Download the Journal - Transportation Research Forum

Volume 50, Number 1 Spring 2011
Transportation Research Forum
NDSU Dept. 2880
PO Box 6050
Fargo, ND 58108-6050
www.trforum.org

Journal of the Transportation Research Forum
The Transportation Research Forum, founded in 1958, is an independent, nonprofit organization of transportation
professionals who conduct, use, and benefit from research. Its purpose is to provide an impartial meeting ground
for carriers, shippers, government officials, consultants, university researchers, suppliers, and others seeking
exchange of information and ideas related to both passenger and freight transportation. More information on
the Transportation Research Forum can be found on the Web at www.trforum.org.
General Editors: Michael W. Babcock, Kansas State University and Kofi Obeng, North Carolina A&T State
University
Book Review Editor: Jack S. Ventura, Surface Transportation Board
Associate Editors:
Richard Gritta, University of Portland; Robert Harrison, University of Texas; Kevin H. Horn, G.E.C. Inc.;
Wesley Wilson, University of Oregon; Barry Prentice, University of Manitoba; Carl Scheraga, Fairfield
University; and John Bitzan, North Dakota State University.
Disclaimer:
The facts, opinions, and conclusions set forth in the articles contained herein are those of the authors and
quotations should be so attributed. They do not necessarily represent the views and opinions of the Transportation
Research Forum (TRF), nor can TRF assume any responsibility for the accuracy or validity of any of the
information contained herein.
Subscriptions:
The Journal of the Transportation Research Forum (JTRF) is distributed to members of the Transportation
Research Forum and subscribers.
Currently, the JTRF is published three times per year. Annual subscriptions are available to the general public
for $150 and must be purchased for a calendar year.
An electronic version (PDF format) is available for $50 per Journal or single Journal articles for $10 each.
Subscriptions and/or membership applications may be obtained from the following office or online at
www.trforum.org
Copyright 2011
The Transportation Research Forum
All Rights Reserved
ISSN 1046-1469
Published and Distributed by
Transportation Research Forum
NDSU Dept. 2880
P.O. Box 6050
Fargo, ND 58108-6050
P: (701) 231-7766 • F: (701) 231-1945
or apply online at www.trforum.org

Journal of the Transportation Research Forum
Volume 50, Number 1 • Spring 2011
Table of Contents
A Message from the JTRF Co-General Editors 3
Kofi Obeng and Michael W. Babock
ARTICLES
Safety Analysis of Continuous Green Through Lane Intersections 5
Thobias Sando, Deo Chimba, Valerian Kwigizile, and Holly Walker
Transit Passenger Perceptions: Face-to-Face Versus Web-Based Survey 19
Laura Eboli and Gabriella Mazzulla
Methodology for Measuring Output, Value Added, and Employment Impacts of
State Highway and Bridge Construction Projects 37
Michael W. Babcock and John C. Leatherman
Rail Rate and Revenue Changes Since the Staggers Act 55
Ken Casavant, Eric Jessup, Marvin E. Prater, Bruce Blanton, Pierre Bahizi,
Daniel Nibarger, Johnny Hill, and Isaac Weingram
Measuring Bulk Product Transportation Fuel Efficiency 79
C. Phillip Baumel
Demand Analysis for Coal on the United States Inland Waterway System:
Fully Modified Cointegration (FM-OLS) Approach 89
Junwook Chi and Jungho Baek
State of the Art: Centerline Rumble Strips Usage in the United States 101
Daniel E. Karkle, Margaret J. Rys, and Eugene R. Russell
BOOK REVIEW
Understanding the Railway Labor Act 119
Gordon P. MacDougall
On the cover: Barge transportation on the inland waterways plays a large role in the transport of crude oil,
petroleum products, coal, grain, and ores. In “Demand Analysis for Coal on the United States Inland Waterway
System: Fully Modified Cointegration Approach,” Junwook Chi and Jungho Baek examine the dynamic
relationship between demand for coal barge transportation and variables such as barge and rail rates, domestic
coal consumption and production, and coal exports.
1

Transportation Research Forum 121
Statement of Purpose 121
History and Organization 121
Annual Meetings 122
TRF Council 123
TRF Foundation Officers 123
Past Presidents 124
Recipients of the TRF Distinguished Transportation Researcher Award 124
Recipients of the Herbert O. Whitten TRF Service Award 124
Past Editors of JTRF 125
Recipients of the TRF Best Paper Award 125
Guidelines for Manuscript Submission 127
2

A Message from the JTRF
Co-General Editors
The papers in this issue of the JTRF cover topics ranging from passenger to freight transportation
and from highway safety to rail revenue. The specific topics addressed are:
• Transit passenger perceptions from face-to-face and web-based survey
• Safety analysis of continuous green through lane intersections
• The use of centerline rumble strips in the United States
• Methodology for measuring output, value added, and employment impacts of state highway
and bridge construction projects
• Measuring bulk product transportation fuel efficiency
• Demand analysis for coal on the United States inland waterway system
• Rail rate and revenue changes since the Staggers Act.
Eboli and Mazulla use a relatively homogeneous group of respondents–university students–
who commute by bus to collect customer satisfaction and importance rating data from web-based
and face-to-face surveys. Then, using paired t-test, Fisher F-test, and discriminant analysis they
compare responses from these surveys to determine if there are statistical differences between them.
They found that the average importance ratings in the face-to-face surveys were generally higher
than those from the web-based surveys. This led Eboli and Mazulla to surmise that users tend to
give more importance to service characteristics in face-to-face interviews than they do when they
complete a questionnaire. For satisfaction rating they found significant differences between the
samples in terms of service frequency, reliability of runs, facilities at bus stops, and availability of
service frequency. They also found similarities in the ratings of satisfaction and importance in the
web-based survey and recommended asking respondents only satisfaction ratings questions in webbased
survey.
Contrary to the Eboli and Mazulla paper, two papers in this issue deal with safety analysis.
The first by Sando et al. analyzes the levels of injury severity from crashes at intersections with
continuous green through lanes in Florida using paired t-test and ordered probit statistical models.
Among their findings are more rear-ended crashes followed by sideswipe crashes and crashes from
left-turning vehicles. In terms of crash severity, they found that angle crashes and crashes from
lane changing maneuvers were more severe than rear-end crashes. They also found higher injury
severities for drivers 65 years and older and lower injury severity for crashes which occur during
the day. These findings led Sando et al. to suggest advance warning signs for the presence of such
an intersection on a road. The second safety paper is by Karkle et al. and deals with centerline
rumble strips. This paper’s objective is to survey states about their policies and guidelines in using
centerline rumble strips and identify gaps in research along with good practices. They found 11,233
miles of centerline rumble strips throughout the United States (excluding Texas and Colorado) and
did not find many states with written policies for their use.
Babcock and Leatherman develop an input-output methodology involving 11 steps to measure
the output, value added, and employment economic impacts of various types of highway and bridge
improvement programs. They applied the model to the Kansas Comprehensive Transportation
Program and concluded that their method can be used to estimate the impacts of any type of highway
improvement once the contract values of the highway improvement programs are known because
each highway improvement has a different multiplier.
Casavant et al. study changes in the rail rate structure for agricultural commodities since the
implementation of the Staggers Act and compares them to changes in the rates of other products
3

over the same period. They found that rail rates for agricultural products are higher than those of
other commodities and increased more rapidly from 2004-2008. Further, they found that rail rates
are lower for large volume shipments, especially those using multiple rail cars, but the rates for
large shipments increased faster than the rates for smaller shipments. Other findings are that on a per
ton-mile basis, the rates for distances less than 500 miles are larger than those for distances of 750
miles. Additionally, they found railroads have shifted car ownership costs to shippers, and railroads
recover fuel costs through surcharges.
The last two studies deal with bulk shipment transportation. Chi and Baek analyze the demand for
coal barge transportation in the United States using the Phillip-Hansen fully modified cointegration
method to determine the factors affecting coal movement and the substitution effects between rail
and barge carriers. They used the volume of coal transportation as the dependent variable and coal
barge rate, coal export level, domestic coal production, domestic consumption of coal and the
rates of other transportation modes as the independent variables. Their results showed one stable
equilibrium and long run demand for barge transportation that is more responsive to changes in
domestic coal exports and domestic coal consumption than barge and rail rates. In the short run, they
found the demand for transporting coal by barges is responsive to domestic coal production only.
Baumel also examines bulk product transportation, but in terms of fuel efficiency. His analysis
shows that fuel efficiency based only on net ton miles per gallon can produce erroneous results. To
correct this problem he provides alternative measures of fuel efficiency, such as total fuel consumed
from origin to destination by each mode and argues they can improve the accuracy of fuel efficiency
comparisons.
Michael W. Babcock Kofi Obeng
Co-General Editor Co-General Editor
4

Safety Analysis of Continuous Green
Through Lane Intersections
by Thobias Sando, Deo Chimba, Valerian Kwigizile and Holly Walker
This paper examines safety characteristics of continuous green through lane (CGTL) intersections
using paired-t test and ordered probit (OP) statistical models. The results suggest that there is a
significant difference between the proportions of sideswipe crashes in the CGTL direction compared
with the opposite direction. However, the results did not suggest a significant difference between the
proportions of rear-end and right-angle crashes for the CGTL and normal directions. The results
further suggest that angle crashes and crashes involving lane changing maneuvers are significantly
more severe compared with rear-end crashes.
INTRODUCTION
Escalating traffic demands on urban roadways have caused traffic engineers to use various measures
to reduce congestion, especially at signalized intersections. Transportation agencies are using
unconventional measures where conventional measures have been exhausted. One unconventional
low cost design strategy used in parts of Florida is the installation of continuous green through
lanes (CGTLs). These lanes are used to
reduce increasing demand for longer green
times for through movement at intersections
with considerably higher through volumes.
CGTLs are “T” intersections with one
or two through lanes on the mainline leg
receiving a continuous green indication,
i.e., passing without stopping, while the
inside through lane(s) in the same direction
receive conventional green, yellow, and red
indications (Figure 1). Installation of CGTLs
is less costly than intersection widening
alternatives, hence in most cases they
provide a cost effective solution for handling
high through traffic at T-intersections.
Although CGTLs have been used for
Figure 1: An Example of CGTL Intersection
more than three decades in Florida and their operational benefits are evident, they are still considered
a relatively new design alternative which many agencies are reluctant to approve. There have been
mixed reviews of the suitability and effectiveness of CGTL intersections in Jacksonville, Florida.
This is because citizens feel they are not safe, especially for motorists unfamiliar with their design,
and this has led to their removal from several locations while new ones continue to be installed
in other locations. This study evaluates different crash patterns that occur at CGTLs and further
analyzes the influence of roadway, traffic, driver, and environmental conditions on injury severity
for different crash patterns.
The rest of the paper is organized as follows: The next section provides a summary of the
literature on the analysis of crash patterns at intersections. It is followed by a methodology section
which outlines the analytical techniques used in this study. The methodology section is followed by
the results where the findings are discussed and explanations offered for them. The final section of
the paper presents the conclusions and recommendation.
5

Safety Analysis
LITERATURE REVIEW
Many previous studies have evaluated crash patterns at signalized intersections. Wang and Abdel-
Aty (2008) evaluated left-turn crashes occurring at 197 four-legged signalized intersections
(intersections connecting four roadway segments) in Florida and found that traffic flows, the width
of the crossing distance and signal phasing, affect left-turn crashes. Mitra et al. (2002) studied rightangled
and rear-end crashes by maneuver type at four-legged signalized intersections in Singapore.
The results indicated that the presence of uncontrolled left-turn channels, wider medians, higher
approach volumes, and an increase in signal phase are the most important factors that increase
accidents from both types of maneuvers. In an attempt to develop expected conflict value tables for
unsignalized three-legged intersections, Weerasuriya and Pietrzyk (1998) modeled conflict types
at unsignalized three-legged intersections. They divided conflicts into three main groups: same
direction, opposing direction, and cross traffic conflicts, which were further subdivided into 12
crash categories. Traffic conflicts increased as the number of lanes increased. For example, they
observed an average of about 70 rear-end conflicts on three-legged 2 x 6 intersections with six lanes
in the major street and two lanes in the minor street. About 20 rear-end conflicts on average were
observed for three-legged 2 x 2 (two lanes for minor street and two lanes for major street) and 2
x 4 intersections, i.e., intersections with two lanes for minor street and four lanes for major street.
Persaud and Nguyen (1998) developed safety performance models for four-legged intersections
based on 25 specific crash patterns, which were defined by the movements of the accident vehicles
prior to collisions. Out of the 25 patterns, the leading three patterns in property damage only crashes
in the order of importance were rear-end, left-turn versus opposing through traffic, and right-angle
crashes. Additionally, the proportion of crashes that involved left-turn versus opposing through
traffic was the highest, followed by right-angle and rear-end crashes among severe crashes.
Whereas there is a growing body of research about modeling intersection crash data and crash
patterns in particular, there is still a lack of an overall picture of the safety characteristics of CGTL
intersections, particularly crash patterns. The literature on the safety of CGTLs is limited. Hummer
and Boone (1995) investigated possible gains in travel efficiency from three unconventional
strategies, including median U-turn, two different CGTLs, and the North Carolina bowtie
intersection. The Florida and North Carolina versions of the CGTLs provided substantial reductions
in travel time and stops for through volumes less than 700 vehicles per hour per lane. Jarem (2004)
evaluated the safety and cost-and-benefit ratios of five CGTL intersections in Orlando, Florida, and
found that crashes related to CGTLs ranged from 8% to 24% for the five intersections that were
investigated. Most of the crashes were rear-end caused by unexpected stopping of vehicles in the
CGTLs followed by sideswipe crashes caused by erratic lane changes of vehicles from non-CGTLs
to avoid a red light. Few crashes involved left-turn vehicles encroaching on CGTLs, and each of the
five intersections had a different magnitude of each type of crash related to the CGTL.
Although the levels of CGTL related crashes reported by Jarem (2004) differ for the intersections
that were investigated, he does not study the influence of site characteristics on the occurrence of
different types of crashes at CGTLs. The study reported herein was conducted on all CGTLs in
Jacksonville with the purpose of quantifying the effects of site characteristics on the safety of CGTL
intersections.
METHODOLOGY
Data
At the beginning of this study, the city of Jacksonville had a total of 17 known CGTL intersections.
Eight of them (shaded intersections in Table 1) have been converted to traditional intersection
configurations or have had major maintenance or construction work done between 2003 and 2008
and were not considered in this study, leaving only nine (sites one though nine in Table 1) to be
6

JTRF Volume 50 No. 1, Spring 2011
studied. Several data sources, including drawings, condition diagrams, intersection photos, aerial
photographs, and straight line diagrams, were used to examine differences in site characteristics
between the intersections. These sources together with field visits were used to collect data on
intersection characteristics such as configurations, land use proximity and location of driveways,
signs and pavement markings, and number of CGTLs.
Table 1: List of CGTL Intersections in Jacksonville, Florida
No. Intersection Install Date Removal Date
1 US 17 @ Ortega Forest Dr. 2/3/1987 -
2 US 17 @ Entrance to Roosevelt Mall 10/13/1986 -
3 US 17 @ Park Street South 9/29/1983 -
4 US 17 @ Long Bow Rd. 5/1/1972 -
5 US 17 @ Baisden Road 2/1/1991 -
6 US 1 @ 45th Street 10/4/1978 -
7 Normandy @ Country Creek 5/1/1985 -
8 Normandy@ I-295 - -
9 SR 13 @ Beauclerc Rd. 10/12/1973 -
10 US 17 @ I-295 South 2/23/1972 11/9/2003
11 US 17 @ Plymouth 8/19/1986 10/2/2004
12 A1A @ Marlin Street 3/1/1987 4/15/2002
13 A1A @ Ponte Vedra Lakes Blvd. 5/1/1987 5/2/2002
14 US 17 @ I-295 North 11/16/1995 -
15 US 17 @ Heckscher Drive 2/1/1987 -
16 US 17 @ Emerson Park Blvd. 5/17/1992 -
17 US 17 @ US 19 3/26/1992 -
Table 2 summarizes the characteristics of the nine sites, including driveway code, number of
CGTLs, and separator type. Categorical values were used to describe differences in these three basic
site characteristics. For driveways, a zero code represents absence of driveways within 250 feet of
an intersection, one represents intersections which do not have driveways in a non-CGTL direction,
and, two, intersections which do not have driveways in the CGTL direction. As far as the number of
continuous green through lanes are concerned, intersections with one CGTL are coded as one while
those with two CGTLs are coded as two. Three main methods were used to separate continuous
green traffic from other movements. These are double white lines (coded as a zero), raised rounded
domes (coded as one), and raised curbs (coded as two). These methods help motorists identify the
special use of CGTLs, provide a buffer between vehicles making left turns from minor roads and
vehicles in the CGTL, and discourage swerving from adjacent lanes as drivers tend to avoid being
stopped by a red light.
7

Safety Analysis
Table 2: Basic Site Characteristics
8
Site Characteristic
Intersection Number*
1 2 3 4 5 6 7 8 9
Driveway code 0 1 0 0 0 0 2 0 1
Number of CGGLs 2 2 2 2 1 2 1 1 2
Separator type 0 0 0 0 0 1 0 2 0
*Intersection numbers (1 through 9) are in the order presented in Table 1.
Crash data were collected from the Florida Department of Transportation (FDOT) database
known as the Crash Analysis Reporting (CAR) System. The data are for 398 crashes that occurred
at nine CGTL intersections from 2003 to 2008. The data categorize the degree of injury severity
as none, possible, non-incapacitating, incapacitating, and fatal. Generally, possible injury and
non-incapacitating injuries represent the same injury severity level, i.e., non-incapacitating injury.
Therefore, they were combined giving a total of four levels of injury severity. Other variables
are site characteristics, traffic, and environmental conditions at the time of the crash. The site
characteristics included whether or not the crash involved vehicles using CGTL traffic lanes, the
number of CGTLs, and if there is a driveway in the vicinity of the intersection. The environmental
conditions are weather and lighting. Speed limit and annual average daily traffic (AADT) are the
traffic factors. Other factors considered are driver age, number of vehicles involved in the accident,
and time of day. All crashes that occurred within 250 feet of the study intersection were assumed to
be influenced by the intersections. The crashes were further screened by examining crash diagrams
to remove those that were not intersection related but within 250 feet of the intersection. Table 3
shows a description of each variable used in the model.
Analytical Techniques
Three methods were used in this study to analyze the data. The first is proportions analysis, which
uses simple percentage calculations to examine crash patterns at CGTLs. The second is comparative
analysis to determine if there is any underrepresentation or overrepresentation of some crash patterns
on CGTL, and the third method is the ordered probit (OP) model, which was used to model injury
severity.
In the comparative analysis, four distinct conflict types are analyzed. They are those due to
lane changes (pattern one versus pattern 10), rear-end crashes (patterns two and three versus pattern
six), angle crashes involving left-turning traffic from a minor street (patterns four and five versus
pattern seven), and other crashes. For each conflict pattern, the proportion of the total intersection
crashes for the CGTL direction was compared with the proportion in the non-CGTL direction using
a paired-t test. This test is appropriate in analyzing samples which have two different treatments,
i.e., paired treatments. In this case, every intersection has two treatments at each mainline direction:
installation of continuous green through lanes (in the CGTL direction) and normal lanes (in the
non-CGTL direction). This method provides the statistic which is used to determine if there is a
significant difference between the proportion means for the CGTL and non-CGTL directions. The
null hypothesis is that the proportions of the aforementioned three conflict types are equal for the
CGTL and non-CGTL directions, while the alternative hypothesis is that the proportions of the
conflict types are not equal for the CGTL and non-CGTL directions. The alternative hypothesis
is accepted only when the data suggest sufficient evidence to support it, hence rejecting the null
hypothesis. All conflict types were tested at the 95% confidence level.

Table 3: Description of the Model Variables
Explanatory Variables Categories Explanation
Injury Severity 0 No injury
1 Non-incapacitating injury
2 Incapacitating injury
3 Fatal
Crash Conflict Group 0 Rear-end (patterns 2, 6, and 9)
1 Angle (patterns 4, 5, and 7)
2 Lane-change (patterns 1 and 10)
3 Left-turn (patterns 8 and 11)
4 All other
On CGTL 0 Not involving vehicles on CGTL
1 Involving vehicles on CGTL
Number of CGTL 0 One CGTL
1 Two CGTLs
Driveway 0 No driveways
1 Driveways on normal direction
2 Driveways on CGTL direction
Separator Type 0 Double white lines
1 Rounded domes
2 Raised concrete curb
Lighting 0 Daylight
1 Dark
Weather 0 Clear and cloudy
1 Rainy
JTRF Volume 50 No. 1, Spring 2011
Time of day 0 Early morning/Late at night (midnight to 6:00 am)
Speed limit 0 45 mph
1 Morning (6:00 am to noon)
2 Afternoon (noon to 6:00 pm)
3 Evening (6:00 pm to midnight)
1 50 mph
Age (years) 0 =65
Number of vehicles Continuous variable
Annual Average Daily
Traffic (AADT)
Continuous variable
9

Safety Analysis
An ordered probit (OP) model was used because injury severity is ordered, i.e., from no injury
(property damage only), possible injury, non-incapacitating injury, incapacitating injury and killed.
Several previous studies have used this method in modeling injury severity (Quddus et al. 2002,
Kockelman and Kweon 2002, Abdel-Aty 2003, Abdel-Aty and Keller 2005). Because the injury
data used in this study are categorical, the use of OP is appropriate as it requires no assumptions
regarding the ordinal nature of the dependent variable (Quddus 2002). The OP model for four
categories of injury severity is given in the following form (Kockelman 2002, Washington et al.
2003):
(1)
Where q n is the observed injury severity (coded as a categorical variable), and µ i values are the
thresholds (cutoffs) that define each q n . The probabilities associated with ordinal outcomes of an OP
model are calculated as:
(2)
Where φ is the standard normal cumulative density function, β is the vector of estimated parameters,
and z a vector of model variables. Predictions from the OP models are done by considering the
thresholds and comparing the predicted probability with the given cutoff probability boundaries and
then classifying injuries based on the cutoffs.
RESULTS
Crash Pattern
This analysis involved careful examinations of crash diagrams to determine distinct crash patterns.
After reviewing crash diagrams and narratives in crash reports, crashes were classified into the 11
distinct patterns shown in Figure 2. Crashes that did not fall into the 11 patterns shown in Figure 2
were combined into pattern 12 as shown in Table 4. This table shows a summary of the percentages
of each of the 12 crash patterns for each intersection. Most of the crashes in pattern 12 occurred on
the side street (minor street direction) while few involved vehicles in the major street. Some crash
types in pattern 12 include rear-end and right-turning crashes from the minor street, run-off the
road crashes, pedestrian crashes, and collisions with vehicles from driveways. These crashes were
combined into one group and included in the comparative analysis.
The data in Table 4 also show that there are more crashes involving lane changing in the CGTL
direction (conflict pattern one) compared with the direction which has traditional through lanes
(conflict pattern 10). Approximately 6.01% of the crashes involved vehicles changing lanes in
the CGTL direction while only 1.78% involved lane changing vehicles in the traditional through
lanes. The percentages of rear end crashes for both traditional through lanes direction (pattern six)
and the CGTL direction (patterns two and three) appear to be approximately equal (23.60% for
continuous through lanes and 23.39% for traditional through lanes). A thorough examination of the
crash diagrams and police report narratives revealed that rear end crashes on the traditional lanes
involved through and right-turning vehicles, mostly caused by right-turning vehicles reducing speed
to perform a right turning maneuver. Crash patterns two and three, which represent rear-end crashes
in the CGTL direction, were mostly caused by motorists who unexpectedly stopped in the CGTL.
10
n
( k)
ϕ ( µ − β z ) − ϕ(
µ − z )
P β
= k+1
n
k
n

Figure 2: CGTL Intersection Crash Patterns Classified by Conflict Types
JTRF Volume 50 No. 1, Spring 2011
Conflict categories four and five represent right angle crashes involving left-turning vehicles
from the minor street and vehicles crossing the intersection from the CGTL direction. The difference
between these two categories is that conflict category four involves right angle crashes with drivers
who are in the non-continuous lane while category five is for right angle collisions that occur on the
CGTL. The main causes of conflict category five crashes are the motorists in the non-continuous
lane who are supposed to stop on red but inattentively cross the intersection by assuming that the
continuous green arrow applies to their lane. Conversely, conflict category five crashes are caused
by left-turning vehicles veering into the CGTL, disregarding lane separation markers. It is observed
from the data in Table 4 that there were no crashes caused by crash conflict pattern five at intersection
eight (Normandy at I-295) due to the use of a curb to separate continuous green through movements
from other movements. Also, there were no crashes caused by conflict pattern one (lane changing
from normal lane to CGTL to avoid stopping at intersection) because the separation curb is extended
to both sides of the intersection.
11

Safety Analysis
Table 4: Proportions of Crashes by Pattern Type for Each Intersection in the Study
12
Conflict
Pattern Conflict Description 1 2 3 4 5 6 7 8 9 Average
Sideswipe between non-CGTL and CGTL
1
5.88% 9.52% 6.56% 7.89% 0.00% 5.41% 17.14% 0.00% 7.14% 6.01%
through traffic
2 Rear-end on CGTL 33.33% 0.00% 0.00% 21.05% 11.76% 13.51% 11.43% 3.08% 12.50% 14.25%
3 Rear-end on non-CGTL 11.76% 14.29% 3.28% 10.53% 5.88% 8.11% 2.86% 6.15% 19.64% 9.35%
Angle collision between left-turn minor street
4
1.96% 0.00% 4.92% 0.00% 0.00% 0.00% 0.00% 0.00% 5.36% 1.78%
and non-CGTL through traffic
Angle collision between left-turn minor street
5
1.96% 0.00% 3.28% 5.26% 11.76% 8.11% 11.43% 0.00% 7.14% 4.68%
and CGTL through traffic
6 Rear-end on conventional approach 17.65% 38.10% 16.39% 18.42% 11.76% 18.92% 5.71% 50.77% 28.57% 23.39%
13.73% 14.29% 0.00% 7.89% 8.82% 2.70% 0.00% 4.62% 5.36% 6.68%
Right-angle collision between through traffic
from conventional approach and left-turn
7
traffic from minor street
0.00% 9.52% 1.64% 5.26% 8.82% 5.41% 2.86% 6.15% 7.14% 4.23%
Angle collision between through traffic from
conventional approach and left-turn traffic
8
from CGTL approach
9 Rear-end collision on left turn lane 0.00% 4.76% 0.00% 0.00% 2.94% 2.70% 2.86% 6.15% 1.79% 2.00%
10 Sideswipe on conventional approach 0.00% 4.76% 1.64% 2.63% 0.00% 0.00% 2.86% 6.15% 0.00% 1.78%
0.00% 0.00% 1.64% 0.00% 2.94% 0.00% 34.29% 0.00% 0.00% 3.12%
Angle collision between left turn traffic from
minor and major streets
11
13.73% 4.76% 60.66% 21.05% 35.29% 35.14% 8.57% 17.92% 5.36% 22.72%
All other conflict patterns not represented by
categories 1 through 11
12
*Intersection numbers (1 through 9) are in the order presented in Table 1.

