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

net/publication/284200030

Alternative Fuel Infrastructure and Customer Location Impacts on Fleet Mix

and Vehicle Routing

Article  in  Transportation Journal · September 2015

DOI: 10.5325/transportationj.54.4.0409

CITATIONS READS

7 644

2 authors:

Joonhwan In John Bell

Chung-Ang University University of Tennessee
11 PUBLICATIONS   313 CITATIONS    41 PUBLICATIONS   1,907 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Crowdsourced Logistics View project

The Crowdbased Logistics Ecosystem and Implications for Omnichannel Distribution View project

All content following this page was uploaded by John Bell on 28 December 2015.

The user has requested enhancement of the downloaded file.

 Fall
2015
Vol. 54. No. 4

TRANSPORTATION JOURNAL™
A Journal of APICS Supply Chain Council

t h e p e n n s y l va n i a s t a t e u n i v e r s i t y p r e s s
Mary Holcomb, Yoshinori Suzuki,
Evelyn Thomchick and Thomas Goldsby, Editors
Juan Carlos Martin Hernández, European Regional Editor
S. I. Ivan Su, Asian Regional Editor
Mary Holcomb, Industry Notes Editor
Richard R. Young, Book Review Editor
Kusumal Ruamsook, Assistant Editor

editorial review board Keith Mason

John Bell John Mello
Yemisi A. Bolumole Paul R. Murphy
David Cantor Robert Novack
Haozhe Chen Richard D. Stewart
Thomas M. Corsi Pete Swan
Michael R. Crum John E. Tyworth
Martin Dresner Bijan Vasigh
Curtis M. Grimm David B. Vellenga
Svienn Gudmundsson W. David Walls
Christian Hofer Robert J. Windle
Milan Janic Kefeng Xu
Carol J. Johnson Richard R. Young
Scott B. Keller Zach Zacharia
Ira A. Lewis Walter Zinn
Robert C. Lieb
2015 Vol. 54 No. 4

TRANSPORTATION JOURNAL ™

iv APICS Introduction: Advancing the Global Supply Chain End-to-End

v Editors’ Introduction

Articles
409 Alternative Fuel Infrastructure and Customer Location Impacts on Fleet
Mix and Vehicle Routing / Joonhwan In and John E. Bell
438 Are We in This Together? The Dynamics and Performance Implications of
Dependence Asymmetry and Joint Dependence in Logistics Outsourcing
Relationships / Adriana Rossiter Hofer
473 Implementing Off-peak Deliveries in the Greater Toronto Area: Costs,
Benefits, Challenges / Jessica McPhee, Ari Paunonen, Taufiq Ramji,
and James H. Bookbinder

Industry Notes
496 Inland Intermodal Terminals Location Criteria Evaluation: The Case of
Croatia / Violeta Roso, Nikolina Brnjac, and Borna Abramovic
516 Safety Culture from an Interdisciplinary Perspective: Conceptualizing
a Hierarchical Feedback-based Transportation Framework / Anjala S.
Krishen, Pushkin Kachroo, Shaurya Agarwal, S. Shankar Sastry,
and Masha Wilson

Book Review
535 Robert E. Gallamore and John R. Meyer. American Railroads—Decline and
Renaissance in the Twentieth Century / John C. Spychalski
The Transportation Journal™ (TJ) is devoted
to the publication of articles that present
new knowledge relating to all sectors of the
supply chain/logistics/transportation field.
These sectors include supply chain/logistics
management strategies and techniques;
carrier (transport firm) and contract logistics
firm (3PL and 4PL) management strategies
and techniques; transport economics; regu-
lation, promotion, and other dimensions of
public policy toward transport and logistics;
and education. Manuscripts published in TJ
undergo a double blind refereed process.
TJ is included in major citation indices and
abstract databases and has been listed in the
Social Sciences Citation Index since 1970. TJ
is also included in SCOPUS.
The Transportation Journal impact factor
for 2010 is 2.348 and the five-year impact
factor is 2.853. TJ ranked second out of
twenty-three journals in the transportation
category for both one-year and five-year
impact factors.
No responsibility for the views expressed
by the authors in this journal is assumed by
the editor, the Editorial Review Board, or the
publisher.
Subscription Information
The Transportation Journal™ (ISSN 0041-
1612; E-ISSN 2157-328X) is published
quarterly by The Pennsylvania State
University Press, 820 N. University Drive,
USB 1, Suite C, University Park, PA 16802.
Non-member subscriptions, claims, and
changes of address should be directed
to our subscription agent, the Johns
Hopkins University Press, P.O. Box 19966,
Baltimore, MD 21211, phone 1-800-548-1784
(outside USA and Canada: 410-516-6987),
jrnlcirc@press.jhu.edu. Subscribers are
requested to notify the Press and their
local postmaster immediately of change
of address. All correspondence regard-
ing permissions and advertisements
should be addressed to Penn State Press,
www.psupress.org.
Copyright
The journal is registered under its ISSN
with the Copyright Clearance Center, 222
Rosewood Drive, Danvers, MA 01923 (www
.copyright.com). For information about
reprints or multiple copying for class-
room use, contact the CCC’s Academic
Permissions Service, or write to The
Pennsylvania State University Press.
Copyright © 2015 APICS. All rights
reserved. No copies may be made without
the written permission of the publisher.
Submission Information
All articles, notes, comments, and book
reviews must be submitted electronically
to www.editorialmanager.com/TransJour.
Articles are received with the under-
standing that they have not been either
published or submitted for review
elsewhere. For more information, please
e-mail the editors, Mary Holcomb at mhol-
comb@utk.edu (or call +1 865-974-1658),
Yoshinori Suzuki at ysuzuki@iastate.edu,
or Evelyn Thomchick at ethomchick@psu.
edu. For questions about book reviews,
please contact Richard Young at rry100@
psu.edu. For questions about industry
notes, please contact Mary Holcomb at
mholcomb@utk.edu.
Adherence to the following guidelines will
facilitate the publication of articles after they
undergo the review process.
1. Manuscripts must be submitted electroni-
cally in Microsoft Word format (Times New
Roman font, size 12, flush left, double spaced).
Printed copy submissions or submissions in
other digital formats will not be accepted.
Pages must be numbered. Length of text (not
including tables, figures, endnotes, and refer-
ences) should be within the range of 3,000–
6,000 words. For a manuscript submitted as an
industry note, length of text should be approxi-
mately 3,000 words. The manuscript must be
accompanied by an abstract of 100–200 words,
which clearly states the principal conclusions
of the article.
2. Documents should not contain automated
formatting. The following represent some exam-
ples of formatting that should be avoided:
A. No function of “Track Changes” should
be in use.
B. The “Style” field should read “Normal”
throughout the text.
C. The article title should be set flush left
(do not tab or format an indent).
D. The “tab” key should be used for para-
graph indentation, not the paragraph indent
function or the space bar.
E. Headings should not have bold type but
should be set off with line spaces (subheads
may be placed in italic to distinguish them
from a full heading).
F. Do not use automated lists. All numbers
or bullets must be keyed.
G. The N-dash should be used for inclusive
numbers, not a hyphen.
H. Epigraphs and extractions from other
published works should be set off with line
spacing. Do not format an indent.
3. A title page must be submitted as a separate
file from the article text. Information of each
co-author should contain, name, job title, affili-
ation, and mailing address. For a corresponding
author, please also provide email address and
phone number.
4. Tables, figures, charts, and photos must be
submitted as a separate file from the article text
with an indication in the text for placement,
i.e., <Table 1>, <Figure 2>, <Photo 3>. They must
be submitted in the original format at the size
the author would like them to appear.
A. Tables should be submitted in Word.
B. Charts and graphs should be submitted
in Excel.
C. Digital images should be submitted in
either .tiff, .eps, or .jpeg files at 300 dpi in
order for the images to be print-worthy.
D. All Digital files should be grey scale.
5. Endnotes are used to elaborate on informa-
tion presented in the article text. If endnotes
are used, they must be numbered consecutively
throughout the article and be indicated by super-
script numerals following the punctuation. The
endnote numbers at the end of the article should
not be superscript text and should be a number
followed by a period. All endnote entries must be
double-spaced at the end of the article and must
appear before references. If it is necessary to cite
a source or sources pertaining to a particular
endnote, citation formation must be identical to
that used for references (see #5 below).
6. Manuscripts must consistently conform to
the current edition of The Chicago Manual of Style
(CMS). References should follow author-date
documentation. Examples of reference citations
for journal articles and books are shown below.
For more examples, please check CMS.
A. Books and journals in the text: Reference
citations in the article must be enclosed in
parentheses and listed in alphabetical order
as follows: (Author 2000, 3); (Murphy 2000,
48–51; Poist and Allen 1999).
B. Books and journals in the reference
list: References citing source materials must
be listed at the end of the article and must
include, in order, the following information:
Journal: Author last name, first name ini-
tial. Year of publication. “Title of article.” Title
of journal Volume(issue):page numbers.
Books: Author last name, first name initial.
Year of publication. Title of Book, city: name of
publisher.
For other types of citations, please check
the CMS.
7. Industry Notes is a section of the Transportation
Journal intended for articles with an emphasis
on transportation, logistics, and supply chain
topics that are managerially relevant and timely.
An Industry Note should be a thesis-driven
exploration of a topic. In most cases, except
for conceptual-based research, the exploration
would involve the collection, compilation, and
analysis of data to make original observations,
to show relationships between the data, and to
make conclusions that would add to the exist-
ing knowledge of the topic/subject. To this
end, a paper submitted as an Industry Note is
expected to utilize a variety of methodological
approaches. Research that examines new areas
and/or the study of traditional approaches in
different perspectives is also appropriate for
an Industry Note. An Industry Note should be
approximately 3,000 words in length and should
follow the style guidelines of the Transportation
Journal.
8. Book reviews should be approximately 750
words in length. The review must be headed by
the following information: title of publication;
name of author or editor; publisher;
publisher’s address; date of publication; num-
ber of pages; ISBN number (if available in both
hardcover and paperback, indicate numbers
for both); and selling price(s). The text of the
review must be double-spaced. The reviewer’s
name, full professional title, employer affilia-
tion, and address must be indicated at the end
of the review.
APICS Introduction: Advancing the Global Supply Chain End-to-End

While the American Society of Transportation and Logistics (AST&L)

officially integrated with APICS in July, both groups have been working
together closely since 2013. Now that we are officially one organization, our
portfolio of education, training, and certification programs are more valu-
able than ever in helping professionals advance and supply chain organiza-
tions thrive in a rapidly evolving environment.
As part of this increasing value, APICS and the APICS Supply Chain
Council are proud to welcome the Transportation Journal into our portfolio
of publications. Whether you are a long-time or first-time reader, whether
you are on faculty or on point as a practitioner, I believe you will find this to
be a valuable resource in understanding the dynamics of the supply chain
from strategy development to realization.
Transportation Journal is one example of how coming together, APICS
and AST&L have more capabilities to address two of the most important
topics in the global economy today—developing talent across the entire
supply chain and elevating supply chain performance. The merger will
deliver unmatched content and subject matter expertise that will enable
individuals, organizations, and learning institutions to meet these and
other key supply chain and logistics challenges. Our goal is simple: to ele-
vate the role of logistics, distribution, and reverse logistics in the supply
chain organization and in the global marketplace.
Although we are excited about the possibilities ahead now that we are
one organization, our main objective is to continue to provide excellent
service to all of our members and customers. Thank you for your support
during these exciting changes.

Abe Eshkenazi, CSCP, CPA, CAE

Chief Executive Officer
APICS

Transportation Journal, Vol. 54, No. 4, 2015

Copyright © 2015 The Pennsylvania State University,
University Park, PA
Editors’ Introduction

In our previous issue we announced some news about the new co-editors
assuming the leadership of the Transportation Journal. In this issue we
announce a change in the ownership of the journal. On July 1, 2015, the
American Society of Transportation and Logistics merged with APICS,
and the Transportation Journal is under the domain of APICS Supply Chain
Council. The editorial objectives of the journal will remain the same
with a focus on transportation and logistics, as relevant to supply chain
management. The editorial staff very much looks forward to serving a
broader academic and practitioner readership.
On a personal note, I will be leaving my co-editor position on December 31,
2015. I feel honored and privileged to have served first as editor-in-chief and
then co-editor for six-and-a-half years. We have accomplished a great deal
in updating the journal’s publishing processes and implementing some
new practices, such as the rotating co-editor positions, establishment of an
Industry Notes editor, and initiating best paper and outstanding reviewer
awards. I use the term “we” because all of the changes reflected a team effort.
I hope I don’t omit anyone in expressing my appreciation to individuals
and groups who worked hard and offered support over the past several years.
The following are deserving of thanks and appreciation: Laurie Denham,
President of the former AST&L and the AST&L Board of Directors for their
continued support of the journal; Kusumal Ruamsook, assistant editor,
whose hard work and attention to detail ensured smooth transitions between
accepted manuscripts and published articles; the Penn State Press leadership
and personnel for their guidance, professional assistance, and patience with
our often late adherence to deadlines; Tom Goldsby, as the first co-editor of the
journal for his thoughtfulness and thoroughness in manuscript processing;
Mary Holcomb and Yoshi Suzuki for their willingness to take on the leadership
for the next phase of the journal; Rich Young, for twice coming to our rescue
to serve as book review editor; and last and definitely not least, the excellent
and hardworking editorial review board that enabled the continued quality
of articles. Of course, we all express our appreciation to the contributors of
articles and the readership of the journal who have maintained the longstand-
ing respect and reputation the Transportation Journal has held.

Best Regards,
Evelyn Thomchick, Co-Editor

Transportation Journal, Vol. 54, No. 4, 2015

Copyright © 2015 The Pennsylvania State University,
University Park, PA
vi / TRANSPORTATION JOURNAL™

In this journal we have a nice variety of article topics. Vehicle routing

has long been a topic of research in transportation. However, the use
of alternative fuel vehicles and the configuration of alternative fuel
infrastructure add complexity to the problem. In the first article, In and
Bell use a mixed integer programming model and logistics regression to
assess the impacts of both the density of the alternative fuel infrastructure
and the patterns of customer locations on the optimal fleet composition.
In the second article, Hofer investigates the mechanisms through which
two dimensions of interorganizational interdependence—dependence
asymmetry and joint dependence—impact a customer’s key attitudinal
and behavioral dimensions in the relationship with a third-party logistics
provider (3PL), as well as associated performance outcomes.
The third article is also involves solving a routing problem. McPhee,
Paunonen, Ramji, and Bookbinder quantify Nestlé’s costs and benefits of
modifying their ice cream supply chain to incorporate nighttime deliv-
eries, while providing a framework for the regulatory, conceptual, and
inertial obstacles to implementation.
In our first Industry Note, Rosa, Brnjac, and Abramovic use the ana-
lytic hierarchy process to evaluate the criteria used for deciding on suitable
locations for intermodal terminals in Croatia. In the second Industry Note,
Krishen, Krachroo, Argarwal, Sastry, and Wilson introduce a feedback
safety model that utilizes and combines disparate data sources and allows
for model-based action to alleviate the problems. Our book review for this
issue is American Railroads—Decline and Renaissance in the Twentieth Century,
authored by Robert E. Gallamore and John R. Meyer and reviewed by John
C. Spychalski.

Mary Holcomb, Yoshinori Suzuki, and

Evelyn Thomchick, Co-Editors
Alternative Fuel Infrastructure and Customer Location
Impacts on Fleet Mix and Vehicle Routing

Joonhwan In and John E. Bell

Abstract
An increase in the use of conventional vehicles in combination with
alternative fuel vehicles requires trucking carriers to make a decision on
when to use a mixed fleet. The purpose of this article is to investigate how
alternative fuel infrastructure and patterns of customer locations influ-
ence fleet compositions when minimizing routing costs. The variable-
route vehicle refueling problem (VRVRP) for the mixed fleet is formulated
as a mixed-integer programming model and solved with an optimization
method. A logistic regression is then employed to assess the impacts of
both the density of the alternative fuel infrastructure and the patterns of
customer locations on the optimal fleet composition. Experimental results
reveal that the right fleet mix decision is impacted by customer location
patterns and available fuel infrastructure. While alternative fuel trucks
were favored over conventional trucks with a dense alternative fuel infra-
structure, the likelihood of mixed fleet use varied across different patterns
of customer demand locations. Further, our analysis suggests that ship-
pers and trucking carriers should understand the relative benefits and con-
straints of different geographical dispersions of customers and densities of
alternative fuel infrastructures when designing optimal delivery strategies.

Keywords
Alternative fuel, vehicle refueling, vehicle routing, mixed-integer
programming, optimization
Joonhwan In Transportation Journal, Vol. 54, No. 4, 2015
University of Tennessee Copyright © 2015 The Pennsylvania State
jin1@vols.utk.edu University, University Park, PA

John E. Bell
University of Tennessee
bell@utk.edu
410 / TRANSPORTATION JOURNAL™

Introduction
With increasing competition and uncertainty in global markets, the
location of customers and facilities is still considered an important strategic
factor in supply chain management decisions. For instance, the complex-
ity of location affects costs of coordination and control in an outsourcing
context (Handley and Benton 2013), while the geography of production,
distribution, and consumption shapes patterns in the global supply
chain (Rodrigue 2012). Along the same lines, firms decide the location of
retail stores and support facilities by balancing the spatial distribution of
demand with the mission to deliver values to customers at the right cost
(Duan and Mela 2009; Lee and Lee 2012). In the area of service supply chains,
Goldstein et al. (2002) examined the performance implications of locations
in a healthcare context, and others have studied the strategic implications
of demand patterns in logistics and delivery (Bell and Griffis 2010). Within
that context, the classic vehicle routing problem (VRP) has gained atten-
tion over several decades in the logistics, operations management, and
mathematics disciplines (Ballou and Agarwal 1988; Bell and Griffis 2010;
Christofides, Mingozzi, and Toth 1979; Laporte 2007). Although there is no
clear consensus on the definition, the VRP in general aims to design the
optimal delivery solution to a set of customers who are geographically dis-
tributed (Laporte 2007). Most previous research in the VRP literature has
focused on meeting customer demand while minimizing total distances or
total costs (e.g., fuel costs) given such constraints as time windows, route
lengths, demand characteristics, and vehicle capacity (Eksioglu, Vural, and
Reisman 2009; Lee and Ueng 1999; Ma and Suo 2006).
In the logistics industry, trucking companies face environmental
issues such as air pollutant and greenhouse gas emissions (Lieb and Lieb
2015). According to Davis, Diegel, and Boundy (2013), 17 percent of the
total US petroleum consumption comes from classes 7–8 heavy trucks.
As one of the largest sources of greenhouse gases (US Environmental
Protection Agency 2013), the transportation industry has had to con-
sider the utilization of alternative fuel vehicles (Lieb and Lieb 2012). For
instance, United Parcel Service (UPS) has more than 5,000 alternative fuel
vehicles (Truckinginfo 2015) and plans to add 1,400 compressed natural
gas vehicles (CNG) over the next year while planning to expand the infra-
structure of CNG fuel stations. The US Postal Service (2014) operates alter-
native fuel delivery vehicles such as those powered by ethanol-blended
gasoline, compressed natural gas, electricity, and propane, and uses more
than 40,000 alternative fuel vehicles that account for about 21 percent of
 In and Bell: Fuel Infrastructure and Customer Location Impacts \ 411

total vehicles (Kemp 2012). Further, a presidential memorandum issued

in 2011 states, “All new vehicles that agencies lease or purchase should
be alternative fuel vehicles by December, 31, 2015.” According to a 2013
federal fleet report, alternative fuel vehicles account for 32 percent of fed-
erally owned vehicles. However, recent statistics show that over half of
the existing alternative fuels refueling stations exist in only eight states:
California, Oregon, Washington, Texas, New York, Florida, Michigan, and
Illinois (Alternative Fuel Data Center 2015). Therefore, as alternative fuels
vehicle fleets continue to grow, there is an immediate need to understand
the available refueling infrastructure to support these vehicles.
In academia, environmentally motivated or green logistics research has
received considerable attention and is aimed at reducing pollutant emis-
sions and improving the efficiency of vehicle routing and scheduling (Lin
et al. 2014). Much research is focused on green vehicle-routing (Erdogan
and Miller-Hooks 2012), pollution routing (Bektas and Laporte 2011), loca-
tion of alternative fuel stops (Wang and Lin 2009), and the impact of alter-
native fuel fleets (Haller et al. 2007).
However, most routing and refueling research in the literature has
considered either conventional (i.e., carbon-based) or alternative fuel
vehicles separately without considering how to optimally deploy and route
a mix of conventional and alternative fuel vehicles to best serve customer
demand with the exception of a few studies (Goeke and Schneider 2015).
For instance, Suzuki (2011) examined the minimization of fuel consump-
tion and pollutant emissions for conventional truck routing, whereas
Erdogan and Miller-Hooks (2012) investigated a green vehicle-routing and
refueling problem for alternative vehicles given a limited refueling infra-
structure. Most recently, Goeke and Schneider (2015) developed a heuris-
tic solution for a mixed fleet of electric and conventional vehicles with
the assumption that a truck driver can recharge an electric vehicle at any
available station.
With regard to alternative fuel vehicles, the selection of vehicle type
is a multidimensional decision that weighs such factors as the fuel avail-
ability and the distribution of alternative fueling infrastructure (Erdogan
and Miller-Hooks 2012). The benefits gained from efficient vehicle rout-
ing, which minimizes vehicle fuel consumption and emissions, may
also depend on transportation network structure (Ahn and Rakha 2013).
Additionally, the quality of a routing solution and the preference for one
vehicle type over another could be influenced by the spatial dispersion of
customer locations (Ballou 1990; Bell and Griffis 2010) as well as the network
412 / TRANSPORTATION JOURNAL™

of fueling infrastructure (e.g., alternative fueling stations). Therefore, the

literature suggests that a properly selected fleet of mixed conventional and
alternative fuel vehicles could perhaps outperform the deployment of a
single type of vehicle, especially under certain customer location patterns
and fueling infrastructure networks.
Geographical location of customers has been examined in the VRP
literature (Eksioglu, Vural, and Reisman 2009). However, most research
has focused on assessing the performance of proposed algorithms given
the categorized locations of customers such as random and urban/rural
area, and emphasized the robustness of proposed algorithms without
investigating how customer location affects the best delivery routes or
optimal mix of vehicles used in a fleet (Ballou 1990; Bell and Griffis 2010;
Zografos and Androutsopoulos 2004).
Therefore, based on these gaps in the literature, the current research
aims to answer the following overarching question: how do the density of
alternative fuel infrastructure and patterns of customer locations affect a
mixed fleet use decision? Hence, this study investigates the optimal fleet
design by considering alternative fuel infrastructure levels and different
customer location patterns given that both conventional and alternative
fuel vehicles are available. The proposed model is formulated and solved as
a mixed integer programming to guarantee the optimality of the solution,
and logistics regression is used to identify the impact of demand patterns
and alternative fueling infrastructure on the optimal fleet mix.
The following sections review the relevant vehicle routing and refuel-
ing literature and explain the proposed analytical model. After the research
methodology and experimental results are provided, research implications
and directions for future research are also discussed.
Theoretical Foundation

Literature Review

The underlying problem of interest for this study is the variable-route

vehicle refueling problem (VRVRP), where the shortest route and the vehi-
cle refueling problems are examined together (Suzuki and Dai 2012). This
problem considers a heterogeneous fleet with respect to different vehicle
driving ranges with time constraints. According to the VRP taxonomy
(Eksioglu, Vural, and Reisman 2009), the current problem is relevant to
subcategories of physical characteristics: geographical location of cus-
tomers and vehicle heterogeneity. The VRP taxonomy by Eksioglu, Vural,
and Reisman (2009) classifies locations of customers into urban, rural, and
 In and Bell: Fuel Infrastructure and Customer Location Impacts \ 413

mixed types, and considers vehicle heterogeneity with respect to vehicle

capacity, but it does not take into account the diversity of vehicle driving
ranges as seen with alternative fuel vehicles and vehicle refueling. Thus,
previous studies on alternative fuel vehicles relevant to the topic of this
study are also surveyed. Table 1 synthesizes selected vehicle routing studies
and highlights the gaps in the literature.
The VRVRP considers routing and refueling policy simultaneously,
not sequentially. Therefore, the solution of the VRVRP outperforms or at
least matches the solution when the vehicle refueling problem is solved
after determining the optimal routing schedule (Suzuki and Dai 2012).
However, the VRVRP has not been thoroughly examined in the litera-
ture with the exception of a few studies (Erdogan and Miller-Hooks 2012;
Khuller, Malekian, and Mestre 2007; S. H. Lin 2008; Suzuki and Dai 2012,
2013). For conventional trucks, Suzuki and Dai (2012) compared solutions
of the VRVRP to those with a sequential routing and refueling approach,
and found empirical evidence that supported the superiority of the simul-
taneous consideration of the VRVRP’s routing and refueling policy over a
sequential approach. Suzuki and Dai (2013) developed a decision-support
approach that has the dual objective of minimizing total costs and vehicle
miles. For alternative fuel vehicles, Erdogan and Miller-Hooks (2012) exam-
ined the VRVRP with constraints on the driving range of alternative fuel
trucks and limited refueling infrastructure. Numerical experiments were
conducted under uniform and clustered customer location conditions
along with two refueling configurations, showing that results of the pro-
posed heuristics varied with customer locations and refueling configura-
tions (Erdogan and Miller-Hooks 2012). In addition, Schneider, Stenger, and
Goeke (2014) focused on addressing the electric vehicle routing problem
that incorporates rechargeable stations, time windows, and vehicle capaci-
ties and proposing a heuristic method. Although the authors considered
customer location patterns such as random, cluster, and a combination of
random and cluster patterns, the focus of their study was on developing a
new heuristic method rather than exploring the implications of customer
location patterns and the density of alternative fuel infrastructure.
In terms of customer location patterns, most research in the VRP lit-
erature examined performance of algorithms under certain patterns of cus-
tomer locations (Bell and Griffis 2010; Dondo and Cerdá 2007; Zografos and
Androutsopoulos 2004) (see table 1). While most research considered one or
two types of customer location patterns with an emphasis on random and
cluster patterns (Dondo and Cerdá 2007; Hiermann, Puchinger, and Hartl
Table 1/Summary of Representative Studies
Problem Location of Customers Vehicle Heterogeneity
Types of Vehicles
Actual Urban/ Vehicle Applied
Study Routing Refueling Data Random Sector Rural Coast Cluster Conventional AFVa Capacity Methods
Current study X X X X X X X X X Optimization
Angelelli and X X X Heuristic
Speranza 2002
Bard, Huang, X X X Optimization/
Dror, and Jaillet heuristic
1998
414 / TRANSPORTATION JOURNAL™

Bell and Griffis X X X X X X X Heuristic

2010
Dondo and Cerdá X X X X X Optimization/
2007 heuristic
Erdogan and X X X X X X Optimization/
Miller-Hooks heuristic
2012
Fleischmann, X X X Real-time
Gnutzmann, and methods
Sandvoss 2004
Fu 2002 X X X Heuristic
Gendreau, X X X Heuristic
Laporte, and
Séguin 1996
Goeke and X X X X X X Heuristic
Schneider 2015
Hiermann, X X X X X Heuristic
Puchinger, and
Hartl 2014
 Problem Location of Customers Vehicle Heterogeneity
Types of Vehicles
Actual Urban/ Vehicle Applied
Study Routing Refueling Data Random Sector Rural Coast Cluster Conventional AFVa Capacity Methods
Jaw, Odoni, X X X X Heuristic
Psaraftis, and
Wilson 1986
Malandraki and X X X Optimization/
Daskin 1992 heuristic
Savelsbergh and X X X X Simulation/
Sol 1998 real-time
methods
Schneider, X X X X X X Heuristic
Stenger, and
Goeke 2014
Suzuki and Dai X X X X Optimization/
2012 heuristic
Suzuki and Dai X X X X Optimization/
2013 heuristic
Trudeau and Dror X X X Heuristic
1992
Zografos and X X X X Heuristic
Androutsopoulos
2004

a
AFV: Alternative fuel vehicles, adapted from Eksioglu, Vural, and Reisman 2009
In and Bell: Fuel Infrastructure and Customer Location Impacts \ 415
416 / TRANSPORTATION JOURNAL™

2014; Schneider, Stenger, and Goeke 2014; Zografos and Androutsopoulos

2004), Bell and Griffis (2010) compared performance of algorithms under
five different patterns of customer locations: random, sector, urban/rural,
coast, and cluster. Their ant colony optimization heuristics performed bet-
ter on random, sector and cluster configurations than on urban/rural and
coast configurations (Bell and Griffis 2010), suggesting that performance of
algorithms can vary with spatial configurations of truck stops (Ballou and
Agarwal 1988; Erdogan and Miller-Hooks 2012).
When it comes to vehicle heterogeneity, some empirical studies devel-
oped vehicle routing algorithms for a heterogeneous fleet, especially with
respect to the vehicle capacity differences (Choi and Tcha 2007; Dondo and
Cerdá 2007; Hiermann, Puchinger, and Hartl 2014; Koç et al. 2015). In terms
of alternative fuel vehicles, Hiermann, Puchinger, and Hartl (2014) devel-
oped a model that incorporates different electric vehicle types with respect
to capacity and costs, and tested the performance of the proposed heuris-
tic by considering multiple customer location patterns such as random,
cluster, and a combination of both. However, it is believed that no prior
research has examined the impact of alternative fuel infrastructure and
customer location patterns on the heterogeneous fleet mix determination
in the VRVRP with respect to variance in driving ranges of different vehicle
types (i.e., a fleet of vehicles run on different fuels).

Problem Definition and Formulation

The VRVRP with the heterogeneous fleet for this study extends models
introduced in Erdogan and Miller-Hooks (2012) and Suzuki and Dai (2012).
The proposed model of this study is defined on a complete graph G = (V, E)
where vertex set V is a combination of customers, the depot, standard fuel
(e.g., diesel) stops, and alternative fuel stops. Whereas n customers are rep-
resented by node 1 through n, node 0 represents both the starting node and
end node (i.e., depot). The edge (i, j) represents the connection between
nodes i and j. Each edge is associated with the distance ( dij ), and the dis-
tance matrix is symmetric (i.e., dij = d ji ). Additionally, Rij = {1, 2, ..., Kij}
refers to a set of standard fuel stops in the edge (i, j), whereas
ARij = {1, 2, ..., Lij} represents a set of alternative fuel stops in the edge (i, j).
Thus, the size of vertex set V is equal to 1 + n + K ij + Lij .
The proposed model aims to find the optimal fleet mix, routing, and
refueling of conventional and alternative fuel trucks minimizing the total
fueling costs with the assumption that maintenance costs of conventional
 In and Bell: Fuel Infrastructure and Customer Location Impacts \ 417

trucks (i.e., diesel trucks) are similar to those of alternative fuel trucks (i.e.,
compressed natural gas [CNG] trucks). Some companies operate CNG or
other alternative fuel–based class 8 trucks; however, the information on
costs is not enough to compare maintenance costs of diesel trucks with those
of alternative fuel trucks (Krupnick 2011). Further, the evidence for transit
buses and long-haul trucks is mixed as to maintenance costs (Krupnick 2011).
Hence, the proposed model does not take into account maintenance costs in
the objective function by assuming that maintenance costs of alternative fuel
trucks are comparable to those of conventional trucks. Table 2 shows nota-
tions of parameters used in formulating the problem of interest.

