and Kearns, 1985) TODIM (Gomes, 1989) and ELECTRE (Roy and Hugonnard,
1982) share a similar underpinning: the notion that decisions entailing multiple objectives
are aided by specifying quantified outcomes and preference-based weighting schemes. The
quantitative ratings generated by these methods are seen as tools for stakeholders and
decision makers to compare simultaneously more alternatives and attributes than they would
have been capable of otherwise, given the complexity of the information, The “divide
and conquer” approach represents philosophies, articulated by a set of axioms, and method-
ologies based on these axioms, for analyzing decision alternatives and describing ranges
of solutions to multiattribute problems. Generically, these methods proceed according to
five basic steps:
(1) Identify relevant participants in the decision process. These can be a single decision
maker, multiple decision makers, or most broadly “stakeholders” - the decision
makers plus those groups affected by the decision (Edwards and Newman, 1982).
(2) Identify the dimensions, criteria or goals that will characterize the alternatives.
(3) Generate preference-based weighting schemes (or in the case of ELECTRE,
threshold analyses).
(4) Develop measures by which each of the alternative projects is assessed along each
of the relevant dimensions.
(5) Rank or rate alternatives based on measured outcomes and groups’ preferences, and
perform analyses. These frequently include marginal analysis of costs and outcomes
between alternatives, sensitivity analysis in which changes in the assumptions are tested
for their capacity to alter final outcomes, and intergroup differences that compare
the utility of different options to different groups and seek satisfactory solutions.
Multiattribute studies have been traditionally used in transportation applications to select
desired transportation improvements from among a series of proposed projects or as an
Goals for intelligent transportation system planning 99

approach to predicting individual level behavior under alternative policy options (Srinivasan
et al., 1981; Bunch et al., 1993). Somewhat less common is the use of multiattribute analysis
to assess tradeoffs between broader goals of a transportation planning process (deNeufville
and Keeney, 1972).
The various approaches to multiattribute analysis are fairly similar regarding the first two
stages delineated above, particularly as these tend to be contextually linked to the decision
and interest groups at hand. In contrast, for the third step - generation of preference-based
weights - a number of distinct alternatives are available. In the simplest instance, the analyst
may merely assume a preference structure (sometimes assuming all the weights associated
with all the attributes to be equal), or series of alternative preference structures, and embody
these in weighting schemes (Giuliano, 1985). Sensitivity analyses may be performed to
determine the extent of change in assumptions that would be required to alter outcomes.
Methods under which preference-based weighting schemes are estimated on the basis
of studies of the decision makers or stakeholder groups themselves can be considerably
more complex, often involving intricate survey instruments or other experiments. When
the alternatives are simple and concrete, e.g. a consumer’s choice between different packaged
soups, this does not tend to pose much of a problem. However, in higher level policy
situations, wherein respondents are asked to compare alternatives that entail quite abstract
economic, environmental and transportation impacts, responses become considerably more
complex and the subject’s task more onerous. At this point the random component of
the respondent’s answers may begin to grow as he or she struggles to cope with the task
at hand. The limiting factor of this family of analytical approaches is principally their
capacity to elicit meaningful responses from respondents.
For these reasons, an approach that would request information regarding preferences on
and tradeoffs between transportation system attributes without requiring the respondent
to consider and compare complex bundles of attributes simultaneously may be suited to
the task of comparison between alternative transportation policies. One such approach is that
employed by the Analytic Hierarchy Process (AHP) (Saaty, 1980) and related techniques
(Gomes, 1989) in which the respondent is presented pairs of attributes and asked to rate
the pair in terms of the relative importance of each of its elements.
In the rating scale developed by Saaty (1980) values range from l/9 (for a case where a
goal is of extremely low importance compared to the other in the pair) through 1 (in cases
where the goals are seen as equal) to 9 (in cases where a goal is of extremely high importance
relative to the other in the pair). Classical AHP attaches verbal descriptors to the values,
such as “moderate importance of one over another” or “demonstrated importance” (Saaty,
1976). We judged these terms to be not especially meaningful to respondents; they were
abandoned here in favor of a simple row of boxes indicating a range from “equal importance”
to “much more importance” without verbal descriptors in between.
In AHP-based analysis all possible pairs of goals must be presented, but since the rating
of goal “A” versus “B” is the inverse of the rating of goal “B” versus “A,” the number
of outcome pairs equals

[KX (K- 1)]/2,

where K equals the number of goals to be evaluated. This phenomenon greatly limits the
complexity of the questionnaires required to implement AHP. The responses are then used to
complete a matrix of K X K elements in which aU(the relative preference for goal i when
compared to goal j) is the inverse of uji, and entries along the main diagonal are set to
unity. The weight assigned to a particular goal is equal to the geometric mean of the values in
that goal’s row, normalized by the sum of the row means. The sum of all criteria weights
will equal unity, and the weights themselves are interpreted as a ratio scale measure of
the importance an individual assigns to their associated goals.
In AHP, an individual’s responses may be gauged for consistency under the logic that,
under perfect consistency,

uik = a, X ujkfor all elements i, j and k.