Comparative Analysis
JTRF Volume 50 No. 1, Spring 2011
The data show that there are more crashes from lane changing maneuvers in the CGTL direction
(pattern one) than in the non-CGTL direction (pattern 10). The results in Table 5 indicate that there
is a significant difference between the proportions of lane changing crashes in the CGTL and non-
CGTL directions (p-value = 0.038) at the 95% confidence level. The high proportion of lane changing
crashes might be due to motorists who suddenly swerve to the CGTLs to avoid being stopped by
the red light on non-CGTLs. The data suggest a slightly higher average proportion of rear-end
crashes in the non-CGTL direction (0.256% of all crashes) than in the CGTL direction (0.229% of
all crashes). Rear-end crashes on CGTLs are most probably caused by motorists who are unfamiliar
with how the CGTLs operate and who unexpectedly stop in the CGTL by mistakenly observing a
red light meant for non-CGTLs. However, the results in the table show that this difference is not
significant at the 95% confidence level (p-value = 0.736). The observed average proportion of rightangle
crashes involving left-turns from the minor street and vehicles in the CGTL direction was
slightly higher (0.074) than for the non-CGTL direction (0.071). The crash diagrams revealed that
right-angle crashes involving CGTLs are mostly caused by motorists turning left from the minor
street and veering into the CGTLs instead of turning to the non-CGTLs. Furthermore, Table 5 shows
there is no significant difference in the observed proportions of angle crashes (patterns four, five, and
seven) between CGTL and non-CGTL directions.
Table 5: Comparative Analysis Results
Conflict
type
Lane
changing
Rearending
Angle
Direction
Proportion
mean
Standard
Deviation
CGTL 0.072 0.055
Non-CGTL 0.023 0.027
CGTL 0.229 0.135
Non-CGTL 0.256 0.164
CGTL 0.074 0.053
Non-CGTL 0.071 0.059
Injury Severity at CGTL Intersection Crashes
Degrees
of
freedom
t-value p-value Reject
null?
8 2.475 0.038 yes
8 -0.350 0.736 No
8 0.102 0.321 No
The STATA statistical package was used for the ordered probit model runs. Two injury severity models
were estimated as in Abdel-Aty and Keller (2005). The first describes the relationship between injury
severity with different crash conflict patterns while the second explains the relationship between
injury severity and intersection characteristics, environmental conditions, and traffic characteristics.
Table 6 shows the coefficients of the first model. The results indicate that right-angle crashes (crash
conflict group one) and lane changing crashes (crash conflict group two) are significant predictors
of injury severity at CGTL intersections. The level of injury severity is higher for conflict categories
one and two compared with rear-end crashes (crash conflict group zero in Table 3).
13

Safety Analysis
Table 6: Ordered Probit Model for Crash Conflict Groups
Variable Coefficient Standard Error Z P>z
Involving Continuous Green
Through Lane Traffic
Crash Conflict Group
0.2096 0.2135 0.98 0.326
Angle (patterns 4, 5, and 7) 0.4796 0.2401 2.10 0.036
Lane-change (patterns 1 and 10) 0.5035 0.3038 1.99 0.046
Left-turn (patterns 8 and 11) 0.4416 0.3385 1.31 0.192
All other -7.2223
Thresholds
0.0000 0 1
µ 1 -0.1312 0.1054
µ 2 0.6819 0.1087
µ 3 1.4721 0.1274
µ 4 2.3822 0.2086
The results of the second model in Table 7 indicate crashes that take place during the time
categories of 6:00 a.m. in the morning to noon and noon to 6:00 p.m. result in lower injury severity.
The results also suggest that drivers 65 years and older have higher injury severity levels. Also, as
the table shows, all the other variables in the model had statistically insignificant coefficients. These
include speed limit, rounded domes, raised concrete curbs, number of vehicles, and annual average
daily traffic.
CONCLUSIONS AND RECOMMENDATIONS
This study was conducted to examine the safety characteristics of unconventional continuous
green through lanes at nine sites in Jacksonville, Florida. A thorough review of crash data
resulted in 11 distinct crash conflict patterns that were used to examine the influence of CGTLs
on the safety characteristics of the study intersections. Three analysis methods were used: general
proportions analysis, comparative analysis, and injury severity ordered probit modeling. Based
on the proportions analysis, there are three common types of crashes that involve CGTL traffic:
(1) sideswipe crashes caused by motorists weaving from adjacent through lanes to avoid having to
stop for the red signal indication, (2) angle crashes caused by motorists turning left from a minor
street and swerving into the CGTL by disregarding the “do not change lane” barriers such as double
white lines and rounded domes, and (3) rear-end crashes caused by motorists who unexpectedly
stop in the CGTL. The results of the proportions analysis show that on average the proportion
of sideswipe crashes in the CGTL was 6.01% (conflict pattern one) compared with 1.78% in the
opposite direction (conflict pattern 10). Also, on average, 4.68% of all crashes were caused by leftturning
vehicles from the minor direction (conflict pattern five) crossing to the CGTLs. Typically,
conflict pattern five is caused by inattentive drivers or motorists who are not familiar with the
presence of CGTL. It is also worth mentioning that on average there were more rear-end crashes on
continuous green through lanes (conflict pattern two, 14.25%) compared with normal lanes (conflict
pattern 3, 9.35%).
The results of the comparative analysis which employed a paired t-test indicate that there is a
significant difference between the proportions of sideswipe crashes in the CGTL direction compared
with the opposite direction. On the other hand, the paired-t test results did not suggest a significant
difference between the proportions of rear-end and right-angle crashes for the CGTL and normal
directions.
14

JTRF Volume 50 No. 1, Spring 2011
Table 7: Injury Severity Results Based on Site and Traffic Characteristics and
Environmental Conditions
Variable Coefficient Standard Error Z P>z
Annual average daily traffic 1.8E-05 0.0000 1.43 0.152
Number of continuous green through
lanes 0.2852 0.3187
0.9 0.371
Number of vehicles
Traffic Involved
0.0727 0.1482 0.49 0.624
Involving vehicles on CGTL
Presence of driveways
-0.2008 0.2258 -0.89 0.374
Driveways on normal direction -0.1636 0.3536 -0.46 0.644
Driveways on CGTL direction
Separator type
-0.3025 0.5080 -0.6 0.552
Rounded domes 0.3887 0.5293 0.73 0.463
Raised concrete curb
Lighting
0.2079 0.5741 0.36 0.717
Dark
Weather
0.4913 0.4173 1.18 0.239
Rainy
Time of day
-0.3694 0.5263 -0.7 0.483
Morning (6:00 am to noon) -0.9012 0.4577 -1.97 0.049
Afternoon (noon to 6:00 pm) -0.8451 0.3395 -2.49 0.013
Evening (6:00 pm to midnight)
Speed Limit
-0.5480 0.4598 -1.19 0.233
50 mph
Age (years)
0.1241 0.3955 0.31 0.754
25 to 64 0.4094 0.2277 1.8 0.072
>=65 0.8517
Thresholds
0.3399 2.63 0.012
µ 1 -1.8869 0.5142
µ 2 0.7630 0.4335
µ 3 1.5785 0.4359
µ 4 2.3698 0.4400
Two different ordered probit models were developed: one based on crash pattern types and
another considering site conditions, environmental factors, and traffic conditions. The results of
the first model indicate that angle crashes and crashes involving lane changing maneuvers are
significantly more severe than rear-end crashes. For the second model, only time of day and age
of driver were found to be significant in predicting injury severity level. Lower injury severity was
observed for crashes that occurred during the day, i.e., between 6:00 a.m. and 6:00 p.m. Crashes that
involved drivers who were 65 years or older had higher injury severity level.
15

Safety Analysis
Based on the observations of this study, the following design features are recommended as
they may improve the safety of CGTL intersections: advance warning signs and highly visible
raised separators. Advance warning signs provide guidance to motorists as to the purpose of the
continuous through lanes and lane use instructions. This is particularly helpful to non-commuters
who are not familiar with continuous green through lanes. Providing highly visible raised separators,
in lieu of double white lines and raised rounded domes, creates a distinct separation between the
continuous through traffic and the adjacent lanes. This separation will prevent lane changing caused
by motorists crossing the double white lines.
Further research is needed to study the influence of the factors which were not included in this
study, such as type of left-turn restrictions (protected versus permitted), downstream and upstream
traffic conditions, advance warning signage, and typical driver population, among other factors.
Efforts are underway to conduct a comparative analysis between CGTL intersections and traditional
“T” intersections. There are also plans to increase the dataset to include CGTL intersections in other
parts of Florida. It is recommended that some specific site characteristics such as signage and lane
markings, left-turning restrictions, and other pertinent variables be included in the analysis. Finally,
because the study used one locality, further studies of similar intersections elsewhere are required to
permit generalizations of the results in this paper.
References
Abdel-Aty, M. and J. Keller. “Exploring the Overall and Specific Crash Severity Levels at Signalized
Intersections.” Accident Analysis and Prevention 37, (2005): 417–425.
Abdel-Aty. M. “Analysis of Driver Injury Severity Levels at Multiple Locations Using Ordered
probit Models.” Journal of Safety Research 34, (2003): 597– 603.
Hummer, J.E. and J.L. Boone. “Travel Efficiency of Unconventional Suburban Arterial Intersection
Designs.” Transportation Research Record: Journal of the Transportation Research Board 1500,
(1995): 153-161.
Jarem, E.S. “Safety and Operational Characteristics of Continuous Green Through Lanes at
Signalized Intersections in Florida.” Presented at ITE Annual Meeting and Exhibit, Orlando, FL,
August 2004.
Kockelman, K.M. and Y. Kweon. “Driver Injury Severity: An Application of Ordered Probit
Models.” Accident Analysis and Prevention 34, (2002): 313–321.
Mitra, S., H.C. Chin and M.A. Quddus. “Study of Intersection Accidents by Maneuver Type.”
Transportation Research Record: Journal of the Transportation Research Board 1784, (2002): 43-
50.
Persaud, B. and T. Nguyen. “Dissagregate Safety Performance Models for Signalized Intersections
on Ontario Provincial Roads.” Transportation Research Record: Journal of the Transportation
Research Board 1635, (1998): 113-120.
Quddus, M.A., R.B. Noland and H.C. Chin. “An Analysis of Motorcycle Injury and Vehicle Damage
Severity Using Ordered Probit Models.” Journal of Safety Research 33, (2002): 445– 462.
Wang, X. and M. Abdel-Aty. “Analysis of Left-turn Crash Injury Severity by Conflicting Pattern
Using Partial Proportional Odds Models.” Accident Analysis and Prevention 40, (2008): 1674–1682.
Washington. S.P., M.G. Karlaftis, and F. L. Mannering. Statistical and Econometric Methods for
Transportation Data Analysis. Chapman & Hall/CRC, Florida, 2003.
16

JTRF Volume 50 No. 1, Spring 2011
Weerasuriya, S.A. and M.C. Pietrzyk. “Development of Expected Conflict Value Tables for
Unsignalized Three-Legged Intersections.” Transportation Research Record: Journal of the
Transportation Research Board 1635, (1998): 121-126.
Thobias Sando is an assistant professor at the University of North Florida in Jacksonville, Florida.
He teaches and conducts research in the area of transportation engineering. His research interests
include modeling of highway safety data, intermodal facility design, intelligent transportation
systems, operational analysis of bicycle and pedestrian facilities, and traffic simulation.
Deo Chimba is an assistant professor at the University of Tennessee in Nashville. He teaches
transportation engineering and conducts research in highway safety modeling, transportation
planning, and traffic simulation.
Valerian Kwigizile is an assistant professor of civil engineering at the West Virginia University
Institute of Technology where he teaches courses in transportation and traffic engineering. His
research interests include intelligent transportation systems, highway safety, roadway design, and
multimodal transportation systems.
Holly Walker works in the Quality Assurance Section/Exceptions and Variations in the Roadway
Design Office in Tallahassee, reviewing and analyzing technical engineering design documents for
recommendations to the state roadway design engineer. Additionally, she is involved in developing
and delivering as appropriate, training on crash analysis procedures.
17

Transit Passenger Perceptions: Face-to-Face
Versus Web-Based Survey
by Laura Eboli and Gabriella Mazzulla
In this paper, face-to-face and web-based survey methods of collecting transit passenger perception
data are compared using two transit customer satisfaction survey tools. Multivariate statistical
analyses are applied to determine the differences between the two surveys. Some differences in
behavior and attitudes of web survey respondents compared with those from a face-to-face survey
are found. The results can help transit agencies manage their bus services to improve passenger
satisfaction and service quality.
INTRODUCTION
Customer satisfaction surveys are tools for capturing consumer perceptions of service. To meet
customer requirements, it is fundamental to provide good basic public services, such as public
transport and social security, which are subject to different conditions and performance standards
than private sector companies. Capturing passenger perceptions and evaluating customer satisfaction
allow transit agencies to improve service quality and maintain passenger loyalty. Moreover, in
a regulated market such as public transport, good management cannot be based only on service
efficiency and effectiveness (e.g., fare revenues and the number of passengers), but, most of all,
service quality as measured by different service attributes. In the first place, a transit service is
characterized by frequency, travel time, and route characteristics such as length, number of stops,
distance between stops and accessibility to stops, and reliability in terms of schedule adherence.
Other important transit service attributes are information provided to users about departure and
arrival scheduled times, boarding/alighting stop location, fares, climate control, seat comfort, ride
comfort – including the severity of acceleration and braking – odors, and vehicle noise. Cleanliness
of vehicles, terminals and stops, safety, and security are also important quality of service measures.
Still others are the fare, personnel appearance and helpfulness, environmental protection, and
customer services such as ease of purchasing tickets and administration of complaints. Each service
attribute plays a part in determining the level of quality of service. As a consequence, passengers’
perceptions of the overall service depend on how they perceive the different service attributes.
This paper focuses on the analysis of transit passenger satisfaction regarding an extra-urban bus
service used by university students. Data gathering was based on traditional face-to-face interviews
and the more innovative Web-based surveys. A comparison of these two different data collection
methods is made to highlight the advantages and disadvantages of both surveys. Although Internet
surveys using online panels are common, there are few studies that compare the two different
surveys, and even fewer studies regarding transit passenger perceptions. Our work aims at the
comparison between face-to-face and web-survey interviews, thus filling a gap in the literature.
The relevance of this paper is certainly the lack of studies about the topic in the transport sector. In
addition, the findings resulting from the comparison of the two surveys can be very useful for transit
agencies to manage bus services.
LITERATURE REVIEW
Traditionally, surveys are carried out by three main methods: face-to-face surveys where the
interviewer conducts a personal interview by asking questions of the respondent; telephone surveys
where an interviewer conducts a survey by contacting respondents by telephone; and mail surveys
19

Transit Passenger Perceptions
where questionnaires are mailed to sampled individuals who complete and return them by mail
(Fricker et al. 2005). Face-to-face and telephone surveys are interviewer-administered methods
whereas mail surveys are self-administered (Biemer and Lyberg 2003). In addition, there are new
technologies developed in the last decade for communicating and interfacing with respondents in
their homes, at work, and during travel (Nicholls et al. 1997). Each method has advantages and
disadvantages and its selection is often complex and depends on the objective of the survey, its
characteristics, design and methodological issues, and the financial resources available (Biemer and
Lyberg 2003).
Face-to-face interviews provide for the maximum degree of communication and interaction
between the interviewer and the respondent. Therefore, it is often associated with good quality
data and it is preferred by many researchers because it allows long and complex interviews to be
conducted, and it is characterized by flexible questions. Owing to the presence of the respondent, the
interviewer can gain cooperation, obtain personal information, make direct observations during the
interview, record spontaneous reactions, and ensure that the respondent’s answers are not affected by
the presence of other persons. These surveys are characterized by relatively high response rates and
elevated coverage of the general population. Its disadvantage is the tendency of respondents to be
more concerned about the interviewer than in providing accurate answers. In fact, interviewers are
an important error source in such surveys and tend to affect respondents in different ways. Another
disadvantage is “misbehavior by interviewers” (Kiecker and Nelson 1996) and refers to activities
that are dishonest. A face-to-face interview is usually more expensive than the other methods of data
collection since it requires the interviewer to visit or meet the respondents at home, work, or public
places. This fact usually requires more time and personnel resources.
A telephone interview has not always been accepted as a good data collection method for
social and economic research. The increased interest in this approach, however, is its lower cost
and the increased coverage of the targeted population (Biemer and Lyberg 2003). Groves and Kahn
(1979) show that it can provide comparable quality data to those from face-to-face surveys. Indeed,
both face-to-face and telephone interviews have very similar characteristics in that they can create
interviewer variance and social desirability bias which describes the tendency of respondents to
reply in a manner that will be viewed favorably by others. However, the literature suggests that
these effects are somewhat less in telephone surveys than in face-to-face interviews and that social
desirability bias might be less in telephone interviews than in face-to-face interviews because of the
anonymity of the interviewer. Also, telephone interviews are less complex and considerably shorter
than face-to-face interviews, with most lasting 30 minutes or less. Typically, their response rates are
lower than in face-to-face surveys of comparable type and size (Biemer and Lyberg 2003).
Today, face-to-face and telephone interviews are increasingly conducted using CAI (Computer-
Assisted Interviewing) technology and its variants, CAPI (Computer-Assisted Personal Interviewing)
and CATI (Computer-Assisted Telephone Interviewing), where the interviewer asks questions and
enters the respondent’s answers using a computer program. As discussed by Groves and Tortora
(1998), the theoretical and logical advantages associated with CAI are not always supported by
data. Nevertheless, studies show clear reductions in indicators of measurement error and item nonresponse
rates in such methods.
The major advantage of the presence of interviewers is that it ensures respondents understand
the questions and a uniform interpretation of the question leads to more accurate responses (Conrad
and Schober 2000). However, its major disadvantage is the possibility of having biased results
(Beatty 1995) unless each interviewer handles and interprets each question in exactly the same
manner. Interview methods such as mail and Web-based surveys that do not use interviewers have
different features. For example, in a mail survey, because there is no interviewer, the questionnaire
and instructions are made easy to understand. To a much greater extent, the quality of data from noninterviewer
surveys hinges on the quality of the questionnaire design. However, mail surveys may
have some advantages in terms of lower cost and reduced risks of social desirability bias associated
with self-administration caused by the privacy involved in completing the questionnaire. In addition,
20

JTRF Volume 50 No. 1, Spring 2011
the response rate to mail surveys can vary considerably depending upon the experience, skill, and
knowledge of the survey organization. Also, the response rates are lower than in interviewer-assisted
surveys, they have a greater risk of considerable item non-response rates, and require a long time
to get acceptable response rates. In addition, it is not possible to ensure that the intended people
complete the questionnaire or that the respondent does not collaborate with others.
The increasing popularity and wide availability of World Wide Web technologies provide
researchers with a new data collection method called web survey. This method uses the internet
to collect data from sampled populations (Al-Subaihi 2008) by interactive interviews or by
questionnaires purposefully designed for self-completion. For example, electronic one-to-one
interviews can be conducted via e-mail or chat rooms. Questionnaires also can be administered
by e-mail (e.g., using mailing lists), postings to newsgroups, or using fill-in forms (Eysenbach and
Wyatt 2002) on the Internet. Over the last 10 years, Web-based surveys have become widely used in
the social sciences and educational research (Couper 2000), and a further increase is expected since
it allows access to a large number of potential respondents (Couper 2000, Loosveldt and Sonck
2008).
Web survey design focuses more on programming ability and web page design rather than
traditional survey methodology (Couper 2001). As Al-Subaihi (2008) reports, the effects of variables
related to web survey on response rates and data accuracy have been of interest to researchers
and applied statisticians and continue to receive considerable attention in the survey methodology
literature. (See, for example, Coomber 1997, Cook et al. 2000, Couper 2000, Dillman and Bowker
2001, Ganassali 2008, Converse et al. 2008.) Web surveys, however, have been suggested to be
far from perfect (Gorman 2000). That is, their non-response rates and coverage errors may be high
(Couper 2000) and respondents may falsify their demographic information. The use of panels
specifically recruited for online research though can mitigate these weaknesses (James 2003). Like
mail surveys, they are cheaper to do and less time consuming than interviewer-administered surveys.
In addition to web-surveys, a number of computerized versions of self-administered interviews
have been developed, such as Disk By Mail (DBM) and Electronic Mail Survey (EMS), Touchtone
Data Entry (TDE) and Voice Recognition Entry (VRE), Computer-Assisted Self-Interviewing
(CASI) with its variants Audio CASI (or ACASI), and Telephone Audio CASI (T-ACASI). A
description of these methods is in Ramos et al. (1998) and an extensive literature review of web
surveys is reported in Schonlau et al. (2002). Al-Subaihi (2008) also presents an interesting literature
review based on technical factors (method of presentation, graphics, or colors), methodological
factors (cost, coverage sampling, and validity), and social factors (social behavior variables such as
age, gender, ethnicity, level of education).
Some studies compare different survey methods. For example, Bonnel and Le Nir (1998)
compare face-to-face and telephone interviews; telephone and mail surveys are compared in Walker
and Restuccia (1984) and Coderre et al. (2004). Al-Subaihi (2008), Braunsberger et al. (2007),
and Fricker et al. (2005) compare telephone interviews and Web surveys while Cobanoglu et al.
(2001) and McDonald and Adam (2003) compare mail interview and Web surveys. While these
comparisons provide useful information, except Heerwegh and Loosveldt (2008) and Bayart and
Bonnel (2008), little research has been done to compare Web-based and face-to-face interview
surveys. And the only such work regarding transport services is by Elmore-Yalch et al. (2008) who
analyzed passenger perceptions collected by telephone interviews and compared them with similar
data collected by Web surveys. Because not much has been done on this comparison in transport,
this study fills a gap in the literature.
21