Table 2/Model Parameters

General Parameters
dij The distance of the edge (i, j)
Tmax Travel-time constraint
Parameters for Conventional Vehicles
ccij Conventional vehicle’s fuel economy (mpg) in the edge(i, j)
cpijk The unit standard fuel price at the kth refueling point in the edge (i, j)
ρc The minimum refueling quantity for conventional vehicle
Qc The tank capacity of conventional vehicle

μc The minimum fuel level constraints for conventional vehicle

α The fixed amount of refueling duration per conventional fuel stop

θcorigin The amount of fuel (gallons) in tank at the origin (starting fuel level),

θcdest The required amount of fuel in tank at the final destination (ending fuel level)

vc The average speed of the conventional vehicle

Parameters for Alternative Fuel Vehicles (AFV)
gcij AFV’s miles per equivalent gallon (mpeg) in the edge(i, j)
gpijk The unit alternative fuel price at the kth refueling point in the edge (i, j)

ρaf The minimum refueling quantity for AFV

Qaf The tank capacity of AFV

μaf The minimum fuel level constraints for AFV

β The fixed amount of refueling duration per alternative fuel stop

θaforigin The amount of alternative fuel (gallons) in tank at origin (starting fuel level),

The required amount of alternative fuel in tank at the final destination (ending
θafdest
fuel level)
v af The average speed of the alternative fuel vehicle
418 / TRANSPORTATION JOURNAL™

Except for fuel economy and time window parameters, all other
parameters are relevant to refueling activities: fuel price, minimum refu-
eling quantity, tank capacity, refueling duration, and minimum fuel level
at the origin and destination (i.e., depot). Those parameters account for
unique features of conventional and alternative fuel vehicles, which in
turn impact the choice of a fleet composition. For each type of vehicles,
two types of decision variables are used: (1) binary decision variables that
determine the route to use and the refueling location; and (2) continuous
decision variables that indicate the refueling quantity at the fuel stop. For
the convenience and modeling purpose, additional decision variables are
used that indicate the number of vehicles selected and the return time of
vehicles (see table 3).
Based on parameters and decision variables defined above, the model
is formulated with some assumptions. Regardless of the type, trucks start
at and return to the depot after serving all customers within a pre-specified

Table 3/Decision Variables for the Proposed Model

Decision Variables for Conventional Vehicles
xij Binary variable such that it is 1 if node j is visited after node i by a conven-
tional vehicle (0 otherwise)

δ ijk Binary variable indicating the refueling location of carbon-based fuels

rijk The fuel level of conventional vehicles at the truck stop ijk
rijo The standard vehicle’s remaining fuel level when departing node i

φijk The refueling quantity of carbon based fuels at the truck stop ijk

ncv The number of conventional vehicles deployed

c
a ij
The arrival time of the conventional vehicle at node j if node j is visited after
node i
Decision Variables for Alternative Fuel Vehicles
zij Binary variable such that it is 1 if node j is visited after node i by an alternative
fuel vehicle (0 otherwise)

εijk Binary variable indicating the refueling location of alternative fuels

sijk The fuel level of alternative fuel vehicles at the truck stop ijk
sijo The standard vehicle’s remaining fuel level when departing node j
ψ ijk The refueling quantity of alternative fuels at the truck stop ijk

nafv The number of alternative fuel vehicles deployed

The arrival time of the alternative fuel vehicle at node j if node j is visited after
aijaf
node i
 In and Bell: Fuel Infrastructure and Customer Location Impacts \ 419

time limit (Barbarosoglu and Ozgur 1999; Cordeau et al. 2002; Cordeau,
Laporte, and Mercier 2004; Erdogan and Miller-Hooks 2012; Privé et al.
2006). Such an assumption is appropriate because a route duration con-
straint is often used to meet hours-of-service regulations or to reduce
overtime labor costs (Bektas and Laporte 2011). A route duration is often
used as a constraint for modeling such problems as soft-drink distribution,
hazardous materials distribution, and catering problem (Bräysy, Dullaert,
and Nakari 2009; Privé et al. 2006; Zografos and Androutsopoulos 2004). In
addition, each customer is served by either a conventional truck or alterna-
tive fuel truck. Constraints on the remaining fuel level, fuel-stop locations,
refueling amount, and fuel-tank capacity are imposed on the model to
incorporate differences in characteristics between conventional and alter-
native fuel trucks. Additionally, the truck capacity in the current model is
left uncapacitated. This assumption helps reduce complexity and is a rea-
sonable assumption in practice when inventory holding costs are large rela-
tive to demand rates, or delivery sizes are small relative to the truck capacity
(Anily and Federgruen 1990). For example, trucks can usually meet all of the
delivery or pickup demand of small-sized products such as express docu-
ments in one route (C. K. Y. Lin 2008). Finally, for simplicity this study also
assumes that both conventional and alternative fuel trucks are always avail-
able at the depot. With these known assumptions, the proposed mixed-
integer programming model based on Suzuki and Dai (2012) and Erdogan
and Miller-Hooks (2012) can be expressed as shown in the appendix.
The objective (1) seeks to minimize total fueling costs for conventional
and alternative fuel vehicles. The first term in the objective represents total
fueling costs for conventional vehicles, whereas the second term refers to
total fueling costs for alternative fuel vehicles. Constraints (2) to (4) ensure
that a single vehicle, regardless of the vehicle type, can serve each customer.
Additionally, only fuel stops on the selected route can be accessible to
deployed vehicles, constraints (7) and (8). Next, constraints (9) to (16) guar-
antee that the purchase quantity of a conventional vehicle is greater than
the minimum refueling quantity, and that its particular fuel level should
be maintained throughout the tour until the deployed vehicle returns to
the depot. The same fuel level constraints are imposed on alternative fuel
vehicles in constraints (17) to (24). Finally, constraints (25) to (31) specify
that regardless of the vehicle type, all vehicles deployed should return to
the depot no later than Tmax (route duration constraint) after serving a sub-
set of customers.
420 / TRANSPORTATION JOURNAL™

Research Methodology

Overview

The VRVRP with a heterogeneous fleet and delivery time window is an NP-
hard problem as it is a variant of the standard VRP, which means that due
to the mathematical complexity of the problem there is no known optimi-
zation algorithm available that is guaranteed to solve all instances of this
problem in practical time. It also implies that heuristic approaches such
as tabu search, genetic algorithms, simulated annealing, and ant colony
optimization are often preferred for larger problems (Griffis, Bell, and
Closs 2012), because optimization techniques can usually solve only the
small-sized instances of a given problem (Ballou 1989). However, in this
initial study of the problem optimization techniques are used for small-
sized instances. Specifically, the proposed model is implemented by using
the IBM-ILOG CPLEX Concert Technology (version 12.5) in Java because
exact solutions can be obtained and compared in the experimental design.
Further, the results of this study can provide a baseline for future com-
parison with the results of different approaches (e.g., metaheuristics).
Additionally, some practical applications such as routing of supply vessels
(Aas et al. 2007) have been shown to have fewer than ten stops, making the
current problem size still within the practical range.
In terms of customer locations, Ballou, and Agarwal (1988) compared
the performance of heuristics under five spatial configurations: random,
cluster, West Coast, sector, and urban/rural. Bell and Griffis (2010) com-
pared the performance of an ant colony optimization heuristic with that
of traditional methods under these five spatial dispersions of customers.
These customer configurations are believed to represent more typical
customer locations in reality (Ballou and Agarwal 1988). Following the lit-
erature, this study generated experimental problems for the VRVRP with
customer locations based on location coordinates used in Bell and Griffis
(2010). Distributions of distances between the customer locations in the
problems studied here are shown in figure 1. These distance matrices show
differences in the demand patterns based on the underlying pattern of geo-
graphic customer location dispersion (Schilling, Rosing, and Revelle 2000).
To provide realistic implications, multiple data sources from the US
Department of Energy (DOE), US Environmental Protection Agency (EPA),
and prior studies were referred to, in order to select appropriate parameter
values (Suzuki and Dai 2012). The parameters used in the experiment are
displayed in table 4.
Figure 1 Distribution of Customer Locations
422 / TRANSPORTATION JOURNAL™

Table 4/Parameters Used in the Experiment

Diesel Truck CNG Truck
Parameters Symbol Value Symbol Value Source

Time constraint Tmax 5 Tmax 5 -

Miles per gallon cc ij 6.0 mpg gc ij 5.1 mpg Kargul 2012

US Depart-
$ 3.91/ gpijk $ 2.09/
Fuel price cpijk ment of
gallon gallon
Energy 2013
Minimum ρc 0 ρaf 0 -
refueling quantity
Suzuki
2008/ Cas-
Fuel-tank capacity Qc 200 gallon Qaf 190 gallon cadia 113
(Freightliner
n.d.)

Minimum fuel μc 0 gallon μaf 0 gallon -

level

α β Suzuki and
Refueling duration 0.5 hour 0.5 hour
Dai 2013
Fuel level at the
θcorigin 100 gallon θaforigin 100 gallon -
origin
Fuel level at the
θafdest 100 gallon θafdest 100 gallon -
destination

vc 40 miles/ v af 40 miles/
Average speed -
hour hour
Average fuel-stop
- 10 miles - 60/ 80 miles -
distance

In the experiment, conventional trucks refer to US class 8 diesel trucks,

whereas alternative fuel trucks refer to class 8 CNG trucks. Experiments of
the current study consider demand patterns composed of six customers.
Suppose that two trucks are deployed and serve three customers, respec-
tively. On average, it takes 5.05 hours including refueling time (0.5 hour) for
two trucks to serve customers and return to the depot because the average
distance between customers for generated test instances is 45.6 miles, and
the average speed of a truck is set to 40 miles/hour. Thus, the time constraint
(Tmax ) is set to five hours to prevent solutions from deploying only one vehi-
cle. If the time window is not set tightly, mixed-integer programming solu-
tions tend to dispatch a single vehicle. The minimum refueling quantity for
both types of vehicles is set to zero to prevent a situation in which the mini-
mum refueling quantity becomes a binding constraint. As of 2014, there are
674 public CNG stations in the United States, which is less than 1 percent
of gasoline stations (US Department of Energy 2015). If we used the current
 In and Bell: Fuel Infrastructure and Customer Location Impacts \ 423

fueling infrastructure network, the imbalance in the fueling infrastructure

would most likely favor the deployment of diesel trucks over alternative
fuel trucks, regardless of customer locations. According to the report by
American Natural Gas Alliance (2012), a CNG industry can be competitive if
it develops a fueling infrastructure equivalent to 10–20 percent of gasoline
stations. If the alternative fueling infrastructure is as dense as 10–20 percent
of gasoline stations, alternative fueling infrastructure may not be consid-
ered a barrier to the adoption of alternative fuel vehicles (Yeh 2007). This
implies that if the distance between stations falls, then mixed fleets and CNG
fleets may become more cost effective. We started our pretesting by assum-
ing that a major industry may be operating in an area where the current CNG
infrastructure has an average distance between stations of 80 miles. In pre-
testing it was noticed that changes in this distance may impact the mixed
fleet decision. Therefore, our experiment uses two levels of alternative fuel
infrastructure with 80 (i.e., low density) and 60 miles (i.e., medium density)
between stations. These distances are not meant to represent any particular
area in the United States and are not based on an empirical sample, but are
only meant to show the ability of the modeling methodology to test for dif-
ference in the fleet mix decision. Additionally, the average distance between
diesel fuel stops are set to 10 miles for both scenarios.

Design of Experiment

Individual test instances with six customer demand locations are generated
based on the data used in the previous studies (Ballou 1990; Ballou and Agarwal
1988; Bell and Griffis 2010). These locations are randomly selected from the list
of the location coordinates in the first problem of random, sector, urban/rural,
cost, and sector configurations from Bell and Griffis (2010). Additionally, the
experiment considers the two types of alternative fueling infrastructures.
Each of 10 cases (5×2) is replicated 100 times with new randomly selected cus-
tomer locations in each replication. The distribution of distances from all of
the test problems was previously shown in figure 1 for the different problem
types. This figure shows that each problem type differs from the others, as
previously shown by Schilling, Rosing, and Revelle (2000), with the random
problem set having a more normal distribution of distances.
A mixed-integer programming optimization model is solved by using
JAVA along with IBM-ILOG CPLEX Concert Technology (ver. 12.5). The time
limit in CPLEX is set to 1,800 seconds with the optimal tolerance of 10−6.
The experiment is run on an AMD CPU 1.9 GHz processor, a 64-bit platform
with 4 GB RAM.
424 / TRANSPORTATION JOURNAL™

Results
In table 5, which provides a comparison of fleet deployment results
across customer locations and alternative fueling infrastructure, it can
be observed that there were dramatic differences in the number of exper-
imental solutions that used all diesel, all CNG, or mixed fleets in their
best solution. For instance, the mixed fleet use was dominant in a cluster
demand pattern, whereas fleets of all CNG trucks were mostly deployed
for a sector demand pattern. For a cluster demand pattern, diesel trucks
were predominantly deployed at the low density of CNG stops, whereas
the mixed fleet became dominant at the medium density. Thus, these
differences appear to be influenced by both the amount of alternative
fuel infrastructure available and by the type customer location pattern.
In general, it can be seen that at the low alternative fuels infrastructure
level the use of all diesel and mixed fleets dominate. However, as the
infrastructure is changed to a medium level, the use of mixed fleets and
all CNG alternatives increases greatly. However, this increase is more
pronounced for certain customer location patterns.
To examine the consistency of impacts of alternative fueling
infrastructure density, we conducted a post-hoc analysis by testing a
“high” density case where the average distance between CNG stops and
the average distance between diesel fuel stops are set to 40 miles and 10
miles, respectively. For most instances, CNG trucks were selected and
deployed by the model when the density of CNG fuel stops was high
(see figure 2). The results suggest that as the density of CNG fuel stops
increases, the likelihood that either CNG trucks or the mixed fleet is

Table 5/Comparison of the Fleet Deployment

Alternative Fueling Infrastructure Random Sector Urban/Rural Coast Cluster
Medium
Deploying only diesel trucks 9% 6% 21% 7% 1%
Deploying the mixed fleet 50% 5% 38% 42% 76%
Deploying only CNG trucks 41% 89% 41% 51% 23%
Low
Deploying only diesel trucks 87% 33% 74% 36% 97%
Deploying the mixed fleet 13% 55% 21% 63% 3%
Deploying only CNG trucks 0% 12% 5% 1% 0%

Note: The optimal fleet mix solution represents the best feasible solution obtained by IBM-ILOG CPLEX
with the time limit of 1,800 seconds and optimal tolerance of 10−6. Additionally, all test instances are
solved within the time limit.
Figure 2 Experimental Results
426 / TRANSPORTATION JOURNAL™

deployed increases. However, the impacts of the density of CNG fuel

stops on fleet compositions show different patterns across spatial dis-
persions of customers.
Following the initial analysis in table 5, a logistic regression is used
to statistically assess whether or not the likelihood of using a mixed
fleet significantly varies with patterns of customer locations and the
density of alternative fuel infrastructure. Using the logistic regression
approach, the probability of using a mixed fleet is defined as a function
of customer locations and density of network configuration of alterna-
tive fuel stops. As shown in the following equation, the two predictors
are categorical variables. The specification of this logistic regression
model is as follows,

⎛ p(the use of mixed fleet for instance i) ⎞

ln ⎜ = β0 + β 1 Densityi + β 2 CustLoci
⎝ 1 − p(the use of mixed fleet for instance i) ⎟⎠
where
Densityi = 1, if density of alternative fueling infrastructure = medium; 0
otherwise.
CustLoci = Patterns of customer locations: random (base category), sector,
urban/rural, coast, and cluster;

The logistic regression of the experimental results was used to test the
impacts of the density of CNG fuel stops and patterns on the mixed fleet
deployment. The first set of analyses (model 1) examines the impacts of
the density of CNG fuel stops on the selection of fleet deployment com-
positions. The impacts of the density of CNG fuel stops ( β = 0.486, odds
ratio = 1.625, p-value = .0002) is significant at p-value < .01, as seen in table 6.
The empirical results provide the evidence that an increase in the number
of CNG fuel stops contribute to the choice of the mixed fleet deployment.
The second set of analyses (model 2) examines the impacts of cus-
tomer locations on the choice of a mixed fleet deployment while control-
ling for the impacts of the density of CNG fuel stops. Analysis reveals
significant positive associations between cluster customer locations and
mixed fleet use (β = 0.356, odds ratio = 1.427, p-value = 0.093), and between
coast customer locations and mixed fleet use (β = 0.890, odds ratio = 2.435,
p-value < .0001) (see table 6). In table 5, the clustered location pattern saw
the highest level of mixed vehicle usage when infrastructure was at the
medium level. Additionally, when customers were clustered in certain
 In and Bell: Fuel Infrastructure and Customer Location Impacts \ 427

Table 6/Logistic Regression Model Estimation Results

Model 1 Model 2
Coeffi- Odds Odds
Variable cient SE Ratio P-value Coefficient SE Ratio P-value
Density 0.486 0.133 1.625 0.0002*** 0.502 0.135 1.652 0.0002***
of CNG
stations
Customer
locations
Cluster 0.356 0.212 1.427 0.093+
Coast 0.890 0.210 2.435 <.0001***
Sector −0.071 0.218 0.931 0.744
Urban −0.096 0.219 0.909 0.662
Constant −0.800 0.097 <.0001*** −1.166 0.174 <.0001***
Number of 1000 1000
observa-
tions
−2 Log 1300.030 1267.301
likelihood
Akaike 1304.030 1279.301
information
criterion
(AIC)
Schwarz 1313.846 1308.747
Bayesian
criterion

Note: †, *, **, and *** denoting the statistical significance at the 10%, 5%, 1%, and 0.1% levels for two-
tailed tests

areas, the odds of using the mixed fleet increased by 42.7 percent accord-
ing to the results in table 6. For the coastal pattern, the odds of using a
mixed fleet increased by 143.5 percent; however, according to table 5, the
coastal pattern may lead to even higher levels of all CNG usage when alter-
native fuels infrastructure moves to the medium level. Finally, the asso-
ciations between the sector customer location pattern and mixed fleet use
( β = −0.071 , odds ratio = 0.931, p-value = 0.744) and between urban/rural
customer locations and mixed fleet use ( β = −0.096 , odds ratio = 0.909,
p-value = 0.662) were not significant at p < .05. However, it can be seen in
table 5 that at the medium level of alternative fuels infrastructure an all-
CNG solution is found to be the best 89 percent of the time for the sector
pattern. Overall, the findings support the notion that alternative fueling
infrastructures and patterns in the spatial dispersion of customers both
influence the choice of mixed fleet usage.
428 / TRANSPORTATION JOURNAL™

Discussion
Using an analytical model for the extended VRVRP, this study investigates
how the optimal deployment of conventional and alternative fuel vehicles
changes across different densities of alternative fueling infrastructure
and types of spatial dispersion of customers. Solutions for the model are
generated with optimization techniques and important theoretical contri-
butions and managerial insights are obtained.
First, our analyses demonstrate the importance of the role of alterna-
tive fueling infrastructure levels and customer locations in determining
the right fleet composition of conventional and alternative fuel vehicles. As
the density of alternative fueling infrastructure increases, the mixed fleet
and alternative fuel trucks are more frequently deployed than diesel trucks
in the best fleet mix solutions. For the low density of alternative fueling
infrastructure, the mixed fleet deployment has been a dominant approach
in minimizing the total costs when customer locations are shaped more
linearly (i.e., coastal) or pie-shaped (i.e., sector). These findings are con-
sistent with Ahn and Rakha (2013), in which transportation network con-
figuration affected benefits of eco-routing systems. Therefore, it can be
inferred from our results that alternative fueling infrastructure and cus-
tomer locations define the relative benefits of deploying alternative fuel
vehicles, which helps to determine the right use of a mixed fleet. When the
density of alternative fueling infrastructure was high, most cases favored
alternative fuel trucks over the mixed and diesel fleet. These results sup-
port the notion that the limited driving range of alternative fuel vehicles
is a barrier to the adoption of alternative fuel vehicles (Erdogan and Miller-
Hooks 2012; Wang and Lin 2009). Most important, the method and results
of this research for the first time provide a method for measuring exactly
where these barriers exist in a routing network. In other words, compa-
nies can use the methods developed here with their own empirical data to
determine at what level of alternate fuel infrastructure to make changes
to the mix of vehicles in their fleets. Similarly, policymakers may be able
to use this approach to determine how much alternative fuel infrastructure
is needed to inspire companies to switch to more sustainable and environ-
mentally friendly alternative vehicles. In addition, our analyses may also
imply that deploying alternative fuel stops at the right locations can be
more important than just increasing their number because the spatial dis-
persion of customer demand locations themselves can affect the relative
benefits of using a mixed fleet.
 In and Bell: Fuel Infrastructure and Customer Location Impacts \ 429

This study also makes several theoretical research contributions.

First, it is the first known study to examine a heterogeneous fleet of con-
ventional and alternative fuel vehicles in the VRVRP. As such, this study
is able to demonstrate for the first time that the choice of a mixed fleet
depends on the alternative fueling infrastructure and spatial dispersions
of customers. Second, most research in the vehicle routing literature con-
siders customer locations as a given condition, focusing on the robust
performance of proposed algorithms across different spatial configura-
tions. However, this study is one of the few in the literature to consider
spatial dispersions of customers an important factor in determining the
fleet deployment. In doing so, it analyzes almost 1,000 distinct problem
instances representing five types of spatial dispersion. Considering the
small number of studies that have examined vehicle routing and vehicle
refueling problems, this study offers an additional contribution to the
vehicle routing literature.
Finally, as previously alluded to, the findings of this study provide
managerial insights to practitioners and policymakers. Truckload carri-
ers operating both conventional and alternative fuel trucks need to take
into account patterns of customer locations with alternative fueling infra-
structure when determining the routes in order to reduce costs and carbon
emissions. When it comes to alternative fuel vehicles, the alternative fuel
infrastructure has been considered an important factor in adopting alter-
native fuel vehicles (Erdogan and Miller-Hooks 2012; Wang and Lin 2009;
Yeh 2007). However, this research implies that policymakers should strate-
gically take into account where to locate alternative fuel stops rather than
just increase the number of alternative fuel gas stations (Kuby, Kelley, and
Schoenemann 2013) by understanding patterns of customer dispersions.
In addition, they should consider that both alternative fuel infrastructure
and customer locations can determine the relative benefits of using mixed
fleets with alternative fuel vehicles.

Limitations and Conclusions

The findings of this study raise some questions about the current limita-
tions of the proposed model and provide an opportunity for future research.
This study focused on exploring the impacts of fuel station infrastructure
and customer locations on the optimal fleet mix and solved the small-sized
problems by using an optimization technique. The experimental results
of this study can provide guidance as to the optimal fleet mix given that
430 / TRANSPORTATION JOURNAL™

multiple types of trucks are operating in different regions. However, this

study assumed that the maintenance costs of conventional vehicles are
similar to those of alternative fuel vehicles. For this idea to be applicable in
practice, researchers should further investigate various contextual factors
that reflect the complexity of the real situations (e.g., network complexity,
maintenance costs of alternative fuel vehicles, and fixed purchase costs of
alternative fuel vehicles). Therefore, future studies are needed to solve the
large-sized problems of 50 to 150 stops by using metaheuristic search meth-
ods with the data from companies, and to quantify the benefits by consid-
ering various contextual factors. The relative impacts of contextual factors
and their interrelationships can be assessed by employing statistical meth-
ods (e.g., ANOVA and ANCOVA). However, the present article serves as an
initial effort to open this line of inquiry and provides a baseline for com-
parison of future results. Another question stems from the assumption of
vehicle capacity. This study assumes uncapacitated vehicles. There is a need
for studies to investigate heterogeneous fleets with respect to varying vehi-
cle capacity and driving ranges in conjunction with other constraints that
exist in practice (e.g., alternative fuel types). Additionally, environmental
mechanisms such as carbon taxes and carbon markets are being imple-
mented worldwide (Chaabane, Ramudhin, and Paquet 2012). Those mecha-
nisms will impact the transportation industry that is considered a major
contributor to emissions of greenhouse gases (Spychalski 2011), which will
influence trucking companies in determining the use of conventional and
alternative fuel vehicles. Therefore, moving forward research should aim to
further investigate how such environmentally related mechanisms change
a fleet deployment and mix strategy for a heterogeneous fleet.

Appendix

Model Formulation

Minimize

∑ ∑ ∑ cpijk φijk + ∑ ∑ ∑ gpijk ψ ijk (1)

i∈N \{0} j∈N \{0,i} k∈Rij i∈N \{0} j∈N \{0,i} k∈ARij
Subject to:
Inflow & outflow balance equations

∑ j∈N \{0,i} xij + ∑ j∈N \{0,i} zij = 1, ∀ i ∈N \ {0} (2)

 In and Bell: Fuel Infrastructure and Customer Location Impacts \ 431

∑ i∈N \{k} xik = ∑ j∈N \{k} xkj , ∀ k ∈N (3)

∑ i∈N \{k} zik = ∑ j∈N \{k} z kj , ∀ k ∈N (4)

Number of vehicles deployed

∑ j∈N \{0} x0 j = ncv (5)

∑ j∈N \{0} z0 j = nafv (6)

Constraints on locations of fuel stops on the route

δ ijk ≤ xij , ∀i ∈N , ∀j ∈N , i ≠ j, k ∈Rij (7)

εijk ≤ z ij , ∀i ∈N , ∀j ∈N , i ≠ j, k ∈ARij (8)

Constraints on the refueling amount, fuel level, and fuel tank capacity for
standard fuels
φijk ≥ δ ijk ρc , ∀i ∈N , ∀j ∈N , i ≠ j, k ∈Rij (9)

φijk ≤ δ ijk Q c , ∀i ∈N , ∀j ∈N , i ≠ j, k ∈Rij (10)

rijk + φijk ≤ Q c , ∀i ∈N , ∀j ∈N , i ≠ j, k ∈Rij (11)

(rijk )
− μc xij ≥ 0, ∀i ∈N , ∀j ∈N , i ≠ j, k ∈Rij (12)

⎧ dk
⎪ r − ij , if k = 1
⎪ ij0 ccij
⎪
rijk = ⎨ , ∀i ∈N , ∀j ∈N , i ≠ j, k ∈Rij (13)
⎪
+ φ −
(
dijk − dijk− 1 )
, if k ≠ 1
⎪ ijk− 1 ijk− 1
r
⎪⎩ ccij

⎡ d ⎤
rij0 = ∑ xij ⎢ rhi 0 + ∑ φhik − hi ⎥ , ∀i ∈N \ {0}, ∀j ∈N , i ≠ j
cchi ⎥⎦
(14)
h ∈N \{i} ⎢ ⎣ k∈Rhi

r0 j0 = θ corigin , ∀j ∈N \ {0} (15)

⎡ d ⎤
ri 00 + xi 0 ⎢ ∑ φi 0 k − i 0 ⎥ ≥ θ cdest , ∀i ∈N \ {0} (16)
⎢⎣ k∈Ri 0 cci 0 ⎥⎦
432 / TRANSPORTATION JOURNAL™

Constraints on the refueling amount, fuel level, and fuel-tank capacity for
alternative fuels

ψ ijk ≥ ε ijk ρaf , ∀i ∈N , ∀j ∈N , i ≠ j, k ∈ARij (17)

ψ ijk ≤ ε ijk Q af , ∀i ∈N , ∀j ∈N , i ≠ j, k ∈ARij (18)

sijk +ψ ijk ≤ Q af , ∀i ∈N , ∀j ∈N , i ≠ j, k ∈ARij (19)

(s ijk )
− μaf z ij ≥ 0, ∀i ∈N , ∀j ∈N , i ≠ j, k ∈ARij (20)

⎧ dk
⎪ s − ij , if k = 1
⎪ ij 0
gcij
⎪
sijk = ⎨ , ∀i ∈N , ∀j ∈N , i ≠ j, k ∈ARij (21)
⎪ (
dijk − dijk− 1 )
⎪ sijk− 1 +ψ ijk− 1 − , if k ≠ 1
⎪⎩ gcij

⎡ d ⎤
sij0 = ∑ z ij ⎢ shi 0 + ∑ ψ hik − hi ⎥ , ∀i ∈N \ {0}, ∀j ∈N , i ≠ j
gchi ⎥⎦
(22)
h ∈N \{i,n + 1} ⎢ ⎣ k∈ARhi

s0 j0 = θ aforigin , ∀j ∈N \ {0} (23)

⎡ d ⎤
si 00 + z i 0 ⎢ ∑ k∈AR ψ i 0 k − i 0 ⎥ ≥ θ afdest , ∀i ∈N \ {0} (24)
⎣ i0 gci 0 ⎦

Time constraints

aic0 ≤ Tmax xi 0 , ∀i ∈N \ {0} (25)

⎧ ⎛ ⎞
⎪ dij
⎜ ∑aki + + ∑ δ ijk α ⎟ xij , if i ≠ 0
c
⎪ ⎜⎝ k≠i vc k∈Rij ⎟⎠
⎪
aijc = ⎨ , ∀i ∈N , ∀j ∈N (26)
⎪ ⎛ d0 j ⎞
⎪ ⎜ + ∑ δ 0 jk α ⎟ xij , if i = 0
⎪ ⎜⎝ vc k∈R0 j ⎟⎠
⎩
c
Cmax ( )
≤ max aic0 , ∀i ∈N \ {0} (27)

aiaf0 ≤ Tmax z i 0 , ∀i ∈N \ {0} (28)

 In and Bell: Fuel Infrastructure and Customer Location Impacts \ 433

⎧ ⎛ ⎞
⎪ dij
⎜ ∑aki + + ∑ ε ijk β ⎟ z ij , if i ≠ 0
af
⎪ ⎜⎝ k≠i vaf k∈ARij ⎟⎠
⎪
aijaf = ⎨ , ∀i ∈N , ∀j ∈N (29)
⎪ ⎛ d0 j ⎞
⎪ ⎜ + ∑ ε 0 jk β ⎟ z ij , if i = 0
⎪ ⎜⎝ vaf k∈ARij ⎟⎠
⎩
af
Cmax ( )
≤ max aiaf0 , ∀i ∈N \ {0} (30)

c
Cmax ≤ Tmax , Cmax
af
≤ Tmax (31)

xij ∈{0,1}, z ij ∈{0,1}, δ ijk ∈{0,1}, ε ijk ∈{0,1} (32)

φijk , ψ ijk ≥ 0

ncv, nafv : integer

References
Aas, B., I. Gribkovskaia, Ø. Halskau Sr., and A. Shlopak. 2007. “Routing of Supply Vessels
to Petroleum Installations.” International Journal of Physical Distribution and
Logistics Management 37 (2): 164–79.
Ahn, K., and H. A. Rakha. 2013. “Network-wide Impacts of Eco-routing Strategies:
A Large-scale Case Study.” Transportation Research Part D-Transport and
Environment 25:119–30.
Alternative Fuel Data Center. 2015. “Alternative Fueling Station Counts by State.” http://
www.afdc.energy.gov/fuels/stations_counts.html.
American Natural Gas Alliance. 2012. “Compressed Natural Gas Infrastructure.”
Angelelli, E., and M. G. Speranza. 2002. “The Periodic Vehicle Routing Problem with
Intermediate Facilities.” European Journal of Operational Research 137 (2): 233–47.
Anily, S., and A. Federgruen. 1990. “One Warehouse Multiple Retailer Systems with
Vehicle Routing Costs.” Management Science 36 (1): 92–114.
Ballou, R. H. 1989. “Heuristics: Rules of Thumb for Logistics Decision Making.” Journal of
Business Logistics 10 (1): 122–32.
———. 1990. “A Continued Comparison of Several Popular Algorithms for Vehicle
Routing and Scheduling.” Journal of Business Logistics 11 (1): 111–26.
Ballou, R. H., and Y. K. Agarwal. 1988. “A Performance Comparison of Several Popular
Algorithms for Vehicle Routing and Scheduling.” Journal of Business Logistics
9 (1): 51–65.
Barbarosoglu, G., and D. Ozgur. 1999. “A Tabu Search Algorithm for the Vehicle Routing
Problem.” Computers and Operations Research 26 (3): 255–70.
434 / TRANSPORTATION JOURNAL™

Bard, J. F., L. Huang, M. Dror, and P. Jaillet. 1998. “A Branch and Cut Algorithm for the
VRP with Satellite Facilities.” IIE Transactions 30 (9): 821–34.
Bektas, T., and G. Laporte. 2011. “The Pollution-Routing Problem.” Transportation Research
Part B: Methodological 45 (8): 1,232–50.
Bell, J. E., and S. E. Griffis. 2010. “Swarm Intelligence: Application of the Ant Colony
Optimization Algorithm to Logistics-oriented Vehicle Routing Problems.”
Journal of Business Logistics 31 (2): 157–75.
Bräysy, O., W. Dullaert, and P. Nakari. 2009. “The Potential of Optimization in Communal
Routing Problems: Case Studies from Finland.” Journal of Transport Geography
17 (6): 484–90.
Chaabane, A., A. Ramudhin, and M. Paquet. 2012. “Design of Sustainable Supply Chains
under the Emission Trading Scheme.” International Journal of Production
Economics 135 (1): 37–49.
Choi, E., and D. W. Tcha. 2007. “A Column Generation Approach to the Heterogeneous
Fleet Vehicle Routing Problem.” Computers and Operations Research 34 (7):
2,080–95.
Christofides, N., A. Mingozzi, and P. Toth. 1979. “The Vehicle Routing Problem.” In
Combinatorial Optimization, 315–38. Chichester: John Wiley.
Cordeau, J. F., M. Gendreau, G. Laporte, J. Y. Potvin, and F. Semet. 2002. “A Guide to
Vehicle Routing Heuristics.” Journal of the Operational Research Society 53 (5):
512–22.
Cordeau, J. F., G. Laporte, and A. Mercier. 2004. “Improved Tabu Search Algorithm for
the Handling of Route Duration Constraints in Vehicle Routing Problems with
Time Windows.” Journal of the Operational Research Society 55 (5): 542–46.
Davis, S. C., S. W. Diegel, and R. G. Boundy. 2013. Transportation Energy Data Book: Edition
32 (No. Ornl-6989). Washington, DC: US Department of Energy.
Dondo, R., and J. Cerdá. 2007. “A Cluster-based Optimization Approach for the Multi-
Depot Heterogeneous Fleet Vehicle Routing Problem with Time Windows.”
European Journal of Operational Research 176 (3): 1,478–507.
Duan, J. A., and C. F. Mela. 2009. “The Role of Spatial Demand on Outlet Location and
Pricing.” Journal of Marketing Research 46 (2): 260–78.
Eksioglu, B., A. V. Vural, and A. Reisman. 2009. “The Vehicle Routing Problem: A
Taxonomic Review.” Computers and Industrial Engineering 57 (4): 1,472–83.
Erdogan, S., and E. Miller-Hooks. 2012. “A Green Vehicle Routing Problem.” Transportation
Research Part E–Logistics and Transportation Review 48 (1): 100–114.
Fleischmann, B., S. Gnutzmann, and E. Sandvoss. 2004. “Dynamic Vehicle Routing Based
on Online Traffic Information.” Transportation Science 38 (4): 420–33.
Freightliner. N.d. “Cascadia 113 Natural Gas Truck.” http://www.freightlinertrucks.com/
Trucks/Alternative-Power-Trucks/Natural-Gas/Cascadia-113-NG.
Fu, L. 2002. “Scheduling Dial-a-Ride Paratransit under Time-Varying, Stochastic
Congestion.” Transportation Research Part B: Methodological 36 (6): 485–506.
Gendreau, M., G. Laporte, and R. Séguin. 1996. “A Tabu Search Heuristic for the Vehicle
Routing Problem with Stochastic Demands and Customers.” Operations
Research 44 (3): 469–77.
 In and Bell: Fuel Infrastructure and Customer Location Impacts \ 435