100 Jonathan Levine and Steven E. Underwood

For example, if criterion i is seen as twice as important asi, andj is four times as important
as k, i should be eight times as important as k. Needless to say, individuals do not demonstrate
such perfectly consistent ratios in their responses. AHP methodology thus develops a metric
to gauge the distances of an individual’ s responses from randomness. A value of 0.1, or
sometimes 0.2 on this metric (in which 0 indicates “perfectly consistent” and 1 signifies
“random”), is seen as reflecting adequate consistency to rely on results as reflecting an indi-
vidual’s considered judgment (Saaty and Vargas, 1982). When the consistency ratio is higher
than this, the analyst typically offers the respondent the opportunity to clarify his or her
assessments.

3. STUDY METHODOLOGY

The current study was performed as a component of the evaluation of the FAST-TRAC
ITS demonstration project in Oakland County, Michigan, U.S.A. FAST-TRAC (Faster
and Safer Travel through Traffic Routing and Advanced Controls) entails an integration
of adaptive traffic controls with real time route guidance (Underwood, 1994) in a major ITS
operational field test. The stakeholder preferences investigated here will ultimately be used
to: (1) determine the value of FAST-TRAC as perceived by various stakeholder groups; (2)
provide a basis of comparison among the stakeholder groups; and (3) identify possible ranges
of responses of stakeholder groups to alternative system designs. The approach emphasizes
the assessment of preferences from more than one perspective in order to identify opportunities
for improving the architecture of the FAST-TRAC system.
Subjects for the current study were drawn from identified stakeholder groups to the
transportation planning process. Individuals were identified and selected within seven
principal groups, as listed in Table 1. Initial individual and group contacts were developed
through consultation with the Road Commission of Oakland County, the project’s sponsor.
In addition, each informant was asked to provide the names of other individuals and groups
that would be likely to have interest in transportation issues in Oakland County. The
groups and their representatives were selected for their likely interest in transportation
policy, rather than for their representativeness of the population at large. Thus no attempt
will be made in this study to infer to a larger population; the results represent the views
of the identified population of stakeholders to the FAST-TRAC planning process only.
In all, 56 individuals were interviewed from the seven groups listed below.
Respondents were asked to complete a questionnaire that entailed two sections: a relatively
detailed set of outcomes and a reduced set of system goals. In the first section, the question-
naire listed specific transportation system outcomes (e.g. “increase trips taken on city arterials”,
“reduce construction of new roads”) that individuals were to rate on a nine-point Likert
scale ranging from “very undesirable” through “neither desirable nor undesirable” to “very
desirable” (Table 2). The purpose of this section was two-fold. First, it elicited responses
to specific outcomes, as opposed to the broader goals in the AHP-style component of the
questionnaire. Second, it allowed the subjects to respond to outcomes that were not goals
of the FAST-TRAC system but potential side effects, such as increased traffic volumes,
or increased use of neighborhood streets. This approach had the benefit of enabling a

Table 1. Stakeholder groups to the FAST-TRAC evaluation process

Group 1: Public sector transportation Public sector transportation (highway and transit)
professionals, school transportation personnel
Group 2: Private sector transportation Delivery services. trucking companies, automotive industry
Group 3: Education/media Educational administrators, university, newspaper
Group 4: Safety Emergency vehicle firms, police, fire departments
Group 5: Other business Chambers of commerce, companies, real estate developers
Group 6: Public administration City managers, mayors, planning departments
Group 7: Citizen groups Homeowner/community groups, environmental groups,
seniors/disabled
Goals for intelligent transportation system planning 101

fairly detailed set of questions that included both potentially positive and negative outcomes
of a FAST-TRAC system. The weakness of this style of analysis is its inability to force
tradeoffs; indication of one outcome as “very desirable” does not detract from the respondent’s
ability to rate another outcome with a similar degree of enthusiasm. The AHP style of analysis
referred to above, while handling fewer outcomes, compels respondents to reveal their
preferences through competition between system goals.
The notion of a tradeoff between system goals is central to the design of this study, as ITS
goals and outcomes can have mutually reinforcing or counteracting effects. For example,
prevention of accidents through the use of ITS technologies is also a highly effective tool
in reducing transportation system delays. On the other hand, increases in system capacity
and the increases in travel volumes they can potentially draw, can partially or entirely negate
air quality gains achieved through congestion reduction.
In order to elicit information on stakeholders’ valuations of such tradeoffs, a modified AHP
approach is employed in this study in which the respondent was presented with pairs of
goals, such as:
10% emissions reduction . . . . . . . . . . . . . . . . . . . . 10% individual travel time reduction
The “10%” improvement was used as a standard for all of the goals listed in Table 3. This
quantitative specification of the goals is a departure from classical AHP. The latter approach
would typically present two attributes or goals - for example, “emissions” and “travel time”
- and ask the respondent to consider the goals’ relative importance. The respondent
might be troubled by such a question: “How much emissions reduction, versus how much
travel time savings?” he or she might wonder in formulating an appropriate response
(Harker and Vargas, 1988). The specification of 10% improvements - chosen simply for
roundness and ballpark-reasonableness - in all goals is an attempt to reduce this ambiguity