Transit Passenger Perceptions
METHODOLOGY
Survey
In this paper, customer satisfaction data about transit are collected by face-to-face and Web-based
interviews. A face-to-face survey was conducted in 2006 using a sample of users of an extra-urban
bus line connecting some towns in the province of Cosenza located on the Tyrrhenian coast with
the University of Calabria in Cosenza, South Italy. Bus service is supplied by one of the largest
transit agencies operating in the province. The bus line covers a distance of about 103 km, and
the route has about 40 stops. The service spans 12 hours, from 6:00 a.m. till 6:00 p.m. and service
frequency is less than one run per hour. The price of a one-way ticket varies with distance, from
a minimum of about 1.50 Euros to a maximum of about 4.50 Euros. Rail transit services are not
available in the study area and mode choice is very much inclined toward the private car. In 2006
the transit agency sold about 280,000 tickets and 2,400 weekly or monthly travel cards. About 1,000
University students daily reach the campus from the Tyrrhenian coast by bus service.
The Web-based survey was conducted in 2008 and was addressed to all students of the
University of Calabria who lived in the province of Cosenza and used the extra-urban bus services
to access the campus. While some students used these transit services daily, others used them to go
home on weekends.
Questionnaire Design
The questionnaire is made up of about 50 items grouped into three sections (see the Appendix). The
first section aims to collect some socio-economic data about the passengers interviewed, such as
age, gender, major course of study, post graduate classification, place of residence, family income,
number of family members, car driving license and number of owned cars, car availability, etc.
The second section collects data about boarding/alighting, access/egress transport mode, access/
egress travel time, waiting time, time on board, bus ticket and fare. In the last section, respondents
were asked to rate the importance of and their satisfaction with 16 service attributes, in addition
to a request for them to rate their satisfaction of the overall service. The service attributes are
availability of a bus stop near home, route, service frequency, reliability of runs in terms of schedule
adherence, reliability of runs in terms of on-time service, availability of shelter and benches at bus
stops, availability of seats, cleanliness of vehicle interior, seats, and windows. Others are ticket
cost, availability of schedule/maps at bus stops, availability of service information by phone or
Internet, vehicle reliability, competence of drivers, security against crimes at bus stops, personnel
helpfulness, administration of complaints, and the physical conditions of bus stops.
In the face-to-face survey, an interviewer administered a paper questionnaire to a sample of
150 users at the bus terminal at the university campus. The questionnaire was completed in five
to eight minutes by each respondent. In a second survey, an invitation to complete a Web-based
questionnaire was sent to 9,900 students using the e-mail addresses provided by the university. Of
these, 329 responded giving a response rate of 3.32%. This low rate is because many of those not
responding did not use transit services or their e-mail addresses provided by the university were
wrong. Of the 329 responding, 251 (76.3%) completed the questionnaire well enough for their
responses to be included in the study. The other 78 (21.7%) could not be considered because they
did not specify the bus service used. Some of the interviews (92 out of the 251 who participated
in the Web-based survey) were completed by passengers traveling on the same bus lines as those
interviewed in the face-to-face survey.
The descriptive statistics in Table 1 show the two samples are similar with most of the
respondents being female, younger than 22 years old, and belonging to middle income-class
families. The average number of people in a respondent’s family is about four (4.35 for face-to-face
respondents and 4.07 for online ones) while the average number of people with drivers licenses is
22

JTRF Volume 50 No. 1, Spring 2011
about three (3.24 for face-to-face respondents and 3.18 for online ones). Finally, the average number
of cars per family is 1.9 in both cases.
Table 1: General Characteristics of the Respondents
Characteristic Value
Face-to-face survey
(150 respondents)
Item
(%) response
rate (%)
Gender Male 37
100.0
Age up to 22 years 43
Female 63 66
from 22 to 24 years
from 24 to 27 years
39
11
100.0
28
9
over 27 years 7 8
Family size 4.35 4.07
Family income level * Lower
lower-middle
Middle
upper-middle
Upper
20
16
52
9
3
100.0
19
27
44
7
3
Faculty Arts
Economics
Engineering
14
29
23
15
20
20
Math., Phys. and Nat. Science 13 100.0 18
Pharmacy
Politics
11
8
11
10
Inter-faculty 2 6
Family members
Members with driving
license
3 or less
4
5 or more
2 or less
3
4 or more
17
43
40
25
35
40
100.0
100.0
27
45
28
21
46
33
Number of cars per family 1
2
3 or more
29
51
20
100.0
24
60
16
Average 3.24 3.18
Car driving license
ownership
did not own car driving license
own car driving license
3
97
100.0
3
97
Car availability did not own car
own car
53
47
100.0
74
26
Ticket/card one-way ticket
one-day travel card
weekly travel card
0
99
0
100.0
22
73
0
monthly travel card 1 5
Travel time minutes 48 52
Travel time including
access and egress times
minutes 73 57
Web-based survey
(92 respondents)
Item
(%) response
rate (%)
34
93.5
55
90.2
91.3
92.4
92.4
92.4
90.2
92.4
90.2
100.0
* The lower level is to 1,000 Euros, the lower-middle from 1,000 to 2,000 Euros, the middle from 2,000 to 4,000
Euros, the upper-middle from 4,000 to 5,000 Euros, the upper is over 5,000 Euros. The classes of income refer
to net monthly income of a family unit.
23

Transit Passenger Perceptions
Transit users were asked to specify their travel times to the university by bus. The reported
average travel time in the face-to-face survey in Table 1 is about 48 minutes, while it is about 52
minutes for those who completed the Web-based survey. Although these times are comparable,
there are notable differences, such as the average total travel time including waiting and access/
egress times, which are about 73 minutes for those in the face-to-face survey and 57 minutes for
the others in the Web-based survey. The travel times of those in the face-to-face interviews are
reliable compared with the travel times of those in the Web survey. This is because those in the Web
survey did not understand the difference between total travel time and on-board travel time, given
that many of them gave similar values for both. Examining the item responses the face-to-face
survey does not produce loss of information, while for each item statement there are 6%-10% nonrespondents
in the Web-based survey.
Analytical Techniques
To determine the differences between the samples in the face-to-face and Web-based surveys, three
analytical techniques were employed. Paired t-test was used to compare the means of the variables
in the two surveys. Specifically, it was used to test the significance of the differences between the
sample means in the face-to-face and Web-based surveys regarding the importance and satisfaction
ratings of the service quality attributes. The null hypothesis is that there are no differences between
the two observations. If the probability associated with a t value is low (< 0.05), there is evidence
to reject the null hypothesis.
Next, the Fisher F-test is used to compare the variances in the two surveys by testing the null
hypothesis that the different populations responding to the surveys have the same variance. More
specifically, this method tests the significance of the differences between the sample variances of the
ratings expressed by users in the face-to-face and Web-based surveys. Finally, discriminant analysis
is used to predict membership in the groups responding to the face-to-face and the Web-based
interviews on the basis of a linear combination of some of the variables in Table 2. This multivariate
statistical technique allows the variables that discriminate between the two surveys to be identified.
FACE-TO-FACE IN COMPARISON TO WEB SURVEY
In both surveys, the respondents expressed their feelings about level of service by rating its importance
and their levels of satisfaction with the service. The survey used 16 service quality variables and a
numerical scale ranging from one to 10. The service characteristics and their descriptive statistics
are in Table 2. While the face-to-face survey provided ratings by all the passengers interviewed, in
the Web-based survey there were six non-responses on average for each service attribute both for
importance and satisfaction rating. The attribute, administration of complaints, is the least rated in
terms of importance with an item response rate of 88%. The averages of the ratings of importance
in the face-to-face survey are higher than the ratings in the Web-based survey. In the face-to-face
survey, the attribute with the lowest importance rating is “availability of service information by
phone and internet” and the highest rated item is “vehicle reliability and competence of drivers.”
The overall average rating of importance from the face-to-face survey is 8.62 compared with 7.68
for the Web-based survey. For each service attribute there is almost a difference of one point between
the rating based on the face-to-face and the Web-based survey.
The t-test test shows that the averages of the importance ratings are dissimilar in both surveys
except for reliability of runs and information through telephones and the Internet etc. (See Table
3.) From Table 2 there are some attributes whose satisfaction ratings are higher in the face-to-face
survey than in the Web survey. These include security and personnel helpfulness. The attributes for
which the satisfaction ratings are lower in the face-to-face interviews are reliability in terms of ontime
performance and availability of schedules or maps at bus stops. In both surveys the average
rating of satisfaction is about 6.5. However, in the face-to-face survey, the range of the average
24

JTRF Volume 50 No. 1, Spring 2011
satisfaction rating is from 3.63 (“availability of shelter and benches at bus stops”) to 8.49 (“security
against crimes at bus stops”). In the Web-based survey the range is from 4.74 (“physical condition
of bus stops”) to 7.46 (“vehicle reliability, competence of drivers”). The service attributes, schedule
adherence, vehicle reliability, and competence of drivers are rated similarly in both surveys. A t-test
of differences of means confirms the same average satisfaction ratings in both surveys. Other service
attributes with similar average satisfaction ratings as the overall average are the availability of a bus
stop near home, service frequency and reliability of runs, bus cleanliness, vehicle reliability and
driver competence, the administration of complaints, and the physical condition of bus stops.
Table 2: User Perceptions of Services
Face-to-face survey Web-based survey
Importance Satisfaction Importance Satisfaction
rates
rates
rates
rates
Service attributes mean
st.
dev.
mean
st.
dev.
mean
st.
dev.
mean
st.
dev.
Availability of bus stop near home 8.99 1.25 6.53 2.69 7.66 2.69 6.27 2.66
Path 8.42 1.40 7.32 1.79 6.91 2.24 6.74 2.29
Service frequency 8.93 1.09 7.50 1.62 8.14 2.46 7.21 2.18
Reliability of runs that come on
schedule
8.56 1.23 5.81 1.69 8.53 1.95 7.40 1.96
Reliability of runs that come on time 8.83 1.12 6.46 2.16 8.20 2.11 6.48 2.37
Availability of shelter and benches
at bus stops
8.21 1.41 3.63 2.14 7.17 2.69 4.87 2.48
Availability of seats 8.85 1.12 7.31 2.11 7.92 1.89 6.25 2.31
Cleanliness of interior, seats and
windows
9.08 1.06 6.88 2.01 8.21 1.91 7.33 2.12
Ticket cost 8.63 1.10 7.24 1.64 7.45 2.39 6.56 2.15
Availability of schedule/maps at bus
stops
8.27 1.30 3.74 2.19 7.18 2.42 6.17 2.67
Availability of service information
by phone, internet
7.75 1.38 5.43 2.20 7.44 2.27 6.43 2.43
Vehicle reliability, competence of
drivers
9.61 0.76 7.45 1.67 8.65 1.86 7.46 1.95
Security against crimes at bus stops 9.32 1.24 8.49 1.59 7.98 2.37 7.00 2.58
Personnel helpfulness 8.42 1.27 7.93 1.71 7.47 2.32 6.72 2.39
Administration of complaints 7.94 1.32 6.31 1.60 7.25 2.35 5.88 2.39
Physical condition of bus stops 8.11 1.38 5.10 2.07 6.68 2.48 4.74 2.44
Overall service 7.24 0.97 6.95 1.54
25

Transit Passenger Perceptions
Table 3: Tests of Differences of Means and Equality of Variance
t-test of differences of
means
26
Fisher F-test of the equality
between the variances
Service attributes Importance Satisfaction Importance Satisfaction
rates * rates * rates ** rates **
Availability of bus stop near home 5.23 n.s. * 0.22 1.02
Path 6.48 2.20 0.39 0.61
Service frequency 3.46 n.s. * 0.20 0.55
Reliability of runs that come on schedule n.s. -6.70 0.40 0.74
Reliability of runs that come on time 3.04 n.s. * 0.28 0.83
Availability of shelter and benches at bus
stops
3.92 -4.11 0.27 0.74
Availability of seats 4.81 3.65 0.35 0.84
Cleanliness of interior, seats and windows 4.56 n.s. * 0.31 0.90
Ticket cost 5.20 2.79 0.21 0.58
Availability of schedule/maps at bus stops 4.55 -7.70 0.29 0.67
Availability of service information by
phone, internet
n.s. -3.31 0.37 0.82
Vehicle reliability, competence of drivers 5.62 n.s. * 0.17 0.73
Security against crimes at bus stops 5,78 5.54 0.27 0.38
Personnel helpfulness 4.10 4.59 0.30 0.51
Administration of complaints 2.94 n.s. * 0.32 0.45
Physical condition of bus stops 5.73 n.s. * 0.31 0.72
Overall service - n.s. * 0.40
(*) Not significant at a level of 5% (t=2.10); (**) not significant at a level of 5% (F=1.27, df (num) = 91, df (den) = 149)
Additional information can be obtained by analyzing the variability of the ratings of importance
and satisfaction. This analysis shows that user ratings are nearly the same in the face-to-face survey,
and the variance of the importance ratings of all the 16 service attributes is higher for the web-based
survey than for direct interviews (Table 2). This is similar to what was obtained for the satisfaction
ratings of the attributes except availability of bus stop near home. The different levels of similarity
in user perceptions are also confirmed by the tests of equality of the variances of the two samples
using the Fisher F-test (Table 3). The obtained values of the test suggest that the equality of variance
test is not significant. Therefore, we surmise that the variances of the two groups of respondents are
significantly different.
Discriminant analysis was used to provide more statistically accurate results to support the
findings. We applied discriminant analysis to both the importance and satisfaction ratings expressed
by the users about the service quality attributes. A summary of the statistical tests regarding the
canonical discriminant function is shown in Table 4. This function shows an eigenvalue of 1.266 for
the analysis based on the importance ratings and 2.079 for the analysis based on satisfaction ratings.
An eigenvalue compares between group variance to within group variance. So, a large eigenvalue
is associated with a model that explains a large proportion of between group variance compared
to within group variance. The canonical relation represents a correlation between the discriminant

JTRF Volume 50 No. 1, Spring 2011
scores and the levels of the dependent variable (importance ratings of face-to-face and Web-based
surveys). The values of correlation obtained are 0.748 and 0.822 respectively, which are high and
show that the function discriminates well between the two survey methods.
Wilks’ Lambda is the ratio of within-groups sums of squares to the total sums of squares. This
is the proportion of the total variance in the discriminant scores not explained by differences among
groups. A Lambda value of one is obtained when the observed group means are equal (i.e., all the
variance is explained by factors other than the differences between those means), while a small
Lambda occurs when within-groups variability is small compared with the total variability. A small
Lambda indicates that group means appear to differ. The associated significance value indicates
whether or not the difference is significant. We obtained a Wilks’ Lambda of 0.441 for importance
ratings and of 0.325 for satisfaction ratings, which are significant at the 0.000 level. Thus, the
average importance ratings significantly differ between the two samples as well as the average
satisfaction ratings.
The canonical discriminant function coefficients show the standardized independent variables
included in the discriminant equation. The results indicate that the differences between the samples
are more evident for the importance ratings of routes, reliability of runs, ticket cost, and availability
of schedules/maps at bus stops, availability of service information by phone, internet, vehicle
reliability, and security against crimes at bus stops. The most discriminating variables among the two
samples of respondents expressing satisfaction are service frequency, reliability of runs, availability
of shelter and benches at bus stops, availability of schedule/maps at bus stops, availability of service
information by phone, internet, security against crimes at bus stops, personnel helpfulness, and
administration of complaints.
The top and bottom parts of Table 5 summarize the numbers and percentages of those in the
face-to-face interview and the Web-based survey correctly and incorrectly classified. The results
show that 91.3% of the grouped cases are correctly classified into “face-to-face” or “online” groups
(importance ratings). Face-to-face respondents were classified with better accuracy (98.0%) than
online respondents (79.0%). Regarding satisfaction ratings, the percentage of correctly classified
cases increases to 95.6% (Table 5). In this case, face-to-face respondents were classified with
slightly better accuracy (97.3%) than online respondents (92.1%). A total of 21 observations in the
analysis based on the ratings of satisfaction and 16 based on the ratings of importance were excluded
because of lack of at least one discriminant variable.
Other results can be obtained from analyzing the difference between the importance and
satisfaction ratings of each service attribute. It is found that this gap is higher for the data collected
by personal interviews and that the ratings of satisfaction expressed by the Web survey participants
are closer to the ratings of importance they expressed. Passenger perceptions about the overall
service are very similar between the two types of surveys. In fact, the passengers interviewed in
the face-to-face survey expressed an overall satisfaction of 7.24, while those interviewed in the
Web survey gave overall satisfaction of 6.95 (Table 2). According to the t-test this difference is
statistically significant at a level of significance of 95%.
Another overall measure is provided by the Customer Satisfaction Index (CSI), which is an
index of service quality calculated as the sum of the average satisfaction rating of each attribute
weighted by the respective average importance rating. A similar index, the Heterogeneous Customer
Satisfaction Index (HCSI), takes into account the heterogeneity in user perceptions by means of the
variance of the importance and satisfaction rates (Eboli and Mazzulla 2009). CSI is similar for the
two types of surveys: for face-to-face survey it is 6.49, whereas it is 6.51 for the Web-based survey.
On the contrary, HCSI is 7.56 for the face-to-face survey data, and 7.26 for the Web-based survey.
This difference can be explained by the heterogeneity of user perceptions.
27

Transit Passenger Perceptions
Table 4: Summary of the Results About Canonical Discriminant Function
28
Importance rates
Eigenvalues Eigenvalue
% of
Variance
Cumulative %
Canonical
correlation
1.266 100.0 100.0 0.748
Test of function Wilks’ Lambda Chi-square Df Sig.
Discriminant variables
0.441 176.3 7 0.000
Standardized
coefficients
Correlation
values
Path 0.443 0.404
reliability of runs that come on schedule -0.287 0.015
ticket cost 0.222 0.388
availability of schedule/maps at bus stops 0.307 0.224
availability of service information by phone,
internet
-0.874 -0.261
vehicle reliability, competence of drivers 0.469 0.489
security against crimes at bus stops 0.475 0.448
Satisfaction rates
Eigenvalues Eigenvalue
% of
Variance
Cumulative %
Canonical
correlation
2.079 100.0 100.0 0.822
Test of function Wilks’ Lambda Chi-square Df Sig.
Discriminant variables
0.325 233.9 8 0.000
Standardized
coefficients
Correlation
values
service frequency -0.271 -0.079
reliability of runs that come on schedule 0.214 0.275
availability of shelter and benches at bus stops 0.355 0.422
availability of schedule/maps at bus stops 0.381 0.416
availability of service information by phone,
internet
0.272 0.317
security against crimes at bus stops -0.603 -0.540
personnel helpfulness -0.385 -0.457
administration of complaints 0.244 0.098

Table 5: Classification Results
Importance rates *
Predicted Group Membership
Group variable online face-to-face total
Count online 64 17 81
face-to-face 3 147 150
% online 79.0 21.0 100.0
Satisfaction rates **
face-to-face 2.0 98.0 100.0
Predicted Group Membership
Group variable online face-to-face total
Count online 70 6 76
face-to-face 4 146 150
% online 92.1 7.9 100.0
face-to-face 2.7 97.3 100.0
(*) 91.3% of importance grouped cases correctly classified
(**) 95.6% of satisfaction grouped cases correctly classified
CONCLUSION
JTRF Volume 50 No. 1, Spring 2011
The aim of this research is to determine significant differences or similarities in behavior when
passengers are asked to provide their perceptions about their use of transit services through two
types of surveys. These perceptions were collected from face-to-face and Web-based surveys
addressed to users of an extra-urban bus service. No particular differences regarding socioeconomic
characteristics were observed in the two samples. These results contrast others where
those interviewed in the Web survey were generally younger and had higher household incomes
than those interviewed by traditional survey methods. However, our findings are different because
both samples are university students belonging to middle income families.
More interestingly, there is a significant difference in the judgments of importance in the surveys.
For almost all the service attributes analyzed, the average rating of importance in the face-to-face
interview survey is higher than the average rating in the self-administered survey. These results
suggest that when users personally express their judgments of importance to an interviewer, they
tend to give more importance to service characteristics than when they complete the questionnaire
alone. Also, users have a different threshold of importance depending on the type of survey. These
differences were not observed for the satisfaction judgments. In fact, there are some attributes for
which satisfaction ratings are higher in the face-to-face interview and others for which they are
lower. Moreover, passenger satisfaction about overall service is very similar between the two types
29

Transit Passenger Perceptions
of surveys. These results contrast the findings of Elmore-Yalch et al. (2008) that respondents of Web
surveys are more satisfied with service than respondents of telephone surveys.
Further results highlight the most significant discriminating service attributes between the two
different types of surveys, confirming some results of the descriptive analysis. From this analysis
there are important differences between the two samples of passengers. In fact, it emerges that the
differences between the samples are more evident for the importance ratings of service aspects like
route, reliability of runs, ticket cost, and availability of service information. For the satisfaction
ratings the differences between the samples regard service frequency, reliability of runs, facilities
at bus stops and availability of service information. Another important finding is that the ratings of
satisfaction expressed by the Web survey are closer to the ratings of importance. Perhaps users did
not understand the difference between importance and satisfaction, or their ratings of importance
and satisfaction are mutually influenced. This is also shown by observing that for the data collected
by Web survey the classification of the service attributes according to the ratings of importance is
similar to the classification according to the ratings of satisfaction.
From the analysis of heterogeneity in user judgments, the data collected from the face-to-face
surveys could be considered more reliable than those from the self-administered interviews. These
results can be considered useful contributions to the analysis of the differences in behavior and
attitudes of respondents depending on the type of survey. Despite the Web survey being cheaper
and less time consuming to conduct than the face-to-face survey, the data collected by the faceto-face
survey are more accurate owing to the presence of interviewers who ensured respondents
understood the questions. Based on the findings, we recommend asking users only satisfaction
ratings when a Web-based data collection method is adopted in customer satisfaction surveys. The
Web-based survey, however, can be considered a valid and convenient alternative to traditional faceto-face
interviews, especially when customer satisfaction surveys are addressed to groups of people
belonging to public or private corporations like universities.
30

APPENDIX: The Questionnaire
JTRF Volume 50 No. 1, Spring 2011
31

Transit Passenger Perceptions
32

JTRF Volume 50 No. 1, Spring 2011
33

Transit Passenger Perceptions
References
Al-Subaihi, A.A. “Comparison of Web and Telephone Survey Response Rates in Saudi Arabia.” The
Electronic Journal of Business Research Methods 6 (2), (2008): 123-132.
Bayart, C. and P. Bonnel. “Comparison of Web and Face-to-Face Household Travel Survey-
Application to Lyon Case.” Proceedings of the European Transport Conference, Leeuwenhorst
Conference Centre, The Netherlands, October 6-8, 2008.
Beatty, P. “Understanding the Standardized/non-standardized Interviewing Controversy.” Journal
of Statistics 11, (1995): 147-60.
Biemer, P.P. and L.E. Lyberg. Introduction to Survey Quality. Wiley, New York, 2003.
Bonnel, P. and M. Le Nir. “The Quality of Survey Data: Telephone versus Face-to-Face Interviews.”
Transportation 25, (1998): 147-167.
Braunsberger, K., H. Wybenga, and R. Gates. “A Comparison of Reliability Between Telephone and
Web-based Surveys.” Journal of Business Research 60, (2007): 758-764.
Cobanoglu, C., B. Warde, and P.J. Moreo. “A Comparison of Mail, Fax, and Web-based Survey
Methods.” International Journal of Market Research 43 (4), (2001): 441-52.
Coderre, F., A. Mathieu, and N. St-Laurent. “Comparison of the Quality of Qualitative Data Obtained
Through Telephone, Postal and Email Surveys.” International Journal of Market Research 46 (3),
(2004): 347-357.
Conrad, F.G. and M.F. Schober. “Clarifying Question Meaning in a Household Telephone Survey.”
Public Opinion Quarterly 64 (1), (2000): 1-28.
Converse, P.D., E.W. Wolfe, Huang Xiaoting, and F.L. Oswald. “Response Rates for Mixed-Mode
Surveys Using Mail and E-mail/Web.” American Journal of Evaluation 29 (1), (2008): 99-107.
Cook, C., F. Heath, and R.L. Thompson. “A Meta-Analysis of Response Rates in Web-or Internetbased
Surveys.” Educational and Psychological Measurement 60, (2000): 821-836.
Coomber, R. “Using the Internet for Survey Research.” Sociological Research Online 2 (2), (1997):
14-23.
Couper, M.P. “Web Surveys: A Review of Issues and Approaches.” Public Opinion Quarterly 64
(4), (2000): 464-494.
Couper, M.P. “Web Surveys: The Questionnaire Design Challenge.” Proceedings of the 53rd Session
of the ISI, Seoul, South Korea, August 22-29, 2001.
Dillman, D.A. and D.K. Bowker. “The Web Questionnaire Challenge to Survey Methodologists.”
U.D. Reips and M. Bosnjak eds. Dimensions of Internet Science. Lengerich, Germany: Pabst Science
Publishers (2001): 159-178.
Eboli, L. and G. Mazzulla. “A New Customer Satisfaction Index for Evaluating Transit Service
Quality.” Journal of Public Transportation 12 (3), (2009): 21-37.
Elmore-Yalch, R., J. Busby, and C. Britton. “Know thy Customer? Know thy Research!: A
Comparison of Web-based and Telephone Responses to a Public Service Customer Satisfaction
34