Goeke, D., and M. Schneider. 2015. “Routing a Mixed Fleet of Electric and Conventional
Vehicles.” European Journal of Operational Research 245 (1): 81–99.
Goldstein, S. M., P. T. Ward, G. K. Leong, and T. W. Butler. 2002. “The Effect of Location,
Strategy, and Operations Technology on Hospital Performance.” Journal of
Operations Management 20 (1): 63–75.
Griffis, S. E., J. E. Bell, and D. J. Closs. 2012. “Metaheuristics in Logistics and Supply Chain
Management.” Journal of Business Logistics 33 (2): 90–106.
Haller, M., E. Welch, J. Lin, and S. Fulla. 2007. “Economic Costs and Environmental
Impacts of Alternative Fuel Vehicle Fleets in Local Government: An Interim
Assessment of a Voluntary Ten-Year Fleet Conversion Plan.” Transportation
Research Part D–Transport and Environment 12 (3): 219–30.
Handley, S. M., and W. C. Benton. 2013. “The Influence of Task- and Location-Specific
Complexity on the Control and Coordination Costs in Global Outsourcing
Relationships.” Journal of Operations Management 31 (3): 109–28.
Hiermann, G., J. Puchinger, and R. F. Hartl. 2014. “The Electric Fleet Size and Mix Vehicle
Routing Problem with Time Windows and Recharging Stations.” Working paper,
1–33. http://prolog.univie.ac.at/research/publications/downloads/Hie_2015_638.pdf.
Jaw, J.-J., A. R. Odoni, H. N. Psaraftis, and N. H. Wilson. 1986. “A Heuristic Algorithm for
the Multi-Vehicle Advance Request Dial-a-Ride Problem with Time Windows.”
Transportation Research Part B: Methodological 20 (3): 243–57.
Kargul, J. J. 2012. “Efficient Use of Natural Gas Based Fuels in Heavy-Duty Engines.”
Paper read at Directions in Engine-Efficiency and Emissions Research (DEER)
Conference, October, at Dearborn, MI.
Kemp, J. 2012. “US Post Office Struggles with Alternative Fuel Delivery.” Reuters.
http://www.reuters.com/article/2012/05/28/us-column-kemp-us-postal-
service-idUSBRE84R0GM20120528.
Khuller, S., A. Malekian, and J. Mestre. 2007. “To Fill or Not to Fill: The Gas Station
Problem.” Paper read at Algorithms–ESA 2007, Berlin Heidelberg.
Koç, Ç., T. Bekta , O. Jabali, and G. Laporte. 2015. “A Hybrid Evolutionary Algorithm
for Heterogeneous Fleet Vehicle Routing Problems with Time Windows.”
Computers and Operations Research 64:11–27.
Krupnick, A. J. 2011. “Will Natural Gas Vehicles Be in Our Future?: Resources for the
Future.” http://www.rff.org/Publications/Resources/Pages/181-Natural-Gas-
Vehicles.aspx.
Kuby, M. J., S. B. Kelley, and J. Schoenemann. 2013. “Spatial Refueling Patterns of
Alternative-Fuel and Gasoline Vehicle Drivers in Los Angeles.” Transportation
Research Part D–Transport and Environment 25:84–92.
Laporte, G. 2007. “What You Should Know about the Vehicle Routing Problem.” Naval
Research Logistics 54 (8): 811–19.
Lee, J. M., and Y. H. Lee. 2012. “Facility Location and Scale Decision Problem with
Customer Preference.” Computers and Industrial Engineering 63 (1): 184–91.
Lee, T. R., and J. H. Ueng. 1999. “A Study of Vehicle Routing Problems with Load-
Balancing.” International Journal of Physical Distribution and Logistics Management
29 (10): 646–57.
436 / TRANSPORTATION JOURNAL™

Lieb, R. C., and K. J. Lieb. 2012. “The North American Third-Party Logistics Industry in
2011.” Transportation Journal 51 (3): 353–67.
———. 2015. “The North American Third-Party Logistics Industry in 2013: The Provider
CEO Perspective.” Transportation Journal 54 (1): 104–21.
Lin, C., K. L. Choy, G. T. Ho, S. Chung, and H. Lam. 2014. “Survey of Green Vehicle Routing
Problem: Past and Future Trends.” Expert Systems with Applications 41 (4):
1,118–38.
Lin, C. K. Y. 2008. “A Cooperative Strategy for a Vehicle Routing Problem with Pickup and
Delivery Time Windows.” Computers and Industrial Engineering 55 (4): 766–82.
Lin, S. H. 2008. “Finding Optimal Refueling Policies in Transportation Networks.”
Paper read at Algorithmic Aspects in Information and Management, Berlin
Heidelberg.
Ma, H., and C. Suo. 2006. “A Model for Designing Multiple Products Logistics Networks.”
International Journal of Physical Distribution and Logistics Management 36 (2):
127–35.
Malandraki, C., and M. S. Daskin. 1992. “Time Dependent Vehicle Routing Problems:
Formulations, Properties and Heuristic Algorithms.” Transportation Science 26
(3): 185–200.
Privé, J., J. Renaud, F. Boctor, and G. Laporte. 2006. “Solving a Vehicle-Routing Problem
Arising in Soft-Drink Distribution.” Journal of the Operational Research Society
57 (9): 1,045–52.
Rodrigue, J. P. 2012. “The Geography of Global Supply Chains: Evidence from Third-Party
Logistics.” Journal of Supply Chain Management 48 (3): 15–23.
Savelsbergh, M., and M. Sol. 1998. “Drive: Dynamic Routing of Independent Vehicles.”
Operations Research 46 (4): 474–90.
Schilling, D., K. E. Rosing, and C. S. Revelle. 2000. “Network Distance Characteristics
That Affect Computational Effort in P-Median Location Problems.” European
Journal of Operational Research 127 (3): 525–36.
Schneider, M., A. Stenger, and D. Goeke. 2014. “The Electric Vehicle-Routing Problem
with Time Windows and Recharging Stations.” Transportation Science 48 (4):
500–520.
Spychalski, J. C. 2011. “Transportation Policy: Precedent-Breaking Choices over Five
Decades.” Transportation Journal 50 (1): 10–22.
Suzuki, Y. 2011. “A New Truck-Routing Approach for Reducing Fuel Consumption and
Pollutants Emission.” Transportation Research Part D–Transport and Environment
16 (1): 73–77.
Suzuki, Y., and J. Dai. 2012. “Reducing the Fuel Cost of Motor Carriers by Using Optimal
Routing and Refueling Policies.” Transportation Journal 51 (2): 145–63.
———. 2013. “Decision Support System of Truck Routing and Refueling: A Dual-
objective Approach.” Decision Sciences 44 (5): 817–42.
Truckinginfo. 2015. “Ups to Deploy 1,400 Cng Vehicles, Fueling.” Truckinginfo. http://
www.truckinginfo.com/news/story/2015/04/ups-to-add-1-400-cng-vehicles-
fueling.aspx.
 In and Bell: Fuel Infrastructure and Customer Location Impacts \ 437

Trudeau, P., and M. Dror. 1992. “Stochastic Inventory Routing: Route Design with
Stockouts and Route Failures.” Transportation Science 26 (3): 171–84.
US Department of Energy. 2013. “Clean Cities Alternative Fuel Price Report.” Washington,
DC: US Department of Energy.
US Environmental Protection Agency. 2013. “Inventory of US Greenhouse Gas Emissions
and Sinks: 1990–2011.” Washington, DC: US Environmental Protection Agency.
US Postal Service. 2014. “U.S. Postal Service Fleet Alternative Fuel Vehicle Program
Report for Fiscal Year 2013.” February. http://about.usps.com/what-we-are-
doing/green/news/afv-program-2013-report.htm.
Wang, Y.-W., and C.-C. Lin. 2009. “Locating Road-Vehicle Refueling Stations.”
Transportation Research Part E: Logistics and Transportation Review 45 (5): 821–29.
Yeh, S. 2007. “An Empirical Analysis on the Adoption of Alternative Fuel Vehicles: The
Case of Natural Gas Vehicles.” Energy Policy 35 (11): 5,865–75.
Zografos, K. G., and K. N. Androutsopoulos. 2004. “A Heuristic Algorithm for Solving
Hazardous Materials Distribution Problems.” European Journal of Operational
Research 152 (2): 507–19.
Are We in This Together? The Dynamics and
Performance Implications of Dependence
Asymmetry and Joint Dependence in Logistics
Outsourcing Relationships

Adriana Rossiter Hofer

Abstract
This study investigates the mechanisms through which two dimensions of
interorganizational interdependence—dependence asymmetry and joint
dependence—impact a customer’s key attitudinal and behavioral dimen-
sions in the relationship with a third-party logistics provider (3PL), as well
as associated performance outcomes. Results of a survey with customers
of a large 3PL in Brazil indicate that customers embedded in relationships
with high levels of joint dependence exhibit higher levels of long-term ori-
entation toward and operational information exchange with the 3PL, which
positively impact their logistics outsourcing performance. In addition, it
is found that the direction of the customer’s dependence asymmetry—
advantage or disadvantage—in the relationship with the 3PL enhances or
hinders the effects of joint dependence. Specifically, while a customer’s
dependence disadvantage positively impacts its long-term orientation, a
customer’s dependence advantage negatively impacts its long-term orien-
tation, which can be detrimental to its logistics outsourcing performance.

Keywords
Third-party logistics, dependence, power, relationships, performance

Adriana Rossiter Hofer Transportation Journal, Vol. 54, No. 4, 2015

University of Arkansas Copyright © 2015 The Pennsylvania State
ahofer@walton.uark.edu University, University Park, PA
 Hofer: Are We in This Together? \ 439

Introduction
Embedded in the philosophy of supply chain management is the notion
that firms depend on other supply chain members to attain their goals of
achieving a high level of customer service while minimizing total costs
(Petersen et al. 2008). While firms within supply chains often collaborate
to maximize supply chain performance (Cai and Yang 2008), conflicts may
arise when individual supply chain members have inherently different
goals (Rinehart and Closs 1991; Taylor and Jackson 2000). One key factor
that impacts a firm’s behavior in an interorganizational relationship and,
as a result, the relationship dynamics of cooperation and conflict between
supply chain members is their relative interdependence (Ganesan 1994).
A key research stream in supply chain management has investigated the
effects of a firm’s dependence and power on firm behavior, relationship
characteristics, and performance (Cox 2001a; Hofer, Knemeyer, and Dresner
2009; Petersen et al. 2008).
Prior studies on the role of dependence and power in supply chain
relationships have often produced inconsistent results. While most stud-
ies find that dependence on a business partner leads to exertion of power
or coercive behavior by the most powerful party (Corsten and Felde 2005;
Gassenheimer and Ramsey 1994; Rajagopal 2009), other studies claim that
this behavior is counterproductive because the most dependent party
develops a relational orientation and is less likely to be opportunistic
(Belaya, Gagalyuk, and Hanf 2009). Similarly, while some studies advocate
that asymmetric relationships result in conflict (Ryu and Kim 2010), others
argue that no asymmetries are alike and the weaker party has a “zone of
tolerance” in which it still maintains the commitment to the most powerful
party (Davis and Mentzer 2006; Thomas and Esper 2010).
This study proposes that such inconsistencies occur because stud-
ies have considered solely a firm’s dependence on or power over its part-
ner (Hofer, Knemeyer, and Dresner 2009; Rajagopal 2009), or have focused
on the asymmetry only (Davis and Mentzer 2006). Emerson’s (1962)
foundational premises of power-dependence relations state that two
dimensions—dependence asymmetry and joint dependence—need to be
considered simultaneously for an accurate depiction of the role of power
and dependence in explaining a firm’s behavior in an interorganizational
relationship. Dependence asymmetry refers to the difference in the depen-
dence levels of the actors in a dyad, placing a given actor in a condition of
dependence advantage or disadvantage in the relation (Emerson 1962). Joint
dependence, on the other hand, is the sum of actors’ dependence on each
440 / TRANSPORTATION JOURNAL™

other (Gulati and Sytch 2007). Based on Emerson (1962), this study proposes
that the focal firm with the same level of dependence on different part-
ners might engage in different behaviors in those relationships, given the
potential different levels of these partners’ dependence on the focal firm.
Joint dependence, as well as the magnitude and direction of dependence
asymmetry assessed altogether, therefore, should be more appropriate pre-
dictors of firm behavior.
Specifically, this study combines Emerson’s (1962) theory of power-
dependence relations with premises of social exchange theory (Blau 1964;
Thibaut and Kelley 1959) and investigates the effects of a customer firm’s
perceived dependence asymmetry advantage or disadvantage and joint
dependence on its attitudes and behaviors in the relationship with a third-
party logistics provider (3PL), as well as associated logistics outsourcing
performance. This study addresses the following research questions: what
are the effects of a customer’s dependence advantage or disadvantage, and
joint dependence on its long-term orientation and information exchange in
the relationship with a 3PL? Which force prevails in shaping a customer’s
behavior: the dynamics of dependence asymmetry or the dynamics of joint
dependence? What conditions of a customer’s dependence asymmetry and
joint dependence enable higher performance outcomes?
The logistics outsourcing industry provides a relevant context to inves-
tigate the dynamics of interdependence from a customer perspective. On
the one hand, as 3PLs enhance the sophistication of their service offerings
(Armstrong and Associates 2013), customers increasingly depend on 3PLs
as they play a strategic role in their operations (Larson and Gammelgaard
2001; Zacharia, Sanders, and Nix 2011), and serve as a source of learning and
expertise (Alcacer and Oxley 2014). On the other hand, functions mostly
outsourced are still transactional, standardized, and repetitive (Langley
and Capgemini Consulting 2012), and the average contract length remains
short, reflecting customers’ desire to maintain flexibility in their 3PL
arrangements (Eyefortransport 2012).
The contribution of this article is threefold. This study contributes to
existing supply chain management literature that has acknowledged that
both facets of interdependence co-exist (Cox et al. 2004; Hingley 2001; Li
and Dant 2001; Mahapatra, Narasimhan, and Barbieri 2010) by providing
a theoretical foundation and empirically testing the relationship and per-
formance impacts of joint dependence and dependence asymmetry. It also
extends recent work in management (Gulati and Sytch 2007) by theorizing
that both dimensions of interdependence simultaneously impact relational
 Hofer: Are We in This Together? \ 441

dimensions and performance. This study, in particular, contributes to

the logistics outsourcing literature by responding to Maloni and Carter’s
(2006) and Marasco’s (2008) calls for theory testing in logistics outsourcing
research.
The structure of the article is as follows. First, an overview of the
dimensions of interdependence, along with a conceptual framework, is
provided. Next, the hypotheses are developed. The methodology section
follows with a thorough description of the steps taken in the measurement
of the constructs in the theoretical model, data collection, and analysis. The
remaining sections present the results, discussion of the findings, manage-
rial implications, as well as limitations and directions for future research.

Interdependence in Supply Chain Relationships

Prior studies in supply chain management have mostly focused on the
effects of a firm’s dependence or power toward a partner on its behavior.
Findings from research have been somewhat inconsistent (Kumar, Scheer,
and Steenkamp 1995). A significant body of research, labeled by some as
an “opportunistic perspective” (Gundlach and Cadotte 1994), has linked
dependence to control and coercive behavior by the most powerful party
(Corsten and Felde 2005; Knoppen and Christiaanse 2007), encouraged to
divert profits from the weaker party (Corsten and Felde 2005; Cox 2001a). In
this school of thought, dependence is viewed as a liability to be managed
(Cox 2001c; Cox, Sanderson, and Watson 2001) and as a source of conflict
(Kumar, Scheer, and Steenkamp 1995). Dependent parties are less satisfied
and committed to the powerful parties (Rajagopal 2009) and experience
lower performance and innovation (Corsten and Felde 2005).
Gundlach and Cadotte (1994) highlight, however, that there is a posi-
tive view on power—a “benevolent” perspective. Power is claimed to have a
positive role in promoting effective coordination and harmony in exchange
relationships (Anderson and Weitz 1989, 1992), improving efficiency and
positively impacting performance outcomes (Belaya, Gagalyuk, and Hanf
2009). In this view, the benefits of long-term dependence are thought to
temper or override opportunistic tendencies, encourage cooperation, and
enhance efficiency. Thus, the use of coercion by the most powerful party
is counterproductive. Dependence becomes, therefore, an enabler of
increasing depth of economic, social, and operational interaction between
exchange partners, jumpstarting a stronger relational orientation (Hofer,
Knemeyer, and Dresner 2009). In this vein, several studies have found that
dependence nurtures cooperation and adaptation in relational exchanges,
442 / TRANSPORTATION JOURNAL™

thus contributing to partner commitment (Dwyer, Schurr, and Oh 1987;

Knemeyer, Corsi, and Murphy 2003; Morgan and Hunt 1994).
Hingley (2001) points out that business-to-business relationships
work with these two apparently contradictory ideas: power-dependency
and quasi-cooperation. This study concurs with Hingley and proposes that
the apparent inconsistencies in the arguments and findings can be under-
stood by looking back at the seminal work of Emerson’s (1962) theory on
power-dependence relations, in which these two views coexist. As opposed
to focusing solely on the dependence of one party, Emerson discusses the
dynamics of “interdependence,” in which the reciprocal dependence is also
taken into consideration. The central tenet of Emerson’s work is that inter-
dependence is composed of two constructs: power (or dependence) asym-
metry and cohesion, the latter also named as mutual dependence or joint
dependence.
Emerson (1962) defines dependence asymmetry as the difference
between the dependence level of a focal firm on a business partner
and the reciprocal dependence, which can be either positive or negative.
If the net result is negative, the focal firm is the least dependent party
and this asymmetry is labeled “dependence advantage” (Emerson 1962;
Gulati and Sytch 2007). Conversely, if the net result is positive, the focal
firm is the most dependent party and this asymmetry is termed “depen-
dence disadvantage.” Joint dependence refers to the sum of both parties’
dependence levels.
Figure 1, adapted from Casciaro and Piskorski (2005) and applied to
the logistics outsourcing context, helps to explain the interplay of these
two dimensions of interdependence in greater detail. The vertical axis
represents a customer’s dependence levels on a 3PL, ranging from low (1)
to high (3). The horizontal axis represents a 3PL’s dependence levels on a
customer, ranging from low (1) to high (3). For any level of a customer-
and-3PL-provider dependence, a power-dependence relation can be charac-
terized by varying levels of dependence asymmetry and joint dependence.
Three important points with strong implications for empirical research on
dependence and power in supply chain management relationships can be
highlighted.
First, configuration 1 and configuration 9 represent dependence-
balanced relationships, but configuration 9 has a high level of joint depen-
dence. For these configurations there is no asymmetry, but the dependence
of a customer on a 3PL is not eliminated because of the identical level of
3PL dependence on the customer (i.e., dependence is a non-zero sum game).
 Hofer: Are We in This Together? \ 443

Figure 1 Configurations of Dependence Asymmetry and Joint Dependence

In configuration 9, a customer perceives that the 3PL is more criti-

cal to achieve its goals (and vice versa), as compared to configuration 1.
Second, joint dependence is not restricted to dependence-balanced dyads.
Configurations 2 and 6, for example, illustrate that dyads might have simi-
lar levels of dependence asymmetry and different levels of joint dependence.
Adversarial behaviors by the customer (who has a dependence advantage)
are less likely to be observed in configuration 6 because there is a stron-
ger level of joint dependence. Conversely, as Ryu and Kim (2010) point out,
given a constant level of joint dependence, a greater level of dependence
asymmetry results in a greater use of power by the party with a dependence
advantage than does dependency symmetry (Bacharach and Lawler 1981).
Third, a customer with the same level of dependence toward different 3PLs
might exhibit different behaviors in those relationships, which can poten-
tially lead to conflicting results in empirical studies that only measure the
customer’s dependence. Looking at configurations 5 and 6, for example, a
customer exhibits the same level of dependence toward two 3PLs. In config-
uration 6, this customer has a dependence advantage (i.e., is more powerful
than the 3PL), so it might impose pressure on this 3PL to pursue certain
strategies and, as a result, trigger conflict in the relation. This customer
is less likely to engage in such behavior with the 3PL in configuration 5,
characterized by a power-balanced relation. Therefore, measuring just a
customer’s dependence on or dependence asymmetry with a 3PL as predic-
tors of behavior leads to an incomplete depiction of the dynamics of such
relationship and can lead to seemingly inconsistent results.
In sum, the main contention of this article is that a customer’s atti-
tudes and behaviors in the relationship with a 3PL will be simultaneously
influenced, sometimes opposed, by forces triggered by the customer’s
444 / TRANSPORTATION JOURNAL™

dependence asymmetry (advantage or disadvantage) and joint dependence.

In addition, following the premises of social exchange theory (Thibaut
and Kelley 1959), this study proposes that such relationship characteristics
are the key antecedents of relationship outcomes, such as performance
(Lambe, Wittmann, and Spekman 2001).
Two relational behavior dimensions are investigated: a customer’s
long-term orientation toward the 3PL and customer’s operational informa-
tion exchange with a 3PL. In a logistics outsourcing context, a customer’s
long-term orientation can be defined as the perceived importance of and
dedication toward the continuity of the relationship with the 3PL. When
a customer exhibits a long-term orientation toward a relationship with
the 3PL, it focuses on the long-term benefits that it expects to receive from
the relationship (Ganesan 1994), and is willing to make sacrifices for the
3PL in the present because it believes it will be compensated in the future
(Lambe, Wittman, and Spekman 2001). In fact, long-term orientation has
been described as commitment (Anderson and Weitz 1992), and has been
proposed as a better proxy of closeness in relationships than relationship
duration (Chung, Sternquist, and Chen 2006; Ganesan 1995).
Operational information exchange can be defined as the timely,
accurate, concise, and usable day-to-day information transfers (Gardner,
Cooper, and Noordewier 1994). Information exchange facilitates respon-
siveness in a supply chain relationship, allowing partners to react to envi-
ronmental changes or new market developments (Kim, Cavulsgil, and
Calantone 2006). Hence, operational information exchange is a dimen-
sion of 3PL-customer partnering behavior (Hofer, Knemeyer, and Dresner
2009), as the breadth and depth of information exchanged between
firms strengthen supply chain partnerships (Lambert, Emmelhainz, and
Gardner 1996).
Figure 2 depicts the conceptual model. The following sections explore
the logics of dependence asymmetry and joint dependence individually and
discuss their effects on a customer’s relational behavior and logistics out-
sourcing performance.

Hypotheses Development

Dependence Asymmetry and Relational Exchange

The first dimension of interdependence relates to the unequal or asym-

metric dependence distribution in buyer-seller relationships (Cox 2001a;
Williams 1994), leading one of the parties to be in a position of relative power.
 Hofer: Are We in This Together? \ 445

Figure 2 Conceptual Model of Effects of Dependence Asymmetry and Joint

Dependence

If a focal firm perceives that the business partner is the most dependent
party, the resulting net-negative dependence corresponds to the focal firm’s
dependence advantage (Emerson 1962; Gulati and Sytch 2007). Conversely,
if a focal firm perceives to be the most dependent party, the net dependence
is positive, representing the focal firm’s dependence disadvantage—thanks
to the partner’s position of relative power in the relationship. As noted in the
previous section, this study argues that the isolated effects of dependence
asymmetry on a focal firm’s attitudes and behaviors will differ depending
on its dependence advantage or disadvantage position.
The party with a dependence advantage is motivated by self-interest
in pursuit of value appropriation (Gulati and Sytch 2007; Gundlach and
Cadotte 1994). Cox, Sanderson, and Watson (2001) define value appropria-
tion as the “amount of money that is retained by any organization from
participating at a particular stage in the supply chain” (p. 31). According
to Cox, Sanderson, and Watson (2001), value appropriation is “the suffi-
cient condition” of success in business. According to this value appropria-
tion perspective, a firm possessing a dependence advantage will have an
incentive to enforce the use of power to attain its objectives, even at the
expense of the least powerful party. In addition, it might increase its use of
adversarial tactics because of a sense of entitlement to do so (Gassenheimer
446 / TRANSPORTATION JOURNAL™

and Ramsey 1994; Gundlach and Cadotte 1994; Maloni and Benton 2000) or
because a lack of fear of retaliation, thereby, positioning itself to capture
greater value in the relationship (Gulati and Sytch 2007). For example, par-
ties with a dependence advantage might require others to make substantial
idiosyncratic investments (Jap and Ganesan 2000) or might exert power
during negotiations (Crook and Combs 2007; Rinehart and Closs 1991).
In addition, they constantly monitor their power positions and maintain
arm’s-length relationships with other companies in order to maintain flex-
ibility and exert leverage (Gulati and Sytch 2007; Mudambi and Helper
1998; Shapiro 1985). From the perspective of the party with a dependence
disadvantage, tolerance over exertion of power by the most powerful party,
including coerciveness, might be needed due to the need to maintain the
relationship as a means of access to relevant resources or due to the lack of
alternatives (Gundlach and Cadotte 1994).
The following subsections discuss the hypotheses development for
the effects of a customer’s dependence advantage and disadvantage and its
effect on a customer’s long-term orientation and information exchange.

Customer Dependence Advantage and Relational Exchange

From a social exchange theory (SET) perspective, a customer with a depen-

dence advantage toward the 3PL perceives that the value extracted from the
3PL (e.g., transportation and warehousing assets, logistics performance, or
expertise) is smaller than the benefits it provides the 3PL in terms of, for
example, revenue generation or relative size of the account vis-à-vis other
customers. Because of its more powerful position, a customer expects to
more easily influence the 3PL’s conduct (Emerson 1962), and that the 3PL
will make a strong effort, and even sacrifice financial outcomes (Gulati and
Sytch 2007) or other customers in order to make sure that this particular
customer has its needs met. As a result, a customer with a dependence
advantage has little incentive to expend resources in the relationship and
make sacrifices for long-term goals (Ryu, Park, and Min 2007), leading to
the following hypothesis:

Hypothesis 1: A customer’s perceived dependence advantage is

negatively related to its long-term orientation toward the relation-
ship with a 3PL.

One of the key resources in a logistics outsourcing relationship refers

to the nature and frequency of information exchange with the 3PL. For a
 Hofer: Are We in This Together? \ 447

successful information exchange with a 3PL, a customer needs to develop

and manage different capabilities related to software and hardware infra-
structure, personnel expertise, and organizational processes in order to
effectively select, structure, and manage the exchange of information with
a 3PL (Kim et al. 2011). When a customer has a dependence advantage, it
is less motivated to devote physical and intellectual resources related to
information exchange with a 3PL (Corsten and Felde 2005) because of the
expectation that the 3PL will do its best to provide the highest possible ser-
vice as to maintain the customer account. This should undermine the qual-
ity and quantity of information exchanged between a customer and its 3PL
(Ganesan et al. 2009). These arguments lead to the following hypothesis:

Hypothesis 2: A customer’s perceived dependence advantage is

negatively related to the exchange of operational information
with a 3PL.

Customer Dependence Disadvantage and Relational Exchange

When a customer has a dependence disadvantage in the relationship with a

3PL, the value extracted and outcomes accrued from working with the 3PL
are perceived to be higher than those provided by this customer to the 3PL.
In this scenario, potential poor logistics performance by the 3PL would have
strong negative implications for this customer’s ability to achieve desired
operational and strategic goals. Because of its more powerful position, the
3PL has less incentive to put in as much effort or to mobilize additional
resources to pursue achieving higher levels of operational performance or
to innovate in its processes and operations toward this particular customer
(Corsten and Felde 2005).
According to Emerson (1962), in this scenario a customer would
engage in balancing or restructuring operations to alter its disadvanta-
geous position. Some strategies employed by the customer would include
changing its dependence position without the 3PL, such as withdrawing
from the relationship or cultivating alternative sources of logistics services
(Emerson 1962). As the dependence disadvantage increases, however, there
are strong environmental limitations in the customer’s ability to work with
a different 3PL or to perform the logistics activities in house (Casciaro and
Piskorski 2005). This is because, as the dependence disadvantage increases,
outcomes accrued by the customer with the focal 3PL are perceived to be
much higher than those provided by this customer to the 3PL. Therefore,
448 / TRANSPORTATION JOURNAL™

a customer with a dependence disadvantage will engage in increasing the

“motivational investment” of the 3PL (i.e., giving status) by offering the 3PL
(i.e., the most powerful organization) other sources of value in the hopes
that the 3PL will reciprocate (Casciaro and Piskorski 2005). A key strategy to
do so refers to demonstrating to the 3PL that it is engaged in making a last-
ing relationship and willing make sacrifices for the relationship as a pledge
of commitment (Anderson and Weitz 1992). The above discussion leads to
the following hypothesis:

Hypothesis 3: A customer’s perceived dependence disadvantage is

positively related to its long-term orientation toward the relation-
ship with a 3PL.

In particular, one powerful mechanism to provide value to the 3PL is

to provide the 3PL with valuable and timely information (Knemeyer and
Murphy 2004). It facilitates the 3PL’s activities because of increased visibil-
ity (Heide and Miner 1992), and allows the 3PL to become more responsive
to operational challenges (Williams et al. 2013). Moreover, the 3PL is better
able to understand the nuances of a customer’s operations and, as a result,
deliver improvements and innovations (Ireland and Webb 2007).
As a result, a customer with a dependence disadvantage will try to
enhance the 3PL’s motivational investment on its account by enhancing the
information exchange with the 3PL, giving access to relevant, and even pro-
prietary, information (Emerson 1962):

Hypothesis 4: A customer’s perceived dependence disadvantage

is positively related to the exchange of operational information
with a 3PL.

Joint Dependence and Relational Exchange

In addition to dependence asymmetry, the second dimension necessary to
understand the dynamics of interdependence in interorganizational rela-
tions is joint dependence, or the sum of the actors’ dependencies on each
other (Emerson 1962). High levels of joint dependence lead actors to forego
immediate self-interest for the sake of the relationship, possibly miti-
gating the negative impacts of dependence asymmetries between actors
(Gulati and Sytch 2007). The mechanism underlying joint dependence
is tied to the concept of embeddedness, primarily found in the strategy
 Hofer: Are We in This Together? \ 449

literature (Granovetter 1985; Uzzi 1996), and to SET, commonly adopted in

marketing and supply chain management research (Blau 1964; Thibaut and
Kelley 1959).
Embeddedness refers to the process through which social interaction
impacts economic action (Uzzi 1996). As Granovetter (1985) states, “behav-
ior and institutions . . . are so constrained by ongoing social relations that
to construe them as independent is a grievous misunderstanding” (482).
The logic of embeddedness is in line with the principles of SET (Blau 1964;
Thibaut and Kelley 1959) in which actors perceive the relationship to be
rewarding and positive actions are reciprocated by the exchange partner,
developing social norms that help govern the relationship. In a situation of
high joint dependence, these social norms support a long-term orientation
toward the relationship (Lambe, Wittmann, and Spekman 2001).
Dwyer, Schurr, and Oh (1987) add that increased levels of joint depen-
dence between a buyer and a seller also foster the reliance on each other for
the performance of various channel functions, increasing the incentive to
make the relationship endure. At the same time, they can be expected not
to act opportunistically because opportunism by one firm is likely to pro-
voke retaliation by the other and jeopardize the firms’ mutual stake in the
relationship (Lusch and Brown 1996). Thus, even in the presence of differ-
ent expectations or conflict in the relationship, this long-term orientation
is critical to the contention that a transformation occurs in the attitudes
of parties toward the relationship. It is expected, therefore, that custom-
ers engaged in logistics outsourcing relationships marked by high levels
of joint dependence exhibit greater long-term orientation toward the 3PL.

Hypothesis 5. A customer’s perceived joint dependence is posi-

tively related to its long-term orientation in the relationship with
a 3PL.

In a situation of a high joint dependence between a 3PL and a customer,

a basis for fine-grained information exchange is set in place. A key behav-
ioral consequence of embeddedness is that the outcomes expected from the
relationship become separate from the narrow economic goals that origi-
nally constituted the interests of the firms in the relationship (Gulati and
Sytch 2007). As a consequence, there is an increase in operational, strate-
gic, and technological linkages (Ireland and Webb 2007). These linkages
facilitate communication, not only of knowledge to enhance the innova-
tiveness but also of knowledge of relational behaviors across the network.
450 / TRANSPORTATION JOURNAL™

As Huber and Daft (1987) note, closer ties result in more frequent and more
relevant information exchanges between partners. Uzzi (1997) found sup-
port for this contention in an ethnographic field analysis of apparel firms.
He found that the information exchanged in embedded relationships was
more proprietary and tacit than the price and quantity data that were traded
in arm’s-length ties. In sum, it is hypothesized that a customer immersed
in higher levels of joint dependence will promote an enhanced exchange of
information with the 3PL.

Hypothesis 6: Joint dependence between a customer and its 3PL is

positively related to the customer’s exchange of operational infor-
mation with the 3PL.

Customer Long-term Orientation and Operational Information Exchange

A customer with a long-term orientation toward the relationship with the
3PL perceives that maintaining the relationship is beneficial, and believes
that the relationship will last a long time (Gardner, Cooper, and Noordewier
1994). Such a customer is willing to work together to meet both companies’
needs and increase mutual profitability (Anderson and Weitz 1992).
As a result, this customer should have a stronger incentive to exchange
timely and useful information that both parties consider critical for opera-
tional success (Sheu, Yen, and Chae 2006). When a customer has a long-
term orientation, it is more willing to share even proprietary information,
because of the understanding that such information is needed for the 3PL
to learn about its operations and perform better. It is expected that, in this
scenario, the frequency and quality of information exchanged would be
enhanced (Huber and Daft 1987).
Previous research has found support for this contention. Huber and Daft
(1987), for example, found that closer relationships result in more frequent
and more relevant information exchanges between high performing partners.
Similarly, Sheu, Yen, and Chae (2006) found that a firm long-term orienta-
tion toward a retailer impacted several activities involved in the supply chain
architecture between a buyer and a retailer, such as the quality of information
shared, inventory systems development, and coordination structure.
From the above discussion, it is hypothesized that:

Hypothesis 7: A customer’s long-term orientation toward a rela-

tionship with a 3PL is positively associated to the exchange of
operational information with the 3PL.
 Hofer: Are We in This Together? \ 451

Relational Exchange and Logistics Outsourcing Performance

The relational exchange dimensions of customer long-term orientation and
operational information exchange with the 3PL should positively impact
the performance of a logistics outsourcing relationship. The following sub-
sections discuss the individual rationales in detail.

Customer Long-term Orientation and Logistics Outsourcing Performance

When a customer has a long-term orientation toward the 3PL, a cus-

tomer will cooperate with the 3PL (Chung, Sternquist, and Chen 2006;
Lusch and Brown 1996), and work to enhance the coordination structure
of the relationship (Min, Mentzer, and Ladd 2007; Sheu, Yen, and Chae
2006). Closer collaboration enables the 3PL to enhance its service per-
formance (Min, Mentzer, and Ladd 2007). Moreover, because a customer
with a long-term orientation is willing to make sacrifices for the con-
tinuation of the relationship, it gives the 3PL more confidence to invest
managerial expertise in finding ways to improve a customer’s perfor-
mance (Hofer, Smith, and Murphy 2014). These argument lead to the fol-
lowing hypothesis:

Hypothesis 8: A customer’s long-term orientation toward the rela-

tionship with a 3PL is positively related to its logistics outsourcing
performance.