Table 2. Transportation system goals and outcomes to be rated

Major ITS goals

Individual and commercial travel times:


Reduce travel times on major city streets
Reduce travel times on freeways
Reduction in tax costs:
Reduce the need for construction of new roads
Reduce tax costs of building and operating streets and highways
Reduction in driving difficulty:
Smooth traffic flow along major city streets
Reduce drivers’ stress
Route drivers around congested areas
Help drivers find their way in unfamiliar areas
Provide traffic information to drivers to help ensure on-time arrival at their destinations
Collision reduction:
Reduce the number and seriousness of accidents on streets and highways
Energy savings:
Reduce total fuel consumption
Emissions:
Reduce total automobile emissions

Other potential system outcomes

Increase the number of car trips taken on major city streets


Increase the number of car trips taken on freeways
Increase the number of car trips taken on neighborhood residential streets
Increase the use of the private car for commuting
Increase the use of ridesharing/carpooling for commuting
Increase the use of public transportation for commuting
Increase the use of biking or walking for commuting
Increase noise along major city streets
Improve mobility for people who don’t drive
Reduce vehicle operating costs
102 Jonathan Levine and Steven E. Underwood

Table 3. Identified FAST-TRAC system goals and their definitions, as presented to respondents

Policy goal Explanation


Reduce by 10%

Reduce individuals’ travel time Roadway travel time by individuals (including commuting,
shopping, recreation, etc.)
Reduce commercial travel time Roadway time required for business related travel, fleets,
trucking and buses
Reduce tax costs The local, state and federal costs of the transportation system
Reduce driving difficulty The difficulty of the driving task through in-vehicle navigation
assistance, trallic smoothing, or other means
Reduce collisions The number and seriousness of roadway accidents
Save energy Total gasoline and diesel consumption by cars, trucks and
buses
Reduce emissions Total emissions from cars, trucks and buses

to the respondent. It is important to note, however, that while the meaning of 10% reduction
in travel times should, in fact, be unambiguous, the precise interpretation of the meaning of
other goals (e.g. a 10% reduction in driving difficulty) was necessarily left to the individual’s
judgment.
The goals were defined using dry, non-evocative terms as much as possible. For example,
there was no “safety enhancement” goal, but only “collision” reduction. Similarly, the
respondent was not asked to rate a goal pertaining to reduction in “air pollution,” but rather
“emissions.” It is important to emphasize that the results presented in this study are entirely
dependent on the specification of system goals above. The addition or deletion of trans-
portation goals would undoubtedly have altered the value of the importance of the remaining
goals. Even given the particular goals listed, their definition largely determines their out-
come. For example, had “reduce individuals’ travel time” and “reduce commercial travel time”

Table 4. Examples of questionnaire format

Changes in transportation systems can have different outcomes. Please rate each of the transportation outcomes
described below on a “very -desirable” to “very undesirable” scale, according to your own preferences.

Transportation outcome

0 ciclo q 000 q
Very Neither Very
desirable desirable undesirable
nor
undesirable

Increase the number of 0 000 cl 000 0


car trips taken on major Very Neither Very
desirable desirable undesirable
nor
undesirable

Which is more important, A or B? Goal B

0 0000000 Bclh 0000000 Q Reduce


A is B IS
much are much travel
more equally more time
mportant important importan

q 0000000 Btlh 0000000 q Reduce


A is B is driving
much are much difficulty
more equally more
mportant important importan
Goals for intelligent transportation system planning 103

been combined into a single goal, the weightings of the latter would probably not equal the
sum of the weightings of the former two. The definition of goals is hence a critical stage
in the AHP process; those defined above were designed to reflect the principal objectives
of the FAST-TRAC system.
Respondents completed written questionnaires (Table 4) in a face-to-face setting.
While individuals were selected on the basis of their association with the stakeholder
groups defined above, they were offered no special instructions regarding representation
of their group’s interests in responding to the questionnaires. After the data were tabulated,
each respondent was sent a pie chart indicating the initial calculation of his or her
weights regarding the seven system goals described in Table 3. The cover letter requested
that the individual review the pie chart for consistency with his or her valuation of system
goals, and indicated that researchers would be calling with follow up questions. All
respondents, regardless of their initial consistency statistic, were called back in order both
to gauge the individual’s satisfaction with the initial calculation of weights, and where
necessary, attempt to improve the consistency of overall response. Fifty-three (of 56) respon-
dents achieved a 0.2 consistency statistic or greater; the rest of the study focuses on these
individuals.
In order to improve individuals’ consistency of responses, a metric was developed to rank
the contribution of each individual answer to the overall inconsistency present in the
questionnaire. This statistic was embedded into a spreadsheet such that the interviewer
at any time was able to identify the question that contributed most to the respondent’s
inconsistency; ranks were updated and recalculated in real time in order to assist the
interviewer. The inconsistency statistic associated with each individual question was
Max [(a, X Wj / W,), (Wi / W, X ~$1,
where
ai, = the raw response of the individual on the question comparing goals i and&
W, = the overall weighting for goal i;
W, = the overall weighting for goal j.
The logic of this measure is as follows: under perfect consistency, the ratio expressed in each
individual question would be equal to the ratio of that particular pair of goals overall. The
magnitude of the deviation of this measure from unity is an indicator of the contribution of
each question to the overall inconsistency observed in the analysis. Since the a, value may be
greater or less than the ratio of the overall weights ( W,l Wj), the ratio of the two terms - or its
inverse - is taken as the indicator of contribution of inconsistency. Finally, the spreadsheet
function RANK, which returns the rank of a particular cell’s value within a range of values,
was used to rank the entire set of questions in terms of their contribution to overall inconsis-
tency, and assist the interviewer in focusing on the most inconsistent questions in follow up.
Care was taken during the follow up calls not to lead the respondents to any particular
conclusion. For example, interviewers were instructed not to say: “Here you said “X”, and
here you said “Y”. You’d better say “Y” in both places to be consistent”. Instead, a probing
style was pursued in which interviewers identified inconsistencies and left their resolution
up to the respondents.