JTRF Volume 50 No. 1, Spring 2011
Survey.” Proceedings of the 87th Annual Meeting of the TRB, Washington, D.C., January 13-17,
2008.
Eysenbach, G. and J.Wyatt. “Using the Internet for Surveys and Health Research.” Journal of
Medical Internet Research 4 (2), (2002).
Fricker, S., M. Galesic, R. Tourangeau, and T. Yan. “An Experimental Comparison of Web and
Telephone Surveys.” Public Opinion Quarterly 69 (3), (2005): 370-92.
Ganassali, S. “The Influence of the Design of Web Survey Questionnaires on the Quality of
Responses.” Survey Research Methods 2 (1), (2008): 21-32.
Gorman, J.W. “An Opposing View of Online Surveying.” Marketing News 34 (9), (2000): 48-49.
Groves, R.M. and R.L. Kahn. Surveys by Telephone: A National Comparison With Personal
Interviews. Academic Press, San Diego, CA., 1979.
Groves, R.M. and R.D. Tortora. “Integrating CASIC into Existing Designs and Organizations: A
Survey of the Field.” M.P. Couper, R.P. Baker, J. Bethlehem, C.Z.F. Clark, J. Martin, W.L. Nicholls
and J.M. O’Reilly eds. Computer Assisted Survey Information Collection. New York: Wiley (1998):
45-61.
Heerwegh, D. and G. Loosveldt. “Face-to-Face Versus Web Surveying in a High-Internet-Coverage
Population: Differences in Response Quality.” Public Opinion Quarterly 72 (5), (2008): 836-846.
James, D. “The Future of Online Research.” Mark News 1, (2003): 1.
Kiecker, P. and J.E. Nelson. “Do Interviewers Follow Telephone Survey Instructions?” Journal of
the Market Research Society 38 (2), (1996): 161-76.
Loosveldt, G. and N. Sonck. “An Evaluation of the Weighting Procedures for an Online Access
Panel Survey.” Survey Research Methods 2 (2), (2008): 93-105.
McDonald, H. and S. Adam. “A Comparison of Online and Postal Data Collection Methods in
Marketing Research.” Marketing Intelligence & Planning 21 (2), (2003): 85-95.
Nicholls, W.L., R.P. Baker, and J. Martin. “The Effect of New Data Collection Technologies on
Survey Data Quality.” L. Lyberg, P. Biemer, M. Collins, E. De Leeuw, C. Dippo, N. Schwarz and
D. Trewin eds. Survey Measurement and Process Quality. New York: Wiley (1997): 221-248.
Ramos, M., B.M. Sedivi, and E.M. Sweet. “Computerized Self-Administered Questionnaires.” M.P.
Couper, R.P. Baker, J. Bethlehem, C.Z.F. Clark, J. Martin, W.L. Nicholls and J.M. O’Reilly eds.
Computer Assisted Survey Information Collection. New York: Wiley (1998): 389-408.
Sconlau, M., R.D. Fricker, and M.N. Elliott. “Conducting Research Surveys via E-mail and Web.”
MR1480, RAND Corporation, 2002.
Walker, A.H. and J.D. Restuccia. “Obtaining Information on Patient Satisfaction with Hospital
Care: Mail versus Telephone.” Health Services Research 19 (3), (1984): 291-306.
35

Transit Passenger Perceptions
Laura Eboli received her Ph.D. in environmental planning and technologies from the University
of Calabria, Italy. She is an assistant professor of transportation engineering in the Faculty of
Engineering at the University of Calabria, where she undertakes research in transportation planning
and service quality in public transport. Her articles on service quality and customer satisfaction
measures have been published in such journals as Transportation Planning and Technology, Journal
of Public Transportation, International Journal of Management Cases, Transport Reviews, EuroMed
Journal of Business.
Gabriella Mazzulla received her Ph.D. in road infrastructure and transportation systems from
University Federico II in Naples, Italy. She is an assistant professor of transportation engineering
in the Faculty of Engineering at the University of Calabria, Italy, where she teaches urban and
metropolitan transport. Her research interests focus on transportation planning, specifically analysis,
modeling, and estimation of travel demand. Her publications have appeared in transportation
journals, including Transportation Planning and Technology, Journal of Public Transportation,
European Transport, European Journal of Operational Research, and Transport Reviews.
36

Methodology for Measuring Output, Value Added,
and Employment Impacts of State Highway and
Bridge Construction Projects
by Michael W. Babcock and John C. Leatherman
The purpose of this paper is to present a methodology to measure some of the economic impacts of
state highway programs. State departments of transportation (DOTs) need such a methodology for
a variety of reasons, including long-term highway planning as well as advising state policymakers
concerning the economic impacts of highway programs. The specific objectives of this study are:
(1) describe a procedure to measure the output, value added, and employment impacts of specific
types of highway and bridge improvement, and (2) illustrate an application of the model using data
from Kansas.
The objectives of the research are accomplished with input-output modeling. An 11-step
procedure is described for adjusting the Kansas IMPLAN input-output model so that it is capable
of measuring economic impacts for specific types of highway and bridge improvement. The model is
illustrated using data from a recently completed study of the Kansas Comprehensive Transportation
Program (CTP), which included expenditure of $5.24 billion on state highway system projects. Data
from this study are used to demonstrate the calculation of output, value added, and employment
impacts for five different highway and bridge improvement categories.
INTRODUCTION
In 2008 and 2009, nearly every U.S. state significantly reduced their budgets as a result of the
2008 financial crisis followed by severe recession in 2009. In 2010, nearly 10% of the U.S. labor
force was unemployed. Since states can’t have deficits in their budgets, legislators have difficult
decisions to make in allocating diminishing funds to state programs. Although politics and policies
will always be the primary factor in government budget decisions, in recent years, benefit-cost
analysis has been given some weight in these decisions. While there is near unanimous agreement
among economists that government programs whose benefits exceed their costs are an efficient use
of resources, practical application of this technique has been limited by the difficulty in measuring
the benefits of government programs. The purpose of this paper is to present a methodology to
measure some of the benefits of state highway programs.
State Departments of Transportation (DOTs) need such a methodology for a variety of purposes.
State policymakers often request DOTs to supply economic impact information for highway
programs. Currently, many DOTs are unable to supply this information. In the current financially
austere environment, state DOTs attempt to justify their budget requests by estimating the economic
benefits of their proposed highway programs. The model presented in this paper could be used to
measure some of these benefits. State DOTs also need economic benefit and impact information for
long-term highway planning, and the model in this paper could be helpful for this purpose, although
the model is not intended to be a substitute for an in-depth analysis of future transport demands.
A very large literature exists in the area of the impact of highways on regional output, income,
and employment, as well as the role highways play in regional economic development. A partial
list of these studies from the 1970s include Dodgson (1974), Humphrey and Sell (1975), Kuehn and
West (1971), and Miller (1979). Numerous contributions to the literature in this area occurred in
the 1980s, including Allen et al. (1988), Briggs (1981 and 1983), Carlino and Mills (1987), Eagle
and Stephanedes (1988), Isserman et al. (1989), Lichter and Fuguitt (1980), Politano and Rodifer
37

Highway and Bridge Construction Projects
(1989), Stephanedes (1989), Stephanedes and Eagle (1986a, 1986b, 1987), and Wilson et al. (1982).
Highways and their effect on economic development continued to attract research interest in the
1990s, including Allen et al. (1994), Babcock et al. (1997), Babcock and Bratsberg (1998), Boarnet
(1995), Brown (1999), Crane and Leatham (1993), Garrison and Souleyrette II (1996), Holleyman
(1996), Holtz-Eakin and Schwartz (1995), Khanam (1996), Midwest Transportation Center (1990),
Mullen and Williams (1992), Singletary et al. (1995), Repham and Isserman (1994), and Talley
(1996). More recent contributions to this area of the literature include Babcock et al. (2010), Chi
and Cleveland (2006), Chandra and Thompson (2000), Van de Vooren (2004), and Peterson and
Jessup (2008).
The consensus of this literature is that highways have their greatest economic impact during the
construction phase with a smaller lagged impact over the long run (Eagle and Stephanedes [1988],
Stephanedes [1989], and Stephanedes and Eagle [1986a, 1986b, and 1987]). Economic impacts
vary widely by region, industry, and time period. Highways have economic impacts in rural areas
but good highways do not guarantee economic development if the region lacks other resources that
are necessary for growth.
This study differs from most of the previous studies in the literature which addressed whether
highways affected growth and which provided an aggregate estimate of how much impact highways
had on regional growth. In contrast, this study describes how state DOTs can estimate output,
income, and employment impacts of specific types of road and bridge improvements.
This study doesn’t measure other important benefits of highway investment, each of which is
a formidable research task. For example, the study doesn’t measure reductions in congestion that
result in lower vehicle operating costs such as maintenance, fuel, tires, and depreciation. It does not
measure the benefits of lower accident costs fostered by road improvements. Also, the study does
not directly measure the benefit of travel time savings resulting from highway investment (Allen,
Baumel, and Forkenbrock 1994). These latter benefits are indirectly measured since it is the lower
transport costs (time) generated by the highway investment that leads to economic growth (impacts)
in the affected region. Nevertheless, the study does measure some important impacts of different
types of highway and bridge improvement. The specific objectives of this paper are to:
1. Describe how to measure the output, value added, and employment impacts of specific
types of highway and bridge improvements.
2. Illustrate an application of the model using data from the state of Kansas.
INPUT-OUTPUT METHODOLOGY
The objectives of the research are achieved with input-output modeling. An input-output model
is a quantitative framework of analysis for examining the complicated interdependence within the
production system of an economy. There are three components to the standard input-output model:
an interindustry transactions matrix; a direct requirements matrix; and a direct, indirect, and induced
requirements matrix. Each of these can be explained with the aid of a simple illustrative example
from Professor Steven Deller (Deller and Williams 2009).
The transactions matrix describes the flow of goods and services between all individual
industries of the economy in a given year. The columns show purchases by a particular industry from
all other industries. For example, in the highly simplified example of an input-output transaction
matrix appearing in Table 1, the data in the Agriculture sector column show that, in order to produce
its $50 million output, that sector purchases $10 million from farm enterprises, $4 million from
manufacturing firms, and $6 million from service establishments. Agriculture firms also made
purchases from non-processing sectors of the economy, such as the household sector ($16 million)
and imports from other regions ($14 million). Purchases from the household sector represent value
added or income to people in the form of wages, salaries, and investment returns. The data in the
Agriculture sector row indicate that Agriculture sold $10 million to farm enterprises, $6 million
to manufacturing, $2 million to services, and the remaining $32 million was sold to households
38

JTRF Volume 50 No. 1, Spring 2011
within the region or exported out of the region. In this case, $20 million was sold to households
within the region and $12 million was sold to firms or households outside the region. Note that total
agriculture output (sum of the row) is exactly equal to Agriculture purchases (sum of the column),
or demand equals supply. This is the case for each sector.
The transactions table is significant because it provides a quantitative framework for the region’s
economy. Not only does it show the total output of each sector but also the interdependencies
between sectors. It also reveals the degree of “openness” of the region through imports and exports.
More open economies have a high percentage of total expenditures devoted to imports and, thus,
smaller multipliers.
The direct requirements matrix indicates the input requirement from each industry for a
particular industry to produce an average $1 of output. These purchase coefficients are obtained
by dividing purchase data in each industry column of the transactions matrix by the corresponding
output value for that industry. The resulting purchase coefficients, or input ratios, may be thought of
as production recipes for a particular product. From the data in the simplistic transactions matrix in
Table 1, a direct requirements matrix can be calculated (Table 2). As an example, the first column
(Agriculture) shows that to produce an average $1 of output, the Agriculture sector buys $.20 from
farming enterprises, $.08 from manufacturing firms, and $.12 from services firms. The Agriculture
column also shows that the sector makes payments of $.32 to households and $.28 to imports.
Households and imports are referred to as final payments sectors.
Table 1: Illustrative Input-Output Transactions Matrix (Millions of Dollars)
Purchasing Sectors (Demand) Final Demand
Processing Sectors
Total
(Sellers) Agr. Mfg. Serv. HH Exports Output
Agriculture 10 6 2 20 12 50
Manufacturing 4 4 3 24 14 49
Services 6 2 1 34 10 53
Households 16 25 38 1 52 132
Imports 14 12 9 53 0 88
Total Inputs 50 49 53 132 88 372
Table 2: Illustrative Direct Requirements Matrix
Purchasing Sectors
(Demand)
Processing Sectors (Sellers) Agr. Mfg. Serv.
Agriculture 0.20 0.12 0.04
Manufacturing 0.08 0.08 0.06
Services 0.12 0.04 0.02
Households 0.32 0.51 0.72
Imports 0.28 0.24 0.17
Total Inputs 1.00 1.00 1.00
39

Highway and Bridge Construction Projects
The direct and indirect requirements matrix is one of the two matrices that measure the
interaction among industries. The other, the direct, indirect, and induced requirements matrix, is
similar but includes the effects of household income and spending in addition to the interindustry
interaction. It is referred to as the total requirement matrix. The data in the columns of Table 3 for
each industry indicate the total requirements of all industries necessary for that industry to deliver
$1 of output to final demand. As an example, for the Agriculture sector to increase output to final
demand by $1, it must increase its overall output by $1.28 (including the initial $1 increase), the
Manufacturing sector must increase its output $.12 and the Services sector must increase its output
$.16. The total output increase of Agriculture in this simplistic economy is the sum of these three
values, or 1.56 times larger than the initial output expansion in Agriculture. The corresponding
values for Manufacturing and Services are 1.35 and 1.16, respectively. These numbers are output
multipliers.
Table 3: Illustrative Total Requirements Matrix
Purchasing Sectors (Demand)
Processing Sectors (Sellers) Agr. Mfg. Serv.
Agriculture 1.28 0.17 0.06
Manufacturing 0.12 1.11 0.07
Services 0.16 0.17 1.03
Total Inputs 1.56 1.35 1.16
PROCEDURES
Measurement of the economic impacts of specific types of highway improvement requires the
following 11-step procedure, developed by the authors, which is illustrated with Kansas data
(Babcock et al. 2010).
1. Establish objectives
2. Conduct secondary data search
3. Tabulate the population of construction firms eligible to obtain state highway contracts
4. Select highway improvement types
5. Measure the total state expenditure for each highway improvement type
6. Select highway construction firm samples
7. Design the questionnaire
8. Conduct the survey
9. Perform consistency check
10. Calculate output, value added (income), and employment multipliers for each highway
improvement type
11. Calculate output, value added (income), and employment impacts
Any study must start with clear objectives to provide a framework for the research effort. In
this type of study, the objectives are determined by the information needed by the state DOT. In a
study recently completed for the state of Kansas, the following objectives were established by the
Kansas Department of Transportation (KDOT):
1. Measure the direct output, value added, and employment impacts by highway improvement
type of the Kansas Comprehensive Transportation Program (CTP).
2. Measure the indirect and induced output, value added, and employment impacts by highway
improvement type of the Kansas Comprehensive Transportation Program (CTP).
The Kansas CTP extended from July 1999 to July 2009 and included expenditures of $5.24
billion for state highway system projects, including interstate highways (Babcock et al. 2010).
40

JTRF Volume 50 No. 1, Spring 2011
A secondary data search helps provide insight regarding the potential outcome of the research
project. For example, the U.S. input-output model (Minnesota IMPLAN Group 1999) was examined.
The model contains input data for construction of new highways as well as maintenance and repair
of highways. Analysis of this data indicated the types of inputs and approximate cost structure that
should emerge from the state survey of highway contractors.
The third step is to tabulate the population of construction firms eligible to obtain state highway
contracts in order to draw a sample for the survey. Every state DOT maintains a list of construction
firms eligible to bid on state highway construction contracts. In the Kansas CTP study, the list of
firms supplied by KDOT was supplemented by a directory published by the Kansas Contractors
Association (KCA). These directories are likely available in the great majority of states.
The next step is the selection of highway and bridge improvement categories. The general
principle with respect to classification of highway and bridge improvement types is homogeneity.
Highway and bridge improvement types with similar cost structures and requiring similar inputs
should be placed in the same class. For the Kansas CTP study, the research team and KDOT
categorized and selected the following highway and bridge improvement types for analysis. 1
Category Highway Improvement Type
1 Resurfacing
2 Restoration and Rehabilitation; Reconstruction and Minor Widening
3 New Bridges and Bridge Replacement
4 Major and Minor Bridge Rehabilitation
5 New Construction; Relocation; Major Widening
6 Safety/Traffic Operations/Traffic Systems Management; Environmentally
Related; Physical Maintenance, Traffic Services
The fifth step is measurement of total expenditure for each highway and bridge improvement
type during the time frame of the study. These data are needed to expand the state survey sample data
to population totals. Every state DOT has records that tabulate total annual expenditure by highway
and bridge improvement type. For the Kansas CTP study, the value of construction contracts by
highway and bridge improvement type is as follows:
Value of CTP Construction Contracts,
Category July 1999-July 2009 (Millions of Dollars)
Resurfacing $1,240.9 (23.7%)
Restoration and Rehabilitation;
Reconstruction and Minor Widening $2,684.8 (51.3%)
New Bridges and Bridge Replacement $439.1 (8.4%)
Major and Minor Bridge Rehabilitation
New Construction; Relocation; Major
$199.8 (3.8%)
Widening $503.2 (9.6%)
Safety/Traffic Operations/Traffic Systems
Management; Environmentally Related;
Physical Maintenance; Traffic Services $169.2 (3.2%)
Grand Total $5,237.0
The sixth step is selection of survey samples for each highway and bridge improvement type.
The general principle is to concentrate research efforts on the construction firms with the largest
highway contracts simply because they are the firms that account for the majority of the construction
activity. For example, suppose there are 10 firms performing a certain type of highway work in a
particular year, and total state spending on this highway improvement type is $50 million. Also
41

Highway and Bridge Construction Projects
assume that one firm has a contract for $25 million, while the remaining $25 million is split equally
among the other nine firms. If random sampling is employed, there is only one chance in 10 that the
firm with the $25 million contract will be selected. A better and more useful strategy is to select the
firms with the large contracts. These samples were drawn from state DOT records which tabulate
each construction contract by amount, highway improvement type, and name of construction firm.
The research team selected the construction firm samples by highway and bridge improvement type
from these records.
KDOT and the research team selected the highway contractors who obtained Kansas CTP
highway construction contracts during the period January 1, 2004, to December 31, 2007. The
total value of the sample contracts let during the 2004-2007 period was $1.98 billion, 37.9% of the
total CTP contract value of $5.24 billion (Babcock et al. 2010). Of the $1.98 billion of contracts
let in the sample period, $1.42 billion was obtained from sample contractors, which was 71.4% of
the $1.98 billion and 27% of the 10-year CTP total contract value of $5.24 billion (Babcock et al
2010). Members of the research team conducted personal interviews with the highway contractors
who received the larger contracts in each highway and bridge improvement category. In the CTP
study it was not uncommon for the large construction firms to be in the sample for more than one
highway improvement category.
The seventh step is designing the questionnaire. The general principles are brevity and clarity.
It should have any necessary explanatory notes as well as definitions and examples of each of the
industry sectors in the model. The questionnaire for the Kansas CTP study had four pages (see
Appendix A). The first page lists the highway contracts for which purchase and cost information
is requested, and space is provided for the firm to provide the final contract amount and total
labor hours for each contract. The contract amount is required for a consistency check since the
sum of the purchases and retained earnings from the second page of the questionnaire must equal
the total contract amount of the first page. Total labor hours are needed to calculate the average
wage per hour for each improvement type. The second page of the questionnaire pertains to input
purchases and other costs of the highway construction firms. On this page the firm lists the name of
each supplying industry, the total purchases from that industry, and the percent of total purchases
supplied by firms located in the state. 2 This page of the questionnaire also requests amounts of other
expenditures such as amounts paid to subcontractors, wages and salaries, taxes, and depreciation.
The latter figure is combined with retained earnings to preserve profit confidentiality, a necessary
ingredient in obtaining the cooperation of the sample construction firms. The third and fourth page
of the questionnaire contains definitions and examples of the input supplying sectors which the
construction firms are requested to use in classifying their purchases.
The next step is to conduct the survey, which begins with sending a letter to the presidents of
the sample construction firms explaining the objectives of the study and how the research project
could benefit the company. The letter is followed by a call to the presidents of the firms to discuss
the project. During the call, explain the research objectives thoroughly, emphasize confidentiality,
and ask for an appointment. At the interview they explain the questionnaire in detail and answer all
questions.
The ninth step is performing consistency checks. After receiving the questionnaires, they should
be checked for errors, inconsistencies (such as the value of the contract not matching the contractrelated
expenditures), and omitted data. Call the respondent to clarify any problems. After resolving
any data problems, the input-output tables were constructed.
The next step is the calculation of output, value added (income), and employment multipliers for
each highway and bridge improvement type. Output multipliers are a good indicator of the degree
of economic interaction between each state industry sector and the rest of the state economy as well
as exports to and imports from other regions. The output multipliers are calculated by summing the
columns corresponding to each highway improvement type of the total requirements matrix.
Value added (income) is the sum of employee compensation (total payroll including value of
benefits), proprietors’ income (payments to self employed individuals), property income (such as
42

JTRF Volume 50 No. 1, Spring 2011
rents, royalties, dividends, and corporate profits), and indirect business taxes (excise taxes, property
taxes, licenses, fees, and sales taxes paid by business). Value added multipliers indicate the total
value added generated from the construction projects of a given highway or bridge improvement
type, including direct value added within the construction industry as well as the indirect value
added of the industries that supply the construction industry with materials, goods, and services.
The value added multiplier also includes the induced value added in various consumer markets
produced by the increased spending of people employed both directly and indirectly as a result of
state construction projects.
The employment multipliers for each highway or bridge improvement type represent the sum of
three effects: the direct employment within the construction industry itself, the indirect employment
in the input supplier industries, and the induced employment in various consumer markets generated
by increased consumer spending of people employed directly and indirectly on state highway
construction projects.
The output, value added, and employment multipliers for five highway and bridge improvement
types are displayed in Table 4. 3 Using resurfacing as an example, the multipliers are interpreted in the
following manner. In the case of the output multiplier, for every $1 increase in resurfacing contract
value, total Kansas production increases by $1.74 (including the initial $1 increase). With respect
to the value added multiplier, for every $1 increase in value added in the Kansas construction firms
performing road surfacing, total Kansas value added increases by $1.78 (including the initial $1
increase). In the case of the employment multiplier, for every job generated in Kansas construction
firms involved in resurfacing work, total Kansas employment increases by 1.91 jobs (including the
initial one job increase).
Table 4: Output, Value Added, and Employment Multipliers of the Kansas CTP Study
Output Value Added Employment
Highway Improvement Type
Multiplier Multiplier Multiplier
Resurfacing 1.74047 1.78454 1.90737
Restoration, Rehabilitation, Reconstruction and
Minor Widening
1.60446 1.68217 1.89845
New Bridges and Bridge Replacement, Major
and Minor Bridge Rehabilitation
1.50128 1.64579 1.69373
New Construction; Relocation, Major Widening 1.52279 1.62144 1.76020
Safety/Traffic Operations/ Traffic Systems
Management; Environmentally Related,
Physical Maintenance, Traffic Services
1.54372 1.63245 1.89454
Total 1.61929 1.69779 1.86215
Source: Babcock et al. (2010).
The multipliers were calculated with a 345 sector input-output model calibrated to 2006, roughly
the mid-point of the project investment period, using the IMPLAN modeling system originally
developed by the U.S. Forest Service (Minnesota IMPLAN Group, Inc. 1999). IMPLAN creates
a detailed model of the economy that charts the financial flows between all production sectors and
institutions, i.e., government, households, and capital. The purchase and cost information from the
contractor survey was placed in the appropriate sector of the 90 sectors most closely associated
with highway construction. The national production function of highway, street, bridge, and tunnel
construction was used for this purpose. This production function provides a national average
43