Operational Information Exchange and Logistics Outsourcing Performance

When the formal and informal exchange of timely, accurate, concise,

and usable day-to-day information (Gardner, Cooper, and Noordewier
1994) takes place between a 3PL and a customer, a platform for fine-
grained communication is created. Such communication engenders
enhanced coordination and responsiveness in a customer’s operations
(Kim, Cavulsgil, and Calantone 2006; Morash 2001), as well as the cre-
ation of knowledge that enables innovation (Ireland and Webb 2007).
This is because the 3PL is more likely to better identify opportunities for
improvement in the customer operations, mitigate potential risks and
thus improve the performance of the outsourcing relationship. Indeed,
operational information exchange is a dimension of 3PL-customer part-
nering behavior (Hofer, Knemeyer, and Dresner 2009) as the breadth and
depth of information exchanged between firms strengthen supply chain
452 / TRANSPORTATION JOURNAL™

partnerships (Lambert, Emmelhainz, and Gardner 1996). Accordingly, it

is hypothesized that:

Hypothesis 9: A customer’s operational information exchange with

a 3PL is positively related to its logistics outsourcing performance.

Research Methodology

Sample Selection

The data used to test the model were collected through an Internet-based
survey of customers of a large Brazilian 3PL. By surveying customers of
a single 3PL, the model controls for the 3PL’s strategic orientation, high-
lighted in prior studies as an important antecedent of firm behavior (Min,
Mentzer, and Ladd 2007), above and beyond interorganizational conditions.
This 3PL is an asset-based company with broad geographical coverage in
Brazil and serves customer firms from a broad variety of industries, such as
apparel, automotive, and electronics. The 3PL provides an array of logistics
services, ranging from simple transportation and warehouse management
to customized distribution solutions. Moreover, it has been consistently
selected by customers as a finalist or winner of an annual best 3PL competi-
tion conducted in Brazil by the Logistics and Supply Chain Institute (ILOS).
Based on its solid reputation, logistics capabilities, and diverse customer
base, this focal 3PL was determined to be an appropriate source of informa-
tion to empirically test the hypotheses of this study.
Consistent with prior research on interdependence (Gulati and Sytch
2007; Kumar, Scheer, and Steemkamp 1995), only the customers’ percep-
tions were surveyed. Because the objective was to identify any interdepen-
dence effects on a customer’s behavior toward the 3PL, both the perception
of customer dependence on a 3PL and the 3PL dependence on a customer
were collected from a customer’s perspective. A customer’s perception of
both its dependence on a 3PL and the reciprocal dependence should impact
its behavior in the relationship.

Measures

The construct scale items were adapted from the existing literature. All
scales were measured in a seven-point Likert scale (1 = strongly disagree
to 7 = strongly agree). A detailed list of all scale items is presented in the
appendix.
 Hofer: Are We in This Together? \ 453

Customer dependence advantage, disadvantage, and joint dependence. Two

steps were followed for the calculation of these measures. Initially, the con-
structs “Customer dependence on 3PL” and “3PL dependence on customer”
were measured with items adapted from Ganesan (1994). Sample items
included: “We do not have a good alternative to [3PL name]” and “We are not
a major customer for [3PL name].” Because these two constructs have differ-
ent measurement items, it was necessary to average the scores for customer
dependence and 3PL dependence measurement items in order to calculate
the dependence advantage, disadvantage, and joint dependence variables.
Next, in line with Emerson’s (1962) original proposition and more recent
management studies (Casciaro and Piskorski 2005; Gulati and Sytch 2007),
the measure of joint dependence was calculated as the sum of the scores
of 3PL dependence on the customer and customer dependence on the 3PL.
For the customer dependence advantage and customer dependence disad-
vantage variables, a spline specification was used (Gulati and Sytch 2007;
Johnston and Bonoma 1984). The spline specification enables the measure-
ment of the extent and direction of the dependence asymmetry of the cus-
tomer on the 3PL. The spline measure of a customer dependence advantage
was created by subtracting customer dependence from 3PL dependence. All
negative values were recoded to zero. The spline measure of customer dis-
advantage was calculated by subtracting 3PL dependence from customer
dependence, recoding all negative values to zero. For every observation in
the sample, there were three interdependence values: (1) one value of joint
dependence (the sum of customer and 3PL dependence); (2) one value of
customer dependence advantage (3PL dependence minus customer depen-
dence); and (3) one value of customer dependence disadvantage (customer
dependence minus 3PL dependence), noting that one asymmetry value was
positive and the other was recoded to zero.
Logistics outsourcing performance. The items for this construct capture
the perceptions of the customer’s operational performance improvements
accrued from the relationship with the 3PL. The scale was adapted from
Knemeyer and Murphy (2004). Sample items include “This relationship has
improved our logistics system responsiveness” and “This relationship has
enabled us to implement changes faster/better.”
Customer’s long-term orientation. The measures for the long-term ori-
entation were adapted from Gardner, Cooper, and Noordewier (1994) and
captured the extent to which the customer expects the relationship with
the 3PL to last a long time. Sample items include “Maintaining a long-term
relationship with [3PL name] is important to us.”
454 / TRANSPORTATION JOURNAL™

Customer’s Operational Information Exchange. The measures for this

construct were also adapted from Gardner, Cooper, and Noordewier (1994),
which capture the extent to which operational information is exchanged
with the 3PL. Sample items include “We use software dedicated to our rela-
tionship with [3PL name].”
Survey Development and Implementation

Once the measures for the constructs were defined, the survey instrument
was pretested through think-aloud interviews with several scholars who are
highly knowledgeable about the logistics outsourcing industry. The survey
instrument was then revised and translated into Portuguese by two differ-
ent researchers fluent in the language. Finally, two other bilingual industry
professionals reviewed the survey and translated it back into English. Based
on this exercise, minor adjustments were made to the final questionnaire.
A pretest was conducted with a subsample of the 3PL’s customer data-
base. Four hundred customers were randomly selected from the database
and were sent e-mails with invitations to visit a website and provide their
names and e-mail addresses if they were willing to participate in the survey.
Only 282 firms received the invitation, with the remaining e-mails returned
due to incorrect or outdated email addresses. Forty-three customers agreed
to participate in the project, provided their contact information, and
received the link to the survey website. Sixteen customers completed the
questionnaire, corresponding to a response rate of 5.67 percent. No prob-
lems were encountered during this pretest. Thus, modifications were not
made to the survey instrument.
The survey implementation consisted of three basic steps. Due to con-
fidentiality concerns, the research team did not have access to the 3PL’s cus-
tomer database. The 3PL e-mailed key representatives of each customer in
its database (2,649 firms), informing them of the study and inviting them to
access the study website and provide their e-mail addresses in order to par-
ticipate. Three hundred and thirty-five customer firms accepted the 3PL’s
invitation, provided their contact information, and received the link to the
website. A second e-mail with a link to the survey website was sent to the
335 firms. In total, 265 firms completed the survey, representing a response
rate of 79.1 percent of those companies that had accepted the 3PL’s invita-
tion and 10 percent of the customers that received the initial invitation.
Two standard methods were used to test for nonresponse bias in this
study. First, 13 key nondemographic questions were selected from the survey
and used to compare early and late respondents. The second method followed
Lambert and Harrington’s (1990) suggestions by selecting a random sample of
 Hofer: Are We in This Together? \ 455

the nonrespondents to answer the same set of questions used when examining
the early and late respondents. An e-mail with the link to a shortened version
of the survey was sent to the customers who did not respond to the initial invi-
tation to participate in the survey. Seventy-five responses were collected. In
both tests, a MANOVA using a two-group Hotelling t-squared test showed no
statistical significance, and the absence of response bias was concluded.

Final Sample and Respondents Characteristics

As outlined earlier, 16 customers completed the survey during the pre-
test phase, and 265 customers completed the survey during the survey
implementation phase. Given that no modifications were made in the sur-
vey instrument between the phases, and that the pretest and survey were
implemented during two consecutive months, both sets of responses were
combined for a total of 281 observations.
The respondents served in a variety of managerial positions: logistics
managers (18.9%), logistics supervisors (12.5%), general managers (10.3%),
CEOs (6.4%), and partners (6.8%). Considering these individuals were the 3PL’s
customer contacts for coordinating activities, and that the respondents were
generally professionals at a management level, it appears that the respondents
had sufficient knowledge to describe their company’s relationship with the
3PL. Thus, key informant bias is not expected to be a concern in this study.
The firms represented a variety of industries, such as apparel (18.5%),
health care (6.4%), automotive and auto parts (5.7%), electronics (5.7%), cos-
metics (5.3%), telecommunications (4.3%), and food and beverage (5.0%).
Almost 75 percent of the sample was composed of small firms with fewer than
250 employees. Larger firms, with more than 1,000 employees, represented
less than 10 percent of the sample. The small size of the firms in the sample
is also reflected in their annual sales distribution. Of the respondent firms,
18.5 percent had annual sales of less than US$ 0.5 million, 31.3 percent had
annual sales that ranged from US$ 0.5 to US$ 4.3 million, and 10.3 percent
had annual sales ranging between US$ 4.3 and 11.4 million. The remaining
respondents had annual sales greater than US$ 11.5 million, of which only 6
percent had annual sales greater than US$ 120 million.
Regarding the respondents’ logistics outsourcing practices, it is evi-
dent that the vast majority outsourced only one or two logistics functions
(52.7% and 13.5% of the sample, respectively), while about 5 percent of the
firms outsourced six or more functions. Transportation operations, consid-
ered to be the strongest capability of the 3PL, formed the most contracted
category. Other outsourced functions included transportation planning,
freight consolidation, and order fulfillment (see table 1).
456 / TRANSPORTATION JOURNAL™

Table 1/Respondents’ Outsourced Logistics Functions

Activity Number of Activity Number of
Firms Firms
Transportation operations 216 Traffic control (distribution) 14
Freight consolidation 48 Transportation planning 13
Final consumer 43 Network/route 8
distribution optimization
Freight bill payment 34 Inventory management/ 7
control
Warehousing 26 Order management 7
Reverse logistics 22 Cross-docking 6
IT systems 14 Traffic control (supply) 5
EDI capability 14

Data Analysis
Following guidelines proposed by Garver and Mentzer (1999), the first
step in the analysis was to test each construct for unidimensionality, con-
vergent validity, and reliability to ensure that the measurement scales
accurately represented the latent constructs. To obtain unidimensional-
ity for reflective multi-item constructs, item-to-total correlations were
calculated for each item, taking one scale at a time. All items exhibited
item-to-total correlations greater than 0.35, the cutoff suggested by Saxe
and Weitz (1982). Next, for each construct, factor analysis with Varimax
rotation was used as a test for convergent validity. All items loaded on the
respective factors, and the variance extracted for all constructs was above
the 0.5 cut-off (table 2), indicating that a significant amount of variance
is captured by the construct. Finally, scale reliability was tested using the
coefficient of maximal reliability, known as coefficient H, developed by
Hancock and Mueller (2001). All values were found to be at or above the 0.7
threshold (table 2).
To control for common method bias, a concern in survey research
where the same respondent provides information for both dependent
and independent variables, a Harman’s single factor test was conducted
(Podsakoff et al. 2003). The test resulted in a clear multifactor solution,
with the most influential factor explaining less than 20 percent of the
data variance. Thus, common method bias is not a significant concern in
this study.
Because the measures of customer dependence and 3PL dependence
were aggregated into manifest measures of customer dependence advan-
tage, disadvantage, and joint dependence, path analysis was the statisti-
cal technique selected for the analysis. The measures for the remaining
 Hofer: Are We in This Together? \ 457

Table 2/Convergent Validity and Reliability Results

Construct Convergent Validity Construct Reliability
(Variance Extracted) (Maximal Reliability H)
Customer’s long-term orientation 0.551 0.863
Customer’s operational information 0.520 0.843
exchange
Logistics outsourcing performance 0.616 0.940
Customer dependence 0.939 0.990
3PL dependence 0.944 0.983

Table 3/Descriptive Statistics

Variables Mean SD 1 2 3 4 5 6
1. Joint 8.144 2.410 -
dependence
2. Customer 1.991 1.577 −0.069 -
dependence
advantage
3. 3PL dependence 0.135 0.433 0.028 −0.416* -
advantage
4. Customer’s 5.635 1.114 0.445* −0.100 0.137* 0.573* -
long-term
orientation
5. Customer’s 3.558 1.731 0.465* 0.002 0.077 0.357* 0.408* -
operational
information
exchange
6. Logistics out- 4.769 1.426 0.510* −0.153* 0.167* 0.554* 0.526* 0.524*
sourcing
performance

Note: The values with an * are statistically significant at the 0.05 level.

constructs were also calculated with the averages of the scales of each con-
struct employed and the analysis was conducted with EQS software. The
means, standard deviations, and correlation table are presented in table 3.
The final step was to assess the fit of the path model. The robust fit indi-
ces obtained for the structural model were: χ2 = 19.289 (df = 3), CFI = 0.958,
SRMR = 0.038, NNFI = 0.952, and RMSEA = 0.139. The χ2 value is statistically
significant and, with the exception of RMSEA, the values for the CFI, NNFI
and SRMR indices fall within acceptable ranges (Kline 2005). While the
RMSEA fit index is not significant, the remaining fit indexes meet the joint
criteria to retain a model proposed by Hu and Bentler (1999). According to
Hu and Bentler, a joint criteria for target values to retain a model include two
alternatives: (1) NNFI, CFI and SRMR, or (2) SRMR and RMSEA. Because our
model obtained an acceptable NNFI, SRMR, and CFI, it can be argued that
458 / TRANSPORTATION JOURNAL™

the model has an acceptable fit. Moreover, the model explains 39.2 percent
of the variance in the dependent variable. With the model presenting an
acceptable fit, the structural paths were estimated. Figure 3 provides an
overview of the standardized estimation results of the structural model.
In addition, the indirect and total effects on the independent and medi-
ator variables on the logistics outsourcing performance were estimated
with EQS software (table 4). The results indicate that joint dependence and
customer dependence disadvantage have a significant positive indirect
effect on logistics outsourcing performance, providing evidence that a cus-
tomer’s long-term orientation and information exchange partially mediate
the relationship between these two interdependence dimensions and per-
formance. A customer long-term orientation fully mediates the relation-
ship between customer dependence advantage and performance.
To ensure the robustness of the model results, the model was tested
controlling for customer firm size (in terms of sales) and number of func-
tions outsourced since it has been suggested that a firm’s logistics out-
sourcing strategy influences its relationship with a 3PL (Bolumole 2003).
Both controls were, however, found to be insignificant, and their inclusion
did not impact the magnitude or the direction of the other path coefficients.
This highlights the relevance of interorganizational conditions, rather
than organizational characteristics, as drivers of a customer’s attitude and
behavior in the relationship with a 3PL.

Figure 3 Structural Model Results

 Hofer: Are We in This Together? \ 459

Table 4/Standardized Indirect and Total Effects on Performance

Predictors Indirect Effects Total Effects
Customer long-term orientation 0.090* 0.076*
Customer operational information exchange - 0.049*
Customer dependence advantage −0.057 −0.053
Customer dependence disadvantage 0.073* 0.089*
Joint dependence 0.360* 0.033*

Note: The values with an * are statistically significant at the 0.05 level.

Results
The model results generally support the premise that dimensions of
interdependence influence a customer’s attitude and behavior in the
relationship with the 3PL, and indirectly impact logistics outsourc-
ing performance. The first set of hypotheses examines the effects of a
customer’s dependence advantage and disadvantage on its relational
behavior in the relationship with a 3PL. In line with hypothesis 1, customer
dependence advantage is found to have a statistically significant nega-
tive effect on the customer’s long-term orientation. This supports the
contention that a customer with a dependence advantage is less inclined
to nurture the relationship with the 3PL. Consistent with hypothesis
3, there is a positive and statistically significant effect of a customer’s
perceived dependence disadvantage on a customer’s long-term orienta-
tion toward the relationship with a 3PL. This result indicates that, after
controlling for levels of joint dependence, customers who perceive the
3PL to be more powerful are likely to make sacrifices for the continuation
of the relationship because they believe they need the 3PL to achieve their
goals. The relationships between customer dependence advantage and
customer dependence disadvantage on information exchange were found
to be nonsignificant (H2 and H4 not supported).
The statistical results for the effect of joint dependence on dimensions
of relational exchange are found to be consistent with the predictions of
SET and the embeddedness perspective. That is, the results indicate that
joint dependence has a positive and significant effect on the customer’s
long-term orientation and operational information exchange. Thus, both
hypothesis 5 and hypothesis 6 are supported.
There is evidence of a positive and significant relationship between a
customer’s long-term orientation and a customer’s operational informa-
tion exchange, providing support to hypothesis 7. The customer opera-
tional information exchange is also directly impacted by a customer’s joint
460 / TRANSPORTATION JOURNAL™

dependence, and indirectly by a customer’s dependence advantage and

disadvantage through a customer’s long-term orientation. It is relevant to
note that model results indicate that the positive influence of joint depen-
dence on dimensions of relational behavior is much stronger than the nega-
tive influence of customer dependence advantage.
The final set of hypotheses highlights the predictive power of social
exchange theory. Both a customer’s long-term orientation and informa-
tion exchange were identified as strong predictors of a customer’s logis-
tics outsourcing performance (hypotheses 8 and 9, respectively). Moreover,
these relational behaviors partially mediate the relationship between the
dimensions of interdependence on performance, which demonstrate the
importance of relational behavior in shaping the performance of logis-
tics outsourcing relationships (Leuschner et al. 2014). An interesting take
away from the results is that customers with a dependence advantage are,
contrary to their expectations, potentially not extracting the highest per-
formance levels possible. The results indicate a customer’s dependence
advantage has a negative effect on its long-term orientation which, in turn,
reduces the positive effect of long-term orientation on its logistics out-
sourcing performance.

Discussion
While studies generally advocate that a firm’s power over a business part-
ner is associated with higher conflict (Kumar, Scheer, and Steemkamp 1995)
and the use of coercion (Gassenheimer and Ramsey 1994), others claim that
power leads to more stable relationships because the dependent party is
less likely to be opportunistic (Anderson and Weitz 1992). Indeed, several
studies hypothesize and find support for the relationship between a firm’s
dependence on a partner and its relational behavior (Hofer, Knemeyer, and
Dresner 2009). How can we explain the potentially divergent behaviors
by firms with a dependence advantage? How can a more dependent party
develop a relational behavior in a relationship embedded in conflict?
This study argues that dependence in an interorganizational relation-
ship is reciprocal because each actor in a dyad depends, at some level, on
each other to achieve desired goals. In addition, dependence (and associ-
ated power) is not a zero sum game and is not neutralized in the relation-
ship (Emerson 1962). Thus, the interdependence dimensions for a given
actor in dyad, including the joint dependence and dependence asymmetry
advantage or disadvantage, should be assessed for a more complete inves-
tigation of potential behaviors of a focal firm in an interorganizational
 Hofer: Are We in This Together? \ 461

relationship. In the context of logistics outsourcing relationships, this

study found that each dimension of interdependence impacts a customer’s
attitude and behavior through different mechanisms. Joint dependence
is driven by the desire to strengthen the relationship. Joint dependence
increases the perception that the relationship is beneficial and that it will
last a long time, positively influencing a customer’s long-term orientation
and the exchange of information with a 3PL. This effect can be enhanced or
reduced depending on the direction of a customer’s dependence asymme-
try. A customer’s dependence disadvantage triggers the customer’s pursuit
to enhance a 3PL’s motivational investment in the relationship, positively
impacting a customer’s relational behavior. A customer dependence advan-
tage, however, is driven by the pursuit of self-interest, even at the expense
of the 3PL, leading to lower levels of relational behaviors. The combined
influence of the interdependence dimensions on a customer’s long-term
orientation and information exchange with the 3PL will ultimately impact
its logistics outsourcing performance.
This study provides a number of considerations to the supply chain
literature and to the logistics outsourcing research stream, in particular.
First, the latter has typically focused on the dependence of one actor only
as a predictor of its relational behavior. The findings provide evidence that,
for the same level of dependence, a customer might behave differently
depending on the perceived degree of the 3PL’s dependence. In line with
prior research that tested both dimensions of interdependence (Gulati and
Sytch 2007; Kumar, Scheer, and Steenkamp 1995), the findings of this study
indicate that joint dependence is expected to serve as a more appropriate
predictor of relational behavior (Xia 2011).
Second, studies that focus on power or dependence asymmetry
should consider not only the magnitude of such asymmetry, but its direc-
tion as well, given that dependence advantage or disadvantage condi-
tions impact behavior in opposite ways. Kumar, Scheer, and Steenkamp
(1995), for example, measured the absolute value of asymmetry (disre-
garding whether the focal firm was the most or least powerful party), and
found a negative relationship between asymmetry and trust and com-
mitment. One relevant takeaway of the present study, however, is that
while customer with a dependence advantage is less inclined or moti-
vated to develop a long-term orientation toward the 3PL (consistent with
Kumar, Scheer, and Steenkamp’s [1995] study), a customer’s dependence
disadvantage positively contributes to its relational behavior, positively
impacting performance.
462 / TRANSPORTATION JOURNAL™

This leads to the second point regarding dependence asymmetry and is

related to the importance of controlling for joint dependence. The results
of this study indicate that, for the same level of asymmetry advantage, cus-
tomers might ultimately engage in different behaviors according to their
level of joint dependence. According to the statistical findings, a customer
with a dependence advantage might see the negative impact of asymme-
try mitigated if the joint dependence is sufficiently high. Therefore, a more
powerful party might not engage in adversarial behavior, which is the tra-
ditional argument in the literature.
The final point is related to performance effects of interdependence.
Following a value appropriation logic, the traditional argument is that a
firm with a dependence advantage should extract the higher value in the
relationship at the expense of the weaker party (Cox 2001a). This rationale
was also adopted by Gulati and Sytch (2007), who hypothesized a positive
direct effect between a manufacturer dependence advantage and its perfor-
mance in a procurement relationship. Contrary to Gulati and Sytch’s expec-
tations, they found a negative effect. This concurs with the findings of this
study in which, after controlling for joint dependence, a most powerful
customer is not necessarily extracting the highest logistics outsourcing
performance. This highlights an important difference between purchasing
a product versus purchasing a service. A powerful customer might be pay-
ing a lower price for the service, but they are not accruing higher perfor-
mance levels. In logistics outsourcing, both parties need to be involved and
give feedback to achieve higher levels of performance.

Managerial Implications
“On several occasions I refused to accept big accounts from potential cus-
tomers because they would take over 10 percent of my business,” a CEO of a
large US 3PL affirmed during interviews conducted for this study. His strat-
egy was to avoid working with a customer who would have the potential
to be the most powerful player in the relationship. The results from this
study indicate that this 3PL might be losing opportunities if following such
a strategy. Relationships with higher level of joint dependence experienced
the highest levels of logistics outsourcing performance. A customer’s
dependence advantage had a negative impact on a customer’s long-term
orientation and exchange of information, but a much weaker effect than
joint dependence. Third-party logistics have the greatest potential for
higher performance levels if working toward increasing joint dependence
and reducing a customer’s dependence advantage. According to Emerson
 Hofer: Are We in This Together? \ 463

(1962), the latter can be achieved by enhancing its customer’s motivational

investment on the relationship (i.e., dependence), which can be accrued by
investing in and demonstrating dedication to this particular relationship
(Anderson and Weitz 1992).
Similarly, customers who perceive themselves to be in control while
working with 3PLs are not necessarily achieving the operational and stra-
tegic benefits they would expect to accrue from the outsourcing effort. In
some instances, powerful customers take for granted that 3PLs, because
of their dependence on maintaining the account, will take extraordinary
measures to perform to the best of their abilities. This mindset might lead
the companies’ representatives to impose their power by means of request-
ing increased capacity or emergency shipments on very short notice or
requesting the 3PL to adopt certain technology or to make a transaction-
specific investment. Exertion of power is actually more likely to nega-
tively impact the performance of the logistics outsourcing arrangement
and not substantially improve customer service levels or reduce logistics
costs. Being a powerful player in an interorganizational relationship does
not necessarily mean that the customer will accrue the highest possible
performance.
It is also relevant to note that relationships are dynamic: expectations
and perceptions of dependence change over time. Very commonly, 3PLs—
being aware that current customers are satisfied—invest their intellectual
energy toward efficiency and innovation with new customers. Current cus-
tomers, as a result, might change their perception of dependence on the
3PL, which could alter their expectations of continuity (long-term orienta-
tion) and exchange of information with the 3PL.
It is important, therefore, to develop a joint dependence between
3PLs and customers. This joint dependence may be a critical factor in
improving the services provided because it means that each partner
will contribute to the relationship by focusing on future conditions and
exchanging relevant information, which will allow the outcomes of the
outsourcing relationship to be improved. Even for small customers,
developing a relationship with the 3PL representatives and showing a
willingness to develop a stronger working relationship with the 3PL can
increase the likelihood of experiencing stronger performance improve-
ments in operations. More dependent 3PLs can strive to learn about
customers’ business and proactively pursue operational improvements.
This should increase the customer’s perceived dependence on the 3PL,
reducing asymmetries.
464 / TRANSPORTATION JOURNAL™

Limitations and Future Research

Because the focus of the study was on identifying the effects of interdepen-
dence on a customer’s attitude and behavior in a relationship with a 3PL, the
survey was conducted with 3PLs’ customers only. Studies have shown that
perceptions often differ between parties (Zhou, Zhuang, and Yip 2007), so
it is expected that a customer’s own perception of its dependence on a 3PL
and its perception of the 3PL’s dependence will ultimately impact its behav-
ior. While such design is in line with other studies on interdependence and
logistics outsourcing (Hofer, Knemeyer, and Murphy 2012; Kumar, Scheer,
and Steemkamp 1995), a dyadic survey would capture a more holistic pic-
ture of such differences in perceptions, and how such differences might
impact goal congruency in the relationship and the behaviors of both cus-
tomer and 3PL. As Mahapatra, Narasimhan, and Barbieri (2010) note, under
conditions of interdependence, outcomes of one firm are also impacted by
joint behaviors of both parties.
In this study, customers of a single 3PL were surveyed. The strength of
this design is that there is some control for the 3PL strategic orientation. A
firm’s strategic orientation (e.g., market orientation, supply chain orien-
tation) has been shown to impact its strategies and supply chain relation-
ships (Min, Mentzer, and Ladd 2007). Although the 3PL’s behaviors were
not explicitly controlled for in the model, it is expected that the 3PL, based
on its rankings in the Brazilian market and on interviews with large and
smaller customer firms, promotes a positive relationship climate with cus-
tomers. Future studies could consider the impact of various 3PLs’ customer
management strategies on dependence asymmetry, joint dependence, and
the relationship outcomes.
The dependence asymmetry and joint dependence measures were
calculated from adapted dyadic dependence scales from Ganesan (1994).
Because the number and items differed for the “customer dependence on
3PL” and “3PL dependence on customer” scales, they had to be averaged
for the calculation of manifest variables and structural equation modeling
could not be used as the statistical technique of the study. Future studies
could explore developing symmetric scales in order to enhance the sophis-
tication of the data analysis.
Performance measures in this study were captured through the cus-
tomers’ perceptions, and no secondary data regarding the actual per-
formance of operations were collected. It was not possible to collect 3PL
performance indicators, such as on-time delivery, due to confidentiality
concerns by the 3PL. Considering that the respondents are managers from
 Hofer: Are We in This Together? \ 465

customers responsible for managing the 3PL, these managers’ perceptions

of performance improvements can be a good predictor of the success of the
relationship. Davis and Mentzer (2006) propose, however, that expectations
of performance might vary depending on whether a firm is more or less
dependent. Supply chain management theory and practice could benefit
from addressing the relationship between the embeddedness characteris-
tics addressed here and specific quantitative logistics performance metrics
from secondary data.
Reflective of the current status of the logistics outsourcing industry
worldwide, over 66 percent of the study sample outsourced one or two
functions. Future research should test this model with a larger sample of
customers that outsource more strategic functions. Also, it would be rel-
evant to test the model with younger relationships and compare them to
older relationships. A final limitation of the sample in the study was that
only 20 percent of the observations (55 out of 281) related to customers that
perceived themselves as less powerful than the 3PL. While both constructs
of asymmetry (including when a customer is more dependent) had signifi-
cant relationships with long-term orientation, a more balanced sample in
terms of asymmetry could have led to stronger results for the customer
dependence disadvantage construct.
To build on the theoretical contribution of this study, resource depen-
dency theory could be used as a theoretical basis to identify whether
customer or 3PL’s environmental conditions, in terms of industry-level
uncertainty or competitive intensity, impact the usage of adversarial actions
under conditions of dependence asymmetries or collaborative behavior due
to joint dependence. In addition, research in strategy has indicated that
very high levels of joint dependence can be detrimental to relationships,
because there are more opportunities to behave opportunistically and less
incentive to innovate or deliver valuable feedback (Uzzi 1996). There may
be optimal level of asymmetry that would lead to better relationships and
performance compared to highly joint-dependent, highly trusted relation-
ships. Future research could investigate the appropriate levels of interde-
pendence that 3PLs and customers should pursue in order to experience the
greatest benefits possible.
A final avenue for future research would be to incorporate the struc-
tural conditions of the relationship, such as contracting, in the model. One
relevant type of dependence balancing mechanisms by the most dependent
party refers to the adoption of long-term contracts (Casciaro and Piskorski
2005) in which the most dependent organization can guarantee the access
466 / TRANSPORTATION JOURNAL™

to resources provided by a 3PL. Future research could investigate the inter-

dependence conditions that lead to the adoption of long-term contracts
and the effects of different types of contracting strategies (e.g., perfor-
mance-based contracting) in the relationship between interdependence
and behaviors by customers and 3PLs.
This study has only scratched the surface in understanding the
dynamics of both dimensions of interdependence—dependence asym-
metry and joint dependence—in buyer-supplier relationships in the con-
text of logistics outsourcing. Power differentials will always exist and
shifts in power balance will occur, so both 3PLs and customer firms can
benefit from understanding the underlying dynamics and outcomes of
interdependence.