4. PREFERENCES OF INDIVIDUALS AND STAKEHOLDER GROUPS

Depending on one’s view of decision-making processes, it is possible to view the respondents


in this study either as individuals or as members of relevant stakeholder groups. This study
will adopt both approaches. First, information on the individuals comprising the population
as a whole will be presented to examine ranges of prevailing opinions on transportation
planning goals among the population studied. Next, individuals will be examined for their
tendency to represent groups of common interests. This second approach will take two
forms. First, the stakeholder groups identified above will be analyzed for any similarity
of values in the transportation area. Finally, cluster analysis will be used to reconstitute interest
groups on the basis of common views, and the membership of these clusters will be examined.
104 Jonathan Levine and Steven E. Underwood

4.1. Analysis of individual results


The output of the modified AHP approach described above is an index of the relative
importance that each individual implicitly attaches to each goal in the study. The values for
an individual sum to unity; i.e. “1” represents the total importance placed on all goals
together. Though the primary result of such an analysis is a set of weights for each individual
in the study, aggregation of scores is presented as a convenience to avoid presentation of
what would otherwise be an indigestibly large amount of data. Means, rather than medians
are generally used to ensure that weightings for groups, like those pertaining to individuals,
sum to unity.
Figure 1 presents the means, 25th and 75th percentile weights for each of the seven
goals across all respondents. For example, the mean weight of the “reduce emissions”
goal was 0.12; 25% of the respondents had weights under 0.06 for this goal, while 75%
had weights under 0.17. Immediately apparent is the nearly universal importance placed
on collision reduction as a desired transportation system goal. So important was this goal
that its 25th percentile weighting was greater than the 75th percentile weightings of all
the other goals.
This result must be interpreted carefully. It can be argued that the stated preference
for safety enhancement is less indicative of a true valuation of system goals than revealed
preference; in practice, many individuals speed to save time, thus ostensibly contradicting
the results of this study. While it may be that some of the preference for collision reduction
is, in fact, an artifact of the research design - perhaps an attempt to tell the interviewers
what the respondents believed they wanted to hear - we argue that there is more to these
results than that. First, risk-taking behavior in transportation is in part a function of class;
respondents here, being largely professional and managerial - as well as closely involved
with transportation and transportation safety - may well act to reduce transportation
risks in their personal behavior. However, perhaps more importantly, studies such as this are
designed to elicit preference for collective, rather than individual action. On an individual
level, and at a given moment, an individual may fail to perceive the riskiness of his or her
behavior. There is no necessary reason why this lapse per se should invalidate the person’s
view that transportation safety is a problem that requires greater attention at the collective
level. Individual level action is, at best, a flawed indicator of collective preference; if it
were not, all taxation could be voluntary!
Other goals tended to be much more comparable in their mean computed weights, with
tax saving rating slightly lower than other goals. This may indicate an overall willingness
to pay taxes for transportation purposes to the extent that these are used to support goals
with high perceived importance.
It is important to note that the mean individual and commercial travel time weights together
account for just 21.4% of the importance the respondents assigned to all goals together.
Traditional transportation planning and engineering practices often treated automotive travel
0.5050

0.40 4

Fig. I. Weights for system goals: means, 25th and 75th percentiles
Goals for intelligent transportation system planning 105

time reduction as the main goal behind the development of transportation improvements;
these results indicate that, for this group of decision leaders in Oakland County, travel time
reduction is just one goal of many. Values estimated here are subject to error associated
with the interrelatedness of goals; respondents undoubtedly had a mental picture of relation-
ships among, for example, travel time, driving difficulty and collision reduction outcomes.
Yet, the format of the original questionnaire was designed to allow them, to the fullest extent
possible, to isolate goals and indicate preferences among competing goals.
Figures 2 and 3 present similar data for the detailed questions regarding transportation
system outcomes, with Fig. 2 presenting those outcomes whose median ratings were negative
or neutral. Negative ratings were associated with outcomes potentially associated with
the traffic inducing or diverting capacity of ITS technologies; increased noise along arterials,
increased automobile trips through neighborhoods and, to a lesser extent, an increase in the
use of the automobile for commuting and increased trips along major city streets. As evidenced
by the width of the interquartile band (the range between the 25th percentile and the 75th
percentile responses), there is relatively little controversy about the undesirability of the first
two outcomes. A wider range of responses surrounds questions regarding the increased
use of the private automobile for commuting and increased traffic along arterials, indicating
greater divergence of opinion in these areas.
Particularly controversial is the potential for increased traffic along freeways; this is the
only goal whose interquartile band ranges from positive to negative. Apparently, while
some respondents were concerned about negative externalities associated with additional

Fig. 2. Ratings of transportation system outcomes (for median <= 0).