Highway and Bridge Construction Projects
“production recipe” of inputs required to produce the output associated with highway construction.
Thus, we can apportion total reported spending to each of the industry sectors associated with
highway construction activity.
The first step in this process was to bridge the 43 sector contractor survey expenditures and
the 90 IMPLAN highway industry input sectors that were present in Kansas. For example, Kansas
contractors reported spending for non-metallic minerals such as sand and gravel. In the IMPLAN
system, non-metallic minerals is represented by two sectors, stone mining and quarrying, and sandgravel-clay
and refractory mining. To determine how much of the Kansas contractor spending on
non-metallic minerals that each IMPLAN sector gets, we examined the national production function
and observe that stone mining gets 77% of the non-metallic minerals spending while sand-gravelclay
and refractory mining gets 23%. Thus, if a total of $10 million had been spent on non-metallic
minerals, $7.7 million was assigned to stone mining and $2.3 million was assigned to sand-gravelclay
and refractory mining. There is a direct correspondence between agricultural services and
agriculture and forestry support activities. Thus, the IMPLAN sector for agriculture and forestry
support activities gets 100% of the spending reported by the highway contractors for agriculture
services. In this way the inter-industry input patterns for Kansas highway construction closely
approximate national input patterns yet retains the distinctive expenditure distribution reported by
Kansas contractors.
Following this procedure, total in-state spending for each of the construction categories was
input into the 90 highway construction input sectors represented in the Kansas input-output model.
With the introduction of the spending into the model, IMPLAN calculates the associated indirect and
induced impacts associated with the activity. At the same time, it generates the various economic
multipliers reported in Table 4 that summarize the total economic activity associated with the new
highway spending.
CALCULATION OF OUTPUT, VALUE ADDED, AND EMPLOYMENT IMPACTS
The last step of the procedure is calculation of impacts. Once the various multipliers have been
determined, the output, value added, and employment impacts of state highway programs can be
easily calculated. To compute output impacts, the researcher obtains from the state DOT the value
of construction contracts by highway and bridge improvement type awarded during the time frame
of the study. The values of contracts spent in the state by highway improvement type are multiplied
by their respective multipliers to obtain the output impacts. The output impacts by highway and
bridge improvement types for the Kansas CTP study are in Table 5. Examination of the table reveals
that the $5.24 billion Kansas CTP generated a total output impact of $6.7 billion (includes the $5.24
billion direct impact).
Value added impacts by highway and bridge improvement type are calculated by multiplying
direct value added (value added generated within the construction industry) by their respective
multipliers. In the Kansas CTP study, the total Kansas direct value added of $1.83 billion resulted
in a total value added impact of $3.11 billion (Table 6).
Employment impacts by highway and bridge improvement type are obtained by multiplying
direct employment by the appropriate employment multiplier. The Kansas IMPLAN inputoutput
model doesn’t measure employment within the Kansas highway construction industry
(direct employment), so it had to be estimated manually. To estimate Kansas direct construction
employment, we used total state output and employment in the Kansas highway construction sector
(Babcock et al. 2010). In 2006, Kansas highway construction output totaled $840,275,000 and
highway construction employment was 8,100 workers. Thus, each $1million in highway construction
outputs was associated with 9.64 workers (8,100/840,275). Multiplying Kansas total construction
spending by highway improvement type for the 10-year CTP era by 9.64 yields an estimate of direct
employment in highway construction companies. The results are in Table 7.
44

JTRF Volume 50 No. 1, Spring 2011
The employment impact by highway and bridge improvement type is obtained by multiplying
the direct employment from Table 7 by the appropriate multiplier (Table 8). In the Kansas CTP
study, the direct employment of 50,483 generated a total of 94,007 jobs (including the direct
employment of 50,483).
Table 5: Kansas CTP Output Impact by Highway Improvement Types
(1)
Highway
Improvement
Type
(2)
Value of CTP
Contracts
(3)
Proportion of
Contracts Spent
Outside Kansas
(4)
Value of
Contracts Spent
in Kansas
(5)
In-state Output
Multiplier
(6)
Output Impact
1 $1,240,934,211 15.77% $1,045,226,437 1.74047 $1,819,192,241
2 $2,684,791,829 18.84% $2,178,896,284 1.60446 $3,495,973,173
3-4 $638,865,242 34.09% $421,063,094 1.50128 $632,136,472
5 $503,152,560 29.46% $354,900,856 1.52279 $540,441,136
6 $169,224,802 18.39% $138,107,886 1.54372 $213,200,167
Total $5,236,968,645 20.98% $4,138,194,557 1.61929 $6,700,943,189
Column (6) is the product of Columns (4) and (5), although not exactly due to rounding of the multiplier. All
data reported in dollars are measured in 2009 dollars.
Source: Babcock et al. (2010).
Table 6: Kansas CTP Value Added Impact by Highway Improvement Type
(1)
Highway
Improvement
Type
(2)
Value of CTP
Contracts
(3)
Direct Value
Added
(4)
Value Added
Multiplier
(5)
Value Added
Impact
1 $1,240,934,211 $463,875,286 1.78454 $827,808,555
2 $2,684,791,829 $973,537,493 1.68217 $1,637,661,853
3-4 $638,865,242 $174,278,818 1.64579 $286,828,010
5 $503,152,560 $157,080,456 1.62144 $254,696,910
6 $169,224,802 $61,054,547 1.63245 $99,668,973
Total $5,236,968,645 $1,829,826,600 1.69779 $3,106,664,301
Column (5) is the product of Columns (3) and (4), although not exactly due to rounding of the multiplier.
Data measured in dollars is reported in 2009 dollars.
Source: Babcock et al. (2010).
CONCLUSION
State highway policymakers are usually very interested in the labor impacts of highway investment.
The model suggested in this paper is capable of measuring some of these impacts. Suppose
policymakers want to know how many jobs would be generated by annual spending of $105 million
on new construction and major widening. Using data from Table 7, direct employment is 1,012 jobs
(105 x 9.64). Multiplying 1,012 by the employment multiplier in Table 8 of 1.76020 results in a
total impact of 1,782 jobs. According to the contractor survey of the CTP study the average wages,
salaries, and benefits per hour in the New Construction and Major Widening category is $34.25
(Babcock et al. 2010). Assuming 2,000 annual work hours results in direct annual wages, salaries,
45

Highway and Bridge Construction Projects
and benefits per worker of $68,500. When this figure is multiplied by 1,782 workers, the result is
total annual direct wages, salaries, and benefits of $122.1 million.
The model in this paper can be employed by state DOTs to advise highway policymakers
regarding the economic impacts of alternative highway programs. If policymakers supply state
DOTs with the proposed total contract value of the various highway improvement types (i.e., column
[2] of Table 5), the state DOT can use the output multipliers to estimate the output impact of any
proposed highway program. Even if the total contract value of alternative highway programs is the
same, the economic impacts will different because the multipliers of the various highway and bridge
improvement types are different.
The highway construction impacts analyzed in this paper represent only a part of the benefits of
highway investment. A more comprehensive view of the benefits would combine three categories
of benefits that would include construction, highway user, and regional economic growth benefits.
Once a highway is built it becomes an asset that yields benefits to highway users over the useful life
of the asset. Many of the user benefits have been identified in the literature and include logistics cost
reductions for firms using truck freight, accident cost reductions, travel time savings for motorists,
and lower vehicle operating costs. However, measurement of these benefits, especially on a regional
basis, is a formidable research task (Allen et al. 1994).
Table 7: Estimated Direct Construction Contractor Employment by
Highway Improvement Type
(1)
Highway
Improvement Type
46
(2)
Value of CTP Contracts
(Millions of dollars)
(3)
Direct Employment per
Million Dollars
(4)
Direct
Employment
1 $1,240.9 9.64 11,962
2 $2,684.8 9.64 25,881
3-4 $638.9 9.64 6,158
5 $503.2 9.64 4,850
6 $169.2 9.64 1,631
Total
Source: Babcock et al. (2010).
$5,236.9 9.64 50,483
Table 8: Kansas CTP Employment Impact by Highway Improvement Type
(1)
Highway
Improvement Type
(2)
Indirect
Employment
(3)
Direct
Employment
(4)
Employment
Multiplier
(5)
Total Employment
Impact
1 10,854 11,962 1.90737 22,816
2 23,253 25,881 1.89845 49,134
3-4 4,272 6,158 1.69373 10,430
5 3,687 4,850 1.76020 8,537
6 1,459 1,631 1.89454 3,090
Total 43,524 50,483 1.86215 94,007
Column (5) is the product of Columns (3) and (4).
Source: Babcock et al. (2010).

APPENDIX
JTRF Volume 50 No. 1, Spring 2011
HIGHWAY CONTRACTOR SURVEY FORMS FOR PURCHASE-COST INFORMATION
KDOT Contractor No.: SAMPLE
AND TOTAL LABOR HOURS
HIGHWAY ECONOMIC IMPACT PROJECT
PRIME CONTRACT SURVEY
Person answering questionnaire _______________________________
We request your purchase and cost information on the highway projects listed below.
These contracts deal only with:
KDOT KDOT Final Total
Contract Project Contract Labor
Numbers Route Number Let Date Amount (if avail.) Hours
92000001 K-490 K 1000-01 10/20/04 ____________ _______
93000001 U-220 K 2000-01 8/28/06 ____________ _______
PQ1
TOTALS ____________ _______
47

Highway and Bridge Construction Projects
48
HIGHWAY ECONOMIC IMPACT PROJECT
KDOT Contractor No.: SAMPLE Respondent _____________________
Please provide your firm’s purchases by supplying industry on only the projects on the previous
page, which were let from January 1, 2004 to December 31, 2007.
Provide figures from all the projects as though they were one project.
PURCHASES:
Supplying Industries
- brief description
(See attached list of industries.)
Other Expenditures:
Paid to Subcontractors
Wages and salaries (include both direct and
overhead salaries)
Taxes - Federal
- State
- Local
Depreciation and Retained Earnings
TOTAL EXPENDITURES
Total Purchases
(Include both direct
and overhead costs)
($ or %)
PQ1
Percent Supplied
by Producers in
Kansas

1. Agricultural Services - landscaping, grass seeding
2. Non-Metallic Minerals - rocks, stone, sand, dirt, aggregates
3. Other Mining - Oil, gas, coal, other minerals
JTRF Volume 50 No. 1, Spring 2011
4. Construction Maintenance and Repair - maintenance and repair of capital assets including
construction machinery and vehicles
5. Heavy Construction - general contractors engaged in the construction of highways, streets,
and bridges. Doesn’t include payments to subcontractors. Only includes purchases from
other construction firms.
6. Special Trade Contractors - plumbing, plastering, painters, carpenters
7. Paper and Allied Products - paper bags, boxes, all types of paper
8. Printing and Publishing - brochures, reports, any type of published material
9. Industrial Chemicals - basic industrial chemicals such as industrial gases, pigments, dyes, etc.
10. Agricultural Chemicals - fertilizer, pesticides
11. Other Chemicals - explosives, paint, cleaning preparations, glue, ink
12. Petroleum and Coal Products - asphalt, lubricating oils, and greases
13. Rubber and Plastic Products - tires, cold plastic and thermal plastic pavement markings,
plastic cones and barrels
14. Cement and Concrete Products - hydraulic cement, concrete products like pipe, pre-stressed
beams, drilled shaft casings
15. Stone, Clay, and Glass Products - lime, gypsum, abrasives, cut stone products, glass products,
flat glass, bricks
16. Primary Metal Products - iron, steel, aluminum, copper, iron pipe
17. Fabricated Structural Metal - rebar, structural steel, corrugated metal pipe, signs, sign
supports, guard rail
18. Other Fabricated Metal - tools, containers, fasteners, wire, nuts, bolts, valves
19. Farm Machinery and Equipment - tractors, combines, bailers
20. Construction and Industrial Machinery - construction machinery parts, equipment, and rentals.
Does not include construction machinery repairs (see sector 4)
21. Electrical Machinery - air conditioning, refrigeration, materials handling machines, power
driven hand tools, lighting fixtures, electric motors, generators, batteries
22. Other Machinery - engines, turbines, machine tools
23. Motor Vehicles and Equipment - purchases of cars and car parts
24. Other Transport Equipment - railroads, boat, aircraft equipment and parts
25. Other Manufacturing - lumber and wood products, furniture, leather products, scientific
instruments, metal filing cabinets, miscellaneous manufacturing
26. Railroad Transportation - transport by railroad
27. Motor Freight Transportation - transport by truck
28. Other Transportation - transport by air, water, or oil pipeline
29. Communications - phones, cell phones, internet connection fees, anything involving oral or
visual communication
49

Highway and Bridge Construction Projects
30. Electric, Gas, and Sanitary Services - expenditures for electricity, natural gas, water, garbage
collection
31. Wholesale Trade, Machinery, and Equipment - purchases from wholesalers of machinery,
equipment, and supplies
32. Other Wholesale Trade - purchases from wholesalers other than for machinery, equipment,
and supplies
33. Gasoline Service Stations - purchases of gas or diesel fuel
34. Eating and Drinking Places - restaurant purchases
35. Other Retail Trade - all other purchases from retail stores, except fuel and food
36. Banking - interest payments on bank loans
37. Other Financial Institutions - interest on all non-bank loans
38. Insurance and Real Estate - performance bonds, liability insurance, employee health
insurance, building or other rental payments
39. Lodging Services - payments to hotels and motels
40. Personal Services - services involving care of the person or person’s clothing
41. Business Services - licenses, filing fees, advertising, data and word processing, professional
and legal services, consulting, vehicle rental or leasing, accounting, tax preparation
42. Medical and Health Services - payments for medical or surgical services to persons. Doesn’t
include health insurance for employees (see sector 38)
43. Other Services - payments for all other services not enumerated above like automotive repair,
entertainment, education
Endnotes
1. The highway improvement categories of the Kansas CTP study are combinations of Federal
Highway Administration (FHWA) highway and bridge improvement types.
2. The questionnaire employed to obtain the purchase and cost data for the various highway
improvement types requires the respondent to report the percent of each input purchase type that
is supplied by in-state producers. This is done to net out imports that have no impact in the state
and are thus not included in the study. For example, if a construction firm purchases $10 million
of cement and 80% is purchased from an in-state supplier, the study measures only the impact of
the $8 million that increases cement production within the state. The other $2 million increases
cement production in an out-of-state location and is not included in the impacts measured by the
study.
3. The relatively small amount of Category 4 (Major and Minor Bridge Rehabilitation) contract
value in returned contractors’ questionnaires resulted in combing Category 3 (New Bridges and
Bridge Replacement) with Category 4.
50

References
JTRF Volume 50 No. 1, Spring 2011
Allen, B.L., W. Butterfield, A. Kazakov, M.L. Kliman, A.A. Kuburst, and J.D. Welland. “Measuring
Economic Stimulation from Capital Investment in Transportation.” Presented at the 67 th Annual
Meeting of the Transportation Research Board, Washington, D.C., 1988.
Allen, Benjamin C., C. Phillip Baumel, and David J. Forkenbrock. “Expanding the Set of Efficiency
Gains of a Highway Investment: Conceptual, Methodological, and Practical Issues.” Transportation
Journal 34 (1), (1994): 39-47.
Babcock, Michael W., M. Jarvin Emerson, and Marvin Prater. “A Model Procedure for Estimating
Economic Impacts of Alternative Types of Highway Improvement.” Transportation Journal 36 (4),
(1997): 30-43.
Babcock, Michael W., John C. Leatherman, Mark Melichar, and E. Dean Landman. Economic
Impacts of the Kansas Comprehensive Transportation Program (CTP) Highway Construction and
Maintenance Activities. Kansas Department of Transportation, Topeka, Kansas, July 2010.
Babcock, Michael W. and Bernt Bratsberg. “Measurement of Economic Impacts of Highway and
Bridge Construction.” Journal of the Transportation Research Forum 37 (2), (1998): 52-66.
Boarnet, Marlon G. “New Highways and Economic Growth: Rethinking the Link.” Access:
Research at the University of California Transportation Center No. 7, (Fall 1995): 11-15.
Briggs, R. “The Interstate Highway System and Development in Nonmetropolitan Areas.”
Transportation Research Record 812, (1981): 9-12.
Briggs, Ronald. “The Impact of the Interstate Highway System on Nonmetropolitan Development,
1950-1980.” R.H. Platt and George Machinko eds. Beyond the Urban Fringe. Minneapolis:
University of Minnesota Press (1983): 83-105.
Brown, Dennis M. Highway Investment and Rural Economic Development: An Annotated
Bibliography. Food and Rural Economics Division, Economic Research Service, U.S. Department
of Agriculture, Washington, D.C., 1999.
Carlino, Gerald A. and Edwin S. Mills. “The Determinants of County Growth.” Journal of Regional
Science 27 (1), (1987): 39-54.
Chandra A. and E. Thompson. “Does Public Infrastructure Affect Economic Activity? Evidence
from the Rural Interstate Highway System.” Regional Science and Urban Economics 30, (2000):
457-490.
Chi, Guangqing and C.J. Cleveland. “Highway Effects on Economic and Population Growth:
Regional Economic Explanation.” C.J. Cleveland ed. Encyclopedia of Earth. Washington, D.C.:
National Council for Science and the Environment (2006).
Crane, Laurence M. and David J. Leatham. “Distributed Lag Effects of Transportation Expenditures
on Rural Income and Employment.” The Review of Regional Studies 23 (2), (1993): 163-182.
Deller, Steven C. and David Williams. “The Contribution of Agriculture to the Wisconsin Economy.”
Department of Agricultural and Applied Economics, University of Wisconsin-Madison, 2009.
Dodgson, J.S. “Motorway Investment, Industrial Transport Costs and Subregional Growth. A Case
Study of M62.” Regional Studies 8, (1974): 75-91.
51

Highway and Bridge Construction Projects
Eagle, D. and Y.J. Stephanedes. “Dynamic Highway Impacts on Economic Development.”
Transportation Research Record No. 1125, (1988): 1-14.
Garrison, William L. and R.R. Souleyrette II. “Transportation Innovation and Development: The
Companion Innovation Hypothesis.” The Logistics and Transportation Review 32 (1), (1996): 63-
76.
Holleyman, Chris. “Industry Studies of the Relationship Between Highway Infrastructure and
Productivity.” The Logistics and Transportation Review 32 (1), (1996): 93-118.
Holtz-Eakim Douglas and Amy Schwartz. “Spatial Product Spillovers from Public Infrastructure:
Evidence from State Highways.” International Tax and Public Finance 2 (3), (1995): 459-468.
Humphrey, C.R. and R.A. Sell. “The Impact of Controlled Access Highways in Population Growth
in Pennsylvania Nonmetropolitan Communities, 1940-1970.” Rural Sociology 40, (1975): 332-
343.
Isserman, Andrew M., T. Rephann, and D.J. Sorenson. “Highways and Rural Economic
Development: Results from Quasi-Experimental Approaches.” Paper Presented at the Annual
Meeting of the American Collegiate Schools of Planning, October, 1989.
Khanam, B.R. “Highway Infrastructure Capital and Productivity Growth: Evidence from Canadian
Goods-Producing Sector.” The Logistics and Transportation Review 32 (3), (1996): 251-268.
Kuehn, John A. and J.G. West. “Highways and Regional Development.” Growth and Change 2,
(1971): 23-28.
Lichter, Daniel T. and Glenn V. Fuguitt. “Demographic Response to Transportation Innovation:
The Case of the Interstate Highway.” Social Forces 59 (2), (1980): 492-511.
Midwest Transportation Center and the University of Iowa Public Policy Center. Road Investment
to Foster Local Economic Development, Iowa City, IA, 1990.
Miller, James P. “Interstate Highways and Job Growth in Nonmetropolitan Areas: A Reassessment.”
Transportation Journal 19, (1979): 78-81.
Minnesota IMPLAN Group, Inc. IMPLAN Professional (version 2.0) Users Guide, Analysis Guide,
Data Guide. Stillwater, MN, 1999.
Mullen, J. and M. Williams. “The Contribution of Highway Infrastructure to States’ Economies.”
International Journal of Transport Economics 19 (2), (1992).
Peterson, Steven K. and Eric L. Jessup. “Evaluating the Relationship Between Transportation
Infrastructure and Economic Activity: Evidence from Washington State.” Journal of the
Transportation Research Forum 47 (2), (2008): 21-40.
Politano, A.L. and C.J. Roadifer. “Regional Economic Impact Model for Highway Systems
(REIMHS).” Transportation Research Record 1229, (1989): 43-52.
Repham, T.J. and A.M. Isserman. “New Highways as Economic Development Tools: An Evaluation
Using Quasi-Experimental Matching Methods.” Regional Science and Urban Economics 24 (6),
(1994): 723-751.
52

JTRF Volume 50 No. 1, Spring 2011
Singletary, Loretta, Mark Henry, Kerry Brooks and James London. “The Impact of Highway
Investment on New Manufacturing Employment in South Carolina: A Small Region Spatial
Analysis.” Review of Regional Studies 25 (1), (1995): 83-90.
Stephanedes, Y.J. Transportation and Economic Development: The Link Between Highway
Investment and Economic Development: A Time Series Investigation. Minnesota Department of
Transportation, Report No. MN/RC, 1989.
Stephanedes, Y.J. and D.M. Eagle. “Highway Expenditures and Nonmetropolitan Employment.”
Journal of Advanced Transportation 20 (1), (1986): 43-61.
Stephanedes, Y.J. and David M. Eagle. “Time Series Analysis of Interactions Between Transportation
and Manufacturing and Retail Employment.” Transportation Research Record No. 1074, (1986):
16-24.
Stephanedes, Y.J. and D.M. Eagle. “Highway Impacts on Regional Employment.” Journal of
Advanced Transportation 21, (1987): 369-389.
Talley, Wayne. “Linkages Between Transportation Infrastructure Investment and Economic
Production.” The Logistics and Transportation Review 32 (1), (1996): 145-154.
Van de Vooren, F.W.C.J. “Modeling Transport in Interaction with the Economy.” Transportation
Research Part E 40 E (5), (2004): 417-437.
Wilson, F.R., Albert M. Stevens, and Timothy R. Holyoke. “Impact of Transportation on Regional
Development.” Transportation Research Record No. 851, (1982): 13-16.
Michael W. Babcock is a full professor of economics at Kansas State University. In his 38-year
career at KSU, he has published over 80 articles in professional journals, along with numerous
monographs and technical reports. His research has been cited in over 75 books, the transportation
press, and professional journals. He has been principal investigator or co-principal investigator on
33 federal and state research grants worth a total of $2.3 million.
Dr. Babcock has received numerous national awards for his transportation research. He has
won five best paper awards from the Transportation Research Forum (TRF) for outstanding research
in transportation economics. He has received the E.S. Bagley award five times from the Economics
Department of KSU for outstanding achievements in transportation economics research. In 1998 he
was awarded the ISBR Senior Faculty Award for research excellence in The Social and Behavioral
Science from KSU. In 2005, he received the Herbert O. Whitten TRF Service Award (the highest
honor bestowed by TRF) for professional contributions to TRF, becoming only the eighth person to
receive the award in the 50 year history of TRF. He has served as general editor of the Journal of
the Transportation Research Forum since 2000.
Dr. John Leatherman is a professor in the Department of Agricultural Economics at Kansas State
University and local government specialist with K-State Research & Extension. He also is the director
of the Office of Local Government, an educational outreach and technical assistance program for
city and county governments in Kansas. He has worked with the Cooperative Extension Service
since 1984 in areas of community and regional economic development policy and practice, local
government and public finance, and natural resources management. Research interests include
economic and fiscal impact analysis, regional economic modeling, rural development and policy
analysis, local public finance, and natural resource management and environmental protection.
53