Appendix
Table A1/Scale Items for Measurement Constructs and EFA Loadings
Construct Source Scale Items EFA Loadings
Dependence Ganesan 1. #3PL name# is crucial to our performance. 0.968
on 3PL 1994 2. #3PL name# is important to our business. 0.970
3. If our relationship with #3PL name# were discon-
tinued, we would have difficulty in performing
its services. 0.975
4. It would be difficult for us to replace #3PL
name#. 0.976
5. We are dependent on #3PL name#. 0.975
6. We do not have a good alternative to #3PL
name#. 0.950
3PL depen- Ganesan 1. We are important to #name#. 0.981
dence 1994 2. We are a major customer for #name# in our
trading area. 0.977
3. We are not a major customer for #name# (R). 0.957
Customer’s Gardner, 1. We conduct many transactions electronically
operational Cooper, and with #3PL name#. 0.835
information Noordewier 2. We exchange operational information with #3PL
exchange 1994 name#. 0.815
3. We use software dedicated to our relationship
with #3PL name#. 0.721
Customer’s Gardner, 1. We expect our relationship with #3PL name# to
long-term Cooper, and last a long time. 0.739
orientation Noordewier 2. We are very loyal to #3PL name#. 0.718
1994 3. We believe that over the long run our relation-
ship with #3PL name# will be profitable. 0.743
4. Maintaining a long-term relationship with #3PL
name# is important to us. 0.801
5. We focus on the long-term goals of this
relationship. 0.708

(Continues)
 Hofer: Are We in This Together? \ 467

Construct Source Scale Items EFA Loadings

Logistics Knemeyer The next set of statements is related to how your
operations and relationship with #3PL name# has improved your
performance Murphy 2004 company
performance. Please indicate your level of
agreement.
This relationship has . . .
improved our logistics system responsiveness. 0.813
improved our logistics system information. 0.805
reduced our operational risk. 0.786
Has provided us specialized services. 0.755
improved our product/service availability. 0.816
allowed us to achieve logistics costs reduction. 0.632
improved our information technology. 0.758
enabled us to implement changes faster/better. 0.854
provided us more specialized logistics expertise. 0.826

References
Alcacer, J., and J. Oxley. 2014. “Learning by Supplying.” Strategic Management Journal
35 (2): 204–23.
Anderson, E., and B. Weitz. 1989. “Determinants of Continuity in Conventional Industrial
Channel Dyads.” Marketing Science 8 (4): 310–24.
———. 1992. “The Use of Pledges to Build and Sustain Commitment in Distribution
Channels.” Journal of Marketing Research 29 (1): 18–34.
Armstrong and Associates, Inc. 2013. “Global Third-party Logistics Buying, Brands, and
Benefits.” Council of Supply Chain Management Professionals Conference
proceedings.
Bacharach, S. B., and E. J. Lawler. 1981. “Power and Tactics in Bargaining.” Industrial and
Labor Relations Review 34 (2): 219–33.
Belaya, V., T. Gagalyuk, and J. Hanf. 2009. “Measuring Asymmetrical Power Distribution
in Supply Chain Networks: What Is the Appropriate Method?” Journal of
Relationship Marketing 8 (2): 165–93.
Blau, P. M. 1964. Exchange and Power in Social Life. New York: John Wiley and Sons.
Bolumole, Y. A. 2003. “Evaluating the Supply Chain Role of Logistics Service Logistics
Providers.” International Journal of Logistics Management 14 (2): 93–107.
Cai, S., and Z. Yang. 2008. “Development of Cooperative Norms in the Buyer–Supplier
Relationship: The Chinese Experience.” Journal of Supply Chain Management
44 (1): 55–70.
Casciaro, T., and M. J. Piskorski. 2005. “Power Imbalance and Interorganizational
Relations: Resource Dependence Theory Revisited.” Administrative Science
Quarterly 50 (June): 167–99.
Chung, J., B. Sternquist, and Z. Chen. 2006. “Retailer-Buyer Supplier Relationships: The
Japanese Difference.” Journal of Retailing 82 (4): 349–55.
Corsten, D., and J. Felde. 2005. “Exploring the Performance Effects of Key Supplier
Collaboration: An Empirical Investigation into Swiss Buyer-Supplier
468 / TRANSPORTATION JOURNAL™

Relationships.” International Journal of Physical Distribution and Logistics

Management 35 (6): 445–61.
Cox, A. 2001a. “Managing with Power: Strategies for Improving Value Appropriation
from Supply Relationships.” Journal of Supply Chain Management 37 (2): 42–47.
———. 2001b. “The Power Perspective in Procurement and Supply Management.” Journal
of Supply Chain Management 37 (2): 4–7.
———. 2001c. “Understanding Buyer and Supplier Power: A Framework for Procurement
and Supply Competence.” Journal of Supply Chain Management 37 (2): 8–15.
Cox, A., J. Sanderson, and G. Watson. 2001. “Supply Chains and Power Regimes: Toward
an Analytic Framework for Managing Extended Networks of Buyer.” Journal of
Supply Chain Management 37 (2): 28–35.
Cox, A., G. Watson, C. Lonsdale, and J. Sanderson. 2004. “Managing Appropriately in
Power Regimes: Relationship and Performance Management in 12 Supply
Chain Cases.” Supply Chain Management 9 (5): 357–71.
Crook, T. R., and J. G. 2007. “Sources and Consequences of Bargaining Power in Supply
Chains.” Journal of Operations Management 25 (2): 546–55.
Davis, B. R., and J. T. Mentzer. 2006. “Logistics Service Driven Loyalty: An Exploratory
Study.” Journal of Business Logistics 27 (2): 53–73.
Dwyer, F. R., P. H. Schurr, and S. Oh. 1987. “Developing Buyer-Seller Relationships.”
Journal of Marketing 51 (2): 11–27.
Emerson, R. 1962. “Power-dependence Relations.” American Sociological Review 26:31–41.
Eyefortransport. 2012. “3PL Selection and Contract Renewal—North American Focus.”
http://www.supplychain247.com/paper/3pl_selection_contract_renewal_
north_american_focus/eyefortransport.
Ganesan, S. 1994. “Determinants of Long-term Orientation in Buyer-Seller Relationships.”
Journal of Marketing 58 (2): 1–19.
Ganesan, S., M. George, S. Jap, R. W. Palmatier, and B. Weitz. 2009. “Supply Chain
Management and Retailer Performance: Emerging Trends, Issues, and
Implications for Research and Practice.” Journal of Retailing 85 (1): 84–94.
Gardner, J. T., M. C. Cooper, and T. Noordewier. 1994. “Understanding Shipper-Carrier
and Shipper-Warehouse Relationships: Partnerships Revisited.” Journal of
Business Logistics 15 (2): 121–43.
Garver, M. S. and J. T. Mentzer. 1999. “Logistics Research Methods: Employing Structural
Equation Modeling to Test for Construct Validity.” Journal of Business Logistics
20 (1): 33–58.
Gassenheimer, J. B., and R. Ramsey. 1994. “The Impact of Dependence on Dealer
Satisfaction: A Comparison of Reseller-Supplier Relationships.” Journal of
Retailing 70 (3): 253–66.
Granovetter, M. 1985. “Economic Action and Social Structure: The Problem of
Embeddedness.” American Journal of Sociology 91 (3): 481–510.
 Hofer: Are We in This Together? \ 469

Grawe, S. J., C. W. Autry, and P. J. Daugherty. 2014. “Organizational Implants and Logistics
Service Innovation: A Relational Social Capital Perspective.” Transportation
Journal 53 (2): 180–210.
Gulati, R., and M. Sytch. 2007. “Dependence Asymmetry and Joint Dependence
in Interorganizational Relationships: Effects of Embeddedness on a
Manufacturer’s Performance in Procurement Relationships.” Administrative
Science Quarterly 52 (1): 32–69.
Gundlach, G. T., and E. R. Cadotte. 1994. “Exchange Interdependence and Interfirm
Interaction: Research in a Simulated Environment.” Journal of Marketing
Research 31 (November): 516–32.
Hancock, G. R., and R. O. Mueller. 2001. “Rethinking Construct Reliability within Latent
Variable Systems.” In Structural Equation Modeling: Present and Future—A
Festschrift in Honor of Karl Joreskog, edited by R. Cudeck, S. du Toit, and
D. Sorbom, 195–216. Lincolnwood, IL: Scientific Software International.
Heide, J. B., and A. S. Miner. 1992. “The Shadow of the Future: Effects of Anticipated
Interaction and Frequency of Contact on Buyer-Seller Cooperation.” Academy
of Management Journal 35 (2): 265–91.
Hingley, M. 2001. “Relationship Management in the Supply Chain.” International Journal
of Logistics Management 12 (2): 57–71.
Hofer, A. R., A. M. Knemeyer, and M. E. Dresner. 2009. “Antecedents and Dimensions
of Customer Partnering Behavior in Logistics Outsourcing Relationships.”
Journal of Business Logistics 30 (2): 141–61.
Hofer, A. R., A. M. Knemeyer, and P. R. Murphy. 2012. “The Roles of Procedural and
Distributive Justice in Logistics Outsourcing Relationships.” Journal of Business
Logistics 33 (3): 196–209.
Hofer, A. R., R. Smith, and P. R. Murphy. 2014. “Spillover Effects of a Firm’s Relationship
Marketing Orientation in the Logistics Triad.” International Journal of Logistics
Management 25 (2): 270–88.
Hu, L., and P. M. Bentler. 1999. “Cutoff Criteria for Fit Indexes in Covariance Structure
Analysis: Conventional Criteria versus New Alternatives.” Structural Equation
Modeling 6 (1): 1–55.
Huber, G., and R. Daft. 1987. “The Information Environment of Organizations.” In
Handbook of Organizational Communication, edited by F. Jablin, L. L. Putnam,
K. H. Roberts, and L. W. Porter, 130–64. Newbury Park, CA: Sage Publications.
Ireland, R. D., and J. W. Webb. 2007. “A Multi-theoretic Perspective on Trust and Power
in Strategic Supply Chains.” Journal of Operations Management 25 (2): 482–97.
Jap, S. D., and S. Ganesan. 2000. “Control Mechanisms and the Relationship Life Cycle:
Implications for Safeguarding.” Journal of Marketing Research 37 (May): 227–45.
Johnston, W. J., and T. Bonoma. 1984. “The Effect of Power Differences on the Outcome
of Consumer Bargaining Situations.” Advances in Consumer Research 11 (1):
170–74.
470 / TRANSPORTATION JOURNAL™

Kim, D., S. T. Cavusgil, and R. J. Calantone. 2006. “Information System Innovations and
Supply Chain Management: Channel Relationships and Firm Performance.”
Journal of the Academy of Marketing Science 34 (1): 40–54.
Kim, G., Shin, B., Kim, K.K., and Lee, H.G. 2011. “IT Capabilities, Process-Oriented
Dynamic Capabilities, and Firm Financial Performance.” Journal of the
Association for Information Systems 12 (7): 487–517.
Kline, R. B. 2005. Principles and Practice of Structural Equation Modeling. 2nd ed. New York:
Guilford Press.
Knemeyer, A. M., T. M. Corsi, and P. R. Murphy. 2003. “Logistics Outsourcing
Relationships: Customer Perspectives.” Journal of Business Logistics 24 (1):
77–109.
Knemeyer, A. M., and P. R. Murphy. 2004. “Evaluating the Performance of Third-party
Logistics Arrangements: A Relationship Marketing Perspective.” Journal of
Supply Chain Management 40 (1): 35–51.
Knoppen, D., and E. Christiaanse. 2007. “Interorganizational Adaptation in Supply
Chains: A Behavioral Perspective.” International Journal of Logistics Management
18 (2): 217–37.
Kumar, N., L. K. Scheer, and J. E. M. Steenkamp. 1995. “The Effects of Perceived
Interdependence on Dealer Attitudes.” Journal of Marketing Research 32
(August): 348–56.
Lambe, C. J., C. M. Wittman, and R. Spekman. 2001. “Social Exchange Theory and
Research on Business-to-Business Relational Exchange.” Journal of Business-to-
Business Marketing 8 (3): 1–37.
Lambert, D. M., M. A. Emmelhainz, and J. T. Gardner. 1996. “Developing and Implementing
Supply Chain Partnerships.” International Journal of Logistics Management 7 (2):
1–17.
Lambert, D. M., and T. C. Harrington. 1990. “Measuring Nonresponse Bias in Customer
Service Mail Surveys.” Journal of Business Logistics 11 (2): 5–25.
Langley, C. J., and Capgemini Consulting. 2012. “2012 Third-party Logistics Study: The
State of Logistics Outsourcing.” http://www.3plstudy.com/.
Larson, P. D. and B. Gammelgaard. 2001. “The Logistics Triad: Survey and Case Study
Results.” Transportation Journal 41 (2/3): 71–82.
Leuschner, R., C. G. Carter, T. J. Goldsby, and Z. S. Rogers. 2014. “Third-party Logistics:
A Meta-Analytic Review and Investigation of Its Impact on Performance.”
Journal of Supply Chain Management 50 (1): 21–43.
Li, Z. G., and R. P. Dant. 2001. “Channel Interdependence: Conceptual and Operational
Considerations” Journal of Marketing Channels 9 (1–2): 33–64.
Lusch, R. F. and J. R. Brown. 1996. “Interdependency, Contracting, and Relational
Behavior in Marketing Channels.” Journal of Marketing 60 (4): 19–38.
Mahapatra, S. K., R. Narasimhan, and P. Barbieri. 2010. “Strategic Interdependence,
Governance Effectiveness and Supplier Performance: A Dyadic Case Study
Investigation and Theory Development.” Journal of Operations Management 28
(6): 537–52.
 Hofer: Are We in This Together? \ 471

Maloni, M., and W. C. Benton. 2000. “Power Influences in the Supply Chain.” Journal of
Business Logistics 21 (1): 49–73.
Maloni, M., and C. R. Carter. 2006. “Opportunities for Research in Third-party Logistics.”
Transportation Journal 45 (2): 23–38.
Marasco, A. 2008. “Third-party Logistics: A Literature Review.” International Journal of
Production Economics 113 (1): 127–47.
Min, S., J. T. Mentzer, and R. T. Ladd. 2007. “A Market Orientation in Supply Chain
Management.” Journal of the Academy of Marketing Science 35 (4): 507–22.
Morash, E. A. 2001. “Supply Chain Strategies, Capabilities, and Performance.”
Transportation Journal 41 (1): 37–55.
Morgan, R. M., and S. D. Hunt. 1994. “The Commitment-Trust Theory of Relationship
Marketing.” Journal of Marketing 15 (2): 135–52.
Mudambi, R., and S. Helper. 1998. “The Close but Adversarial Model of Supplier Relations
in the US Auto Industry.” Strategic Management Journal 19 (8): 775–92.
Petersen, K. J., R. B. Handfield, B. Lawson, and P. D. Cousins. 2008. “Buyer Dependency
and Relational Capital Formation: The Mediating Effects of Socialization
Processes and Supplier Integration.” Journal of Supply Chain Management 44 (4):
53–65.
Podsakoff, P. M., S. B. MacKenzie, J. Y. Lee, and N. P. Podsakoff. 2003. “Common Method
Biases in Behavioral Research: A Critical Review of the Literature and
Recommended Remedies.” Journal of Applied Psychology 88 (5): 879–903.
Rajagopal, A. R. 2009. “Buyer-Supplier Relationship and Operational Dynamics.” Journal
of the Operational Research Society 60 (3): 313–20.
Rinehart, L., and D. J. Closs. 1991. “Implications of Organizational Relationships,
Negotiator Personalities, and Contract Issues on Outcomes in Logistics
Negotiations.” Journal of Business Logistics 12 (1): 123–44.
Ryu, S., J. E. Park, and S. Min. 2007. “Factors of Determining Long-term Orientation in
Interfirm Relationships.” Journal of Business Research 60 (12): 1225–33.
Ryu, S., and E. Kim. 2010. “The Moderating Effect of Long-term Orientation on the
Relationship between Interfirm Power Asymmetry and Interfirm Contracts:
The Cases of Korea and USA.” Journal of Applied Business Research 26 (6): 135–46.
Saxe, R., and B. A. Weitz. 1982. “The SOCO Scale: A Measure of the Customer Orientation
of Salespeople.” Journal of Marketing Research 19 (3): 343–52.
Shapiro, R. 1985. “Toward Effective Supplier Management: International Comparisons.”
Unpublished manuscript, Harvard Business School.
Sheu, C., H. R. Yen, and B. Chae. 2006. “Determinants of Supplier-Retailer Collaboration:
Evidence from an International Study.” International Journal of Operations and
Production Management 26 (1/2): 24–49.
Taylor, J., and G. Jackson. 2000. “Conflict, Power, and Evolution in the Intermodal
Transportation Industry’s Channel of Distribution.” Transportation Journal
39 (3): 5–17.
Thibaut, J. W., and H. H. Kelley. 1959. The Social Psychology of Groups. New York: John Wiley
and Sons.
472 / TRANSPORTATION JOURNAL™

Thomas, R. W., and T. L. Esper. 2010. “Exploring Relational Asymmetry in Supply Chains:
The Retailer’s Perspective.” International Journal of Physical Distribution and
Logistics Management 40 (6): 475–94.
Uzzi, B. 1996. “The Sources and Consequences of Embeddedness for the Economic
Performance of Organizations: The Network Effect.” American Sociological
Review 61 (4): 674–98.
———. 1997. “Social Structure and Competition in Interfirm Networks: The Paradox of
Embeddedness.” Administrative Science Quarterly 42 (1): 35–67.
Williams, L. R. 1994. “Understanding Distribution Channels: An Interorganizational
Study of EDI Adoption.” Journal of Business Logistics 15 (2): 173–204.
Williams, B. D., J. Roh, T. Tokar, and M. Swink. 2013. “Leveraging Supply Chain Visibility
for Responsiveness: The Moderating Role of Internal Integration.” Journal of
Operations Management 31 (7/8): 543–54.
Xia, J. 2011. “Mutual Dependence, Partner Substitutability, and Repeated Partnership:
The Survival of Cross-Border Alliances.” Strategic Management Journal 32 (3):
229–53.
Zacharia, Z. G., N. R. Sanders, and N. W. Nix. 2011. “The Emerging Role of the Third-party
Logistics Provider (3PL) as an Orchestrator.” Journal of Business Logistics 32 (1):
40–54.
Zhou, N., G. Zhuang, and L. S. C. Yip, 2007. “Perceptual Difference of Dependence and its
Impact on Conflict in Marketing Channels in China: An Empirical Study with
Two-sided Data.” Industrial Marketing Management 36 (3): 309–21.
Implementing Off-peak Deliveries in the Greater
Toronto Area: Costs, Benefits, Challenges

Jessica McPhee, Ari Paunonen, Taufiq Ramji,

and James H. Bookbinder

Abstract
Nestle Canada currently uses 32 routes that serve over 4,500 customers in the
Greater Toronto Area (GTA). This study aims to quantify Nestlé’s costs and
benefits of modifying their ice cream supply chain to incorporate night-time
deliveries, while providing a framework for the regulatory, conceptual, and
inertial obstacles to implementation. Employing Nestlé’s customer data set,
we created routing software to determine the proportion of customers who
must be willing to accept deliveries outside of normal working hours so that
the change would be financially feasible. Based upon a literature review we
found that, before proceeding, the following qualitative factors should be
considered: safety, sustainability, regulatory concerns, truck noise, traffic,
and congestion. Reduction of 3–10 percent in the number of routes may result
from switching a suitable proportion of deliveries to night-time, achiev-
ing the minimum fleet size when 50–60 percent of locations are served on
night routes. The operation of both night-time and daytime deliveries would
enable an increase in truck utilization, thus decreasing the number of vehi-
cles required. Recommendations for success of night-time deliveries include
preparation of a safety plan, procurement of plate trucks, noise-abatement
techniques, and the development of a noise-monitoring program.

Keywords
Night delivery, congestion, environment, truck noise, sustainability

Jessica McPhee James H. Bookbinder

University of Waterloo
jessica.mcphee@hotmail.com
Ari Paunonen
University of Waterloo
paunonen.ari@gmail.com
Corresponding Author
University of Waterloo
jbookbinder@uwaterloo.ca
Transportation Journal, Vol. 54, No. 4, 2015
Copyright © 2015 The Pennsylvania State
University, University Park, PA

Taufiq Ramji
University of Waterloo
taufiq.ramji@gmail.com
474 / TRANSPORTATION JOURNAL™

Introduction
Toronto is a rapidly growing city and, for two years in a row, was ranked
the number one most sustainable city in Canada (Marchington 2011). As a
byproduct of urban growth, daytime congestion is increasing as more and
more people are commuting into the downtown core. The volume of cars is
adversely impacting delivery vehicles by increasing travel times, fuel con-
sumption, and accumulated truck-parking violations. In addition to the
nuisance caused by heavy traffic, these vehicles are also a major contribu-
tor of smog and pollution.
One mitigation strategy to address this problem is shifting deliveries
to off-peak hours. Although hardly novel, it is only recently that this con-
cept has been given serious consideration. Several academic papers and
industry initiatives have examined implementing a night-time delivery
model, concluding that significant financial and environmental benefits
are realizable from adopting this approach. (See the “Literature Review”
section below.)
Enticed by these benefits, Nestlé Canada—wholly owned subsidiary of
Nestlé S.A., the world’s largest food company—became interested in modi-
fying its ice cream distribution network to incorporate off-peak delivery.
Specifically, Nestlé wished to quantify the costs and benefits of switch-
ing, as well as a framework for navigating the regulatory, conceptual, and
inertial obstacles to implementation. We have analyzed Nestlé’s ice cream
direct-store delivery network for the Greater Toronto Area (GTA), the sta-
tus quo with respect to off-peak deliveries in the GTA, and contemporary
academic research on numerous factors affecting successful implemen-
tation, including environmental effects (Sathaye, Harley, and Mandanat
2010), noise (City of Toronto By-law No. 111-2003), and safety (Cavar, Kavran,
and Jolic 2011). This article summarizes the methodology and results of the
investigation, and provides recommendations for eventual adoption of an
off-peak delivery schedule.

Literature Review
The concept of delivering goods during the off-peak hours to reduce day-
time traffic congestion has been of interest to policymakers throughout
history. In fact, the first known record of the implementation of night-
time deliveries dates back all the way to ancient Rome, when Julius Caesar
banned all commercial deliveries during the day (H. Dessau, Inscriptiones
Latinae selectae, no. 6085 [1892], cited in Holguín-Veras 2008). It is interest-
ing to note that this move may have been influenced by similar rules in
 McPhee et al.: Off-peak Deliveries in the Greater Toronto Area \ 475

Greek cities, indicating that traffic congestion required legislation even

before the era of Julius Caesar.
Much more recent publications on the topic of night-time deliveries
include Churchill (1970), Organization for Economic Growth (1979), and
Noel et al. (1980). In fact, most of the research and information on this
subject has appeared in the last 15 years. Holguín-Veras et al. (2007, 2008)
study the policies of the carrier and consignee. Verlinde et al. (2010) con-
sider these and other stakeholders. Holguín-Veras (2006, 2008) and Vilain
and Wolfrom (2007) treat the problem as that of the pricing of urban freight.
A survey of restrictions on night-time deliveries has been given by Browne
et al. (2006).
Vilain and Wolfrom (2007) conducted a set of interviews with
management-level personnel at various trucking companies. The ques-
tions pertained to proposed off-peak discounts (or peak-period sur-
charges) for the New Jersey–Manhattan bridges and tunnels. It was found
that “value pricing,” as those authors termed it, would have minimal effect
in encouraging night deliveries. Many consignees specify delivery time
windows to their carriers; the financial penalty for missing a time window
far outweighs any off-hour discount in the bridge tolls. Other findings of
the interviews were that union regulations and/or hours of operation at
consignees worked against avoiding congestion by trying to deliver in the
6:00 PM–6:00 AM interval.
Several intuitive reasons initially attract organizations to consider
implementing night-time deliveries: optimized vehicle utilization, possi-
bly lower truck emissions due to favorable night-time traffic conditions,
and fewer disturbances for shop owners receiving deliveries during peak
sale times (Forkert and Eichhorn n.d.). Although these factors are typically
cited as arguments in favor of switching to night-time deliveries, some ele-
ments (such as environmental and traffic congestion) have wider-spanning
implications that may be difficult to quantify accurately. The impact to
traffic congestion and the environment will be discussed in more detail in
subsequent sections.
A significant aspect of successfully implementing off-peak deliver-
ies is stakeholder buy-in. Three key stakeholders have been identified:
city administration, transport operators, and shop owners (Forkert and
Eichhorn n.d.). The majority of drawbacks associated with night-time
deliveries affect receivers (shop owners); they will now require staff trained
to accept deliveries that arrive during the off-peak (potentially incurring
training costs and wage increases/shift premiums), will probably have
476 / TRANSPORTATION JOURNAL™

higher heating and lighting costs due to additional operating hours, and
may also have increased security and insurance costs (Holguín-Veras
et al. 2005). As such, these stakeholders could pose the biggest roadblock
to implementation. Hence, it has been suggested that the most likely
configuration to successfully implement off-peak deliveries are shippers
who handle their own transportation and receivers who are open during
extended hours, followed closely by the same shippers who handle their
own transportation and receivers who do not currently operate over a
lengthened day, but would be willing to (Holguín-Veras et al. 2005).
Among the most important considerations for implementation of
night-time deliveries is the general effect on society, mainly increased noise
in residential areas during off-peak hours due to delivery activities. Noise
abatement techniques have been widely studied, and will be discussed later
in this article.
Assuming that potential benefits outweigh the drawbacks discussed
above for night-time deliveries, a corporate organization will typically not
make such a significant strategic move without a justified reason to expect
long-term financial benefit. The major cost savings associated with imple-
menting night-time deliveries are achieved through lowered logistical
expenses due to optimized utilization of vehicles and personnel. However,
there are also many financial costs associated with the implementation of
night-time deliveries, such as the potential expense of purchasing quiet vehi-
cles and equipment, increased customer costs due to receiving deliveries at
night, and shift premiums paid to off-shift employees (Forkert and Eichhorn
n.d.). This article will provide a thorough analysis of the major benefits and
drawbacks of implementing night-time deliveries, both financial and other-
wise, to offer a framework for implementation at Nestlé Canada Ice Cream.
Our Approach
From the insights gained through the literature review, this study was
refined to encompass two major subject areas: quantitative factors, includ-
ing the requisite changes to the supply chain and the cost of implementa-
tion, and qualitative factors such as environmental and societal impacts.
The initial quantitative concern was to determine whether the imple-
mentation would be financially feasible. Customized routing software was
created to simulate a heuristic approach that closely approximates Nestlé’s
current routing methods. By varying the proportion of customers receiving
at night and the relative increase in speed for off-peak travel vis-à-vis day-
time travel, the software is able to estimate the number of routes and, ergo,
the number of vehicles required to serve the network.
 McPhee et al.: Off-peak Deliveries in the Greater Toronto Area \ 477

From a qualitative perspective, several conditions must be satisfied

before a night-time delivery network can be implemented. Many of these
factors are specific to the municipality or region. In this article, the noise,
environmental, and societal considerations are explored, particularly with
respect to the GTA. Together, this qualitative analysis and the analysis of
outputs from the software form the framework for implementing night-
time deliveries in the GTA.
A significant element of evaluating the potential impact of night-time
deliveries was the consideration of qualitative factors affecting the feasibil-
ity of their implementation. The following section details the consequence
of night-time deliveries on safety, sustainability, bylaws, truck noise, traffic
congestion, and parking.

Qualitative Factors

Safety

There is not much literature related to safety conditions during night-time

deliveries. Cavar, Kavran, and Jolic (2011) discuss the technologies of intel-
ligent transportation systems for the planning of safe deliveries at night.
Their context is that of “City Logistics,” the picking up and delivering of
goods within urban areas, with the goal that these freight movements be
environmentally friendly. Those authors emphasize advanced traveler
information systems that would enable delivery personnel to better esti-
mate their travel times, and perhaps reroute, in light of the reduced night-
time congestion.
The article by Bjerkan, Sund, and Nordtømme (2014) refers to “Green
Urban Distribution,” different words than employed by Cavar, Kavran, and
Jolic (2011), but with essentially the same context and goals, in this case
for Oslo, Norway. Bjerkan, Sund, and Nordtømme go further in that they
focus on the stakeholders: carriers, consignees, and the municipal authori-
ties. Survey results demonstrated that each of the stakeholder groups faced
obstacles to successful implementation of night deliveries.
The present authors have identified additional safety concerns asso-
ciated with implementation of such deliveries. Suggested resolutions are
given here because they are not consequences of the routing analyses we
performed.
One of the major issues facing implementation is the theft of product
or cash. (Many deliveries of ice cream were to smaller, mom-and-pop con-
venience stores. Their preference for cash payment, a request accepted by
478 / TRANSPORTATION JOURNAL™

Nestlé, may have been somewhat related to various ethnic backgrounds).

Night-time deliveries increase the risk of theft due to darkness and lack
of crowds. As a result, implementation of a low-cash policy, or a switch
entirely to automated transactions for night-time deliveries, should be
instigated. This will minimize the chance of theft and potentially danger-
ous conflict for the driver. Additionally, procedures for unloading must be
developed for the purpose of product safety. Keeping the door of the truck
locked and adopting methodologies for delivery analogous to those devel-
oped by armored transportation companies will increase both driver and
product safety.
Another important concern is associated with driver fatigue. If night-
time deliveries are implemented in a company that previously operated
only one shift, the firm and its distributors (if applicable) will need to
determine how they will schedule the second shift of employees. To lessen
fatigue and ensure a smooth transition to overnight work, organizations
should try to minimize the frequency with which drivers change between
the two shifts. Two common shift schemes that could be adopted are fixed
shifts, where some employees work the day shift and some work the off-
shift, and a rotating-shift schedule, where employees alternate between the
two shifts for a set period of time (i.e., a one- or two-week period). The shift
schemes and rotations should be applied to drivers, warehouse personnel,
and any other relevant support staff.

Sustainability and Environmental Issues

A significant benefit commonly associated with implementation of night-

time deliveries is the potential reduction in truck emissions. By shifting
deliveries from the peak periods during the day, when roads are congested, to
the off-peak periods, when travel is less interrupted, drivers will experience
decreased idling and shorter travel times. Therefore, the amount of emis-
sions produced by the same delivery trucks will be reduced if they switch
to night-time operations (Forkert and Eichhorn n.d.). Additionally, some
case studies have shown that the number of trucks can be diminished if an
organization switches to night deliveries, due to the time savings gained by
avoiding traffic (Finlay 2008). Clearly, fewer trucks performing deliveries
would also lessen the emissions, although some researchers have found that
truck emissions (with the exception of NOx) are likely to noticeably decline
only if the number of trucks is decreased sufficiently (Campbell 1995).
Several studies provide estimates for these emission reductions in
specific contexts (e.g., Sathaye, Harley, and Mandanat 2010). Piecyk and
 McPhee et al.: Off-peak Deliveries in the Greater Toronto Area \ 479

McKinnon (2010) conducted research with a focus group and employed a

large-scale Delphi survey. Based on those opinions, they developed three
possible scenarios concerning freight-related energy consumption and
CO2 emissions. These scenarios were employed in a model to forecast
road-freight movements and the corresponding emissions in 2020. Only if
the “optimistic” scenario prevailed would the motor freight sector in the
United Kingdom be on a trajectory to meeting the 80 percent CO2-reduction
targets set by the government for the overall economy by 2050.
Another area of environmental impact that is frequently over-
looked by publications examining night-time deliveries is the effect on
24-hour-average pollution levels. Using atmospheric modeling, research in
California demonstrated that the night-time stability of the atmospheric
layer is always equal to, or more stable than, daytime stability. The enhanced
night-time stability of the atmospheric layer could trap more pollutant par-
ticles than during the day (Sathaye, Harley, and Mandanat 2010).
However, there is greater probability of environmental benefit when
off-peak policies target a particular time period, for example, the morning
rush hour. Benefits are also more likely when traffic speeds in that peak
period are relatively low, implying high peak-period emission factors
(Sathaye, Harley, and Mandanat 2010). Recommendations for identify-
ing and quantifying environmental issues are offered in our section on
“Future Work.”
Regulatory Concerns and Bylaws

To address potential regulatory worries, the authors reviewed the cur-

rent noise bylaws within the City of Toronto to determine their effect
on off-peak deliveries. In city-designated “quiet zones,” the act of load-
ing, unloading, delivering, packing, unpacking, or otherwise handling
any containers, products, or materials is prohibited from 7:00 PM one
day to 7:00 AM the next day, and all day Sunday and statutory holidays.
A quiet zone is defined as any property within the municipality used as a
hospital, retirement home, nursing home, senior citizens’ residence, or
for other similar use (By-law No. 111-2003). In residential areas, loading/
unloading noise is prohibited from 11:00 PM one day to 7:00 AM the next
day (9:00 AM Sundays and statutory holidays). It is important to note
that the City of Toronto By-law No. 111-2003 specifically defines super-
markets as routine unloading points, and as such, a stationary source of
sound. Deliveries during the night would, therefore, be acceptable at
these locations.
480 / TRANSPORTATION JOURNAL™

For Nestlé, typical residential areas within the city where the majority
of their customers are located are considered “class 1 areas.” These are
defined as “areas with an acoustic environment typical of a major popula-
tion center, where the background noise is dominated by the urban hum”
(By-law No. 111-2003). The maximum acceptable noise levels for a class
1 area within the City of Toronto are defined as 45 decibel (dB) during the
night, between 11:00 PM and 7:00 AM. Refrigeration units, more com-
monly known as “reefer” units for half trucks such as those Nestlé uses,
have been evaluated at 64 dB at 15 meters (Cvercko and Coulter 2009). That
noise is evidently far greater than the acceptable limit during evening
hours. However, Nestlé has begun procuring nonreefer trucks utilizing
cold plate technology; these operate at sound levels within the acceptable
night-time range.
There are, thus, several challenges that an organization will face
regarding noise issues and violations when determining the feasibility of
implementing night-time deliveries. Simple sound-level testing on the com-
pany’s fleet should be completed for more accurate analysis of the impact of
noise bylaws within the urban center. Moreover, even if night routes may be
implemented legally and fall within acceptable noise limitations, residents
might still complain to the city about the continuous disturbance due to
evening deliveries. The city could then launch an investigation to protect
the best interests of its inhabitants. In all cases, it is imperative to advocate
the benefits of this study to the population as a whole.
Truck Noise

Although truck noise is an important roadblock to successful implementa-

tion of night-time deliveries, the noise caused by a firm’s equipment used
for daytime deliveries is not typically of major concern. Current equipment
configurations may, thus, make it impossible to comply with noise bylaws
at night. While delivery vehicles alone may not violate the noise restric-
tions, noise caused by delivery events in addition to the truck noise may
result in a violation. Therefore, it would be prudent for the organization
to consider, where possible, retrofitting the equipment that will be used
for night-time deliveries to minimize the amount of noise caused by those
deliveries.
Research related to noise emission includes SenterNovem (2002) and
Finlay, Byrne, and Grimes (2008). A study on noise-abatement techniques
for night-time deliveries (Finlay 2008) determined the delivery events that
caused the majority of peak sounds: passage of pallets on the floor of the
 McPhee et al.: Off-peak Deliveries in the Greater Toronto Area \ 481

trailer, positioning the tail-lift platform, crossing pavement to the delivery

point, and stacking empty pallets inside the trailer unit.
As a result, equipment associated with the events above should be tar-
geted to maximize noise reduction. Some examples of suggested equip-
ment retrofits are: dampening material for floor and rollers of trucks, a
hush-kit for steel roll-cages, and rubber or polyurethane wheels on pallet
trucks (Finlay 2008).
It is recommended that the company examine the noise levels asso-
ciated with its equipment and processes to determine where the focus of
noise-abatement retrofitting could have the most impact, as those efforts
may greatly increase their ability to successfully implement this project.
Additionally, drivers should receive training that enhances their aware-
ness and provides ways for them to proactively reduce high-noise activities
at night. It is interesting to note that increasing the awareness of noise-
reduction efforts (e.g., by attaching quality labels, indicating “quiet equip-
ment” is used on trucks with low-noise equipment) can result in a positive
impact on the project’s success (Forkert and Eichhorn n.d.).

Traffic Congestion and Parking

For organizations in urban centers, it is of significant concern that park-

ing is scarce for vehicles attempting to deliver. As a result, many trucks
are forced to park illegally to make a delivery; the cost of parking tickets is
accepted by the firm as a cost of doing business (Han et al. 2005). Nestlé cur-
rently estimates their annual expenses for parking fines at several thousand
dollars per delivery truck.
A one-month pilot study to test the feasibility of night-time deliver-
ies was conducted in Manhattan in 2010, in which each truck that operated
between 7:00 PM and 6:00 AM saved about $1,000 in parking tickets, along
with decreases in delivery-route time and improvement in driver working
conditions (Smerd 2010). Similar outcomes are expected from implementa-
tion in other urban areas due to increased parking availability during off-
peak hours.
A final consideration is the aggregate impact of many distribution net-
works switching to night deliveries. If large numbers of trucks are removed
from the road in the daytime, the decrease in congestion is expected to result
in increased daytime travel speed, especially in urban centers. These aggregate
effects may be significant if off-peak delivery becomes the norm. But it could
also happen that, if daytime congestion improves, more people may choose to
drive to the city center, tending to negate some of the benefits of fewer trucks.
482 / TRANSPORTATION JOURNAL™

Routing Simulation and Software

Customized routing software, which applies heuristic methods to achieve
a feasible, “good” solution, was created for Nestlé. The intended use of that
software was to allow Nestlé to estimate the number of routes needed and,
by extension, the cost of operation for night-time delivery for various condi-
tions. The primary focus of development was to provide software that could
take as input a list of locations and as output a list of routes by applying
various heuristics and iterating. For the purposes of analysis, algorithms
were selected to replicate Nestlé’s current manual route-creation method-
ology. External validity of the model was established by comparing outputs
for only daytime routes under standard conditions against empirical data
provided by Nestlé. Models were then formulated to vary two parameters:
the proportion of night routes and the night-time travel-speed multiplier.
These models were run and the outputs analyzed to understand the impacts
of those variables on night deliveries.

Nestlé Methodology

Currently, Nestlé uses mapping software to develop route clusters manu-

ally. The approach is as follows:

1. Find the outermost store location to first define a route.

2. Add nearby locations to the route until that route reaches a maxi-
mum number of locations based on truck capacity.
3. Repeat until all locations have been added to routes.
4. Using expert judgment, manipulate the routes to respect geographi-
cal and contractual constraints.
5. For each route, divide the locations into groups that receive on the
same day based on customer requirements.

It is important to note that Nestlé does not specify a sequence in which

the stores are visited. Rather, drivers are given a list for a particular day, and
are free to deliver to these locations in any order. Therefore, routes—in the
academic sense of the word—are not truly created. The actual sequence is
determined by the driver based on knowledge of the area, traffic conges-
tion, and loading-dock availability.