Fig. 3. Ratings of transportation system outcomes (for median <= 0).


106 Jonathan Levine and Steven E. Underwood

freeway travel, others may have viewed increases in freeway travel volumes as associated
with decreases along city streets, or as facilitating economic activity in the region. Also
controversial is the outcome of reducing the need for new road construction. While
respondents rated this as a positive outcome overall (Fig. 3) the distribution of responses
was among the broadest of all transportation system outcomes. Many people may have been
ambivalent about reducing the need for new road construction because they see such
construction as the principal way to alleviate transportation problems. Similarly controversial
items included assistance with navigation, reduction of emissions and encouraging the use
of biking or walking for commuting.
In contrast, the goal of accident reduction received the highest overall rating, and the
greatest consensus, as evidenced by the narrowness of the interquartile band. Other highly
rated outcomes included travel cost and time-reducing outcomes along city arterials, as
well as increasing mobility for the transportation disadvantaged. Reduction of travel time
along freeways was somewhat less emphasized, as were the information-related options
of providing traffic advisory information and helping drivers find their way in unfamiliar
areas. Encouragement of transportation alternatives (ridesharing, transit and pedestrianism)
were supported, but not as highly as options that directly enhanced the driving experience
or lowered its costs. As expected, this form of analysis had a slight tendency toward extreme
answers, as respondents were not compelled to trade one goal off the other. For this reason,
the AHP tradeoff style of analysis will be utihzed in the remainder of this paper.
4.2. Analysis of identljied stakeholder groups
An implicit assumption of a planning process that explicitly involves stakeholder groups is
one of similarity of interest within identifiable groups. Under this framework, these groups -
as groups, not aggregations of individuals - are major players in the process of policy-making.
This section of the paper will analyze the extent to which the respondents form groups,
and the extent to which these match their identified affiliations.
Figure 4 presents mean weightings for system goals by these seven categories of stakeholder
groups. Each group’s valuation of all goals combined is set to 100% the manner in which each
group divides up this total valuation among the various goals is represented by the heights
of the bar segments in the charts.
Overall, there appears to be more similarity than conflict between the various groups. All
of the groups rate collision reduction highly, though the safety, transportation and other
public groups tend to give it a higher rating than do others. The citizen groups rate emissions
reduction and energy savings highly compared to others. The difference in this area is especially
large between the citizen groups and the private sector transportation interests. Reduction
of driving difficulty was rated similarly between the various groups, while the importance
assigned to travel times - individual and commercial - varied considerably. The private

100

90
Difficulty
80
TU.
70
Emissions
60 Energy
50 Corn. time

40 Ind. time

30 Collisions

20
10

0 L
Public Public Safety Educ.1 Citizen Private Other
admin. rraIs. media groups trans. business

Fig. 4. Valuation of goals by stakeholder groups.


Goals for intelligent transportation system planning 107

Table 5. Analysis of variance on stakeholder groups’ valuation of transportation goals

Variable F Significance of F

Commercial time 5.59 0.000


Energy savings 2.38 0.044
Emissions I.51 0.196
Individual time 1.32 0.265
Tax I.19 0.330
Collisions 0.98 0.447
Driving difficulty 0.44 0.850

Univariate F-tests with (6, 46) degrees of freedom

sector transportation group rated commercial travel time savings as well exceeding individual
time savings in importance, while all the other groups valued the two goals roughly similarly
or valued individual travel time more highly. Valuation of travel time savings was greatest
among the business and private sector transportation groups, with the citizen groups assigning
a mere 15% on the two travel time goals together. Tax reduction was given relatively low
weight by all groups, but particularly by the public sector transportation group, whose mission
revolves largely around solving transportation problems using public expenditures.
The statistical difference between the various groups in terms of their valuation of
transportation goals is quite small. While the 48 individuals included in this study are not
viewed as a sample but a selected population of transportation decision leaders in Oakland
County, an analysis of variance (ANOVA) can be used to gauge the extent to which the
ostensible stakeholder groups form identifiable groups in terms of their valuation of
transportation goals. Had this group been a sample, for only two of the goals (commercial
time reduction, energy consumption) would the conclusion of differing population means
between the groups have been supported; for the other goals, no statistically significant
difference would have been found for a study as small as this one (Table 5). It may be that,
with the exception of a limited set of goals that are closely identified with certain stakeholder
groups, e.g. commercial travel time reduction with trucking and delivery firms, energy
savings with environmental groups, people tend to respond in accordance with their interests
as individuals, rather than as members of identified stakeholder groups.
In order to explore this hypothesis further, stepwise multiple regressions were estimated to
attempt to predict valuation of transportation goals on the basis of individuals’ characteristics,
as well as stakeholder group affiliation. Table 6 presents the results of the three (out of seven)
goals for which at least 20% of the variance could be explained by the demographic
characteristics and group membership in combination. The results tend to support the
findings reported above; for only two of these was stakeholder group affiliation a statistically
significant (p < 0.05) predictor of transportation system goal valuation.