Rail Rate and Revenue Changes
Since the Staggers Act
by Ken Casavant, Eric Jessup, Marvin E. Prater, Bruce Blanton, Pierre Bahizi,
Daniel Nibarger, Johnny Hill, and Isaac Weingram
An examination of the effects of deregulation and the performance of the Surface Transportation
Board (STB) under that deregulation usually includes an analysis of rail rates that have evolved
since implementation of the Staggers Act of 1980 (Staggers Act). This paper examines the rail rate
structure for agricultural commodities and compares it with rates for other commodities. Changes
in agricultural rail rates are evaluated relative to shipment size and distance shipped to understand
how they affect agricultural shippers. Railroads transferred costs to agricultural shippers and overrecovered
fuel costs with surcharges. Shippers question the reasonableness of rail rates in the light
of railroad revenue adequacy data.
INTRODUCTION
Individual agricultural producers in the short run are “price takers” rather than “price makers,” with
little control in the short run over the price they receive for their products (Kohls 1967). Due to
their numbers (many), relative size (small), the nature of their products (homogeneous with many
substitutes), and lack of market power, they have little or no ability to influence the price they
receive for their products, therefore, are usually unable to pass cost increases on to customers (USDA
November 7, 2006). Agricultural producers are unique in that they typically bear the transportation
costs, not the grain elevator, when grain is transported (Kohls 1967). Consequently, increases in
transportation costs result in decreased producer profit (Montana Wheat & Barley Committee et
al. 2006). For agricultural shippers with no cost-effective alternative to rail, and located far from
markets, rail is the only viable transportation available and the rail rate determines the net price the
producer receives (USDA November 2, 2006).
Lower prices and incomes hinder farmers from borrowing funds to purchase fertilizer, seed, and
machinery, reducing economic prosperity in rural areas. Higher transportation costs also affect the
competitive position of U.S. agricultural products in highly competitive export markets. The rates
agricultural shippers pay for rail transportation can facilitate or inhibit American competitiveness in
world agricultural markets (USDA November 2, 2006).
The costs of rail transportation to market represent a significant percentage of the average onfarm
price because grain and oilseeds are bulk commodities with a low value in proportion to their
weight (Figure 1). For example, average rail tariff rates as a percent of the farm price of wheat have
varied from 11.3% in 2007, when wheat prices were high, to 23.1% in 1999, when wheat prices
were low. Rail transportation costs for individual movements of agricultural products have been as
much as 40% of the delivered price (USDA 2005).
Despite these concerns, rates for land transportation of agricultural commodities in the United
States remain among the lowest in the world. Although rail rates for agricultural commodities have
not fallen as much as rates for some other products (such as coal) (GAO 2006), Figure 1 shows that
the rail transportation cost during 2007, as a percentage of the price of a bushel of wheat, was at a
14-year low.
This paper evaluates changes in agricultural rail rates relative to shipment size and distance shipped
to understand how they affect agricultural shippers. Since the Staggers Act, railroads have transferred
costs to agricultural shippers and over-recovered fuel costs with surcharges. Captive agricultural
shippers question the reasonableness of rail rates in light of railroad revenue adequacy data.
55

Rail Rate and Revenue Changes
Figure 1: Wheat–Average Rail Tariff Compared to Average Farm Price*
Average tariff rail rate per bushel
LITERATURE REVIEW
For nearly 100 years, the performance of railroads reflected the constraints put on them by federal
regulation. The Interstate Commerce Commission Act of 1887 (ICC Act) created the Interstate
Commerce Commission (ICC). The ICC implemented the provisions of the ICC Act, working for
“just and reasonable” rates without price discrimination. The regulatory environment created by the
ICC Act and subsequent statutes required railroads to employ cost-of-service pricing and to price at
average cost, with some variation usually allowed by commodity and length of haul.
Pervasive regulation interfered with the ability of railroads to react to competitive situations
and efficiently manage their firms. Rate adjustments were slow, innovations were stymied, and
rationalization of rail infrastructure was expensive and time-consuming (Gallamore 1999). The
unwieldy regulatory framework, along with increased competition from other modes—in part due
to government promotion of competing transportation modes—led to a loss of market share of
intercity freight and the attendant revenue (GAO 1990). The railroads were unable to maintain their
infrastructure, were close to bankruptcy, and were not competitive.
Regulatory reform happened slowly. The most important legislation was the Staggers Act
of 1980. Railroads seized on their new regulatory freedom to actively pursue profits and return
on investment using differential pricing, cost efficiencies, abandonment of un-remunerative rail
lines, mergers with other railroads, and the rate innovations of contracts and multiple-car pricing
(Gallamore 1999).
Railroads have also successfully controlled and reduced costs by abandoning rail lines,
creating short line railroads, reducing labor in operations and administration, making longer hauls,
increasing traffic density on rail lines, and using new technologies imaginatively (Gallamore 1999,
Prater and Klindworth 2000). Increasing shipment and car sizes, running directionally, 1 and sharing
dispatching have also contributed to efficiency.
56
$0.90
$0.85
$0.80
$0.75
$0.70
$0.65
$0.60
$0.55
$0.50
4.55
0.630
1995
4.30
0.613
1996
0.584
1997
3.38
2.65 2.55 2.62
0.572
1998
0.588 0.585
1999
2000
*Marketing year ending May 31.
Sources: STB Confidential Waybill Sample
USDA, NASS, Crop Value Summary
Average rail tariff rate per bushel
National average price of all wheat
2.78
0.573
2001
Year*
3.60
0.577
2002
0.694
3.40 3.40
3.40
0.592
2003
0.639
2004
2005
0.746
2006
4.26
6.48
2007
0.732
0.850
6.80
2008
8.00
7.00
6.00
5.00
4.00
3.00
2.00
1.00
0.00
Average farm price per bushel

JTRF Volume 50 No. 1, Spring 2011
Railroads adopted differential pricing to use their capacity efficiently and recover their high
fixed and common costs. If a railroad charged the same prices to all shippers, some shippers would
find it more profitable to ship by another mode. As these shippers withdrew, the railroad would have
to raise prices on its remaining customers to cover its fixed costs. Differential pricing also gives
railroads the flexibility to react to differences in modal competition (Prater and Klindworth 2000).
Consequently, the variable cost of providing rail transportation serves only as a floor below
which rates should not go and bears little relationship to individual rail rates. Instead, rail rates are
based on the price and service characteristics of competing transportation modes, the railroad’s own
price and service characteristics, and the railroad’s cost.
With differential pricing, shippers are charged different rates for the same service based on the
shipper’s dependence upon rail service. Differential pricing results in unequal rates and revenue-tovariable
cost ratios for different commodities, geographical locations, and producers, even in similar
circumstances. Consequently, with differential pricing, captive shippers bear a higher proportion of
a railroad’s fixed and common costs than non-captive shippers (Prater and Klindworth 2000).
The Staggers Act relies on competition to limit rail rates, but includes rate appeal procedures
to limit the rates railroads could charge captive shippers (who have no competitive transportation
choice). A shipper must meet three conditions to appeal rail rates (USDA and USDOT 2010):
• Shippers may appeal only tariff rates. 2 The STB has no jurisdiction over contract rates
and rates for exempt movements. 3
• The movement must have a revenue-to-variable cost ratio that exceeds 180%.
• The shipper must show that the railroad has market dominance, which is the lack of
effective intermodal and rail-to-rail competition.
In the early years of deregulation, intramodal competition may have been sufficient to yield a
competitive rail grain rate structure. However, rail mergers and line abandonments have reduced
intramodal competition considerably, resulting in an oligopolistic market structure that may allow
railroads considerably more pricing freedom.
Thus, although differential pricing offers shippers the benefit of having viable and stable
rail service, reaction to rail deregulation from shippers has not been all positive. Shippers feel
responsiveness to shipper needs has been lost, rail costs have been shifted to the shipper, overall rail
service and capacity have decreased, rates are generally increasing, and rates have been “unfair and
inequitable” in some corridors and for some commodities. Such shippers often charge that railroads
unreasonably raise their rates to levels that are far beyond those that should be charged (Montana
Wheat & Barley Committee et al. 2005; NGFA 2005; USDA 2005).
Numerous papers (cited below) discuss railroad industry competition and pricing, providing
varying degrees of analysis as to the impact of competition within the industry. Many of these
papers are regional in scope, investigate the impact of deregulation after the Staggers Rail Act of
1980, and the majority were written in the decade after enactment. The following is not a complete
survey of the prior research related to railroad competition, but instead emphasizes the interaction
between railroad competition and rail grain transportation prices.
Babcock, Sorenson, Chow, and Klindworth (1985) investigated the impact of the Staggers Act
on Kansas agriculture. The study found substantial railroad rate reductions in the four-year period
of 1981 through 1984. The pattern of rate changes suggested the presence of both intramodal and
intermodal transportation competition. Tariff rates to the Gulf of Mexico during this period dropped
34% compared with a 64% increase in the four years preceding the Staggers Act.
MacDonald (1987) used regression analysis of the 1983 waybill sample data to examine the
rail rates for corn, wheat, and soybeans. Rates were negatively related to tonnage, distance, and
volume of the shipments. Also, rates were negatively related to increased intramodal competition
(the reciprocal of the Herfindahl Index) and rail rates increased with distance to waterways.
A later study (MacDonald 1989) uses waybill data from 1981 through 1985 to analyze rail rates
and competition for corn, wheat, and soybeans. MacDonald found that when rail service goes from
a monopoly to a duopoly, rail rates decline 18%. The addition of a third competing railroad resulted
57

Rail Rate and Revenue Changes
in an additional 11% decrease in rail rates. Also, he found that shippers located 400 miles from barge
access paid rail rates that were 40% higher than those located 100 miles from barge access. Finally,
MacDonald calculated inverse Herfindahl-Hirschman indices for each crop reporting district (CRD)
and concluded that each CRD was characterized by rail oligopolies.
Chow (1986) studied post-Staggers rail grain rates for the Central Plains region. His analysis
indicated an overall reduction in wheat rail rates of 34.5% in the five-year period after enactment of
the Staggers Act, with the most significant reductions occurring in movements to the export markets.
Kwon, Babcock, and Sorenson (1994) examined the impacts of the Staggers Act in the latter
half of the 1980s and found that railroads practiced differential pricing for intra Kansas and export
shipments of wheat. They discovered substantial differences in the factors affecting the revenueto-variable
cost ratios for intra Kansas wheat movements versus that of Kansas export wheat
movements. Revenue-to-variable cost ratios increased steadily from 1986 through 1989, but this
could have been caused by diminishing export demand.
Fuller, Bessler, MacDonald, and Wohlgenant (1987) found deregulation to have had a significant
effect on rail corridors linking Kansas and Texas with Gulf ports and a relatively modest effect on
the corridor linking Indiana with East Coast ports. Real rail rates declined $.37 per bushel in the
Kansas corridor and $.31 per bushel in the Texas corridor during the 1981-1985 period. In the
Indiana corridor, real rail rates were estimated to decrease $.08 per bushel. Railroad deregulation
had little statistically significant effects on real rail rates from Iowa and Illinois to the Gulf ports.
Koo, Tolliver, and Bitzan (1993) examined railroad pricing behavior in North Dakota, which
is often considered a captive railroad shipping market. The region has unique transportation
characteristics that include limited intermodal competition due to great distances to barge-loading
facilities and to major domestic and export markets. They found that distance, volume, weight per
car, intramodal competition, and intermodal competition had significant negative effects on rail
rates.
Thompson, Hauser, and Coughlin (1990) evaluated the pre- and post-Staggers effect of
competition on railroad revenue-to-variable cost ratios for export shipments of corn and wheat. The
regression results for corn were less significant than those of wheat. There was a lack of identifiable
differences in pre- and post-Staggers pricing, which may be attributable to the close correlation
between changes in operating factors, such as shipment size, and destination opportunity. They
concluded that their results did not indicate a clear effect of the Staggers Rail Act on rail rate
competitiveness.
Wilson and Wilson (2001) examined rail rates for barley, corn, sorghum, wheat, and soybeans
moved by rail. The explanatory variables include commodity ton-miles, commodity prices, average
length of haul, and a non-linear specification of deregulation that allows the effects to phase in over
time. They found that commodity prices have positive effects on rail rates and length of haul has a
strong negative effect. The results indicate a large negative effect on rates from deregulation, which
dissipate with time.
The STB waybill rate data are used in Figure 2 to examine the real revenue per ton-mile for
the period 1985 to 2007. The STB uses the Tornqvist Index to track rail rates. The Tornqvist index
measures the change in prices in commodity categories and assigns a percentage weight to each
category based on its share of total revenue. The index is essentially the weighted average of price
changes within the various commodity categories. Both the prices within the various commodity
categories and the weights assigned to each category can vary (STB 2009).
The downward pressure on rates identified above as a result of railroad efficiency improvements
and competitive pricing is evident. From 1985 to 2004 the rail rate index fell almost continuously,
with only a slight increase being noted in 2002. However, as frequently stated to the STB by
shippers, the years since 2004 have seen rapidly increasing rates for shippers. Starting in 1985, rail
rates dropped about 10% in the first two years, continued dropping at nearly that rate through 1992,
and then declined at a slower rate during the period between 1992 and 2000. Over the next few
58

100.0
94.4
90.5
88.1
83.7
80.8
77.0
72.4
71.8
69.6
67.2
65.9
64.4
JTRF Volume 50 No. 1, Spring 2011
Figure 2: STB Rail Rate Index 1985 to 2007; Real Revenue Per Ton-Mile (1985=100)
Tornqvist Index
100.0
90.0
80.0
70.0
60.0
50.0
40.0
years, the rates hovered in a narrow range, varying both positively and negatively until 2004. From
2004 to 2007 the rate index increased nearly 15%, from 56.8 to 65.5 (STB 2009).
Various studies (GAO 1990, GAO 2006, Christensen 2008) have agreed with the findings that
overall rail rates decreased substantially from the mid-1980s to the early 2000s. The causes of the
decrease included:
• The rationalization of the rail network, with abandonments and creations of short line or
regional railroads decreasing costs while maintaining much of the original traffic.
• The increase in trainload shipments.
• The shifts to larger-capacity rail cars and technology innovations.
The recent STB study of railroad rates from 1985 to 2007 (2009) found that “inflation-adjusted
rates” increased from 2005 to 2007. The STB wrote: “This represents a significant change from prior
years, given that inflation-adjusted rail rates declined in every year but one from 1985 through 2004.”
The STB further elaborated: “In fact, adjusting for the purchasing power of the dollar, shippers spent
$7.8 billion more in 2007 than they would have if the rate levels of 2004 had remained in place.”
The STB rate study (2009) further points out that well over half the increase in rail rates between
2004 and 2007 could be attributed to higher fuel costs. Yet, even after consideration of fuel costs,
railroad rates have been steadily increasing during the last few years (STB 2009).
The Government Accountability Office (GAO) has reported that the percentage of traffic in
tons traveling at rates above a revenue-to-variable cost ratio (R/VC) of 300, which is substantially
above the statutory jurisdiction level of 180, has generally increased from 1985 through 2005 (GAO
2007). The share of tonnage traveling at rates over 300% R/VC increased from 6.1% in 2004 to
6.4% in 2005.
This paper examines the R/VC for railroad grain and oilseed movements and railroad rates
for grain and oilseed movements by shipment size and distance. Railroad fuel surcharges are also
examined and found to exceed the growth in railroad fuel costs. Finally, railroad revenue adequacy
is examined relative to railroad industry costs and merger premiums.
62.4
60.8
58.6
58.1
58.8 57.2
56.8
65.5
64.3
60.8
1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007
Year
Source: STB, Study of Railroad Rates: 1985-2007
59

Rail Rate and Revenue Changes
RECENT RAIL RATE LEVELS
This study calculated average R/VC ratios for grain and oilseed tariff rates by dividing the freight
revenue field in the Confidential Waybill Sample by the variable cost field. The variable costs are
calculated for the waybill sample by the STB using its Uniform Rail Costing System.
Figure 3 shows a slight downward trend from 1988 to 1998 in the percent of grain and oilseed
tonnage traveling above an R/VC of 300%. The increase in the percentage of tons moving at R/VC
greater than 300% began in 1999 and peaked at 7.7% in 2002, then decreased to 2.8% in 2006 and to
2.4% in 2007. The vertical line in figures 3 and 4 denotes the lack of waybill data for 1992 and 1993.
Figure 3: Percent of Grain and Oilseed Tons Moved at Tarriff Rates with R/VC Greater
Than 300%
Percent
In some states, however, a much greater percentage of grain and oilseed tonnage moves at R/VC
ratios greater than 300% (Figure 4). These states include Iowa, Montana, and North Dakota. The
high percentage of rail rates exceeding a R/VC of 300% for Montana shippers from 1998 through
2004 could be due to its distance from intermodal competition and the fact that one railroad handles
95% of the rail movements of grain. However, rates and R/VC ratios for movements of agricultural
commodities can differ from state to state for numerous reasons, and can change significantly from
year to year as Figure 4 shows.
The analysis of rail rates is limited to tariff rates since contract rates are confidential and
unregulated. However, several limitations in Waybill Sample data mean that tariff rates should
be used with some caution. Volume discounts and rebates for use of non-railroad equipment are
not included in tariff rates. Fees for guaranteeing delivery of rail equipment on specific dates, or
“certificates of transportation” payments, are not included. Also, in some instances contract rates
can differ substantially from tariff rates, while in other instances there can be little if any difference
between contract and tariff rates. Thus, the use of Waybill Sample tariff data and costs for the
calculation of R/VC ratios can provide a misleading picture for some comparisons. While these
anomalies can distort the R/VC calculations for some comparisons, the results presented in the rate
analysis for this study are thought to be generally representative of rate trends over the period.
Agriculture Rates are Higher Than Those of Other Commodities
The GAO found that “although rates have declined since 1985, they have not done so uniformly, and
rates for some commodities are significantly higher than rates for others” (GAO 2006) (Figure 5).
Specifically, GAO found that “grain rates declined from 1985 through 1987, but then diverged from
the other commodity trends and increased, resulting in a net 9% increase by 2004.” In 2005, rates
60
10
9
8
7
6
5
4
3
2
1
0
3.7
1988
2.6 2.6 3.0 3.5 2.9 2.1 2.5 2.5
1989
1990
1991
1994
1995
1996
Source: USDA analysis of STB Waybill Sample
1997
1998
Year
5.8 5.4
1999
2000
6.9 7.7 7.6 6.7
2001
2002
2003
2004
4.7
2005
2.8 2.4
2006
2007

JTRF Volume 50 No. 1, Spring 2011
Figure 4: Selected States with Higher Percentages of Grain and Oilseeds Moving at R/VC
Over 300%
Percentage
50
45
40
35
30
25
20
15
10
5
0
1988
1989
1990
1991
1994
for all commodities increased by 9% over 2004 rates, the largest annual increase in 20 years. Rail
rates for grain increased 8.5% over 2004 (GAO 2007).
According to the AAR Freight Commodity Statistics, agricultural rates not only are higher than
those of other commodities, but also have increased more rapidly (see Figure 6). For instance, rail
rates for grain and oilseeds increased to $2,809 per carload in 2008, up 73% from 2003; rates for
all other commodities increased to $1,556 per carload, up 50%. In addition, grain and oilseed rates
during 2008 were 81% higher than those paid by all other commodities, compared with 55% higher
in 1998. The average number of grain and oilseed tons per carload increased 2.2%, from 97.6 tons
per carload in 1998 to 99.7 tons per carload in 2008.
Comparison of Rates by Shipment Size and Distance Shipped
1995
1996
Source: USDA analysis of STB Waybill Sample
1997
1998
The STB waybill sample allows specific analysis of grains and oilseeds, which is presented in detail
in this section. This study did not have access to the unmasked confidential waybill data, which report
the unmasked rail rates for contract movements as well as for tariff movements. Consequently, only
tariff rail rates are analyzed in this section. In addition, samples with fewer than 30 observations are
not included in the figures to increase the statistical reliability of the analysis. The vertical lines in
figures 7 and 8 denote the lack of waybill data for 1992 and 1993.
The rates for grain and oilseeds reflect a significant advantage for large trainload shipments.
As can be seen in Figure 7, rates for all shipment sizes have risen steadily and rapidly since 2003.
The rates for the smallest shipment size have increased 21% since 2003, compared with 25 and
23%, respectively, for 6–49 car and 50+ car shipments, keeping the relative relationships between
shipment size categories about the same over the last five years analyzed. However, since 1988,
the rates for the smallest shipment sizes have increased by only 13%, while the rates for 6–49 car
shipments and 50+ car shipments have increased by 40% and 43%, respectively. This shows that
the rates for larger sized shipments have increased relatively more than for smaller shipments over
the entire period.
Rates for large shipments are nearly 2.1 cents per ton-mile, contrasted with about 3.0 cents for
smaller movements. Rates for large shipments are about one cent or 33% lower than the smallest
1999
2000
2001
2002
2003
2004
2005
2006
Iowa Montana North Dakota
2007
61

Rail Rate and Revenue Changes
Figure 5: Rate Changes for Coal, Grain, Mixed Shipments, and Motor Vehicles, 1985–2005
Source: GAO analysis of STB data
Figure 6: Railroad Average Freight, Revenue per Carload
Dollars per carload
62
3,300
2,800
2,300
1,800
1,300
1,559
Grain and oilseeds up 73%; 2008 over 2003;
All other commodities up 50%; 2008 over 2003.
1,613 1,615 1,608 1,609
1,624
1,786
1,004 1,008 1,017 1,028 1,030 1,039 1,078
2,063
1,171 1,281
2,251 2,369
2,809
1,362 1,556
800
1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008
All other commodities Grain and oilseeds
Source: Association of American Railroads, Freight Commodity Statistics

0.035
0.030
0.025
0.020
0.015
0.010
JTRF Volume 50 No. 1, Spring 2011
Figure 7: Grain and Oilseeds Tariff Revenue (Current $) Per Ton-mile by Shipment Size
Dollars per ton-mile
Source: STB Waybill Samples
1988
1989
1990
1991
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
1-5 cars 6-49 cars 50+ cars
shipment size. In 1988, though, large shipments were 46% less than small shipments. The discount
for medium-size shipments relative to small shipments has decreased substantially—from 23% in
1988 to only 5% in 2007.
A similar situation holds for shorter distance movements, in which rates are consistently about
double the rates for movements over 751 miles in length—4.6 cents versus 2.15 cents per tonmile
in 2007. Rates for short hauls started increasing in 2000, but longer hauls didn’t have sharp
increases until the last four years (see Figure 8).
Figure 8: Grain and Oilseeds Tariff Revenue (Current $) Per Ton-mile by Shipment Distance
Dollars per ton-mile
0.050
0.045
0.040
0.035
0.030
0.025
0.020
0.015
Source: STB Waybill Samples
20-500 miles 501-750 miles 750+ miles
63

Rail Rate and Revenue Changes
This detailed examination of the grain and oilseeds commodity group shows the effect of
distance on railroad rates. As railroads seek to increase the usage and revenue generation of their
rolling stock, it is often in their best interest to give price/rate incentives to shippers with long hauls.
Also, the cost disadvantages in equipment utilization make the short hauls more expensive for the
carriers.
It is widely acknowledged that railroads have used trainload or multiple car rates to encourage
shippers to consolidate shipments, thereby increasing the efficiency of the capital stock, power, and
labor. The analyses above demonstrate that the longer the movement, the lower the rate charged by
the railroads. However, the analysis does not consider the costs that have been shifted to the shipper
so they could access these rates.
TRANSFER OF RAILROAD COSTS TO SHIPPERS
Rail rates have decreased since deregulation in the early 1980s. Inflation-adjusted rates have
decreased by slightly over 30% since 1985. However, a broad and consistent increase in rail rates
over at least the last four years—and for some commodities the last seven years—indicates the
railroads have used rates to achieve profit levels previously unseen in the industry.
Moreover, the overall decrease in revenue per ton-mile for railroads does not reflect the actual
impact on shippers. The logistical cost to shippers, and to the public, has increased over that time.
The Christensen study defined cost-shifting as additional costs incurred by shippers as a
result of changes in railroad operations. Examples of cost-shifting identified in that study include
(Christensen 2008):
• A shift in railcar ownership and its associated expenses, such as maintenance and
insurance, from railroads to shippers or other private firms.
• Increased railcar maintenance standards being required by railroads as necessary to
maintain service and capacity.
• Increases in and additions to accessorial charges, such as finance charges, charges for
faxing versus electronic transmission, higher demurrage charges, private car storage
charges, and car cleaning charges.
• Deterioration in railroad service, causing the increased use of shipper labor to monitor
railroad performance or to unload railcars.
• The use of trucking to transport goods to distant terminals to access multiple-car rates.
• Increased highway congestion and maintenance because of the increased use of trucking.
The average rate per ton-mile has decreased, in part, because all shippers, and especially grain
shippers, are assuming greater responsibility for car supply and other functions that railroads have
traditionally provided. Many shippers, in times of short railcar supply, use guaranteed rail-ordering
systems, paying fees in addition to tariff rates to guarantee car delivery within a specified time
period rather than risking a delay in receiving railcars on a first-come-first-served basis (Prater and
Klindworth 2000).
The attractiveness of unit and shuttle trains due to the railroad’s rate structure has caused
shippers to invest in sidings, inventory, storage capacity, and loading facilities to access these more
cost-effective rail services. Shippers note that, after investing in equipment to handle 50–54-rail-car
shipments, the railroads have changed some rate structures to emphasize 100–110-car shipments,
requiring further investments.
The costs of railcar ownership have shifted from railroads to shippers, adding further to costs
not reflected in tariff rates. As can be seen in Figure 9, private ownership has been the source, in
a steady increase, of new covered hopper railcar capacity. In 1981, private ownership accounted
for 41% of the total covered hopper cars, with the Class I railroads providing 56% and the smaller
railroads contributing 3% of the capacity (AAR, Railroad Equipment Report 1982). By 2008,
hopper car ownership was 70% private, 26% Class I railroad, and 4% smaller railroads (AAR,
Railroad Equipment Report 2008). Another way of looking at rail car ownership is to see that from
64