Software Design

Our software architecture implements a model-view-controller (MVC)

architectural style. The MVC design separates the logic of the user inter-
face from the “business logic” of the model itself. Changes to the way data
 McPhee et al.: Off-peak Deliveries in the Greater Toronto Area \ 483

are manipulated (by the model) are, thus, decoupled from how they are
displayed and stored.
The core of the program is a model object that represents the entire
network of sites to be serviced using a list of locations and a list of routes.
A configuration file defines values of a number of parameters, as well as the
heuristics to be employed to reach a solution.
To run a model, the initialization phase reads input data and creates
clusters of locations. Each cluster is simply an unordered collection of sites
to be served by a single truck on one of six daily trips each week. Next, an
iteration method swaps locations between clusters, sequences the locations
in each cluster to create that day’s route, and resizes those clusters. The num-
ber of sites on a given day is limited by the total trip time. If a cluster cannot
route all locations onto the six trips, the cluster is considered “over capacity”
and marked for reduction. On the other hand, if all locations are assigned to
day trips and the cluster has enough available capacity, the route is marked
for growth. Clusters are resized, and the iteration step is repeated until fea-
sibility is established (or a predetermined limit on iterations is reached). The
last step finalizes route assignments and computes various metrics.

Traffic, Distance, and Time

To incorporate traffic volumes in the GTA, traffic was monitored over the
course of five business days using real-time information from Google Maps
(https://www.google.com/maps). The observed area was segregated into zones
with each location assigned a traffic intensity factor from 1 to 5, with the value
5 representing severe gridlock, and value 1 denoting free-flowing traffic. (The
traffic intensity factor, thus, reflects differences in average speeds achieved
relative to the posted speed limit within the given area.) While an intensity
factor is as much a qualitative as a quantitative measure, the factor assigned
within each zone agreed with information obtained from Google Maps. Where
possible, factors were also verified for consistency with traffic counts sum-
marized by GTA civil engineers. The GTA network of Nestlé customers, with
their assigned intensity factors, was used in the software model to indicate a
preference for a higher-intensity traffic location to be placed on a night route.
In each case, the Euclidean distance in kilometers between any two
points is found using the latitude and longitude of each location and a
scaled conversion factor from degrees to kilometers for the region. For time
calculation, a travel speed is obtained by dividing the base speed of 60km/h
by the average traffic intensity of the endpoints.

speed = baseSpeed / ((trafficFactor location 1 + trafficFactor location 2) / 2)

484 / TRANSPORTATION JOURNAL™

Thus, between two locations with traffic intensity factors of 1, the aver-
age speed of travel is 60 km/h, whereas between two locations with traffic
intensity factors of 5, the average speed of travel is 12 km/h. For travel at
night, the speed is multiplied by a constant night speed-up factor between
1 and 1.5.

Choice of Heuristics

Drawing from academic literature, four heuristics were selected to emu-

late Nestlé’s routing procedure and were customized for simple, efficient
implementation. We refer to the first heuristic employed in the initializa-
tion phase to create clusters as the “far-point” heuristic. It is a variant of
that described by He (2011) and closely approximates Nestlé’s manual pro-
cedure. The (un-assigned) location farthest from the depot is the basis upon
which each new route is created.
Locations are iteratively swapped between clusters using the second
heuristic. It minimizes average distance between each location and the cen-
troid of all locations on that location’s assigned cluster. The swap heuristic,
referred to as “local max savings,” defines a neighborhood of the k clusters
that are closest to the chosen cluster. It considers each pair of locations and
swaps locations to maximize distance savings. This heuristic is nearly iden-
tical to that of Kanungo et al. (2002), which utilizes squared distance rather
than linear distance.
The third heuristic performs the routing between sites. Because
Nestlé’s methodology does not prescribe the sequence to visit locations, it
was decided not to optimize the routes generated. Instead, the unimproved
route distance was used as an approximation of actual distance traveled by
Nestlé drivers. Based on trial runs, the sweep algorithm (Gillett and Miller
1974) was employed to obtain conservative distance estimates.
Finally, a resizing heuristic is applied to dynamically change the num-
ber of locations in a cluster. That heuristic, also responsible for adding or
removing routes, is implemented as follows:

Step 1: If there are both routes with available capacity and routes over
capacity, randomly assign a location from each overcapacity
route to a route with available capacity.
Step 2: If there are routes with available capacity, no routes over
capacity, and the model has not already failed to find a feasible
solution with one route less than the current number of routes,
randomly select a route and move all locations to the other
randomly selected routes.
 McPhee et al.: Off-peak Deliveries in the Greater Toronto Area \ 485

Step 3: If there are routes over capacity, and no routes with available
capacity and the model has not previously found a solution with
the current number of routes, create a new route, randomly
select one location from each other route and move it to the new
route.
Step 4: If there are no routes with available capacity and no routes
over capacity, STOP.

This algorithm is actually “pseudorandom,” in the sense that the prob-

ability of a location or route being chosen is weighted by an appropriate
parameter such as density, distance from centroid, or extraCapacity. This
small modification greatly improves performance.

Night-time Routing

The selection of night routes is conceptually simple: The model forces all
sites that can receive deliveries at night to do so. Night routes could be cho-
sen via the same clustering algorithms employed in the initialization stage,
considering only the feasible night-delivery locations. However, as routes
have already been selected, there is an opportunity to locate night routes
efficiently to obtain a good starting assignment.
The heuristic implemented attempts to develop densely clustered
night routes by creating routes from day routes that have the greatest num-
ber of sites capable of receiving night deliveries. Each location that cannot
receive night deliveries is swapped to the closest day route. This is repeated
until all sites that can receive at night are on a night route.

Validation and Verification

Throughout development, the software and methodology were validated con-

sidering adherence to Nestlé’s current routing methodology, conformance
with academic best practice, and the satisfaction of stakeholders. Early in the
design process, close work with Nestlé route engineers ensured a full under-
standing of the methodology used to create routes. Extensive research was
conducted into contemporary methods used in route creation. The “cluster-
first, route-second” methodology (e.g., Laporte and Semet 2002) was selected
due to the unordered nature of each day’s drops. Over the course of our study,
regular contact was maintained between the authors and Nestlé to ensure the
software accurately represented Nestlé’s routing methods.
To verify the software-generated results, values for the model’s
number of routes and total distance traveled were compared to the val-
ues for Nestlé’s current delivery network. A feasible routing model was
486 / TRANSPORTATION JOURNAL™

created for all customer locations using 31 routes. In comparison, Nestlé

currently serves all customers using 32 routes—a difference of 3.1 percent.
The slight inaccuracy may be attributed to several factors (e.g., speed of
travel or traffic factors used); however, this variation was considered small
enough to lend credibility to the model’s numerical results for routes.
Nevertheless, there was a large difference between Nestlé’s estimate of
total annual distance traveled and the software figure. But two major fac-
tors satisfactorily explain this discrepancy. Firstly, Nestlé builds daily trips
based on the days its customers receive deliveries, whereas the model devel-
ops a day’s trips based on a distance-minimizing heuristic, hence under-
estimating the distance traveled. Second, the software uses Euclidean
distance rather than rectilinear distance.
We briefly remark that the use of Euclidean distance was not consid-
ered to be problematic because:

A large component of the distance driven was on a near-Euclidean

path; the highway line-haul from a distribution center (DC) to
the customer cluster would be close to a straight line.
Actual distances driven would differ from rectilinear distances
because, in many cases, drivers had to circle the block several
times to find parking.

Therefore, to interpret the values obtained for distance in sensitivity

analysis, it was necessary to consider relative changes from the baseline
prediction. The distances were used for comparative purposes. Because we
knew our distance figures were not likely to be 100 percent accurate, the
most important consideration was that the same methodology be used so
that we could compare relative changes.

Results and Recommendations

Models were employed to assess the impact of varying off-peak delivery
adoption rates and the relative increase in travel speed during those hours.
The specific locations that may receive during off-peak hours were ran-
domly chosen. Table 1 summarizes the results of simulations.
In general, results indicate that a reduction of 3–10 percent of total routes
may be attained by switching an appropriate proportion of deliveries to off-
peak hours. The minimum fleet size is achieved for a 50–60 percent adoption
rate, with the exact figure dependent on the relative speed increase. Figure 1
shows the number of routes as a function of the proportion of night deliver-
ies for off-peak travel speeds at 1.25 times the daytime travel speed.
 McPhee et al.: Off-peak Deliveries in the Greater Toronto Area \ 487

Table 1/Sensitivity Analysis of Off-peak Adoption Rates and Increased Speeds

at Night
Night Proportion Night Speed Day Routes Night Routes Total Routes Trucks
Factor Needed
0 1 31 0 31 31
0.2 1 26 7 33 26
0.3 1 22 10 32 22
0.4 1 18 14 32 18
0.5 1 16 16 32 16
0.6 1 13 20 33 20
0.8 1 7 25 32 25
1 1 0 31 31 31
0 1.25 31 0 31 31
0.2 1.25 26 7 33 26
0.3 1.25 23 9 32 23
0.4 1.25 20 12 32 20
0.5 1.25 16 15 31 16
0.6 1.25 13 19 32 19
0.8 1.25 7 24 31 24
1 1.25 0 29 29 29
0 1.5 31 0 31 31
0.2 1.5 26 7 33 26
0.3 1.5 21 10 31 21
0.4 1.5 19 12 31 19
0.5 1.5 17 14 31 17
0.6 1.5 13 18 31 18
0.8 1.5 7 22 29 22
1 1.5 0 28 28 28

The total number of routes increases for very small adoption rates of
off-peak deliveries, then decreases as the proportion becomes relatively
equal. The total number of routes then slightly increases again as the bal-
ance shifts toward a greater number of off-peak deliveries, and tapers off
as the entire network moves to night deliveries. The result is a distinctive
W-shaped curve for the total number of routes because when only a small
number of sites are either on day routes or night routes the average distance
between locations grows significantly. Figure 2 shows a comparison of the
total number of routes for different relative night-time travel speeds.
It is important to note that this does not necessarily rule out the pos-
sibility of operating off-peak deliveries for low adoption rates. Since
488 / TRANSPORTATION JOURNAL™

35
30
Number of Routes

25
20 Daytime Routes
15 Nighttime Routes
10 Total Routes

5
0
0 0.2 0.4 0.6 0.8 1 1.2
Proportion of Night Deliveries

Figure 1 Number of Routes vs. Proportion of Night Deliveries at 1.25x Speed

34

33

32
Total Routes

31
1.0x Speed
30
1.25x Speed
29 1.5x Speed

28

27
0 0.2 0.4 0.6 0.8 1 1.2
Proportion of Night Deliveries

Figure 2 Total Routes vs. Proportion of Night Deliveries

randomness was incorporated in determining which locations accept night

deliveries, those sites are spread out across the network. It may be cost
effective to select a subset of these locations that are close together for a
night route. Such a decision requires inputs such as a fixed cost of operat-
ing trucks, and variable costs per unit distance traveled, as well as an accu-
rate distance calculation.
The most significant effect of implementing off-peak deliveries is the
increase in truck utilization, which results in a decrease in the total fleet
size. Figure 3 demonstrates the number of trucks needed as a function of
the off-peak adoption rate.
There is clearly strong potential for decreasing fleet size by deliver-
ing off-peak. The number of trucks is at a minimum when close to an
 McPhee et al.: Off-peak Deliveries in the Greater Toronto Area \ 489

equal proportion of locations are on day and night routes. Notably, the
optimal proportion does shift closer to 60 percent as the relative speed at
night increases, since more locations can then be served on a given route.
However, this effect is minimal.
Collectively, results of the analysis indicate that there may be potential
for cost savings if the decrease in fixed cost of operating trucks is greater
than the increase in costs incurred per unit distance travelled. The optimal
proportion, therefore, will vary based on the cost structure used. In gen-
eral, it is necessary to have at least a 25 percent off-peak adoption rate if the
number of routes operated is to decrease. Even so, cost savings will not be
realized if the variable cost multiplied by the mean incremental distance is
greater than the fixed cost per truck multiplied by the average decrease in
number of trucks.
Recently, Nestlé surveyed a sample of approximately 180 customers
in order to grasp their attitudes toward receiving night-time deliveries.
Customers were asked whether they would be willing or able to receive
night-time deliveries between the hours of 4 PM–8 PM, 8 PM–12 AM, and
12 AM–4 AM. In total, approximately 40 percent of those surveyed indicated
that they would be willing or able to receive deliveries during one of those
periods. These preliminary results indicate that the organization is well
within its feasible range for implementing off-peak deliveries.
Overall, our findings show some possible cost savings if Nestle decides
to implement night-time deliveries. The subsections below provide an
overview of suggested measures they (or any organization) should follow

33
31
29
Number of Trucks

27
25
1.0x Speed
23
1.25x Speed
21 1.5x Speed
19
17
15
0 0.2 0.4 0.6 0.8 1 1.2
Proportion of Night Deliveries

Figure 3 Number of Trucks vs. Proportion of Night Deliveries

490 / TRANSPORTATION JOURNAL™

to implement evening routes, a summary of qualitative advice, and areas

for future work.

Steps for Implementation

To determine whether off-peak deliveries are a feasible undertaking, the

following will be required to prepare the organization for implementation:

1. Complete a survey of a sample of key customers to assess the interest

in deliveries at night.
2. Estimate the impact of safety, environment, traffic, and local bylaw
regulations on off-peak deliveries.
3. Create a model to simulate evening routes, and do sensitivity analy-
sis to learn the effect of various factors such as traffic speed and per-
centage of night-time customers.
4. If the above three steps indicate positive results, obtain the network
of customers who are willing and able to accept deliveries at night.
5. Specify the day and night shift hours.
6. Renegotiate contracts with distributors (if applicable).
7. Calculate total anticipated cost versus benefits to obtain final figures.
8. Decide whether to proceed with implementation.
9. Define operational strategy for noise management and warehouse
requirements.

Table 2 provides recommendations for mitigating the potential negative

impacts concerning safety, environment, traffic, and noise.

Future Work

Considerations 1–3 below have been identified for potential future work in
areas where the study scope has been limited due to lack of expert knowledge.

1. Research the impact of night-time deliveries on driver safety: There

is currently limited literature available in this area. A possible future
study could examine such areas as driver fatigue, possible traffic
accidents, and risk of theft. (The latter will likely emphasize theft
of product. Nestlé is now developing a system whereby night-time
transactions will be electronic or prepaid.)
2. Study the environmental consequence of evening deliveries on
emissions: Although there is fairly extensive literature pertain-
ing to the general environmental effects of night-time deliveries,
 McPhee et al.: Off-peak Deliveries in the Greater Toronto Area \ 491

Table 2/Recommendations for the Company in Question Concerning Qualitative

Factors
Area of Impact Recommendations
Safety 1. Operators working at night should minimize the amount of cash that
they carry.
2. The organization should consider implementing a no-cash policy for night-time
deliveries by switching all transactions to electronic, credit or prepaid.
3. To increase product safety, a protocol for delivering product that includes risk
identification and locking doors should be developed.
Environment 4. The company should consider investing in delivery vehicles with low-noise
using sustainable energy when increasing fleet size.
5. Impact on the environment from emissions will not be initially significant;
however, the large-scale adoption of night-time deliveries will reduce the
number of trucks on the road during rush hour periods. If the average vehicle
velocity in the downtown core increases, there will be a large-scale positive
impact on both society and the environment.
Traffic 6. Taffic volumes within the customer data set should be considered when creat-
ing night routes in order to focus concentration in the most congested areas.
7. Additional analysis into the seasonal nature of traffic should be undertaken by
the firm.
Noise 8. The organization should identify and examine the bylaws for its urban center.
9. The company should do an internal study of the noise levels associated with its
equipment and processes to determine whether they are bylaw compliant.
10. The firm should attempt to retrofit current equipment in order to prevent
unnecessarily large capital investments.
11. Drivers should receive training to increase awareness and provide ways for
them to proactively reduce high-noise activities at night.

there is a gap on how to quantitatively determine the environmen-

tal impacts. Sathaye, Harley, and Mandanat (2010) provide a frame-
work for a comprehensive environmental assessment that involves
six steps; these anticipate the ways that a given policy may influ-
ence traffic congestion, and the resultant emissions and pollutant
concentrations.
3. Advocate the advantages to society of evening deliveries: As
congestion in urban areas continues to grow at a rapid rate, alter-
native delivery solutions must be explored. To obtain the maximum
benefit from night-time deliveries, awareness must be raised and a
significant number of organizations need to implement the policy.

Summary and Conclusions

The primary objective of this research was to assess the feasibility of imple-
menting off-peak deliveries in an urban center such as the Greater Toronto
Area (GTA). Working with Nestlé Canada, the authors quantified the
costs and benefits of modifying its ice cream supply chain to incorporate
night-time deliveries, and provided a framework for the regulatory, con-
ceptual, and inertial obstacles to implementation. Using Nestlé’s routing
492 / TRANSPORTATION JOURNAL™

methodologies and GTA customer data set as a case study, the authors
developed customized routing software to determine the proportion of cus-
tomers who must be willing to accept deliveries outside of normal working
hours for the implementation to become financially feasible for Nestlé.
The routing software created applies heuristic methods to achieve
feasible solutions for the Nestlé customer network in the GTA. The soft-
ware logic was validated against Nestlé route engineering methodology,
academic literature, and industry best practice. Analysis of a number of
models, with varying proportions of locations assigned to night routes and
night-time travel speeds, was completed.
Our major findings are contained in the sensitivity analysis of table
1. Whether the proportion of evening deliveries is small or is close to
100 percent, there is typically a 3–10 percent savings in the number of
routes. The percentage saving in the number of trucks is often double that:
Operation of a second shift permits a given vehicle to serve that many more
customers. Note that the range of these percentage improvements holds for
each level of the night speed factor (1.0, 1.25, and 1.5).
The total number of routes (day plus night) for each of the three night
speed factors is given in figure 2 in terms of the proportion of night deliver-
ies. These savings are conservative, perhaps slightly underestimated due to
the random locations chosen for evening routes. We remark that some of
the improvements are achievable not just because routes were moved from
day to night, but may also be a consequence of using automated routing
tools rather than a manual algorithm.
Figure 3 shows that the minimum fleet size occurs when around
50 percent of locations are served by day and night routes, respectively.
(From Nestlé’s perspective, this target is preferable to a much higher
percentage of night-time deliveries.) An implementation plan, in terms of
safety, environment, traffic and noise, for any firm considering a switch
toward evening deliveries, is presented in table 2.
Feasibility for implementing night-time deliveries occurs when the
proportion of customers willing to accept such deliveries is approximately
25–80 percent of all customers. When only a small number of locations are
either on day routes or night routes, the average distance between those
sites grows significantly. Trucks have to travel much further between loca-
tions and they spend more time on line-haul as opposed to making deliver-
ies. Therefore, it may be cost effective to select a subset of close-together
sites for an evening route. Locations in the central business district or
nearby could, thus, form a natural cluster.
 McPhee et al.: Off-peak Deliveries in the Greater Toronto Area \ 493

The most significant effect of operating both night-time and daytime

delivery is the increase in truck utilization, resulting in a decrease in the
required number of vehicles. Collectively, results of the analysis indicate
that there may be potential for cost savings if the decrease in fixed cost of
operating trucks is greater than the increase in costs incurred per unit dis-
tance traveled.
Additionally, the authors conducted a thorough literature review in
order to analyze the qualitative factors that must be considered prior to
implementation. The major concerns addressed in this article include sus-
tainability, regulatory issues, and congestion. GTA-specific recommenda-
tions for the factors analyzed are provided to create a general framework for
ensuring the most effective implementation of night-time deliveries. These
recommendations include the development of a safety plan, procurement
of plate trucks, the addition of noise-abatement techniques, and the imple-
mentation of a noise-monitoring program.
We mention in closing that, while improved performance and
reduced costs accrue to the carrier from night deliveries, the consignees
assume an additional cost to accommodate those deliveries. Hopefully,
the societal benefits will tip the overall balance in favor of the new
schedule.

Note
This article was written as part of a fourth year undergraduate design project in
Management Engineering at the University of Waterloo. The project was conducted by
the first three authors, under the guidance of the fourth.

References
Bjerkan, K. Y., B. J. Sund, and M. E. Nordtømme. 2014. “Stakeholder Responses to
Measures Green and Efficient Urban Freight.” Research in Transportation
Business and Management 11 ( July): 32–42.
Browne, M., J. Allen, S. Anderson, and A. Woodburn. 2006. “Night-time Delivery
Restrictions: A Review.” In Recent Advances in City Logistics, edited by
E. Taniguchi, and R. Thompson, 245–58. Amsterdam: Elsevier.
Campbell, J. F. 1995. “Peak Period Large Truck Restrictions and a Shift to Off-peak
Operations: Impact on Truck Emission and Performance.” Journal of Business
Logistics 16 (2): 227–47.
Cavar, I., Z. Kavran, and N. Jolic. 2011. “Intelligent Transportation System and Night
Delivery Schemes for City Logistics.” Computer Technology and Application
2:782–87.
494 / TRANSPORTATION JOURNAL™

Churchill, J. D. C. 1970. “Operation Moondrop: An Experiment in Out-of-Hours Goods

Delivery.” In The Urban Movement of Goods, Proceedings of the 3rd Technology
Assessment Review, 135–40. Paris: Organization for Economic Cooperation
and Development.
Finlay, H. 2008. “Noise Abatement and Night Deliveries.” Master’s thesis, Paper 31.
Dublin: Dublin Institute of Technology. http://arrow.dit.ie/engmas/31.
Finlay, H., R. Byrne, and J. Grimes. 2008. “Low Noise Solutions for Night Deliveries in
Dublin.” Final Report, Eurocities (EC Network) Working Group on Noise,
Mansion House, Dublin, April. www.eurocities.eu.
Forkert, S., and C. Eichhorn. n.d. “Innovative Approaches in City Logistics: Inner-city
Night Delivery.” Innovative Approaches in City Logistics 7, Niches. http://
www.niches-transport.org/.
Gillett, B., and L. Miller. 1974. “A Heuristic Algorithm for the Vehicle-dispatch Problem.”
Operations Research 22 (2): 340–49.
Han, L. D., S.-H. Chin, O. Franzese, and H. Hwang. 2005. “Estimating the Impact
of Pick-up and Delivery-related Illegal Parking Activities on Traffic.”
Transportation Research Record: Journal of the Transportation Research Board
1906:49–55.
He, Z. 2011. “Farthest-Point Heuristic based Initialization Methods for K-Modes
Clustering.” Department of Computer Science and Engineering, Harbin
Institute of Technology, December. http://arxiv.org/pdf/cs/0610043.pdf.
Holguín-Veras, J. 2006. “Potential for Off-peak Freight Deliveries to Congested Urban
Areas.” Working Paper 294. Albany, NY: Rensselaer Polytechnic Institute.
———. 2008. “Necessary Conditions for Off-hour Deliveries and the Effectiveness of
Urban Freight Road Pricing and Alternative Financial Policies in Competitive
Market.” Transportation Research Part A 42A (2): 392–413.
Holguín-Veras, J., J. Polimeni, B. Cruz, N. Xu, G. List, J. Nordstrom, and J. Haddock. 2005.
“Off-Peak Freight Deliveries—Challenges and Stakeholders’ Perceptions.”
Transportation Research Record: Journal of the Transportation Research Board
1906:42–48.
Holguín-Veras, J., M. Silas, J. Polimeni, and B. Cruz. 2007. “An Investigation on the
Effectiveness of Joint Receiver-Carrier Policies to Increase Truck Traffic in
the Off-peak Hours. Part I: The Behavior of Receivers.” Networks and Spatial
Economics 7 (3): 277–95.
———. 2008. “An Investigation on the Effectiveness of Joint Receiver-Carrier Policies to
Increase Truck Traffic in the Off-peak Hours. Part II: The Behavior of Carriers.”
Networks and Spatial Economics 8 (4): 327–54.
Kanungo, T., D. M. Mountz, N. S. Netanyahux, C. D. Piatko, R. Silverman, and A. Y. Wu.
2002. “A Local Search Approximation Algorithm for k-Means Clustering.”
18th Annual ACM Symposium on Computational Geometry, June. http://www
.cs.umd.edu/~mount/Projects/KMeans/kmlocal-cgta.pdf.
 McPhee et al.: Off-peak Deliveries in the Greater Toronto Area \ 495

Laporte, G., and F. Semet. 2002. “Classical Heuristics for the Capacitated VRP.” In The
Vehicle Routing Problem, edited by P. Toth, and D. Vigo, D., 109–28. Philadelphia:
SIAM.
Marchington, E. 2011. “The 2011 Most Sustainable Cities in Canada.” 2011
Sustainable Cities and Responsible Investing. http://corporateknights.ca/
report/2011-most-sustainable-cities-canada.
Noel, E. C., S. H. Crimmins, N. K. Myers, and P. Ross. 1980. “A Survey of Off-hours
Delivery.” ITE Journal 50 (2): 18–23.
Organization for Economic Growth. 1979. “Requirements and Specifications for Off-
hours Delivery.” Report 123. Organization for Economic Growth. Washington,
DC, 123.
Piecyk, M. I., and A. C. McKinnon. 2010. “Forecasting the Carbon Footprint of Road
Freight Transport in 2020.” International Journal of Production Economics 128 (1):
31–42.
Sathaye, N., R. Harley, and S. Mandanat. 2010. “Unintended Environmental Impacts of
Night-time Freight Logistics Activities.” Transportation Research Part A 44 (8):
642–59.
SenterNovem. 2002. “PEAK, How Innovative Measures Meet Dutch Noise Emission
Standards.” Ministry for Transport and Public Works, Den Haag, Netherlands.
Smerd, J. 2010. “Night-time Truck-delivery Test Called a Success.” Crain’s New York
Business. http://www.crainsnewyork.com/article/20100701/FREE/100709981.
Verlinde, S., C. Macharis, W. Debauche, and A. Heemeryck. 2010. “Night-time Delivery as
a Potential Option in Belgian Urban Distribution: A Stakeholder Approach.”
12th World Conference on Transport Research (WCTR), Lisbon, Portugal.
http://intranet.imet.gr/Portals/0/UsefulDocuments/documents/02535.pdf.
Vilain, P., and P. Wolfrom. 2007. “Value Pricing and Freight Traffic: Issues and Industry
Constraints in Shifting from Peak to Off-peak Movements.” Transportation
Research Record: Journal of the Transportation Research Board 1707/2000:64–72.
Industry Notes

Inland Intermodal Terminals Location Criteria Evaluation:

The Case of Croatia
Violeta Roso
Corresponding author
Chalmers University of Technology
Gothenburg, Sweden
violeta.roso@chalmers.se

Nikolina Brnjac
University of Zagreb
Zagreb, Croatia

Borna Abramovic
University of Zagreb
Zagreb, Croatia

Abstract
Determining a suitable location for an intermodal terminal is a critical
element of the terminal establishment process, a decision on which the
functionality of the entire intermodal freight distribution chain depends.
The purpose of this article is to evaluate the criteria used for deciding on
suitable locations for intermodal terminals in Croatia. First, quality indica-
tors of intermodal terminals were identified: flexibility, safety and security,
reliability, time, and accessibility. Based on these quality indicators, and in
compliance with the European traffic policy and efficient functioning of
intermodal transport, location criteria were formed and evaluated: legis-
lative, environmental, goods flows, spatial, technical-technological, and
organizational. Each criterion is divided further into subcriteria and evalu-
ated using the analytical hierarchy process (AHP) method. The results of
the study show that the criterion of goods flows has the most significant
impact on the selection of the terminal location, followed by the spatial cri-
terion. A position that connects to the European traffic corridor network is
essential for a Croatian terminal’s success.

Transportation Journal, Vol. 54, No. 4, 2015

Copyright © 2015 The Pennsylvania State University,
University Park, PA
 Industry Notes \ 497

Keywords
Intermodal terminal, quality indicators, location criteria, analytic hierar-
chy process, Croatia

Introduction
One of the main objectives of the European Commission’s transport policy
is the development of intermodal transport, since the same is often con-
sidered as a prospective means to alleviate the increasing pressure of road
freight transport and network congestion. However, the competitiveness
of intermodal transport has limiting factors, such as the cost of transship-
ment between transport modes, loss of time and reliability due to these
activities, an increased risk of damage, and loss of control over the ship-
ment that is transferred from operator to operator. Problems not only con-
cern the functioning of the terminal but implementation process as well;
the infrastructure, land use, the environment and regulations are the most
common factors that impede implementation of intermodal terminals
(Roso 2008). One way to tackle intermodal transport issues might be the
construction of new terminals or extending existing intermodal terminals,
with safer and faster freight transshipment, and reorganization of freight
flows in the terminal catchment area.
The role of terminals has increased significantly through the
implementation of intermodal technologies and, over time, terminals have
become increasingly technically and technologically complex systems. The
transportation strategy addresses several questions, such as: which locations
are suitable for new terminals, which type of network is most favorable,
how will they impact the freight flows in the network, and what will be the
attractiveness of the network in comparison to the freight flows. The critical
analysis of the features of the intermodal terminal’s network and logistics
distribution centers should result in the system reorganization concept, that
is, proposal of the methodology of planning and establishing intermodal ter-
minals and logistic distribution centers based on the existing system’s param-
eters, intermodal terminal location criteria, and network evaluation. One of
the most important strategic decisions when implementing an intermodal
terminal concerns its location, and it is essential for both governments and
investors to obtain a well-functioning terminal network. A strategic location
can facilitate transportation, handling, storage, and transshipment of goods
moving in international trade (Regmi and Hanaoka 2013). Many researchers
have covered the strategic or optimal location for intermodal freight facili-
ties from different perspectives, such as Arnold, Peeters, and Thomas (2004),
Macharis and Verbeke (1999), Pekin and Macharis (2007), Racunica and Wynter
498 / TRANSPORTATION JOURNAL™

(2004), and Regmi and Hanaoka (2013). However, research on evaluation of the
criteria used for finding those locations is scarce. Furthermore, according to
Abramović, Lovrić, and Stupalo’s (2012) study, unsatisfactory infrastructure
and unsuitable locations of intermodal terminals are two significant transport
problems in Croatia. Therefore, the purpose of this article is to evaluate
criteria to determine suitable locations for intermodal terminals in Croatia.
To conduct the analysis, the analytical hierarchy process (AHP) method is
applied as an appropriate tool for multicriteria decision-making.

Inland Intermodal Terminals

Intermodal terminals, as important nodes in the transport network, have
gained substantial attention in transportation literature. Considerable
research has been conducted on how to improve the efficiency of road-
rail terminals and, consequently, improve the intermodal transport chain
as a whole (Ballis and Golias 2002, 2004; Kozan 2000; Nierat 1997; Roso,
Woxenius, and Lumsden 2009). Slack (1999), in his research on inland load
centers, shows the importance of the development for intermodal trans-
portation, as well as emphasizes the inland terminal/satellite terminal’s
role in reducing environmental effects, as does Roso (2007). Not only are
transport terminals of importance to the transportation research field,
but they are also at the very center of critical issues in economic, political,
urban, and other geographic subfields (Goetz and Rodrigue 1999).
Modeling container terminals is of importance for many applications,
such as planning and evaluating new facilities, determining the required
resources, improving the operational procedures of existing facilities, and
evaluating alternative equipment systems (Ballis and Abacoumkin 1995).
Many transportation-related models in the literature are used for finding
the optimal location of intermodal terminals, such as in Arnold, Peeters,
and Thomas (2004), Macharis and Verbeke (1999), Racunica and Wynter
(2004), and Regmi and Hanaoka (2013). Pekin and Macharis (2007) use a GIS-
based location analysis model for intermodal terminals, while Kayikci (2010)
applies a fuzzy AHP and artificial neural networks methods in the process
of decision-making to select the most appropriate location for an intermo-
dal freight center. In their survey on localization of freight consolidation
centers, Olsson and Woxenius (2014) investigate the potential to establish
those facilities from an urban transport perspective; whereas Portugal,
Morgado, and Lima (2011) use the AHP method to establish a procedure for
ranking locations that are candidates for cargo terminal facilities.
The basic problem of differentiating between “conventional” ter-
minals and the various types of large-scale intermodal logistics centers
 Industry Notes \ 499

is discussed by Höltgen (1995). Furthermore, the author tries to find a

unique definition for intermodal logistics centers, since the concept varies
from country to country. However, there is a common ground: it should
contribute to intermodal transport, promote regional economic activity,
and improve land use and local goods distribution. The author also dis-
cusses that the complexity of combined (intermodal) transport terminals
due to numerous areas of influence (figure 1) can create bottlenecks in the
freight flow. Even Woxenius (1997) discusses whether or not the terminals
are barriers for intermodality; despite their important role in transport
networks, terminals sometimes impede the development of intermodal
transport with additional transshipment costs at road-rail terminals, or
due to the shippers’ lack of freedom in choosing traffic modes once they
move their business to intermodal freight centers (15).
An ideal intermodal terminal is not a certain physical configuration
of pavement and tracks, but an organization of services integrated with a
physical plant that meets the business needs of a specific marketplace (Bask
et al. 2014; Zimmer 1996). These physical plants may take many forms, and
are influenced by many different factors such as the landscape, proximity
to the seaport or major industrial complex, location relative to the main rail
infrastructure, or distance from the country’s highway network. There is
no general agreement on how to measure terminal performance; however,
overall transit times and, more important, transit-time reliability, are often
cited as key mode choice decisions by customers (Ferreira and Sigut 1995).
According to the authors, there are three main performance areas,
including customer service, operational efficiency, and terminal produc-
tivity. Most performance measures used in terminal monitoring are partial
productivity indicators, with transit times and transit-time reliability being

Figure 1 Combined Transport Terminal Areas of Influence (Höltgen 1995)

500 / TRANSPORTATION JOURNAL™

the key issues (Ferreira and Sigut 1995). Therefore, customer turnaround
times and train departure times are critical indicators of terminal perfor-
mance. From the customers’ perspective, it is also important to be able to
track the shipment. Other terminal performance indicators relate to plant
and human resource productivity and operating costs per unit of output
handled (Ferreira and Sigut 1995). The performance criteria Bontekoning,
Macharis, and Trip (2004) use for evaluating the performance of freight ter-
minals include cost of load unit, train service and sojourn time, utilization
rate of equipment, additional capacity jump, percentage of delayed depar-
ture times, percentage of wrong destinations, and percentage of damaged
load units or rail wagons.