Table 6. Stepwise regressions predicting valuation of transportation goals (/-statistics in parentheses)

Dependent variables
Independent variable Individual time reduction Commercial time reduction Emission reduction

Constant 0.076 (6.74) 0.083 (8.27)


Commute distance 0.006 (5.16)
Age 0.002 (2.72)
Business dummy* 0.068 (2.76)
Public transportation dummy
Private transportation dummy 0.106 (4.80) -0.0.054 (-2. IO)
R’ 0.38 0.39 0.21

*Safety” is the omitted stakeholder group for dummy variables. Other variables tested and not found to be significant
included dummy variables for gender and age, as well as for citizen groups, education/media and public administration.
108 Jonathan Levine and Steven E. Underwood

Consistent with expectations regarding the value of commercial time reduction, higher
ratings in this goal were best predicted by the private transportation and other business
dummy variables. In contrast to the specific interest that private transportation has in commute
time reduction, goals associated with a more general interest - or more particularly a
driver’s interest - appear to receive a rating by individuals based upon the respondents’
personal commuting characteristics more than anything else. Thus, the single significant
predictor of the “individual travel time” reduction goal was commute distance, a predictor
sufficient to explain 38% of the variance in ratings of the importance of travel time reduction.
The only other goal for which stakeholder affiliation appeared as a significant determinant
was emissions reduction; in this case, affiliation with the private transportation stakeholder
group was associated with lower weights for the emissions reduction goal.
4.3. Clustering to assemble groups of similar interests
Given the limited ability of stakeholder group affiliation, as originally defined, to predict
valuation of transportation system goals, the statistical technique of cluster analysis was used
to construct groups of similar views with regard to these objectives. The goals of this clustering
were two-fold: to describe stakeholder group affiliation in a fashion that did not rely on a
priori group definitions, and to ascertain the manner in which preferences tended to move
together, i.e. which goals tended to be closely tied to other goals.
A K-mean clustering procedure was applied, based on the valuation of the seven trans-
portation goals, to develop four distinct clusters. Table 7 describes the clusters and their
transportation preference characteristics.
When a cluster ranks a given goal as the highest of all the clusters the goal is presented
in bold face; low values are presented in parentheses. Using these rankings, it is possible
to characterize the clusters. Cluster 1 is characterized by an unusually high rating given
to collision reduction. Since other weightings are quite close to global averages, this cluster
will be referred to as the “safety” cluster. Cluster 2 focuses on time savings - both indi-
vidual and commercial - considerably above other groups. Cluster 3 is the “environ-
mental” cluster, with particularly high ratings of energy savings and emissions reductions
goals. Cluster 4 is similar to Cluster 2 in that individual travel speeds are highly valued,
though Cluster 4 has higher concerns in reduction in driving difficulty and collisions.
This is interpreted here as the “quality of travel” cluster.
The above analysis can depict families of concerns and the way in which concerns tend to
be grouped. For example, the travel time goals seem to be associated with the tax reduction
goal in the “Time” cluster (and in the “Environment” cluster, in which all these goals were
ranked relatively low). Similarly, rating of the energy goal appears to move with the emissions
reduction goal across clusters.
In addition, it is interesting to compare cluster membership with stakeholder group
affiliation to determine the adequacy of conventional descriptors of stakeholder groups. To
this end, Table 8 presents the ratio of actual to expected populations within each cell of

Table 7. Cluster analysis based on valuation of transportation system goals

Cluster number 4
Cluster size 25 9 6 13
Mean weights: 0.46 (0.18) 0.21 0.28
collision reduction
Individual time 0.08 0.18 (0.06) 0.14
Commercial time 0.09 0.21 (0.08) (0.08)
Driving difficulty 0.10 0.13 (0.05) 0.17
Emissions 0.10 (0.07) 0.23 0.13
Energy (0.09) (0.09) 0.32 0.11
Tax 0.08 0.13 (0.05) 0.10
Cluster characterization Safety Time Environment Quality of travel
Goals for intelligent transportation system planning 109

a four cluster by seven stakeholder group matrix. For example, the representation of the
business stakeholder group within the “Safety” cluster is 61% of that which would be expected
if stakeholder group membership and cluster membership were independent. In this fashion,
it is possible to associate preference-based clusters with stakeholder groups.
Viewing the population of Oakland County stakeholders to the FAST-TRAC planning
process and their assignment to the clusters, several concentrations emerge. The safety
cluster, being the largest, appears to represent broadly-based transportation interests; of all
the clusters this is the one in which the various stakeholder groups are most proportionately
represented. The business and private sector transportation stakeholder groups are consider-
ably over-represented in the “travel time” cluster; these groups have much to gain in the
efficiency of its operations from reduction in commercial travel times. The citizen group
- including representatives of environmental organizations - appears to be predominant
in the “environmental” cluster. Finally, the education/media and public administration groups
are represented beyond their numbers in the “quality of travel” cluster. It may be that
this group focuses on a generalized “quality of life” goal, as opposed to strict travel time
minimization.
Notwithstanding these interpretations, it should be understood that the relationship
between stakeholder group and preference-based cluster is a weak one. The only groups
that are represented at more than twice their numbers are those where transportation
system goals are intimately related to stakeholder groups: private transportation group and
business groups in the “travel time” cluster and the citizen’s group in the “environment”
cluster. Other groups with more average tastes in transportation show little internal consistency
in their transportation preferences.