Figure 9: U.S. Covered Hopper Fleet Ownership, 1982–2008
Cars
500,000
400,000
300,000
200,000
100,000
0
1982
1984
1986
1988
Source: AAR, Railroad Equipment Report
JTRF Volume 50 No. 1, Spring 2011
1981 to 2008 privately owned cars increased from 128,394 to 290,176, or 126%, as Class I railroads
decreased their ownership by 37% (AAR, Railroad Equipment Report 1982-2008). The costs of car
ownership have been shifted to the shippers or their agents.
Fuel Surcharges Versus Fuel Prices
1990
1992
1994
Rates per-ton-mile decreased from the time of deregulation until around 2002. Over the last four
years these rates have significantly increased. Recently, railroad fuel charges have added to the
shipper’s cost burden. These surcharges are designed to allow railroad firms to recover from shippers
the impact on costs caused by abnormally high fuel prices. Basic fuel charges have always been
included in rail rate determination, but the recent spikes and variation in fuel prices caused railroads
to search for ways of recapturing these costs in the near term.
The fuel cost increases were first estimated as a percentage of tariff rates, but shippers felt
any errors in estimation were on the side of the railroad carrier. As fuel prices and the attendant
fuel surcharges were implemented, shippers felt that carriers were using these surcharges as profit
centers, whether the fuel costs were going up or down. They also believed that rate-based fuel
surcharges did not fairly apportion the additional cost of the fuel among shippers. Subsequent to a
regulatory proceeding on rail fuel surcharges, the STB (STB 2007) on January 25, 2007, ruled that:
• Computing rail fuel surcharges as a percentage of a base rate is an unreasonable business
practice because rail rates do not accurately reflect the additional cost of fuel used in
individual movements. The STB reasoned that a rate-based fuel surcharge would result in
shippers who pay higher rail rates also paying higher fuel surcharges.
• The fact that a railroad may not be able to recover its increased fuel costs from some of
its traffic does not provide a reasonable basis for shifting those costs onto other traffic.
• Railroads are prohibited from “double dipping”—charging a fuel surcharge in addition to
increasing rates using an index that includes fuel costs as a component.
• Railroads operating in the United States had until April 26, 2007, to change their fuel
surcharge programs to comply with the STB ruling.
When examining the performance of fuel surcharges in recovering fuel cost increases, wide
differences among fuel surcharge rates cause concern about the accuracy of surcharge formulas.
For instance, during September 2008, when surcharges peaked, they varied among railroads from
1996
1998
2000
2002
2004
Total Hoppers Private Owned
Class I Railroad Small Railroad
2006
2008
65

Rail Rate and Revenue Changes
Figure 10: Growth in Railroad Grain Fuel Surcharges v. Growth in Railroad Fuel
Costs by Quarter
750
650.77
650
1
46.58 cents to 87 cents per car mile, a difference of nearly 87%. The weighted average surcharge
was 59 cents per car mile or $590 per car moving 1,000 miles.
Shippers contend that fuel surcharges should reimburse railroads for only the incremental
increase in fuel costs and not the base, since the base fuel costs are already in the rate. The average
growth in fuel surcharge per grain carload during the 3 rd quarter of 2008 was $650.77, contrasted to
the growth in railroad fuel costs from 2001 until the 3 rd quarter of 2008 of $286.46, a difference of
127% over the incremental increase in the cost of fuel (see Figure 10).
Figure 11 shows that the percentage by which the growth in grain fuel surcharges exceed the
growth in railroad fuel costs since 2004 ranges from -30% to 163%. Note that the percentage growth
in grain fuel surcharges exceeds the growth in railroad fuel costs in all but two quarters. Furthermore,
as fuel costs increase, the difference between the quarterly growth in grain fuel surcharges and the
growth in railroad fuel costs tends to increase. This correlation is not perfect mainly because fuel
surcharges lag fuel cost increases by two months.
RAILROAD REVENUE ADEQUACY
An evaluation of railroad rate reasonableness requires consideration of both the relative profitability
and costs of railroads. Such an evaluation must also consider merger premiums 4 and how they relate
to STB revenue adequacy measures 5 , as well as the revenue adequacy and profitability of Class I
railroads over time.
66
$ per carload
550
450
350
250
150
50
(50)
4Q '01
2Q '02
4Q '02
2Q '03
4Q '03
2Q '04
4Q '04
2Q '05
4Q '05
2Q '06
4Q '06
2Q '07
4Q '07
2Q '08
286.46
4Q '08
Growth in Grain Fuel Surcharges (since Qtr. 4, 2001)
Growth in Rail Fuel Costs (Since Qtr. 4, 2001)
1 For the 7 Class I railroads operating in the United States. Weighted average fuel surcharges per carload
were estimated by multiplying the average length of haul for grain by the quarterly weighted average fuel
surcharge per carload mile. Weighted average fuel surcharge per carload mile prior to 2Q ’06 was estimated
using mileage-based fuel surcharge formulas for individual railroads.
Source: Class I Railroad quarterly filings to the Security and Exchange Commission.
AAR, The Rail Transportation of Grain (2009).
2Q '09
4Q '09

Figure 11: Percentage Growth in Grain Fuel Surcharges Exceed Growth
in Railroad Fuel Costs
200%
150%
100%
50%
0%
-50%
Merger Premiums and STB Revenue Adequacy
JTRF Volume 50 No. 1, Spring 2011
1Q '04
2Q '04
3Q '04
4Q '04
1Q '05
2Q '05
3Q '05
4Q '05
1Q '06
2Q '06
3Q '06
4Q '06
1Q '07
2Q '07
3Q '07
4Q '07
1Q '08
2Q '08
3Q '08
4Q '08
1Q '09
2Q '09
3Q '09
4Q '09
Source: Class I Railroad quarterly filings to the Security and Exchange Commission.
AAR, The Rail Transportation of Grain (2009).
The STB annually measures the revenue earned from the rate structure against the adequacy of that
revenue stream to infuse capital into the industry. To determine the annual revenue adequacy, the
carrier’s return on net investment (ROI) is compared with the rail industry’s after-tax cost of capital
for that year. If ROI is greater than the cost of capital, revenue is determined to be adequate.
ROI is normally determined by dividing net income from railroad operations by the depreciated
original cost, or book value, of the railroads’ assets. This ROI is then compared with the railroad
industry cost of capital. The STB seeks to ensure that a railroad has the capability to invest in its
infrastructure and provide a reasonable return to its investors.
The costs to be included in determining the critical ROI have been examined in various
proceedings and shipper testimonies. If the depreciated book value, or original cost, is increased,
then calculated ROI decreases and revenue adequacy is negatively affected, allowing railroads to
charge higher rates. In addition, the STB’s Uniform Rail Costing System uses these higher cost
values in the calculation of variable costs, thereby driving down the R/VC calculations for the
railroad’s movements.
Shippers and shipper representatives have become concerned about the premiums being paid
to newly formed railroads when a merger is granted. The ICC/STB has been consistent in allowing
such premiums, usually above the current stock or book price prior to the merger, to be included
in the depreciated cost figure. Shippers have argued that railroads should not be allowed to pay
acquisition premiums if these costs are then used to decrease the railroad firm ROI, which is used
for revenue adequacy determination. This can result in the railroads being allowed to charge higher
rates than would have been possible if the premiums had not been paid, resulting in economic
impact and harm to the shippers.
The extent of these premiums is difficult to determine, but some information is available.
Recent mergers have involved significant premiums paid by the merging railroads. These estimated
premiums range from $1.4 billion for the Union Pacific (UP) purchase of Chicago Northwestern
in 1996 to $2.7 billion for the Atchison, Topeka, & Santa Fe merger with Burlington Northern in
1995, and $3.7 billion for UP’s purchase of Southern Pacific in 1996. Consultants estimate that the
premium paid for Conrail by Norfolk Southern and CSX Transportation was about $6.9 billion.
Other estimates also have been generated, but the relevant point is that these premiums, if added to
the book value of the merger, affect the ROI value used for revenue adequacy purposes.
67

Rail Rate and Revenue Changes
The railroad industry and the STB are the only industry and regulator that use book value for
determining ROI and add merger premiums into the rate base. For example, the Federal Energy
Regulatory Commission will not allow regulated entities to pass through to the customer acquisition
or merger premiums unless the effect of the transaction has a net benefit (typically, a rate reduction)
to the customers of the acquired entity. The net result is that this approach discourages the payment
of large premiums because they are not likely to be permitted to be passed through to customers.
The net effect of merger premiums, which increase both variable and fixed costs of the railroads,
is that some rates that would have been above 180% of variable costs might no longer meet that
criterion and would no longer be subject to STB regulation.
A contrasting opinion on the ROI calculation is offered by the railroad industry and the
AAR. The railroads, through the AAR, have argued that the ROI calculation should be based not
on depreciated value but on the replacement cost of the rail assets used to provide transportation.
Merger premiums also could reflect the net present value of the merger in expected cost savings to
the combined firm.
STB Measures of Rail Revenue Adequacy
Class I railroad revenue adequacy is determined by comparing the ROI with the cost of capital. The
STB determines the cost of capital for each year and determines which Class I railroads are revenue
adequate. The STB used a simple discounted cash flow (DCF) method to determine the industry’s
weighted average cost of capital through 2005. After shippers requested public hearings to examine
the methodology, the STB then changed to a capital asset pricing model (CAPM) for the years 2006
and 2007. After another public hearing, STB decided to use a simple average of CAPM and a multistage
discounted cash flow model (MSDCF) in 2008 and beyond.
Since the Staggers Act, the ROI for the railroad industry has increased from an average of 2.5%
during the 1970s to an average of 10.25% from 2006 through 2008 (AAR, Railroad Facts).
Based upon the CAPM methodology, Figure 12 shows that the Class I railroads have been
revenue adequate during 2005 and 2006 and nearly revenue adequate for the other years since
2002. In contrast, the Christensen (2008) study, which used return on equity, found that the Class I
railroads could be considered revenue adequate since 2001.
Figure 12: Class I Railroad Cost of Capital and Return on Net Investment, 1997–2008
Percent
68
14
12
10
8
6
4
2
0
1997
1998
1999
2000
2001
2002
DCF CAPM ROI CAPM + MSDCF
Sources: AAR Railroad Facts; Surface Transportation Board
2003
2004
2005
2006
2007
2008

Financial Measures of Railroad Profitability
JTRF Volume 50 No. 1, Spring 2011
Whether measured by commonly used financial measures or by STB-determined revenue adequacy
standards, the profitability of the railroad industry has improved considerably since deregulation.
The Christensen (2008) study used various measures of profitability to compare the railroad industry
with other industries and with the Standard & Poor’s 500. Since 2004, railroad profitability was
found to be comparable to that of most other industries (Christensen 2008).
The rapid increase in rail rates since 2004 contributed to the surge in railroad profitability at that
time. The increase in rail rates is the result of aggressive pricing as rail capacity constraints appeared
and the over-recovery of fuel costs. The higher rail rates also reflect higher rail costs since 2004
(Figure 13).
Railroad financial measures of profitability increased at a moderate rate through 2004, and
then surged from 2005 through 2007. Net income, earnings before interest and taxes (EBIT),
and earnings before interest, taxes, depreciation, and amortization (EBITDA) are commonly used
financial measures of profitability. Net income, EBIT, and EBITDA changed 2%, -5%, and 5%,
respectively, over the six-year period from 1998 to 2004. Over the four-year period from 2004 to
2008, net profit, EBIT, and EBITDA increased 183%, 152%, and 102%, respectively.
Figure 13: Class I Railroad Profitability
Billion $
20
18
16
14
12
10
8
6
4
2
0
1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008
Sources: AAR, Analysis of Class I Railroads
Factors Affecting Railroad Industry Costs
Net income EBIT EBITDA
Several factors affecting railroad costs are often overlooked when analyzing those costs. Railroad
management decisions affect some of these factors, which include merger premiums, size of
operation, traffic density, the amount invested in capacity, and successful integration of operations
during mergers. Other factors, such as unusually high fuel costs and extreme weather events, are
factors that railroad management are unable to control.
As discussed in an earlier section, merger premiums can add substantially to the average fixed
and variable costs 6 of the new railroad firm. The effects of these mergers, including increased
costs due to merger implementation difficulties, are visible in Figure 14, showing average railroad
industry costs. Variable costs for the railroad industry increased from 1997 through 2000 and fixed
costs increased from 1995 through 1997.
69

Rail Rate and Revenue Changes
Figure 14: Railroad Industry Average Cost, Variable Cost, and Fixed Cost in Dollars
Per Ton-mile (adjusted for inflation in 2000 dollars)
Source: Laurits Christensen Associates (2008)
Variable and total costs for the merged railroads increased after each of these major rail mergers
or acquisitions:
• The merger of the Atchison, Topeka, & Santa Fe with the Burlington Northern
(September 1995)
• The Union Pacific with the Southern Pacific (implemented December 1996)
• The split of Conrail between CSXT and Norfolk Southern (June 1999)
In each of these mergers, the railroads had difficulties merging operating systems and lines,
resulting in congestion that drove up average total and variable costs for the merging railroads.
A recent study of railroad cost curves concluded that the four largest Class I railroads, BNSF,
CSXT, NS, and UP, may have surpassed the optimal size of operation and may be experiencing
diseconomies of scale (Bereskin 2008). This means that the average costs for those railroads are
higher than they would be if the firms were smaller. Based upon 2005 data, the optimal size of a
railroad was estimated to be slightly less than 21,000 route miles. BNSF and UP operate more than
32,000 route miles, while CSXT and NS operate more than 21,000 route miles. The three smaller
Class I railroads, Kansas City Southern, Canadian National, and Canadian Pacific, all appear to be
operating with constant or increasing returns to scale.
Excess traffic density on the railroad also affects railroad average costs by slowing train speeds
and increasing terminal dwell times. The slower train speeds and reduced terminal efficiency further
reduce the effective capacity of the railroad, compounding the problem. When a railroad has excess
capacity, fixed costs are higher than necessary. As railroads near capacity and capacity constraints
appear, variable costs increase. The effects of railroad capacity constraints, beginning in 2005, are
also visible in the above figure, which shows average railroad industry costs.
Another factor affecting average railroad fixed and variable costs is the amount the railroad
industry invests in rail capacity. From 2004 through 2006, the railroad industry invested heavily
70
$0.035
$0.030
$0.025
$0.020
$0.015
$0.010
$0.005
$0.000
1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006
ATC AVC AFC

JTRF Volume 50 No. 1, Spring 2011
in capacity, which is shown in the average cost data in Figure 14 as increased fixed costs for the
industry after 2004. As capacity bottlenecks are removed, however, variable costs should be reduced
by these investments.
Unusually high fuel costs occurred from 2004, peaking in September of 2008. Fuel is a major
component of railroad costs, so high fuel costs result in increased operating costs. This can be seen
in the figure above, where high fuel costs and capacity constraints resulted in increasing variable
costs in 2005 and 2006.
Extreme weather events can also increase railroad industry costs by adding the costs of repair
and rerouting traffic. For instance, Hurricanes Katrina and Rita resulted in substantial damage to the
rail network in Louisiana and Mississippi. Damages to the CSX coastal line, which had the most
damage, required nearly $250 million to repair (CSX Transportation 2006). Likewise, a massive
mudslide on the UP line between Klamath Falls and Eugene, OR, swept track, ties, and ballast
halfway down the mountain and buried over 3,000 feet of mainline track in 20 feet of mud, snow,
and downed trees (Union Pacific Railroad 2008).
Railroad Industry Revenue Compared to Marginal Costs
Railroad industry revenue per ton-mile decreased slowly through 1996, rose slowly through 2004,
and then increased rapidly in 2005 and 2006. Marginal costs (i.e., the addition to total cost attributable
to the addition of one ton-mile) increased in 2005 and 2006, probably due to rail congestion as
capacity constraints in the rail network and higher fuel costs drove marginal costs up (see Figure
15). Average revenue increased more rapidly than marginal costs in 2005 and 2006, indicating
aggressive pricing due to capacity constraints and over recovery of fuel costs.
Figure 15: Railroad Industry Average Revenue Per Ton-mile and Marginal Costs
$0.035
$0.030
$0.025
$0.020
$0.015
$0.010
$0.005
$0.000
19871988 1989 1990 1991 19921993 19941995 1996 19971998 1999 2000 2001 2002 2003 20042005 2006
RPTM MC
Source: Laurits Christensen Associates (2008)
71

Rail Rate and Revenue Changes
SUMMARY
In both the short and long run, farmers cannot individually raise the prices of their commodities to
reflect rising costs, so any increase in costs reduces their individual profit. Agricultural producers
are unique in that they bear the transportation costs when grain is transported; in the short run the
price the producer receives from the elevator is net of transportation costs. Consequently, increases
in transportation costs result in decreased producer profit.
Rail rates need to be sufficient to provide railroads with sufficient profit to maintain their
equipment and lines, provide reliable service, invest in needed capacity, and provide a reasonable
return on investment. However, unnecessarily high rail rates can damage the economic health of
the farming sector and rural communities, and also make it more difficult for America to compete
in export markets.
The costs of rail transportation to market are important to agricultural producers because
they represent a significant percentage of the average on-farm price; grain and oilseeds are bulk
commodities with a low value in proportion to their weight. Average rail tariff rates as a percent of
the farm price of wheat have varied from 11.3% in 2007, when wheat prices were high, to 23.1% in
1999, when wheat prices were low.
Agricultural producers have become more concerned with rail rates as rail mergers and line
abandonments after implementation of the Staggers Act, with its limited regulation of rates, have
led to less intramodal competition.
This study reported an increase in the percentage of tons moving at R/VC greater than 300%
that began in 1999, peaked in 2002, and decreased until 2007. In Iowa, Montana, and North Dakota,
however, a much greater percentage of grain and oilseed tonnage moved at R/VC ratios greater than
300%. The high percentage of rail rates exceeding an R/VC of 300% for Montana shippers from
1998 through 2004 could be due to its distance from intermodal competition and the fact that one
railroad handles 95% of the rail movements of grain.
Not only are rail rates for agricultural products higher than those for other commodities, but the
rates have increased more rapidly from 2004 to 2008. Rail rates for grain and oilseeds increased to
$2,809 per carload in 2008, up 73% from 2003; rates for all other commodities increased to $1,556
per carload, up 50%.
Railroad rate structures favor large movements due to cost and operational efficiencies. There is
a significant rate advantage for the largest trainload shipments of grain and oilseeds. Rates are 30%
lower for shipments of more than 50 cars; rates for large shipments are about 2.1 cents per ton-mile,
contrasted to about 3.0 cents for smaller movements. The rates for larger sized shipments have
increased relatively more than for smaller shipments over the entire period, thus, factors other than
cost efficiencies may be driving the rate changes.
Rates for long hauls have a similar structure; movements less than 500 miles are about twice the
rates for movements over 751 miles, 4.6 cents versus 2.15 cents per ton-mile in 2007. As railroads
seek to increase the usage and revenue generation of their rolling stock, it is often in their best
interest to give price/rate incentives to shippers with long hauls. Also, the cost disadvantages in
equipment utilization make the short hauls more expensive for the carriers.
Shippers bear increasing responsibility for car supply and other functions historically provided
by the railroads. The costs of railcar ownership have shifted from railroads to shippers, adding
further to costs not reflected in tariff rates. Private ownership has been the source of new covered
hopper railcar capacity; by 2008, hopper car ownership was 70% private, 26% Class I railroad, and
4% smaller railroads. From 1981 to 2008, privately owned cars increased from 128,394 to 290,176,
or 126%, as Class I railroads decreased their ownership by 37%.
Fuel surcharges should reimburse railroads for only the incremental increase in fuel costs, not
the base fuel costs that are already included in the rate. The average growth in fuel surcharge per
grain carload during the 3 rd quarter of 2008 was $650.77, contrasted to the growth in railroad fuel
costs of $286.46, a difference of 127% over the incremental increase in the cost of fuel. Furthermore,
72

JTRF Volume 50 No. 1, Spring 2011
as fuel costs increase, the difference between the quarterly growth in grain fuel surcharges and the
growth in railroad fuel costs tends to increase.
This study also reports that billions of dollars in premiums paid as part of mergers are included
in the determination of railroad revenue adequacy, which is likely to result in higher rail rates for
shippers than otherwise would be the case.
Rail rates have increased rapidly since 2004, resulting in a surge of railroad profitability. The
increase in rail rates is the result of aggressive pricing as rail capacity constraints appeared, and the
over-recovery of fuel costs. The higher rail rates also reflect higher rail costs since 2004.
Railroad financial measures of profitability increased at a moderate rate through 2004, and
then surged from 2005 through 2007. Net income, earnings before interest and taxes (EBIT), and
earnings before interest, taxes, depreciation, and amortization (EBITDA) are commonly used
financial measures of profitability. Over the four-year period from 2004 to 2008, net profit, EBIT,
and EBITDA increased 183%, 152%, and 102%, respectively.
Endnotes
1. Running directionally occurs when two railroads with parallel rail lines share rail lines, allowing
shared use of one line in each direction. This practice eliminates trains waiting on sidings for
trains moving in the opposite direction to pass.
2. Tariff shipments comprise approximately 75% by weight of the grain and oilseed movements
while contract shipments comprise approximately 25%.
3. Some movements have enough competition to limit rail rates and are exempt from regulation.
Exemption of particular movements or exempt commodities can be appealed before the STB,
and the STB may remove the exemption if competition no longer adequately constrains rates.
4. A merger premium is the amount paid for a firm that exceeds its net book value according
to historical accounting methods. Net book value is the historical cost less accumulated
depreciation.
5. As railroad tariff rates are partially deregulated and the financial condition of the railroad
industry in 1980 was poor, the Staggers Act requires the regulatory agency (STB) to evaluate
the revenue adequacy of the major U.S. railroads and the railroad industry. The STB calculates
the cost of capital (weighted average of the cost of debt and cost of capital) for the railroad
industry annually. If the railroad’s return on net investment exceeds the cost of capital, the
carrier is judged revenue adequate, which means that the railroad is able to attract sufficient
capital.
6. Merger premiums add to variable costs when the premiums are paid on assets included in the
calculation of variable costs.
References
Association of American Railroads (AAR). Analysis of Class I Railroads. (1998-2009).
Association of American Railroads (AAR). Freight Commodity Statistics. (1997-2008).
Association of American Railroads (AAR). Railroad Equipment Report. (1982-2008).
Association of American Railroads (AAR). Railroad Facts. (2009): 18-19.
73

Rail Rate and Revenue Changes
Association of American Railroads (AAR). The Rail Transportation of Grain. (2009): 62.
Babcock, Michael W., L. Orlo Sorenson, Ming H. Chow, and Keith Klindworth. “Impact of Staggers
Rail Act on Agriculture: A Kansas Case Study.” Proceedings of the Transportation Research Forum
26 (1), (1985): 364-372.
Bereskin, C. Gregory. Railroad Cost Curves Over Thirty Years: What Can They Tell Us? presentation
at the Transportation Research Forum, Fort Worth, Texas, March 18, 2008.
Chow, Ming H. “Interrail Competition in Rail Grain Rates on the Central Plains.” Proceedings of
the Transportation Research Forum 27(1), (1986): 164-171.
Christensen, Laurits R. Associates, Inc. A Study of Competition in the U.S. Freight Railroad Industry
and Analysis of Proposals That Might Enhance Competition. 2008, 5-14, 5-15, 8-41, 8-46.
CSX Transportation. Webpage several months after Hurricane Katrina (2006).
Fuller, Stephen, David Bessler, James MacDonald, and Michael Wohlgenant. “Effects of Deregulation
on Export-Grain Rail Rates in the Plains and Corn Belt.” Proceedings of the Transportation Research
Forum 28 (1), (1987): 160-167.
Gallamore, Robert E. “Regulation and Innovation: Lessons from the American Railroad Industry.”
José Gómez-Ibáñez, William B. Tye, and Clifford Winston eds. Essays in Transportation Economics
and Policy. Washington, D.C.: Brookings Institution (1999): 493-529.
Government Accountability Office (GAO). Railroad Regulation: Economic and Financial Impacts
of the Staggers Rail Act of 1980. 1990.
Government Accountability Office (GAO). Freight Railroads: Industry Health Has Improved, but
Concerns about Competition and Capacity Should Be Addressed. 2006, 3.
Government Accountability Office (GAO). Freight Railroads: Updated Information on Rates and
Other Industry Trends. 2007, 12.
Kohls, Richard L. Marketing of Agricultural Products. The Macmillan Company, New York, 1967:
147, 153, 305.
Koo, Won W., Denver D. Tolliver, and John D. Bitzan. “Railroad Pricing in Captive Markets: An
Empirical Study of North Dakota.” The Logistics and Transportation Review 29 (2), (1993): 123-
137.
Kwon, Y.W., Michael W. Babcock, and Orlo L. Sorenson. “Railroad Differential Pricing in
Unregulated Transportation Markets: A Kansas Case Study.” The Logistics and Transportation
Review 30 (3), (1994): 223-244.
MacDonald, James M. “Competition and Rail Rates for the Shipment of Corn, Soybeans and
Wheat.” Rand Journal of Economics 18 (1), (1987):151-163.
MacDonald, James M. Effects of Railroad Deregulation on Grain Transportation. U.S. Department
of Agriculture, ERS Technical Bulletin No. 1759, Washington, D.C., 1987.
MacDonald, James M. “Railroad Deregulation, Innovation, and Competition: Effects of the
Staggers Act on Grain Transportation.” Journal of Law and Economics 32, (1989): 63-95.
74