Geographical Coverage
Croatia is the only central European country that neighbors southeastern
Europe, and is also Pannonia-Danube and a Mediterranean country (Steiner
2007). Following its geographical position, Croatia is a country of multi-
directional routes such as the route between western and central Europe,
and the Black Sea region and southeastern Europe. Furthermore, equally
important are the routes from central Europe, the Pannonia and Baltic
regions, and eastern Europe toward the Mediterranean area (figure 2).
Thus, Croatia creates a bridge between western and central Europe with the
Black Sea, and Eastern Europe and the Baltic Sea with the Mediterranean

Figure 2 Croatia in the Network of European Traffic Corridors (Steiner 2007)

 Industry Notes \ 501

(Steiner 2007). Highly important for Croatia are the links with Pan-European
corridors V, VII, and X, and the ports of the Adriatic Basin and the Sava and
Danube rivers. The Adriatic Sea is potentially the shortest and most cost-
effective route that connects Europe with the majority of countries in Asia
and Africa via the Suez Canal.
The North Adriatic ports where the Port of Rijeka is located are called a
multiport gateway region which consists of a broad hinterland that could
be reached by road, rail, or inland waterways and, as such, offers good
inland access to and from the ports, bringing them a certain competitive
advantage (Notteboom 2009). Unfortunately, due to poorly implemented
transport policies, Croatia has a very immature intermodal transport sys-
tem. It was developed without a clearly defined strategy and each mode of
traffic has been considered separately, resulting in heavy investment in
the road sector, while rail and inland waterways were left largely neglected
(Steiner 2007). Consequently, no proper network of intermodal terminals
exists, which should be the key nodes in this system. The main issues that
the terminals face are inadequate infrastructures, long waiting times caus-
ing delays, high terminal costs, and unexpected problems. Thus, termi-
nals are often congested for extended periods of time, resulting in a low
quality of services and price increases for customers (Abramović, Lovrić,
and Stupalo 2012). Recently, some actions were taken toward development
of intermodal transport, particularly the Port of Rijeka’s transport system
(Steiner 2007). Still, there is huge potential and need in terms of the devel-
opment of intermodal transport in Croatia.

Data Collection and AHP

Data was collected through face-to-face interviews during multiple site vis-
its and through video conferences, as well as by a questionnaire sent by e-mail
and conventional mail to relevant governmental institutions and actors of
the transport system in Croatia. A total of 20 respondents were included in
the study, such as the Ministry of Construction and Physical Planning; the
Ministry of Maritime Affairs, Transport and Infrastructure—Department
for Road, Department for Rail and Department for Inland Waterways;
Croatian Railways Cargo; Croatian Railways Infrastructure; AGIT (inter-
modal logistics operator); RCA Rail Cargo Austria-Expressinterfracht
Croatia (intermodal logistics operator); KT Zagreb (terminal operator), Port
of Rijeka; Port of Ploče; Trast (logistics service provider); and CMA CGM
Croatia.
The AHP is a suitable method for solving complex problems that
contain several alternatives and decision-making criteria (Forman and
502 / TRANSPORTATION JOURNAL™

Peniwati 1998; Harker and Vargas 1987). The AHP shows good practical
characteristics that are of crucial importance to decision-making on cer-
tain kinds of infrastructure implementations (Forman and Peniwati 1998).
The multicriteria decision-making model is based on the optimization
of the function of an objective on a set of possible solutions. The process
enables the decision-makers to set the priorities and make decisions for
the case when it is necessary to take into consideration both quantitative
and qualitative characteristics. The complex decision-making process is
reduced to a number of single comparisons between the set goals and crite-
ria, which allows full insight into the decision-making process in order to
select the best scenario (Yang and Lee 1997). Qualitative analysis is based on
the decision-maker’s assessments, experience, and intuition. A significant
feature of the AHP is to quantify decision-makers’ subjective judgments
by assigning corresponding numerical values based on the relative impor-
tance of factors under consideration (Yang and Lee 1997).

Intermodal Terminals’ Quality Performance Indicators

Croatian intermodal terminals have not been subjected to any specific,
independent quality audit, nor any control that would allow customers to
objectively consider their performances (Steiner 2007). The criteria and the
quality standards are lacking to a great extent. In order to increase the qual-
ity and efficiency of terminal services, it is necessary to define and identify
the criteria for the planning of intermodal terminals, as well as to define
the quality indicators and quality designation in accordance with European
evaluation system of the quality of terminals. The quality indicators
have been structured according to different aspects of quality (European
Commission 2000, 2005):

Reliability
Punctuality
Flexibility
Frequency of services
Accessibility
Capacity
Availability of professional human resources
Services with added value
Level of safety and security of intermodal transport

Services provided by intermodal terminals differ in their quality and

vary over time, thus affecting a change in price for the realized service.
 Industry Notes \ 503

Continuous changes in quality and service price stipulate the possibility of

selection for the customers, and the need for the service providers to adapt
to the new situations, which require the existence of a dynamic quality sys-
tem. The establishment of a dynamic quality system insures advantages for
both the service providers and for the users. The service providers can opti-
mize the usage of their resources and the service users can select how much
they want to pay for the provided services for which the service providers
have to offer a guaranteed level of quality.
A terminal is a system whose functioning requires full integration into
the logistics chain in order to fulfill its function completely. The efficiency
and quality of service provided by the intermodal terminal requires an ade-
quate infrastructure, adequate connections with other transport modes,
motivated management, and qualified employees. The quality, efficiency,
and price of terminal services are factors that affect the terminal com-
petitiveness, together with the quality of its access and road/rail interface
(Roso 2008). The latter depends on the behavior of a large variety of actors,
such as terminal operators, freight forwarders, and/or transport operators
(Abramović, Lovrić, and Stupalo 2012; Roso 2008). The identification and
recognition of quality as a standard and, at the same time, a specific char-
acteristic of the provided service, contributes to better positioning of an
intermodal terminal on the market of terminal services, and the creation
of the important assumption of the terminal’s integration and competi-
tiveness in the national and international frames. Users’ satisfaction with
the provided service at the terminal is directly related to their needs and
requirements, and the perception of the service is the most important indi-
cator used for this measurement.

Criteria in Use for Determining the Location of the Intermodal Terminal

Based on the existing literature (e.g., Bontekoning, Macharis, and Trip
2004; Brnjac 2009; Brnjac and Ćavar 2009; Ferreira and Sigut 1995; and
Organisation for Economic Co-operation and Development 2002;), which
describes and evaluates the aspects of quality and productivity of termi-
nals, the necessary criteria to define the intermodal terminal’s network were
selected. In the selection of the criteria, several sources from actual studies
and the literature were analyzed, and a list of the criteria and their subcrite-
ria was created. Based on the critical analysis and the method of surveying
traffic experts (road, rail, water, and air traffic), the relevant criteria and
subcriteria were then selected. The recognized criteria include legislative,
environmental, goods flows, spatial, technical-technological, and organi-
zational. The legislative (table 1) and environmental criteria (table 2) are not
504 / TRANSPORTATION JOURNAL™

evaluated in parallel with the others, since they depend on the legislation of
individual states. Thus, it is first necessary to check whether certain loca-
tions meet the required legislative and environmental criteria to establish
intermodal terminals in these areas. The legislative criteria of the Republic
of Croatia include the following items: the Act on Physical Organization and
Construction (NN 76/07), physical planning strategy and program, physical
plans of the counties, city and districts, and the implementation documen-
tation (urban plan and detailed organization plan). The legislative criteria
attract flows or stipulate bans on land use for terminal development. The
environmental criteria include the Act on Environmental Protection (NN
110/07) and the environmental impact assessment, which includes two doc-
uments, the first being the strategic document of 2007, and the second the
classical assessment of environmental impact. The environmental criteria
significantly impact on the selection of the location for the intermodal ter-
minals, especially in the case of goods that can adversely affect the human
environment.
Goods flows (table 3), spatial (table 4), technical-technological (table 5),
and organizational criteria (table 6) are all further divided into relevant sub-
criteria, that is, structured into a two-level hierarchy. The organizational
criteria have been considered last, since they reflect the characteristics of
the system formed by the integration of all aspects of performance and
contain the system performances. The subcriteria of the organizational cri-
terion reflect the organization of the terminal defined by technical and tech-
nological criteria, which are used to realize the given system performances.

Table 1/Legislative Criteria

L1 Act on physical design and construction NN76/07
L2 Strategy of physical arrangement and program 1999
L3 Physical plan of the county and city
L4 Physical plan of single districts and city
L5 Urban physical plan (implementation document)
L6 Detailed physical plan (implementation document)

Table 2/Environmental Criteria

E1 Act on the Environment
E2 Assessment of impact on the environment (2 types)
1. Strategic (since 2007)
2. Classical assessment of impact on the environment
 Industry Notes \ 505

They are defined as quality indicators, terminal performance indicators,

and the development level of the logistics structures. The characteristics of
an efficient terminal are a location conditioned by freight flows, access to
road and rail traffic routes, accompanying service activities, and technical
equipment. The modifications of the existing terminals are reflected in
their technical, technological, and environmental possibilities. The identi-
fied criteria are applicable for modeling in the region: in the narrow sense
of the Republic of Croatia, and in a wider sense with slight modifications,
the legislative and environmental criteria to any region.

Table 3/Criteria of Goods Flows

RT1 Goods flows in the catchment area of the terminal
RT2 Future goods flows in the catchment area of the terminal
RT3 Possibility of getting a major status in the network
RT4 Power of the economic sector in the catchment area of the terminal
RT5 Flows that are created in the catchment area of the terminal
RT6 International freight flows that pass through the area of the terminal (tonnes per year)
RT7 Ratio of total flows rail / road
RT8 Free zones (SZ) in the catchment area of the terminal
RT9 Safety of planning (SZ) – planning (is it approved for implementation, and in compliance
with national and regional planning?)
RT10 Ownership status / availability (public or private, one or several owners)

Table 4/Spatial Criteria

Pk1 Distance from main industrial zones (km)
Pk2 Distance from river ports (km)
Pk3 Distance from seaports
Pk4 Distance from airports
Pk5 Distance from railway stations
Pk6 Distance from transport and rail companies
Pk7 Accessibility to municipal and commercial centers
Pk8 Distance from agricultural centers
Pk9 Possibility of expanding the area considering future requirements

Table 5/Technical and Technological Criteria

T1 Connection of the terminal with roads significant for international Terminal connections
traffic
T2 Connection of the terminal with roads significant for regional
traffic

(Continues)
506 / TRANSPORTATION JOURNAL™

Table 5/ Technical and Technological Criteria (Continued )

T3 Connection of the terminal with roads significant for local traffic
T4 Connection of the terminal with railway lines significant for
international traffic
T5 Connection of the terminal with railway lines significant for
regional traffic
T6 Connection of the terminal with railway lines significant for local
traffic
T7 Connection with seaports
T8 Connection of the terminal with short sea shipping centers
T9 Connection with airports traffic hubs
Z1 Interoperability Rail network charac-
teristics
Z2 Number of tracks on rail terminals
Z3 Free profile for passage of rail vehicles
Z4 Maximum permitted mass of rail vehicles
Z5 Maximum longitudinal gradient of railway line
Z6 Designed maximum speed of trains
Z7 Minimal length of the main railway tracks at railway stations
I1 Categorization in compliance with European standards IW network charac-
teristics
I2 Level of designation of waterways
O1 Sufficiency of terminal capacities Equipment and activi-
ties at the terminal
O2 Development level of infrastructural network (electricity, water,
sewage . . .)
O3 Defined transshipping and warehouse activities

Table 6/Organizational Criteria

Q1 Time of freight turnover Quality indicators
Q2 Reliability
Q3 Availability
Q4 Flexibility
Q5 Safety and security
PO1 Terminal productivity Terminal performanc-
es indicators
PO2 Productivity of human power
PO3 Consumption of energy per transhipping unit
LS1 Frequency of bottleneck occurrence Development of logis-
tic structures
LS2 Presence of logistic operators
LS3 Support of development in the form of national strategic docu-
ments
 Industry Notes \ 507

Analysis of the Criteria Evaluation

The results of the AHP application show that the criterion of goods flows has
been rated as the most important one, at 56.9 percent of our four main crite-
ria (figure 3). The criterion of goods flows represents the most reliable start-
ing point for the analysis and forecast of the transport of goods—demand on
a traffic route. Regarding the total volume of flows, their structure and orien-
tation have been imposed as imperative and represent the basis in defining
and selecting the locations of intermodal terminals. Huge goods flows meet
at intermodal terminals and pass across them. The goods are transshipped
at the terminals, and every part of these goods flows must be manipulated
at least once. All the activities and all the terminal subsystems are in the
function of the goods flows. Knowledge of the structure and characteristics
of flows that pass through the intermodal terminal is necessary for all the
activities of planning, management, control, and analysis of the system and
process at the terminal. The characteristics of the requirements of the goods
flows have been considered through the structure and characteristics of the
users and the flows they generate by the requirements of individual tech-
nologies of transport chains that are directed to intermodal terminals.
The next most important criterion is spatial with 28.2 percent
(figure 3). The physical selection of the location of the terminal is of utmost
importance. The terminals require a good location and connection to the
transport network (traffic corridors). The location of the intermodal termi-
nal has to be in accordance with the users’ needs—terminals and economic
systems in a narrower and wider area. It has to be at a location with a strong
economy—a logistics environment that will attract goods, transport flows,
and all the other accompanying activities.

Figure 3 Evaluation of Four Main Criteria: Goods Flows, Spatial, Technical-

Technological, and Organizational
508 / TRANSPORTATION JOURNAL™

The importance of the technical and technological criteria accounted for

10.3 percent (figure 3). These are divided into four subgroups (connection
of terminals, characteristics of railway connection, characteristics of the
network of sea paths, and equipment and activities at the terminal). The
connection of the terminal allows consideration of the advantages and
drawbacks of potential locations of intermodal terminals by means of
the connections of terminals with the roads, rail lines of regional/local/
European significance, and the ports and their infrastructure. The subcri-
terion of the railway network characteristic is structured as the number of
railway tracks at the terminal, free profile, greatest allowed mass, longitu-
dinal gradient, vehicle velocity, and length of the main tracks at the rail-
way stations. The free profile represents the end limitation of the space in
the cross-section in relation to the track axis. The velocity is the exclusive
responsibility of the railway, since it is the consequence of its technical
competence. Higher speed means better competitive competence of the
railway in the transport market since less time is spent on traveling. On
the other hand, for the railway, higher commercial speed means a shorter
time for railcar turnover, which also means higher productivity of railcars.
The number of tracks allows for more rational organization of trains at the
railway station. Apart from the number of tracks, their length is also very
important for the traffic of trains. Thus, regarding the track’s length, it is
also possible to organize crossing and overtaking. The relevant track gra-
dient for braking is the value of the longitudinal gradient, which is used
to determine the braking percentage, that is, the necessary braking mass
of the trains on a railway line, that is, the line section. The subcriterion of
waterway network characteristic is a categorization in compliance with the
European standards, since the Republic of Croatia has committed to con-
struct and organize waterways open for international navigation in com-
pliance with the stipulated standards according to navigability categories
from IV to Vb, as well as maintaining an adequate level of designation of
the waterways. The subcriterion of equipment and activity at the terminal
refers to the sufficiency of the terminal capacity, development level of the
terminal infrastructural network, and defined transshipment and ware-
house activities. The terminal capacity sufficiency stipulates its future
development and operation. The development of an infrastructural net-
work includes land areas for servicing the needs of the terminal, areas for
the supply and servicing installations, areas for commercial activities, as
well as areas for terminal industry.
The last criterion, regarding importance, is the organizational criterion,
with 4.6 percent (figure 3). It consists of the following subcriteria: quality
indicators, terminal performance indicators, and the development of the
 Industry Notes \ 509

logistics structures. The quality indicators of the intermodal terminal are

reflected through the time of goods turnover, reliability, availability, flex-
ibility, and safety and security. The terminal performance indicators are
reflected through terminal productivity, productivity of human resources,
and consumption of energy per transshipment unit. The development
of logistics structures consists of bottlenecks, logistics operators, and
national strategic documents. The presence of strong economic subjects
from the field of logistics, management, and information technologies
represents a successful development of the intermodal terminal as well as
quality and terminal performance indicators.
Regarding the goods-flows subcriteria (figure 4), the highest impor-
tance for determining the location has the goods flows in the catchment
area of the terminal with 19.7 percent, followed by the economic sector at
17.9 percent, stipulated by the development of the hinterland. The interna-
tional freight flows account for 16.1 percent, resulting from the current sit-
uation that has minor importance as well as potential future freight flows
and flows that are formed in the catchment area at 10.2 percent. The rest of
the subcriteria are of almost equal importance.
Analysis of the spatial criterion (figure 5) shows that the subcriterion
distance from the main industrial zones has the highest importance, with
18.1 percent. The next highest–ranked subcriteria are the distance from rail-
way stations with 17.4 percent and seaports with 16.4 percent. Criterion in
relation to future requirements for the development of intermodal terminals
is the possibility of expanding the area with 14 percent, and the distance
from transport and railway companies with 12 percent. The least important

Figure 4 Goods Flows Criteria

510 / TRANSPORTATION JOURNAL™

subcriterion, probably due to the insufficient development of intermodal

transport, is the distance from airports with 2.4 percent.
To facilitate evaluation of the technical and technological criteria, the
same has been divided into terminal connections, railway network char-
acteristics, inland waterway network characteristics, and equipment and
activities at the terminal. Their evaluation has resulted in the following
values: terminal connections as the most important subcriterion with as
much as 43.3 percent, railway network characteristics with 30.8 percent,
waterway network characteristics with 16.5 percent, and, as the last subcri-
terion, equipment and activities at the terminal with 9.4 percent (figure 6).
Terminal equipment as the technical and technological criterion rep-
resents the factor of the development of intermodal terminals of minor
importance, since greater intermodal terminal turnover generates higher
investments and specialization. Modern equipment of sufficient capacity
ensures a continuous flow of freight through the intermodal terminal, but
the terminal development depends primarily on the geo-traffic position of
the terminal (position of the terminal in relation to the traffic infrastructure).
Figure 7 shows the importance of different subcriteria belonging to the ter-
minal connections subdivision of the technical and technological criterion,
where the subcriterion, terminal connection by road significant for interna-
tional traffic, shows the highest importance, 21.9 percent.

Figure 5 Spatial Criteria

 Industry Notes \ 511

The organizational criterion is divided into terminal performance

indicators, quality indicators, and the development of logistics structures.
The study identified terminal performance indicators with 55.8 percent as

Figure 6 Technical and Technological Criteria Division into Four Subcriteria

Figure 7 Technical and Technological Criteria-Terminal Connections

512 / TRANSPORTATION JOURNAL™

the most important indicator in this category followed by quality indicators

with 32 percent, and the development of logistics structures accounting for
12.2 percent (see figure 8).
Terminal performances interweave with the terminal quality indica-
tors. Terminal performances and quality indicators depend on the volume
of traffic load and the variability of arrivals of the customers and transport
means. The demand for transport services depends on the development of all
transport modes, their integration in an integral whole, pricing of transport
modes, and the range and quality of services that are performed for poten-
tial customers. Fast, efficient, timely, and high-quality transport of freight
will contribute to the generation of added values and economic growth. In
determining the importance of the quality indicators, the highest importance
belongs to flexibility with 39 percent, safety and security with 25.5 percent,
reliability with 19.2 percent, time with 9.7 percent, and availability with
6.6 percent. The flexibility allows optimal flow of freight and productivity of a
certain terminal. The quality of services at terminals and the efficiency of the

Figure 8 Organizational Criteria Divided into Three Subcriteria

Figure 9 Organizational Criteria-Subcriteria on Terminal Performance Indicators

 Industry Notes \ 513

process, that is, input-output procedures, transshipping, warehousing, and

loading/unloading, significantly affect the quality and the costs of the whole
intermodal transport chain. Figure 9 shows the importance of the terminal
performance indicators subcriteria; the subcriterion of terminal productivity
is recognized as the most important with 58.4 percent.

Conclusion
The results of the research show that traffic characteristics have a significant
impact on the selection of terminal location. A good position and connection
to the European traffic corridor network is essential for a terminal’s
success in Croatia. Considered from the regional economic point of view,
an intermodal terminal on a traffic corridor has development priority.
Intermodal terminals of European significance ensure international access
to the entire European intermodal network of high performances.
Intermodal transport requires cooperation between different actors of
the system under the condition of different activities across the entire geo-
graphic region. Intermodal terminals and interface points are dependently
connected with different modes of transportation. The criteria for defining
the network of intermodal terminals and their efficiency in the processes, for
example, approach procedures, transshipment, warehousing, and loading/
unloading, have great impact on the quality and price of the entire process
of intermodal transport. The criteria of the goods flows and the spatial cri-
teria are crucial for the selection of the terminal location. Within the frames
of the organizational criterion, the quality of the intermodal system ser-
vice is reflected, and it is structured per the indicators of quality, terminal
performances and the development of logistics structures. The terminal
performance indicators have a great impact on the quality and price of trans-
port and on the entire service from the sender to the receiver. The prices of
transshipment at terminals usually do not cover all costs and, therefore, the
remaining difference in price is realized from the funds of the European
countries and very widely spread government funds. Thus, it is in the
interests of the European national governments that intermodal terminals
funded by the government successfully operate with high-quality services.
The main issues faced by Croatia’s intermodal terminals are inad-
equate terminal infrastructures and equipment, or unsuitable locations.
Consequently, there are issues related to long waiting times, reliability in the
preparation when loading units, access problems, damaging loading units,
and paperwork and communication problems between the engaged trans-
port companies. The analysis of the criteria and subcriteria and the imple-
mentation of modeling methods have determined the suppositions for the
construction of new or the extension of existing intermodal terminals.
514 / TRANSPORTATION JOURNAL™

References
Abramović, B., I. Lovrić, and V. Stupalo. 2012. “Analysis of Intermodal Terminals Service
Quality in the Republic of Croatia.” Promet—Traffic and Transportation 24 (3):
253–60.
Arnold, P., D. Peeters, and I. Thomas. 2004. “Modelling a Rail/Road Intermodal
Transportation System.” Transportation Research E 40:255–70.
Ballis, A., and C. Abacoumkin. 1995. “A Container Terminal Simulation Model with
Animation Capabilities.” Journal of Advanced Transportation 30 (1): 37–57.
Ballis, A., and J. Golias. 2002. “Comparative Evaluation of Existing and Innovative Rail-
Road Freight Transport Terminals.” Transportation Research Part A: Policy and
Practice 36 (7): 593–611.
———. 2004. “Towards the Improvement of a Combined Transport Chain Performance.”
European Journal of Operational Research 152:420–36.
Bask, A., V. Roso, D. Andersson, and E. Hämäläinen. 2014. “Development of Seaport-Dry
Port Dyads: Two Cases from Northern Europe.” Journal of Transport Geography
39:85–95.
Bontekoning, Y. M., C. Macharis, and J. Trip. 2004. “Is a New Applied Transportation
Research Field Emerging? A Review of Intermodal Rail-Truck Freight
Transport Literature.” Transportation Research: Part A 38:1–34.
Brnjac, N. 2009. “Identification of Relevant Criteria for Defining a Network of Intermodal
Terminals.” Doctoral diss., University of Zagreb.
Brnjac, N., and I. Ćavar. 2009. “Example of Positioning Intermodal Terminals on Inland
Waterways.” Promet—Traffic and Transportation 21 (6): 433–39.
European Commission. 2000. “IQ—Intermodal Quality Final Report for Publication.”
Transport RTD Programme of the 4th Framework Programme—Integrated
Transport Chain.
———. 2005. “Integrated Services in the Intermodal Chain (ISIC)” Final Report Task D:
Improving Quality of Intermodal Terminals.
Ferreira, L., and J. Sigut. 1995. “Measuring the Performance of Intermodal Freight
Terminals.” Transportation Planning and Technology 17 (3): 268–79.
Forman, E., and K. Peniwati. 1998. “Aggregating Individual Judgments and Priorities
with the Analytic Hierarchy Process.” European Journal of Research 108:165–69.
Goetz, A. R., and J.-P. Rodrigue. 1999. “Transport Terminals: New Perspective.” Journal of
Transport Geography 7:237–40.
Harker, P. T., and L. G. Vargas. 1987. “The Theory of Ratio Scale Estimation: Saaty’s
Analytic Hierarchy Process.” Management Science 33 (1): 1,383–403.
Höltgen, D. 1995. “Terminals, Intermodal Logistics Centres and European Infrastructure
Policy.” Doctoral diss., European Centre for Infrastructure Studies.
Kayikci, Y. 2010. “A Conceptual Model for Intermodal Freight Logistics Centre Location
Decisions.” The Sixth International Conference on City Logistics, Procedia
Social and Behavioral Sciences 2 (6): 297–311.
Kozan, E. 2000. “Optimizing Container Transfers at Multimodal Terminals.” Mathematical
and Computer Modeling 31:235–43.
 Industry Notes \ 515

Macharis, C., and A. Verbeke. 1999. “The Optimal Location of Intermodal Terminals.”
NECTAR Conference, Delft.
Nierat, P. 1997. “Market Area of Rail-Truck Terminals: Pertinence of the Spatial Theory.”
Transportation Research: Part A 31 (2): 109–27.
Notteboom, T. 2009. “Economic Analysis of the European Seaport System.” Report
Serving as Input of the Discussion on the TEN-T Policy. Antwerp, ITMMA—
University of Antwerp.
Olsson, J., and J. Woxenius. 2014. “Localisation of Freight Consolidation Centres Serving
Small Road Hauliers in a Wider Urban Area: Barriers for More Efficient Freight
Deliveries in Gothenburg.” Journal of Transport Geography 34:25–33.
Organisation for Economic Co-operation and Development. 2002. “Benchmarking
Intermodal Freight Transport.” Paris: OECD. http://www.internationaltrans-
portforum.org/pub/pdf/02BenchmarkingE.pdf.
Pekin, E., and C. Macharis. 2007. “A GIS-Based Location Analysis Model for Intermodal
Terminals.” Presented at the International Logistics and Supply Chain
Congress 2007, November 8–9, Istanbul, Turkey.
Portugal, L. da S., A. V. Morgado, and O. J. R. Lima. 2011. “Location of Cargo Terminals
in Metropolitan Areas of Developing Countries: The Brazilian Case.” Journal of
Transport Geography 19:900–10.
Racunica, I., and L. Wynter. 2004. “Optimal Location of Intermodal Freight Hubs.”
Transportation Research Part B: Methodological 39 (5): 453–77.
Regmi, M. B., and S. Hanaoka. 2013. “Location Analysis of Logistics Centers in Laos.”
International Journal of Logistics Research and Applications: A Leading Journal of
Supply Chain Management 16 (3): 227–42.
Roso, V. 2007. “Evaluation of the Dry Port Concept from an Environmental Perspective.”
Transportation Research Part D 12:523–27.
———. 2008. “Factors Influencing Implementation of a Dry Port.” International Journal of
Physical Distribution and Logistics Management 38:782–98.
Roso, V., J. Woxenius, and K. Lumsden. 2009. “The Dry Port Concept: Connecting
Container Seaports with the Hinterland.” Journal of Transport Geography
17:338–45.
Slack, B. 1999. “Satellite Terminals: A Local Solution to Hub Congestion?” Journal of
Transport Geography 7:241–46.
Steiner, S. 2007. “Valorizacija Prometnog Sustava Hrvatske.” Research report. Hrvatska
akademija znanosti i umjetnosti, Znanstveno vijeza promet.
Woxenius, J. 1997. “Terminals—A Barrier for Intermodality?” Nordic Transport Research
Conference on Intermodal Freight Transport, September 22–23, Ebeltolft,
Denmark.
Yang J., and H. Lee. 1997. “An AHP Decision Model for Facility Location Selection.”
Facilities 15 (9/10): 241–54.
Zimmer, N. R. 1996. “Designing Intermodal Terminals for Efficiency.” Transportation
Research Circular 459:99–109.
516 / TRANSPORTATION JOURNAL™

Safety Culture from an Interdisciplinary Perspective:

Conceptualizing a Hierarchical Feedback-based
Transportation Framework
Anjala S. Krishen
Corresponding Author
University of Nevada Las Vegas
anjala.krishen@unlv.edu

Pushkin Kachroo
University of Nevada Las Vegas

Shaurya Agarwal
University of Nevada Las Vegas

S. Shankar Sastry
University of California, Berkeley

Masha Wilson
MBW & Associates, LLC

Abstract
Given the colossal issue of traffic accidents causing serious injuries, fatalities,
and traffic congestion in the United States, the current article introduces a
feedback safety model that utilizes and combines disparate data sources and
allows for model-based action to alleviate the problems. This feedback model
will serve two purposes: (1) it will provide an overarching view of safety cul-
ture that incorporates fundamentally distinct data sources; and (2) it will aid
in the development of targeted messages and control actions at various levels,
especially to the highest-risk public segments, which can highlight the risks
of unsafe driving as well as increase perceived importance for traffic safety.
In order to develop this model, the article begins with the issue of distracted
driving in the main study. The basic premise of the argument is that drivers
who are distracted are more likely to be unsafe (i.e., the behavior is normally
augmentative) and that safety is a culture, not a concept. In essence, curbing
distracted driving should lower vehicle-based accidental deaths. We will then
argue for an overarching safety culture framework to lower distracted driv-
ing from a holistic perspective through gathering multiple data sources and
developing targeted communication mechanisms.

Keywords
Safety culture, feedback control, hierarchical model, traffic safety, trans-
portation framework
 Industry Notes \ 517

Introduction
After six consecutive years of declining motor vehicle crash-related
fatalities, their number increased by 3.3 percent in 2012, from 32,479
fatalities in 2011 to 33,561 in 2012 (National Highway Traffic Safety
Administration 2013). The number of traffic-related injuries saw a simi-
lar increase from 2.22 million in 2011 to 2.36 million in 2012; additionally,
there was a 3.1 percent increase in the total number of police-reported
traffic crashes between 2011 and 2012. High-risk behavior and high-speed
mobility combine to create this externality of the automobile industry
(Williams and Haworth 2007). To make matters worse, when individuals
are faced with highly familiar activities such as driving, research shows
that they tend to underestimate the possibility of a negative outcome
(Douglas 1985).
Today, sustainable and reliable transportation development is one
of the most important and essential investments made by any govern-
ment. Transportation and infrastructure affect the financial and social
life of both individuals and society as a whole and plays a critical role in
the development of a society’s overall economy (Shang, Tjader, and Ding
2004). Our research premise is that to promote a transportation safety cul-
ture, there must be a feedback loop between infrastructure investments
and ideas and individual transportation safety information. Improvement
projects should be based on safety information that, while it may be cur-
rently collected, does not often serve a purpose in terms of planning appro-
priate and solution-oriented approaches to infrastructural modifications.
Unfortunately, safety of transportation systems is one of the major chal-
lenges today. Some of the characteristics that are traditionally related to
safety of any system include guidelines, audits, training programs, qual-
ity checks and assurances, and operating procedures. However, research
indicates that the safety of a large and interconnected system such as a
transportation network may not only depend upon direct factors but also
on indirect factors such as organizational design and practices. Hence,
research exploring the relationships between safety, workplace organiza-
tion, and technology use in the transportation systems context is needed.
Thus, safety as a cultural entity in transportation systems must be studied
and researched.
Given the issue of traffic accidents causing serious injuries, fatalities,
and traffic congestion in the United States, the current article introduces
a feedback safety model that utilizes and combines disparate data sources
and allows for model-based action to alleviate the problems. This feedback
518 / TRANSPORTATION JOURNAL™

model will serve two purposes: (1) it will provide an overarching view of
safety culture that incorporates fundamentally distinct data sources; and
(2) it will aid in the development of targeted messages and control actions
at various levels, especially to the highest-risk public segments, which can
highlight the risks of unsafe driving as well as increase perceived impor-
tance for traffic safety. In order to develop this model, the article begins
with the issue of distracted driving in the main study. The basic premise of
the argument is that drivers who are distracted are more likely to be unsafe
(i.e., the behavior is normally augmentative) and that safety is a culture,
not a concept. In essence, curbing distracted driving should lower vehi-
cle-based accidental deaths. We will then argue for an overarching safety
culture framework to lower distracted driving from a holistic perspective
through gathering multiple data sources and developing targeted commu-
nication mechanisms. Research shows that targeted communications of
social marketing causes can be very effective, such as health-related inter-
ventions (Block and Keller 1995) or taxation perceptions (Krishen, Raschke,
and Mejza 2010).
This article proposes a new unique methodology to study and quan-
tify safety culture in various types of transportation agencies. In this article
safety culture is treated as a latent variable, and then a feedback-hierarchi-
cal-multilevel model shows that safety culture can be analyzed and quanti-
fied at various levels. The main contribution of this article is that it uses
latent variables in a hierarchical setting to utilize big data and build mod-
els at different levels. Different layers of these models are then abstracted
together at a higher level. Salient features of the proposed model are as
follows:

Different layers of modeling

Feedback mechanism
Comprehensive multiscale modeling
Use of latent variables
Use of dynamic as well as static models in a single framework

Previous studies that have investigated this problem are largely

based upon a single model approach. One such study was conducted by
the Federal Motor Carrier Safety Administration (Pratt 2003) to identify
an hours-of-service focus on establishing certain regulations on com-
mercial drivers. Another study based on a single-model approach on
commercial truck drivers and safety-related issues was performed by
 Industry Notes \ 519

(Arboleda et al. 2003). These examples represent just part of an overall

safety culture assessment framework proposed in our article. Another
piece could be the effectiveness of seat-belt campaigns organized by vari-
ous DOTs and the effect of seat-belt usage in injury reduction (Agarwal
et al. 2013). All of these heterogeneous data could be fused together and
modeled in a hierarchical fashion to assess the safety culture in each
organization and provide an overall industry view. Hence our method-
ology is inherently different from the ones proposed earlier and a novel
contribution to the field.