5. CONCLUSIONS

Most transportation preference assessments address either public attitudes toward public
transportation or consumer attitudes toward automobile variations and options. The
public transportation surveys typically focus narrowly on variations in transit service,
and the mostly-proprietary consumer surveys rarely look at broader transportation-related
concerns. In contrast, the current study addressed community preferences for transportation
alternatives in the context of an evaluation of driver information systems and traffic signal
improvements. The emerging literature on ITS has little to say about public preferences
for advanced traveler information systems and advanced traffic management systems. An
up-to-date survey of ITS evaluation results is summarized in MITRE (1995).
Several themes emerge from the current study. First, in the region examined, travel time
reduction is merely one goal among many for transportation planning in general and ITS
planning in particular. Much transportation activity in the past decades was premised on travel
time reductions as a prime benefit. Even many current ITS projects - FAST-TRAC included
- appear in their designs to focus on travel time reduction as an overarching transportation
system goal. Yet other goals accounted for the majority valuation placed on transportation

Table 8. Ratio of observed to expected cluster membership, by stakeholder group

Safety Time Environment Quality of travel

Business 0.6 2.5 1.3 0.6


Citizen 0.8 0.7 3.3 (0.5)
Education/media 0.8 (0.0) 1.1 2.0
Private sector transportation I.1 2.9 (0.0) (0.0)
Public administration I.3 (0.0) (0.0) 1.7
Public sector transportation 1.3 (0.0) 0.9 I.2
Safety I.2 0.8 (0.0) I.2

Note: cells with representation considerably greater than expected (actuakexpected II 1.5) are presented in bold.
Cells with representation considerably less than expected (actual/expected lo 0.5) are presented in parentheses.
110 Jonathan Levine and Steven E. Underwood

system outcomes. For example, energy savings and emissions reduction are considered by
the Oakland County stakeholder groups studied here to be of roughly equal importance
to travel time as transportation system goals. While traditional traffic engineering has
seen little conflict between these goals, the potential for travel time reduction strategies
to induce more peak hour traffic may set these families of goals at odds.
These results are specific to the group studied and will vary between metropolitan areas;
yet, to the extent that transportation system problems vary together, e.g. areas with greater
congestion tend to suffer from more severe air quality problems, the weights may vary less
than one might expect between metropolitan regions.
The results of this study also raise the potential for outcomes associated with traffic diversion
and inducement to raise objections to ITS installations. Increased noise and traffic volumes
in neighborhoods and increased car use for commuting were universally viewed negatively,
while the population split in its view of increased traffic volumes along arterials and
expressways. Little is known regarding peoples’ behavior during traffic diversion; presented
with a route that takes him or her off the freeway, will a driver attempt to improve on it
still by taking short cuts through neighborhoods? Public action regarding traffic diversion
advisories represents a transportation policy, and should be treated as such. Thus, it may be
reasonable for implementing agencies to consider thresholds of congestion below which
diversion advisories are not issued, and even modifying the computer encoding of transporta-
tion networks (e.g. by selectively modifying link impedances) to protect potentially vulnerable
neighborhoods near freeways and arterials.
A reduction in collisions was the goal that virtually all respondents could agree upon
in its primacy. While traffic smoothing through advanced traffic management systems may
indirectly enhance safety through a reduction in speed variance, and advanced traveler
information systems may reduce exposure to risk by minimizing travel time, no net safety
benefits have been empirically proven for either of these technologies. Both have potential
safety drawbacks as well; for advanced traffic management systems, associated increases in
travel speeds and volumes may tend to overcome safety gains, while advanced traveler infor-
mation systems may represent yet another in-vehicle device distracting drivers’ attention.
An ITS policy that was responsive to stakeholders’ concerns regarding safety might emphasize
collision warning technology over navigation and travel time reduction strategies and -
perhaps more immediately - speed limit enforcement technology, both for highways
and for city streets.
This study may also shed some light on the growing focus on stakeholder groups in
planning processes in genera1 and ITS planning in particular. It must not be automatically
assumed that the assignment of an individual to a nominal stakeholder group defined by
the planner, evaluator or analyst will guarantee that an individual’s actions and responses
reflect the interests of that group. In many cases, the respondent may be reacting more
as an individual - or, in the case of a transportation planning process, as a commuter
- than as a member of his or her ostensible group. Where stakeholder groups have clear
interests that impinge on their essence as a group, this statement tends to hold less;
where group interests are more diffuse, individual interests tend to take over as determinants
of valuation of transportation system outcomes. Investigations seeking stakeholder group
values would benefit from approaches that elicit group level responses. At a minimum,
these may take the form of instructions to individuals to respond in their capacity as
stakeholder group members. Alternatively, similar studies may base their inquiries on
collective processes such as focus groups to facilitate the kind of interaction that may be
necessary for definition of group interests.
Finally, the range of valuations of transportation system outcomes revealed are consistent
with the principle stated at the beginning of this study: transportation planning has multiple
goals and multiple constituencies, and a search for a unitary valuation of transportation
outcomes in the style of traditional benefit cost analyses is likely to mask this reality.
Analytical approaches to planning problems of this sort work best when they help individuals
and groups understand their own values and tradeoffs, as well as those of others in a
planning process. While the task is less clearly specified than that of optimization or
Goals for intelligent transportation system planning Ill