Literature Review

Safety Culture

The concept of safety culture originated in its application to organizations;

research shows that both individual and workplace environmental com-
ponents serve as important factors. Watson et al. (2005) show that trust,
safety norms, management safety values, safe work environments, and
at-risk behavior all factor into increases in safety culture. In an attempt
to curb vehicle-related accidental deaths, several governmental organiza-
tions are reaching out to academia in search of interdisciplinary frame-
works to explore safety culture as it relates to transportation (Institute of
Transportation Engineers 2013). For example, the American Automobile
Association (2007) introduced a traffic safety culture index (also often
named the safety culture climate) and through surveys of the general pub-
lic concluded that several important factors (many of which are based on
driver attitudes) contribute to lack of safe behavior. Girasek (2012) argues
that traffic safety culture has five major components, namely priority, dis-
satisfaction, accountability, engagement, and social norms. Developed
through a driver survey, her modified model consists of five key individ-
ual items: (1) Traffic safety is valued; (2) individuals engage in behaviors
that promote traffic safety; (3) policies that promote traffic safety receive
broad public support; (4) traffic safety is monitored; and (5) social institu-
tions are held accountable for traffic safety. A subsequent study finds that
gaining public support for traffic safety remains a cultural challenge in
the United States (Girasek 2013). In essence, safety culture as a measur-
able construct is still in its infancy (Naevestad and Bjornskau 2012). These
authors also indicate that unsafe driving practices occur more in younger
drivers, and that demographic factors are found to be predictive of dis-
tracted driving also.
520 / TRANSPORTATION JOURNAL™

Recently, organizations and governing bodies are beginning to focus

on the concept and terminology surrounding the definition and embodi-
ment of “safety culture,” using this term to capture the essence of safety
as a whole in a system. This idea alludes to a shift from focusing on indi-
viduals and direct accident-causing factors to organizational and indirect
factors related to safety, a movement from studying the problem as a col-
lection of issues to studying it from a systematic perspective. The term
“safety culture” was coined in the aftermath of the Chernobyl disaster. As
such, safety culture is a novel way of formalizing the processes involving
risk management and mitigation in organizational and other complex sys-
tem networks. Safety culture research highlights some of the behavioral
practices that might lead to accidents in high-risk sociotechnical systems.
Such collective research might also prove to be useful in the development
of risk mitigation strategies and serve as a complement to the current safety
assessment practices (Pidgeon 1991).
In general, large organizations are complex and difficult to under-
stand, and are therefore more prone to failure in terms of organizational
practices to prevent accidents in safety-critical systems (Grabowski et
al. 2010). Safety culture is believed to be an important factor regarding
operational safety. Some of the common features of an adequate safety
culture include proper organizational communications, appropriate
organizational learning, and a commitment from senior management to
safety (Sorensen 2002). A safe culture is often seen as a well-informed cul-
ture that is created by an effective reporting structure, within which the
behavioral guidelines are clearly explained and followed (Reason 1998).
For this reason, the formation of a proper safety culture is an important
determinant to the mitigation of latent factors that lead to disasters in
complex interdependent networks of systems.
Grote and Kunzler (1996) propose a sociotechnical systems approach,
termed Total Safety Management, in agreement with current models of
quality assurance. This concept underscores the importance of organiza-
tional design and safety culture that are normally assumed to have an indi-
rect effect on the safety of the overall system. Sorensen (2002) discusses
the evolution of the term “safety culture” and the perceived relationship
between safety culture and safety of operations in nuclear power genera-
tion as well as the realms of hazardous technologies. Reason (1998) analyzes
theoretical and practical aspects of achieving a safe culture. In particular,
some of the discussion points include reasons that an unsafe culture is more
likely to be involved in organizational rather than individual accidents,
 Industry Notes \ 521

such as (1) pathological adaptations; (2) recurrent accident patterns and the
role of cultural drivers; and (3) whether a safety culture can be engineered.
Grabowski et al. (2010) explore the connection between safety culture and
performance in complex safety-critical systems. Moreover, Grabowski and
colleagues investigate a methodology that can provide early warning of
hostile events by identifying leading indicators (LIs) of safety that have a
positive correlation with safety performance. Wiegmann et al. (2004) sum-
marize and integrate various studies that define and analyze safety culture
along with the interrelated concept of safety climate. Such studies seem to
have general disagreement pertaining to the definition of safety culture as
well as the interplay or relationship between safety culture and safety cli-
mate. In doing so, Wiegmann et al. provide a list of key organizational indi-
cators that are predictive of safety culture along with various methods that
can be used to assess these factors. Sachon and Paté-Cornell (2004) propose
a model that helps to determine the strategy for optimal budget allocation
for the development of new technologies and safety-critical systems. Their
research suggests that decomposition of the system’s development process
into multiple development subprocesses or modules can aid in the forma-
tion of designing multiple projects based on their criticality. Using this risk
analysis approach, the balance between probabilities of development and
operational failures can be established, enabling decision makers to allo-
cate optimal budgets.
In addition to the issue of developing a safety culture, the prioritiza-
tion and selection of critical transportation projects from several compet-
ing projects comes under a multi-objective combinatorial optimization
(MOCO) problem. Joshi and Lambert (2007) develop a methodology for
integrating equity metrics with traditional metrics, for planning and pri-
oritization of large- scale transportation projects. This methodology is
helpful to planners and managers in visualizing and comparing measures
of the distributed equity along with cost-benefit tradeoffs.
The metric, technology readiness level (TRL), is a measure of maturity
of an individual technology. Several concerns may arise when this measure
is taken from an individual technology and deployed for a system involving
interplay between multiple technologies that are integrated throughout the
system. Ramirez-Marquez and Sauser (2009) propose a system-focused tech-
nique for managing a system developmental life cycle in an effective and effi-
cient way. They define a system readiness level (SRL) that aids in determining
the current and future readiness of a system. SRL index incorporates both the
current TRL scale and the concept of an integration readiness level.
522 / TRANSPORTATION JOURNAL™

Data Mining, Big Data, Complex Networks

With the advent of modern electronics, which include inexpensive ubiq-

uitous sensing, society is faced with massive amounts of data to process.
The data contains information that should enable firms to make optimal
decisions. However, methods by which to extract this information are not
easily evident. There are many techniques that are used in data mining of
big data (Ratner 2011). Some techniques come from the area of multivariate
analysis such as regression, principal component analysis, factor analysis,
multidimensional scaling, cluster analysis, canonical correlation analysis,
discriminant analysis, and latent structure analysis. There are also other
intelligent systems methods that use techniques such as neural networks,
fuzzy logic, and genetic algorithms. In stochastic methods, Bayesian analy-
sis has been popular as a technique in model building.
When data-mining techniques are applied to a specific area or a spe-
cific problem, prior knowledge about the system can simplify the process.
Applying theoretical or blind techniques on the data in the hopes of finding
some structure can be a daunting task or practically and computationally
intractable. In the area of transportation safety, prior knowledge based on
practitioner insights and theoretical fundamentals related to transporta-
tion, safety, and human behavior should be utilized to perform effective
extraction of knowledge from the data. Thus, existing information about
the system allows for the derivation of a nominal structure of a model;
parameters for the model can be determined by using the data on that
structure, as shown in figure 1.

Figure 1 Data Structure Model

 Industry Notes \ 523

Feedback Control for Modeling

Feedback mechanisms have been used to control systems in the history of

humankind. As far back as 300 BCE, Greeks and Arabs used feedback mech-
anism to keep track of time (Lewis 1992). During the industrial revolution,
steam engines utilized control mechanisms to operate efficiently. Since
then, feedback controllers have been used for machines, vehicles, and mul-
tiple types of systems. In fact, adaptive advertisements employ feedback
mechanisms to determine which promotions should be displayed to con-
sumers based on their previous click-stream behaviors. For instance, feed-
back control has been used to improve consumer decision quality (Krishen
and Nakamoto 2009), and also handle consumer information overload in
choice sets on websites (Kheirandish, Krishen, and Kachroo 2009; Krishen,
Raschke, and Kachroo 2011).
A system can be controlled in an open-loop structure where the output
of the system is not measured or used to make changes to control actions.
However, since it is almost impossible to know the exact model of any
system, an open-loop system cannot be expected to perform with a very
high accuracy. In feedback control, we use sensors to estimate the state
of the system, and the real-time control action uses the measured output
to derive the control action to be performed. To design a feedback control
law, we only need a nominal model of the system that should be simple,
but must include the essential dynamics of the system. In relation to this,
Shang, Tjader, and Ding (2004) explore the potential of applying the ana-
lytic network process (ANP) to evaluate transportation projects in Ningbo,
China. Advantages of using ANP over traditional hierarchical analysis tools
include feedback and interdependence among various decision levels and
criteria. Shang and colleagues argue that the model incorporates a much
wider range of factors, which they classify as benefits, opportunities, costs,
and risks; conventional transportation evaluation methods normally lack
these elements.
Our methodology of using a structured model—static model (e.g., a
structural equation model); a dynamic model (e.g., a Markovian model); or
a stochastic hybrid system model (Lesser and Oishi 2014)—can capture the
known nominal structure of a static or dynamic system. Modeling with
latent variables can make the model much simpler in comparison with
using only measured variables if the real system does in fact have the influ-
ence of those latent variables. In the context of safety culture, since the
system is cyber-physical, human-system interaction and its implication
to safety can be captured by models with latent variables. One advantage
524 / TRANSPORTATION JOURNAL™

of using a nominal model is that it allows for parameter estimation with

measured data and therefore more effective model validation. If the model
is not appropriate, an iterative modification can be applied, followed by
repeating the steps of modeling to derive the new model. These iterations
can continue until we obtain a model within a predefined range of accept-
ability, or until we reach a prespecified upper bound iteration count fail-
ing to find a model. This modeling flow chart is shown in figure 2.
The proposed feedback-based transportation traffic safety framework
can be implemented in a phased manner, as shown in figure 3. During
phase 1, we gathered observational data from 10,343 drivers; the purpose
of this collection was to verify a core driver safety issue. In some cases,
transportation-related marketing campaigns target distracted drivers for
single-infraction offenses such as No Phone Zone (Nevada Department of
Transportation 2014a) and Click it or Ticket (Boatman 2014). On the other

Figure 2 Modeling Flow Chart

 Industry Notes \ 525

Figure 3 Research Methodology

hand, other campaigns are aimed at a combination of multiple unsafe

driving practices, for example Bad Driving (Clark County Nevada 2014) and
Zero Fatalities (Nevada Department of Transportation 2014b). However, a
recent study by Abouk and Adams (2013) finds that drivers change their
behaviors temporarily after law enforcement threats are announced but
eventually return to their dangerous habitual behaviors. For this reason,
the goal of our main study is to verify existing notions regarding dis-
tracted driving.

Main Study

Participants and Procedure

This study utilizes observational data collected from various locations in

the western region of the United States following a complex sample frame-
work (Sancheti, Kachroo, and Amei 2012). Multiple trained data collectors
recorded the following information (based on best approximation): (1)
whether or not the driver was wearing a seat belt; (2) number of passengers
inside the car; (3) whether or not the driver was using a cell phone while
driving; and (4) any demographic information such as age and gender
(approximated). A total of 10,343 observations of cars in southern Nevada
were collected for the dataset.

Results—Main Study

The pertinent data are given in table 1, which consists of the frequency and
percentages for the number of drivers wearing seat belts and whether they
were or were not using cell phones during the observation frame.
Of particular interest are the drivers with and without seat belts who
are using cell phones. The ones who were wearing their seat belts and using
526 / TRANSPORTATION JOURNAL™

Table 1/Main Study Results

Observation Fre- Percent Valid Cumu-
quency Percent lative
Percent
Driver wearing Valid Driver not using a cell phone 9,491 98.4% 98.4% 98.4%
belt
Driver using a cell phone 158 1.6% 1.6% 100.0%
Total 9,649 100.0% 100.0%
Driver not wear- Valid Driver not using a cell phone 648 97.2% 97.2% 97.2%
ing belt
Driver using a cell phone 19 2.8% 2.8% 100.0%
Total 667 100.0% 100.0%
Don’t know Valid Driver not using a cell phone 27 100.0% 100.0% 100.0%

a cell phone consisted of 158 out of 9,649 (1.6%) and the ones who were not
wearing their seat belts and using their cell phones consisted of 19 out of
667 (2.8%). The question we ask about these data is: knowing that the total
number of drivers wearing belts was 9,649 and those not wearing belts was
667, is there a significant difference when comparing 158 belted drivers who
used phones and 19 unbelted drivers who used cell phones in the second
case? In order to calculate this, a Fisher’s exact test can be performed (see
http://www.langsrud.com/fisher.htm), which essentially is used to examine
the significance of the association (contingency) between the two kinds
of classification (see http://www.physics.csbsju.edu/stats/exact.html). To
determine the probability that there is a correlation between cell-phone
usage and seat-belt usage, application of Fisher’s exact test shows that the
proportion of drivers using cell phones without seat belts is significantly
higher than those using cell phones with seat belts (p = 0.03).
Thus, this observational study shows that those who are unsafe in
terms of seat-belt usage are more likely to be unsafe in terms of cell-phone
usage as well. Whereas this is a not a causal relationship, the correlation
does appear to be significant. This finding warrants a framework to define a
traffic safety indicator that is derived from multiple data sources.

Proposed Hierarchical Feedback-based Transportation Traffic Safety

Framework
Interdisciplinary systems must be designed that can create adaptive safety
communications, determine appropriate monetary levels for various trans-
portation issues, and employ safety mechanisms to improve the overall
safety culture in various geographic locations. Of utmost importance is
the ability to take data, which exists in abundance, and overlay a feedback
 Industry Notes \ 527

control model around it to improve safety performance. As shown in figure

4, phase 2 begins by identifying a framework in which multiple data sources
should be synergized to create a safety performance index (SPI) in a feed-
back model design. Next, the feedback model for the safety performance
index will derive the important latent variables for a structural equation
model of salient safety constructs (see figure 5). Finally, the model can be
systematically implemented by targeted communications to increase per-
formance measures at various stages to increase traffic safety, as in figure 6.
Our conceptualization begins with figure 4, which depicts the basic
feedback control loop–based structure of our framework. The figure shows
the flow of data that is obtained from various data sources and sensors.
That data is then processed in software and performance measures are
estimated using the appropriate mathematical formulation. These mea-
sures are then fed into the model of the system from which the latent safety

Figure 4 Safety Performance Index (SPI) Model (Phase 2)

528 / TRANSPORTATION JOURNAL™

Figure 5 SPI Feedback (Phase 2)

Figure 6 SPI Hierarchical System Implementation (Phase 3)

 Industry Notes \ 529

performance index is computed. Following this, the index can be used to

perform the appropriate control action for the system. Figure 5 augments
figure 4 by providing more details of the data sources and the model. To fur-
ther define the available sources, the data have been divided into multiple
possible inputs; these include mobile data, system data, people data, and
agency data. Mobile data can be obtained from many sources, for example,
the data from GPS enabled smartphones, or from cellular network data.
Additionally, it can come from transportation-related social media apps, or
GPS probes such as the ones used in transit vehicles. The system data pro-
vide information about the overall system such as weather data, economic
data, relevant transportation news, and so on. People data are the data that
involve people activities, including survey data, complaint data, manual
data, or data from special events. Finally, there are data that are collected by
transportation agencies, such as that obtained from road sensors, Traffic
Management Center (TMC) data, toll data, and data related to incidents.
The most important part of our framework is the modeling to capture
the safety culture related latent variables of the system. This conceptualiza-
tion is illustrated in figure 5 as an expansion of the model. In the sample
model depicted in the right-hand side of the figure, measured variables,
their error sources, and latent variables are all shown. Moreover, relation-
ships between all of the variables are hypothesized in the model. The data
collected from the system drives the values of these variables which are
then used to compute the system performance index (SPI).
Our overall proposed model is a hierarchical framework that works at
various levels (see figure 6). As an example, safety is important at a signal-
ized intersection where pedestrians and vehicles are involved. The data for
this can come from sources including traffic sensors, complaint systems,
traffic reports, and survey data. Based on all of the data sources as inputs,
a safety model can be derived that will define latent variables and their rela-
tionships, and then use those to provide action or control. There are many
control actions that are relevant to this example. For instance, the pro-
gramming of the signal timing can be changed, more sensors can be added
at the site, or flashing lights can be installed at the location. At a higher
level, we could be interested in controlling all of the intersections and mid-
sections of the road, which would include the signalized intersection of the
previous example. Thus, the safety index of the entire road depends on the
safety indices of each of its components. A higher level of modeling could
involve an even larger network. The total data for the system has measure-
ments that could be relevant to one or more of these levels. The SPI values
530 / TRANSPORTATION JOURNAL™

at a lower level are fed into higher levels to achieve consistent modeling and
a hierarchical computation structure. The control actions at various levels
can also be different. In particular, at the highest level in a system state, the
control action could be a policy decision regarding safety, such as to pass a
law banning texting while driving, or it could be a budget allocation prob-
lem for various projects.

Conclusions and Future Research

Our framework is designed to handle the complexities inherent in com-
plex systems with a phased approach. We contribute to existing theory
on safety culture by providing a feedback mechanism by which to struc-
ture the systematic problem of safety. The most important aspect of our
model is the interdisciplinary nature of our overall safety culture solu-
tion. In effect, by including data from nontraditional sources (e.g. social
networks, blogging sites, product commentaries) in our model and utiliz-
ing existing qualitative techniques to analyze the data (Krishen et al. 2014)
while also including quantitative data with quantitative techniques, we
can maximize the conceptualization of safety culture. Our model aims to
transcend traditional communication mechanisms between transporta-
tion and marketing, such that big data can better inform marketing com-
munications. We therefore propose the following future research based
on our model:

1. Safety Culture Conceptualized as a Subculture, Community, or Tribe

Several concepts regarding groups of individuals, including virtual com-

munities, subcultures of consumption, and consumer tribes, share a theo-
retical underpinning regarding the meaning of membership within them.
For example, for virtual communities, Healy and McDonagh (2013) discuss
seven consumer roles, such as voice, loyalty, exit, entry, and twist, all of
which are represented by various consumer actors. Whereas loyalty as a
role involves attachment to the community or brand, twist entails the use
of a product, symbol, or service in a way that was not originally intended
by the firm. In its original definition, a subculture of consumption is a
subgroup of individuals who voluntarily participate as members, shar-
ing activities, beliefs, and values (Schouten and McAlexander 1995). Cova
and Cova (2002) describe consumer tribes as a network of heterogeneous
individuals, linked by shared passion, and able to create collective action
and advocacy. Whereas some of these conceptualizations, such as subcul-
tures and communities, identify individuals as a homogeneous group;
 Industry Notes \ 531

others like tribes include heterogeneous consumers acting as a collective.

Recent research highlights the nature of a running community or subcul-
ture, postulating that heterogeneity, belonging, and tension characterize
membership (Thomas, Price, and Schau 2013). Following the conceptual-
izations of groups of networked individuals forming shared memberships
and collective action, Krishen et al. (2014) describe two political tribes based
on collective good and self-interested actors, within the domain of trans-
portation taxation communication. Similar to subcultures, communities,
and tribes, the formation of a safety culture will require shared values and
ethos, tension while negotiating group boundaries, and a sense of belong-
ing, in order to enable collective action.

2. Qualitative Data to Quantitative Model

This mechanism proposes the use of data in natural language, such as a

complaint system where people’s complaints have been recorded. Such data
can be processed using natural language-processing techniques and then
models based on elements including sentiment analysis can be obtained.
These models have quantitative information derived from processing qual-
itative language data. These models can be then used in conjunction with
models built out of quantitative data, or ways to synthesize qualitative and
quantitative data can also be developed.

3. Qualitative Research Methods for Building Nominal Quantitative Models

Qualitative research methods normally involve ethnographic studies that

involve deep qualitative analysis of a very small sample size as compared to
statistical methods that require quantitative data of a much larger sample
size. These techniques can provide a starting point for building the nominal
model for the quantitative model-building in our framework. Qualitative
data from interviews and surveys can be used to derive variables and nomi-
nal relationships among them before quantitative data can be used to obtain
parameters of that model in the iterative process of model improvement.
For a complex system we can use the Fuzzy Set Qualitative Comparative
Analysis (fsQCA) method to combine the qualitative knowledge from dif-
ferent parts of a complex system (Ragin 2008; Wu, Yeh, and Woodside 2014).
In order to extract more out of the qualitative data, the fsQCA methodology
can be employed, the outcome of which can be used as the starting nominal
model for the quantitative model building with latent variables such as the
structural equation models for static modeling or hybrid stochastic models
for the modeling of dynamic phenomenon.
532 / TRANSPORTATION JOURNAL™

Note
The first three authors contributed equally in the preparation of this manuscript.

References
American Automobile Association. 2007. “Improving Traffic Safety Culture in the United
States—The Journey Forward.” Washington, DC.
Abouk, R., and S. Adams. 2013. “Texting Bans and Fatal Accidents on Roadways: Do
They Work? Or Do Drivers Just React to Announcements of Bans?” American
Economic Journal: Applied Economics 5 (2): 179–99.
Agarwal, S., A. Sancheti, R. Khaddar, and P. Kachroo. 2013. “Geospatial Framework for
Integration of Transportation Data Using Voronoi Diagrams.” Paper 13-5378.
Transportation Research Board 92nd Annual Meeting, Washington DC,
January 13–17.
Arboleda, A., P. C. Morrow, M. R. Crum, and M. C. Shelley. 2003. “Management Practices
as Antecedents of Safety Culture within the Trucking Industry: Similarities
and Differences by Hierarchical Level.” Journal of Safety Research 34 (2): 189–97.
Block, L. G., and P. A. Keller. 1995. “When to Accentuate the Negative: The Effects of
Perceived Efficacy and Message Framing on Intentions to Perform a Health-
related Behavior.” Journal of Marketing Research 32 (2): 192–203.
Boatman, S. 2014. “Click It or Ticket Campaign in Full Swing on Nevada Roads.”
KRNV-DT Reno MyNews 4. http://www.mynews4.com/news/local/story/
Click-it-Or-Ticket-Campaign-In-Full-Swing-on/uMJxddgKSEOn-0CPxO-1Dw.
cspx?rss=3298.
Clark County Nevada. 2014. “Bad Driving.” http://www.clarkcountynv.gov/depts/bad-
driving/Pages/default2.aspx.
Cova, B., and V. Cova. 2002. “Tribal Marketing: The Tribalisation of Society and Its Impact
on the Conduct of Marketing.” European Journal of Marketing 36 (5/6): 595–620.
Douglas, M. 1985. Risk Acceptability According to the Social Sciences. New York: Russell Sage
Foundation.
Girasek, D. C. 2012. “Towards Operationalising and Measuring the Traffic Safety Culture
Construct.” International Journal of Injury Control and Safety Promotion 19 (1):
37–46.
———. 2013. “Gauging Popular Support for Traffic Safety in the United States.” Accident
Analysis and Prevention 50:1,112–17.
Grabowski, M., Z. You, H. Song, H. Wang, and J. R. Merrick. 2010. “Sailing on Friday:
Developing the Link between Safety Culture and Performance in Safety-critical
Systems.” IEEE Transactions on Systems, Man and Cybernetics 40 (2): 263–84.
Grote, G., and C. Kunzler. 1996. “Safety Culture and Its Reflections in Job and
Organizational Design: Total Safety Management.” International Journal of
Environment and Pollution 6 (4/5): 618–31.
Healy, J. C., and P. McDonagh. 2013. “Consumer Roles in Brand Culture and Value Co-Creation
in Virtual Communities.” Journal of Business Research 66 (9): 1,528–40.
 Industry Notes \ 533

Institute of Transportation Engineers. 2013. “Partnering Across Disciplines for Traffic

Safety Culture Change.” Institute of Transportation Engineers Journal 83 (11): 44.
Joshi, N. N., and J. H. Lambert. 2007. “Equity Metrics with Risk, Performance, and Cost
Objectives for the Prioritization of Transportation Projects.” IEEE Transactions
on Engineering Management 54 (3): 539–47.
Kheirandish, R., A. S. Krishen, and P. Kachroo. 2009. “Application of Optimal Control
Theory in Marketing: What Is the Optimal Number of Choices on a Shopping
Platform/Website?” International Journal of Computer Applications in Technology
34 (3): 207–15.
Krishen, A. S., and K. Nakamoto. 2009. “Improving Consumer Quality Efficiency by
Using Simple Adaptive Feedback in a Choice Setting.” International Journal of
Computer Applications in Technology 34 (3): 155–64.
Krishen, A. S., R. L. Raschke, and P. Kachroo. 2011. “A Feedback Control Approach to
Maintain Consumer Information Load in Online Shopping Environments.”
Information and Management 48 (8): 344–52.
Krishen, A. S., R. L. Raschke, P. Kachroo, M. LaTour, and P. Verma. 2014. “Promote Me
or Protect Us? The Framing of Policy for Collective Good.” European Journal of
Marketing 48 (3/4): 742–60.
Krishen, A. S., R. L. Raschke, and M. Mejza. 2010. “Guidelines for Shaping Perceptions of
Fairness of Transportation Infrastructure Policies: The Case of Vehicle Mileage
Tax.” Transportation Journal 49 (3): 24–38.
Lesser, K., and M. Oishi. 2014. “Reachability for Partially Observable Discrete Time
Stochastic Hybrid Systems.” Automatica 50 (8): 1,989–98.
Lewis, F. L. 1992. Applied Optimal Control and Estimation. Upper Saddle River, NJ:
Prentice-Hall.
Naevestad, T., and T. Bjornskau. 2012. “How Can the Safety Culture Perspective Be Applied
to Road Traffic?” Transport Reviews 32 (2): 139–54.
National Highway Traffic Safety Administration. 2013. “Traffic Safety Facts Research
Note: 2012 Motor Vehicle Crashes Overview.” Washington, DC.
Nevada Department of Transportation. 2014a. “Talking and Texting While Driving is
Illegal.” http://www.nophonezonenv.com/.
———. 2014b. “Zero Fatalities.” http://www.zerofatalitiesnv.com/.
Pidgeon, N. F. 1991. “Safety Culture and Risk Management In Organizations.” Journal of
Cross-cultural Psychology 22 (1): 129–40.
Pratt, S. G. 2003. “Work-related Roadway Crashes: Challenges and Opportunities for
Prevention.” Atlanta, GA: The National Institute for Occupational Safety and
Health, Centers for Disease Control and Prevention. http://www.cdc.gov/niosh/
docs/2003-119/pdfs/2003-119.pdf.
Ragin, C. C. 2008. Redesigning Social Inquiry. Chicago: University of Chicago Press.
Ramirez-Marquez, J. E., and B. J. Sauser. 2009. “System Development Planning via
System Maturity Optimization.” IEEE Transactions on Engineering Management
56 (3): 533–48.
534 / TRANSPORTATION JOURNAL™

Ratner, B. 2011. Statistical and Machine-Learning Data Mining: Techniques for Better Predictive
Modeling and Analysis of Big Data. Boca Raton: CRC Press.
Reason, J. 1998. “Achieving A Safe Culture: Theory and Practice.” Work and Stress 12 (3):
293–306.
Sachon, M., and M. E. Paté-Cornell. 2004. “Managing Technology Development for
Safety-Critical Systems.” IEEE Transactions on Engineering Management 51 (4):
451–61.
Sancheti, A., P. Kachroo, and A. Amei. 2012. “Nevada State Seat Belt Use Survey 2012.”
Transportation Research Center, University of Nevada–Las Vegas.
Schouten, J. W., and J. H. McAlexander. 1995. “Subcultures of Consumption—An
Ethnography of New Bikers.” Journal of Consumer Research 22 (1): 43–61.
Shang, J. S., Y. Tjader, and Y. Ding. 2004. “A Unified Framework for Multi-criteria
Evaluation of Transportation Projects.” IEEE Transactions on Engineering
Management 51 (3): 300–313.
Sorensen, J. N. 2002. “Safety Culture: A Survey of the State-of-the-Art.” Reliability
Engineering and System Safety 76 (2): 189–204.
Thomas, T. C., L. L. Price, and H. J. Schau. 2013. “When Differences Unite: Resource
Dependence in Heterogeneous Consumption Communities.” Journal of
Consumer Research 39 (5): 1,010–33.
Watson, G. W., D. Scott, J. Bishop, and T. Turnbeaugh. 2005. “Dimensions of Interpersonal
Relationships and Safety in the Steel Industry.” Journal of Business and Psychology
19 (3): 303–18.
Wiegmann, D. A., H. Zhang, T. L. Von Thaden, G. Sharma, and A. M. Gibbons. 2004.
“Safety Culture: An Integrative Review.” International Journal of Aviation
Psychology 14 (2): 117–34.
Williams, A. F., and N. Haworth. 2007. “Overcoming Barriers to Creating a Well-
functioning Safety Culture: A Comparison of Australia and the United States.”
ITE Journal-Institute of Transportation Engineers 77 (6): 24–31.
Wu, P. L., S. S. Yeh, and A. G. Woodside. 2014. “Applying Complexity Theory to Deepen
Service Dominant Logic: Configural Analysis of Customer Experience-and-
Outcome Assessments of Professional Services for Personal Transformations.”
Journal of Business Research 67 (8): 1,647–70.
Book Review

Robert E. Gallamore and John R. Meyer. American Railroads—

Decline and Renaissance in the Twentieth Century. Cambridge,
MA: Harvard University Press, 2014. Pp. xiii, 506 pp, ISBN 978-0-
674-72564-5 (Hardback) US $55.00

John C. Spychalski, DLP

Professor Emeritus of Supply Chain Management
The Pennsylvania State University
University Park, PA 16803-3603
jcs2@psu.edu

The stated objective of this thirteen-chapter book, as first conceived by co-

author John R. Meyer, is presentation of a “full sweep” account of the US
railroad industry’s evolution in the twentieth century (Preface, viii). Work
on the book spanned 15 years up to the date of publication. Meyer contrib-
uted to the drafting and/or review of 10 chapters. In 2006, deteriorating
health caused Meyer to enlist Robert Gallamore as co-author. Sadly, Meyer
died in 2009, leaving Gallamore with responsibility for the drafting of three
chapters (1, 12, and 13) and final revision of the entire draft. Both co-authors
brought eminent credentials to the writing effort. Harvard University–
based Meyer had a distinguished academic career as an economist focused
heavily on policy issues in rail and other modes of transport. Additionally,
he served for a time as a board member and vice-chairman of Union Pacific.
Gallamore served Northwestern University as director of its Transportation
Center and as professor of managerial economics and decision sciences
between 2001 and 2006. Before entering the academic sphere, he served in
numerous varied executive and staff analytical positions with several enti-
ties, including Union Pacific, Transportation Technology Center (provider
of railway equipment research and testing services), the Federal Railroad
Administration within the US Department of Transportation, and the
United States Railway Association.
As portrayed in the book, the railroad industry dominated intercity
freight and passenger transport in the United States within the first two

Transportation Journal, Vol. 54, No. 4, 2015

Copyright © 2015 The Pennsylvania State University,
University Park, PA
536 / TRANSPORTATION JOURNAL™

decades of the twentieth century. Onward from 1920 through the 1970s, its
dominance declined steadily in the face of increasingly severe competition
from the entry and steady rise of road, inland waterway, pipeline, and air
transport. By the first half of the 1970s, this downward trajectory reached
crisis level with the fall into bankruptcy of most railroads operating within
the Northeast and a portion of the Midwest. Actions to abate the crisis
between the mid-1970s and the start of the 1980s succeeded in fostering an
economic rebirth for the freight railroad industry that has continued into
the present century.
A prime although not exclusive focus of the book is the relationship
between government policy actions and the condition and performance
of the railroad industry between the beginning and end of the century.
Within this array of actions, the authors place greatest negative criticism
on federal economic regulation of the industry, as it existed between 1887
and repeal of most elements in 1980. They hold that economic regulation of
freight rail transport was ill-founded and economically indefensible from
its inception. Their most intense criticism is centered on how such regula-
tion constrained railway pricing and service offering policies and practices
in ways that greatly handicapped railroads’ ability to compete effectively
against other modes of transport, thus weakening the viability of rail firms
and preventing freight shippers and consignees from obtaining the ben-
efits of using rail service. They provide a foundation for this criticism with
a tutorial on the inherent economic characteristics of railway firms and an
exposition of why such characteristics make unconstrained (by govern-
ment regulation) demand-based differential pricing essential for enabling
freight railroads to remain financially viable without external subsidy.
Government policy toward intramodal rail competition and consolida-
tion of firms within the freight railroad industry is viewed by the authors as
having been unenlightened prior to the later decades of the century, due to
inadequate understanding of the inherent economic characteristics of rail
service operation and market conditions and/or ill-founded pleas of opposi-
tion from parties with vested interest in the status quo. They judge that the
consolidations that occurred largely after the mid-1970s have now yielded
a national freight railroad system with components that strike an adequate
balance between (1) protection of shippers’ interests through preservation
of intramodal competition where sustainable, and (2) achievement of rail-
way operational efficiency and financial viability.
Intercity rail passenger service receives attention in two chapters.
One covers its decline between 1900 and 1970, and the other addresses the
 Book Review \ 537

problems that have arisen under Amtrak’s operation of the service ever
since it began operating in 1971. Suffice to say here that subject matter cov-
erage in both chapters is substantive. However, a reader who believes that
subsidization of intercity rail passenger service is justified by the public
benefits it yields will find no strong support for that belief in what is pre-
sented. The authors appear to hold a philosophical preference for confining
judgments of the economic value of intercity passenger service to private
market place criteria.
A reader might wish that the book’s coverage of the topic of economic
regulation and the positive results of its downsizing had been paralleled by
coverage of the impacts of the great increase in rail safety and environmen-
tal regulation that has occurred since the 1970s. Also, the post-deregula-
tion rise of the regional and short line component of the railroad industry
deserves more than the passing mention that it receives, given the fact that
such smaller railroads play a vital role in serving as originators and deliver-
ers of carload shipments for line-haul Class I railroads. Last but not least,
the US rail network is, functionally, a member of the North American rail
network. Thus, devotion of some attention to the implications of conti-
nent-wide economic and policy issues would have been welcome.
Overall, the authors have provided substantive, well-documented
coverage of economic and policy-related subject matter that is key to an
in-depth understanding of key determinants of the American railroad
industry’s century-long evolution. The book thus deserves attention from
academicians, policy analysts, and business practitioners who are seeking
a deeper understanding of how the US railroad industry has reached its
present state of performance and condition. Lessons learned from that evo-
lution have relevance for parties grappling with current policy issues such
as efforts presently underway by some shipper groups to make the Surface
Transportation Board establish new rules that would place caps on charges
for rail freight service under some conditions.
View publication stats