maximization approaches to transportation planning, analytical frameworks that acknowledge


multiple goals and constituencies are better suited to decision-making realities of the current
transportation environment, and hence may be a necessary link to connect technical
knowledge and decision making outcomes in the transportation realm.

A~knolrledgemenrs-The authors wish to acknowledge the Road Commission of Oakland County for the sup-
port of this research and Gwo-Wei Torng, Richard Wallace and Daniel Shapiro for their invaluable research assis-
tance.

REFERENCES
Bunch D. S., Bradley M., Golob T. F. and Kitamura R. (1993) Demand for Clean-Fuel Vehicles in California:
A Discrete-Choice Stated Preference Pilot Project. Transpn Rex-A 27, 237-253.
Chen K. and Underwood S. E. (1988) Integrative Analytical Assessment: A Hybrid Method for Facilitating
Negotiations. Negofiarion J. April 1988, 183-197.
de Neufville R. and Keeney R. L. (1972) Use of Decision Analysis in Airport Development in Mexico City.
Analysis of Public Systems (Drake A. W., Keeney R. L. and Morst M. P., Eds) MIT Press, Cambridge, MA.
Downs, A. (1992) Smck in TruJic: Coping with Peak-Hour Congestion. Brookings Institution, Washington,
DC, and Lincoln Institute of Land Policy, Cambridge, MA.
Edwards W. and Newman J. R. (1982) Multiattribute Evaluation. Sage Publications, Newbury Park, CA.
Giuliano G. (1985) A Multicriteria Method for Transportation Investment Planning. Transpn Res-A 19, 29-41.
Gomes L. F. A. M. (1989) Multicriteria Ranking of Urban Transportation System Alternatives. J. Adv.
Transpn. 23, 43-52.
Gordon D. (1992) Intelligent Vehicle/Highway Systems: An Environmental Perspective. In Trunspormfion,
Information Technology und Public Policy: Instirutionul and Environmenrul Issues in IVHS. Proceedings of
IVHS Policy: A Workshop on Insrirutional and Environmental Issues (Gifford J. L., Horan T. A. and Sperling
D.. Eds), Asilomar Conference Center, Monterey, California. The Institute of Public Policy, George Mason
University, Fairfax. VA.
Harker P. T. and Vargas L. G. (1988) The Theory of Ratio Scale Estimation: Saaty’s Analysis Hierarchy
Process. Munage. Sci. 33, 1383-1403.
IVHS America (1992) Strategic Plan for Intelligent Vehicle-Highway Systems in the United States. Report
Number IVHS-AMER-92-3.
Keeney R. L. and Raiffa H. (1976) Decisions with Mulriple Objectives: Preferences and Value Trudeofs. Wiley,
New York.
MITRE (1995) Assessment of ITS Benefits: Early Results. Sponsored by Federal Highway Administration,
U.S. Department of Transportation, Washington, DC. Report No.: FHWA/JPO-96/OOI.
Roy B. and Hugonnard J. C. (1982) Ranking of Suburban Line Extension Projects on the Paris Metro System
by a Multicriteria Method. Transpn Res. 16A, 301-312.
Saaty T. L. (1976) Axiomatic Foundation of the Analytic Hierarchy Process. Munage. Sci. 32, 841-855.
Saaty T. L. (1980) The Analytic Hierarchy Process: Planning. Priority Setting. Resource Allocution. McGraw-
Hill, New York, NY.
Saaty T. L. and Kearns K. P. (1985) Anulyrical Planning: The Orgunixtion of’Syxrems. Pergamon Press, New
York, NY.
Saaty T. L. and Vargas L. G. (1982) The Logic of Priorities: Applications in Business, Energy, Health and
Transporfation. Kluwer-Nijhoff, Boston, MA.
Shladover S. (1993) Potential Contributions of Intelligent Vehicle/Highway System (IVHS) to Reducing
Transportation’s Greenhouse Gas Production. Transpn Rex-A 27, 207-216.
Srinivasan V., Flachsbart P. G., Dajani J. S. and Hartley R. G. (1981) Forecasting the Effectiveness of Work-Trip
Gasoline Conservation Policies through Conjoint Analysis. J. Marketing 45, 157-l 72.
Underwood S. (1994) FAST-TRAC: Evaluating an Integrated Intelligent Vehicle-Highway System. In
Proceedings,for Fourth Annuul Meeting oj’lVHS America, 17-20 April 1994, Atlanta, GA.

You might also like