JTRF Volume 50 No. 1, Spring 2011
Montana Wheat & Barley Committee, et al. STB Ex Parte No. 658, The 25 th Anniversary of the
Staggers Rail Act of 1980: A Review and Look Ahead. October 19, 2005.
Montana Wheat & Barley Committee, et al. STB Ex Parte No. 665, Rail Transportation of Grain.
October 30, 2006.
National Grain and Feed Association (NGFA). STB Ex Parte No. 658, The 25 th Anniversary of the
Staggers Rail Act of 1980: A Review and Look Ahead. October 19, 2005.
Prater, Marvin and Keith Klindworth. Long-Term Trends in Railroad Service and Capacity for U.S.
Agriculture. USDA/Agricultural Marketing Service, 2000.
Security and Exchange Commission. Class I railroad quarterly filings. (2001-2007).
Surface Transportation Board (STB). Confidential Waybill Sample. (1985-2007).
Surface Transportation Board (STB). Decisions in STB Ex Parte No. 558, Railroad Cost of Capital.
(1998-2009).
Surface Transportation Board (STB). STB Ex Parte No. 661, Rail Fuel Surcharges. Decision
37341, January 25, 2007.
Surface Transportation Board (STB). Study of Railroad Rates: 1985-2007. 2009.
Thompson, S.R., R.J. Hauser, and B.A. Coughlin. “The Competitiveness of Rail Rates for Export-
Bound Grain.” The Logistics and Transportation Review 26 (1), (1990): 35-53.
Union Pacific Railroad. Letter from Jack Koraleski, www.uprr.com. 2008.
U.S. Department of Agriculture (USDA). Comments to the STB during Ex Parte No. 658, The 25 th
Anniversary of the Staggers Rail Act of 1980: A Review and Look Ahead. October 19, 2005.
U.S. Department of Agriculture (USDA). STB Ex Parte No. 665, Rail Transportation of Grain.
http://www.ams.usda.gov/AMSv1.0/getfile?dDocName=STELPRDC5049155&acct=atpub,
November 2, 2006.
U.S. Department of Agriculture (USDA). STB Ex Parte No. 646 (Sub-No.1), Simplified Standards
for Rail Rate Cases. http://www.ams.usda.gov/AMSv1.0/getfile?dDocName=STELPRDC5049156
&acct=atpubN, November 7, 2006.
U.S. Department of Agriculture (USDA), National Agricultural Statistical Service (NASS). Crop
Value Summary, 2009.
U.S. Department of Agriculture and the U. S. Department of Transportation (USDA/DOT). Study
of Rural Transportation Issues. http://www.ams.usda.gov/AMSv1.0/Rural Transportation Study,
April 2010.
Wilson, Wesley W. and William W. Wilson. “Deregulation, Rate Incentives, and Efficiency in the
Railroad Market.” B. Starr McMullen ed. Transportation After Deregulation. New York: Elsevier
(2001): 1-24.
75

Rail Rate and Revenue Changes
Marvin Prater is a transportation economist specializing in railroad transportation and rural
infrastructure issues and is employed by the Transportation Services Division, Agricultural
Marketing Service, U.S. Department of Agriculture. Marvin earned a B.S. degree in Horticulture,
a masters in business administration, and a Ph.D. in economics from Kansas State University.
Marvin has been a member of the Transportation Research Forum since 1997, presented papers at
the annual forum, and authored papers for the Journal of the Transportation Research Forum and
the U.S. Department of Agriculture.
Ken Casavant is a professor in the School of Economic Sciences and a nationally renowned
transportation economist after spending 35 years at Washington State University. He has received
numerous teaching, research, and service awards throughout his tenure at Washington State
University, including recipient of the “Lifetime Achievement Award” from the Upper Great Plains
Transportation Institute in 2006, the “Washington State University Sahlin Faculty Excellence
Award” in 2004, and being named “Distinguished Scholar” by the Western Agricultural Economics
Association in 2003. Ken also serves as the faculty athletic representative to the president for
Washington State University.
Eric Jessup is an associate professor in the School of Economic Sciences at Washington State
University. His research focus and area of specialty is transportation economics and freight systems
modeling. He is also co-principal investigator for the multi-year, statewide freight research and
implementation project for the state of Washington, the Strategic Freight Transportation Analysis
(SFTA) study.
Bruce Blanton currently serves as director of the Transportation Services Division with the U.S.
Department of Agriculture’s Agricultural Marketing Service. Over the years he has served in various
positions in both government and industry. Past positions held at USDA include deputy assistant
secretary for congressional relations, special assistant to the Farm Service Agency administrator,
special assistant to the Under Secretary for International Affairs and Commodity Programs,
and confidential assistant to the Deputy Secretary. Blanton also worked as a senior analyst for
agriculture for the United States Senate Committee on the Budget. In the private sector Blanton
served as executive director of the National Renderers Association, vice president of rendering
sales for Moyer Packing Company, and as an economic research associate for the American Farm
Bureau Federation. A native of the Southwest, Blanton holds a bachelor’s degree in agricultural
business management from New Mexico State University and a master’s degree in agricultural
economics from the University of Missouri-Columbia.
Pierre Bahizi is an economist with the Transportation Services Division at the U.S. Department
of Agriculture. Prior to working for USDA, Pierre was an economist in the Division of Consumer
Expenditure Surveys at the Bureau of Labor Statistics. He has also worked in the Travel and
Tours Department at the Smithsonian Institution. Pierre graduated from the University of Texas at
Arlington with a B.A. in economics.
Daniel Nibarger is an economist with the U.S. Department of Agriculture, Foreign Agricultural
Service. Daniel is from Gardner, Kansas, and attended Kansas State University in Manhattan. At
Kansas State, Daniel earned a B.S. degree in economics with an international economics overlay,
and an M.S. in economics with an emphasis in international trade. During his graduate coursework,
Daniel spent a summer in Wuhan, Hubei Province, China, teaching English language courses at
Huazhong University. After completing his M.S., Daniel earned a legislative fellowship with the
State of Kansas Legislature. He has also served as a labor economist with the Kansas Department
of Labor researching ways to project employment and unemployment. Daniel has also served
76

JTRF Volume 50 No. 1, Spring 2011
as an economist for the Transportation Services Division of the U.S. Department of Agriculture.
Daniel has presented his research at the 2006 Western Economics Association-International
annual meeting, the 2010 Transportation Research Forum, and the 2009 Caribbean Conference for
Agricultural Trade.
Johnny Hill is an agricultural economist with the Transportation Services Division, Agricultural
Marketing Service, U.S. Department of Agriculture, specializing in grain exports and rail
transportation. Johnny earned a B.S. degree in rural development and an M.S. degree in agricultural
science from Tennessee State University. Johnny has worked at USDA since 1988 and prior to this
position, worked as a graduate research assistant and instructor in the School of Agriculture at
Tennessee State University.
Isaac Weingram is a research assistant at the Federal Reserve Bank of Boston. Previously, Isaac
worked as an agricultural economist assistant in the Transportation Services Division of the U.S.
Department of Agriculture. He graduated from Washington University in Saint Louis with an A.B.
in mathematics and economics.
77

Measuring Bulk Product Transportation
Fuel Efficiency
by C. Phillip Baumel
This paper reviews the literature that compares the fuel efficiencies of bulk commodity transportation
modes. Most studies used net-ton-miles per gallon to compare modal fuel efficiencies. Net-tonmiles
per gallon have traditionally been estimated from aggregate industry data of total net-tonmiles
and total fuel consumed. More recent studies have targeted specific origins, destinations,
products hauled, types and sizes of equipment, backhauls, and miles traveled to estimate total fuel
consumption. This paper shows that fuel efficiency estimates based only on net-ton-miles per gallon
can be erroneous. The paper identifies basic variables and measurement methods that can improve
the accuracy of modal fuel efficiency comparisons.
INTRODUCTION
In 1971, the price of imported petroleum was $2.00 per barrel; it peaked at $147 per barrel in July
2008 and fell to $50 in December 2008 (Baumel 2009). At the time of the writing of this article,
the price of petroleum fluctuated around $90 per barrel. No one knows the precise future prices of
petroleum, but few people expect the long run price to decline.
Major air pollutants from motorized vehicles include hydrocarbons, carbon monoxide, carbon
dioxide, nitrogen oxides, and particulate matter. The Environmental Protection Agency estimates
emissions of these hazardous air pollutants in grams per vehicle mile traveled (Texas Transportation
Institute 2009). Thus, fuel consumption and miles traveled are major factors in estimating air
pollution from freight transportation.
Users and operators of the three major modes of bulk product transportation call for major
infrastructure upgrading. Bulk products include coal, grains, chemicals, aggregates, and liquid fuels.
Increased traffic congestion is evidence of needed highway upgrading. A 2007 study estimated that
the U.S. railroad industry would need to invest $147 billion over the next 35 years in infrastructure
expansion to meet the U.S. Department of Transportation projected 88% increase in rail freight
tonnage by 2035 (Cambridge Systematics Inc. 2007). Barge users and the barge industry have been
urging the U.S. public to invest in upgrading America’s inland waterway locks and dams to “help
keep America green” (Waterways Council Inc. 2010).
Environmental concerns, increasing fuel costs, and needed infrastructure upgrades suggest
the need to improve the fuel efficiency of the bulk product transportation system. This objective
requires that modal fuel efficiencies should be accurately estimated.
Previous research on measuring bulk product transportation fuel efficiency can be grouped into
five categories:
1. Using aggregate data to estimate net-ton miles per gallon (NTMG) by mode
2. Estimating NTMG by river segment or by direction of rail movement
3. Estimating NTMG by operating characteristic
4. Estimating NTMG by product hauled
5. Estimating total fuel consumption using NTMG and miles traveled by mode from
specific origins to specific final destinations
A review of the literature in each of these five categories follows.
79

Transportation Fuel Efficiency
Using Aggregate Data to Estimate Modal NTMG
The majority of past fuel efficiency studies and reports used total fuel consumed and total net
ton miles of freight over thousands of commodities and routes to estimate NTMG. A 1975 U.S.
Department of Transportation advisory report on the replacement of Alton Locks and Dam 26 on
the Upper Mississippi River summarized the results of 19 energy studies (U.S. Department of
Transportation 1975). The 19 studies used aggregate data to estimate rail and/or barge NTMG. The
estimated rail fuel efficiencies ranged from 138.5 to 693.5 NTMG. Barge fuel efficiency ranged
from 243.3 to 639.2 NTMG.
Eastman (1980) estimated the following NTMG: barge, 514; rail, 202; and truck, 59. The
Eastman numbers were frequently used in other reports, i.e., U.S. Department of Transportation
(1994) and were still being reported 30 years later (Bernert Barge Lines Inc. 2010).
The Texas Transportation Institute (TTI) estimated the following NTMG: truck, 155; rail, 413;
and barge, 576 (Texas Transportation Institute 2009). Some writers and organizations use the TTI
estimates to promote barges and short sea shipping as the most fuel efficient modes of bulk product
transportation (National Waterways Foundation 2008, Quigley 2009). The Association of American
Railroads (2010) reported a 2009 U.S. Class I railroad NTMG of 480.
Most of the above reports used their estimated NTMG as the only basis for comparing the fuel
efficiency of the three major modes of freight transportation. TTI assumed that since the miles
traveled by each mode were similar, NTMG could be used to define the fuel efficiency of each of
the three modes.
Estimating NTMG by River Segment and Rail Direction
Using data from four barge companies, Baumel, Hauser, and Beaulieu (1982) estimated barge
NTMG for moving grain from several Mississippi River system origins to New Orleans, Louisiana
(NOLA). At a 25% backhaul, NTMG on the Lower Mississippi River from Cairo, Illinois, to NOLA
was 525, while barges on the Upper Mississippi and Illinois Rivers averaged about 450 NTMG.
Baumel et al. (1985) used daily fuel measurement data from three barge companies to estimate
the NTMG of barges on the Upper Mississippi and Lower Mississippi Rivers. Calibrated steel tape
measurements were used to estimate daily fuel consumption. Fuel meters were not possible because
when one or more propellers were in reverse, the vibrations caused fuel meters to malfunction. At a
35% backhaul, the NTMG for barges on the Upper Mississippi River was 526, while barges on the
Lower Mississippi River obtained 548 NTMG.
Burton (1997) used a Tennessee Valley Authority (TVA) Barge Costing Model to estimate barge
NTMG on six rivers. His estimated NTMG was 694 for the Upper/Middle Mississippi River and
917 for the Lower Mississippi River.
Baumel et al. (1985) used fuel meters to estimate NTMG for three 54-car unit grain trains and
three 75-car unit trains from Iowa to West Coast ports. Four unit grain trains were shipped from
Iowa to NOLA. The four grain trains to NOLA averaged 640 NTMG. This was 46% more than the
437 average NTMG achieved by the six West Coast trains. All West Coast trains had to traverse one
or more mountain ranges to and from the West Coast. This explains most of the difference in the
average NTMG of the two sets of trips.
Gervais and Baumel (1999) used TVA calculated barge NTMG from three segments of the
Mississippi River. The TVA estimated total fuel consumption from actual barge fuel tax collections.
The 1995 estimates of NTMG were as follows: Lower Mississippi River, 646: Mouth of the Missouri
to the Mouth of the Ohio River, 595; and Minneapolis to the Mouth of the Missouri River, 308. All
of the locks and dams are located between Minneapolis and the Mouth of the Missouri River. The
weighted average NTMG of the three Mississippi segments was 420 NTMG.
80

Estimating NTMG by Operating Characteristic
JTRF Volume 50 No. 1, Spring 2011
Burton (1997) used TVA data to estimate NTMG for each of 12 railroad companies. The estimated
NTMG ranged from 118 to 374. The highest NTMGs were for the four largest railroad companies.
Seven of the other eight companies have either merged with the four larger companies, or merged
together to form new companies.
Baumel et al. (1985) used data from three barge companies to estimate the impact of the percent
backhaul on barge NTMG on the Upper and Lower Mississippi rivers. At zero backhaul, NTMG
were estimated to be 420 on the Upper Mississippi and 483 on the Lower Mississippi River. At 50%
backhaul, the NTMG gap between the two rivers narrowed to 578 on the Upper Mississippi and 592
on the Lower Mississippi River. At 100% backhaul, there was little difference in the NTMG on the
two rivers; those estimates were 756 for the Upper Mississippi and 754 for the Lower Mississippi
River.
Gervais and Baumel (1999) used data from computer simulations by two railroad companies
to estimate the impact of the type of locomotive and size of rail car on rail NTMG. Three scenarios
were analyzed. Two were from Central Iowa to NOLA and to Los Angeles. The third was from
western Iowa to Tacoma, Washington. Each trip was simulated with three different locomotives and
two sizes of covered hopper cars. The two types of rail cars were 100-ton and 110-ton capacities.
The 100-ton cars are now 30-40 years old and are being replaced by new 110-ton cars.
The three locomotives were the SD40 that was introduced in the late 1970s and the newer
SD60 and C40-8. Four SD40s are required to pull a 100-car grain train. Only three SD60 or C40-8
locomotives are needed to pull 100-car grain trains. All simulations were at 35 mph.
The SD60s had the highest NTMG and the older SD40s had the lowest. The 110-cars had higher
NTMG than the older 100-ton cars. Finally, the NTMG for the trips to NOLA were 30% higher than
those to Los Angeles and 22% higher than those to Tacoma. The reason for the higher NTMG to
NOLA is that the terrain to NOLA is relatively flat, while each trip to Los Angeles and Tacoma was
over one or more mountain ranges on the loaded and empty legs of each trip.
Gervais and Baumel (1999) also reported computer simulations of state-of-the art semi trucks
obtaining 131 NTMG at a speed of 60 MPH. NTMG declined 16% to 110 NTMG at a speed of 70
MPH.
Baumel et al. (1985) used fuel consumption data for 254 grain hauling ocean vessels taken
from The Journal of Commerce and Commercial (February 1, to July 31, 1983) to estimate NTMG
for grain carrying ocean vessels. Small vessels (

Transportation Fuel Efficiency
a drayage truck, usually older and less fuel efficient than a long-haul truck, operating in congested
conditions.
ICF International (2009) updated the Abacus Technology Corporation (1991) study. ICF
International (2009) used computer simulations of rail and truck movements in 23 competitive railtruck
corridors to compare rail and truck fuel efficiencies. Individual rail movements included in the
analysis were double stack, covered hopper, tank car, trailers on flat cars (TOFC), and automotive
rack trains. Individual truck movements included dry vans, dump, tanker, container, flatbed with
sides and auto haulers. The dominant measure of fuel efficiency was NTMG. The fuel efficiency of
railroads exceeded that of trucks in each of the 23 movements. However, the difference between rail
and truck fuel efficiencies varied widely among the products hauled.
Estimating Total Fuel Consumption Using NTMG and Miles Traveled by Mode from Specific
Origins to Final Destinations
Baumel et al. (1985) estimated total fuel consumption in gallons per short ton (GPT) in shipping
grain from six origins in Iowa to Yokohama, Japan. GPT was estimated by dividing total miles
traveled by each mode by the appropriate NTMG for that mode. The NTMG for railroads were
estimated from metered fuel consumption data provided by five railroad companies. The NTMG for
barges were calculated from daily fuel measurement data provided by three Mississippi River barge
companies. The truck NTMG were calculated from three fuel metered trips hauling grain to barge
loading elevators on the Mississippi River. The Journal of Commerce and Commercial ship fixture
data (February 1-July 31, 1983) on bulk carrier time charters were used to estimate ocean vessel fuel
consumption. Fuel consumed by each mode in each intermodal shipment was added to obtain the
total GPT for each route. The results for shipments to Japan are ranked below in descending order
of total fuel efficiency when similar sized ocean vessels and typical routes are used:
1. Unit trains direct to West Coast ports
2. Unit trains direct to NOLA and the unit train-barge combination with 100% barge
backhaul
3. Unit-train-barge combination with less that 100% backhaul
4. Truck-barge combinations
EVALUATION OF THE ABOVE FUEL EFFICIENCY STUDIES
The major characteristic of the above fuel efficiency studies is the conflicting results and conclusions
among the many studies. There are at least three reasons for these conflicting results. One is the
50-year span over which these studies were conducted. The earlier studies, based on data from the
early 1960s, reported smaller NTMG than those based on late 1990s and 2000 data. Technological
improvements and larger vehicle and vessel sizes have greatly increased fuel efficiency and NTMG
in all modes of bulk product transportation. Some of these technological improvements are reflected
in later study results.
A second reason for the conflicting results is that most of the earlier studies were based on
average data over thousands of commodities and routes for the truck and rail industries. Many
of the later studies were targeted to specific commodities, specific routes, and even alternative
types and sizes of equipment. The Abacus Technology Corporation (1991) and ICF International
(2009) studies estimated NTMG on trains and trucks, each hauling only automobiles, manufactured
products, and liquid or bulk products. Most of the specific product studies focused on the types
and sizes of equipment and miles traveled on routes typically used to transport bulk commodities
including grain. These targeted analyses of actual movements generally allowed the studies to
focus on larger size shipments such as unit trains and ocean vessels and on specific routes, miles,
and direction of shipments that are typical of bulk commodities. These targeted analyses should
provide more accurate estimates of fuel consumption than average NTMG estimates averaged
82

JTRF Volume 50 No. 1, Spring 2011
over different weights, speeds, transportation equipment, terrain, and distances hauled. Speed is
important because aerodynamic resistance of a train increases with the square of the speed (Grervais
and Baumel 1999). Ship company executives indicate that, on average, fuel consumption decreases
about 20% with each 10% reduction in speed (Baumel et al. 1985).
A third, and perhaps the most important reason for different results, is that some of the
later studies have focused on estimating total fuel consumption from specific origins to specific
destinations. The TTI (2009) estimated the following NTMG as measures of fuel efficiency:
Truck 155
Rail 413
Barge 576
These NTMG suggest that barges are 39% more fuel efficient than railroads and almost four
times more fuel efficient than trucks. However, NTMG alone tells only part of the fuel efficiency
story
The following example focuses on grain shipped by railroad and barge from Iowa to export
grain elevators in the NOLA area. Almost all grain shipped by barge must be hauled from inland
elevators, or from farms, to barge-loading elevators. Grain shipments from elevators direct to a
final destination are typically by rail or truck. Usually, no other mode is involved in these transfers.
Similar movements also are typical for coal and some chemicals and liquid fuels. To make correct
comparisons of total fuel consumption of barge versus rail direct to an export port, fuel consumed
by truck or rail to a barge loading facility must be added to the barge fuel consumption. Moreover,
NTMG measures only the miles that one ton of freight is moved by one gallon of fuel. It fails to
measure the total fuel consumed in moving the freight from an origin to a destination.
Bruton (1997) makes the following argument in his paper (p. 7) prepared for the U.S. Army
Corps of Engineers:
“A majority of barge shipments involve a truck movement at one or both ends of the
line-haul move and in some cases, these truck hauls may be of considerable length.
Anecdotal evidence suggests that some grain shipments may be trucked as much as
300 miles for trans-loading to barge. Alternately, most rail movements can be made
rail direct and where additional truck movements are necessary, they are seldom
more than a few miles. Consequently, the appropriate fuel usage comparison is not
between line-haul barge and line-haul rail. Rather, this comparison must be made
over the entire movement. This often means comparing the fuel efficiency of an all
rail movement with that of a truck-barge-truck combination. Quite clearly, such cases
tend to diminish the aggregate efficiency advantage otherwise attributable to barge.”
Table 1 illustrates the impact of distance travelled on total modal fuel consumption. The table
shows the total fuel consumed to move one ton of grain from one Iowa origin to a NOLA destination
by rail and by truck-barge. Rail direct versus rail-barge was not evaluated because almost all grain
shipped by barge from Iowa is now delivered by trucks to barge loading elevators (Van Der Kamp
2010).
The NTMG listed in Table 1 are those estimated by TTI. The Association of American Railroads
(2010) reports that the average NTMG for Class I railroads for 2009 was 480; this is 16% more
efficient than the 413 reported in TTI. Nevertheless, Table 1 uses TTI’s lower rail NTMG of 413.
St. Charles Parish, Louisiana (SCPL), was selected as the destination for both the rail and barge
delivered grain. Three of the 10 grain export elevators in the NOLA area are located in SCPL.
Waterloo, Iowa, located near the center of Black Hawk County, was chosen as the origin of
the grain in Table 1. The railroad miles in Table 1 from Waterloo to SCPL are the average miles for
typical routes of rail shipments of grain from Black Hawk County, Iowa, to SCPL.
The data in Table 1 show that total fuel consumption per ton of grain shipped by truck-barge
combination from Waterloo to SCPL is 6% greater than for the direct rail shipment. If the Association
of American Railroads (2010) rail NTMG of 480 was substituted for the TTI (2009) rail NTMG
in Table 1 the total truck-barge fuel consumption would be 23% greater than for the direct rail
83

Transportation Fuel Efficiency
Table 1: Estimated Total Fuel Consumption to Ship Grain from Waterloo, Iowa, to
St. Charles Parish in Gallons per Ton of Grain
shipment. These results are the opposite of the conclusion derived from NTMG alone. The major
reasons for the different conclusions are:
1. The truck portion of the barge movement adds almost 0.6 gallons to total truck-barge fuel
consumption per ton of grain,
2. The total barge distance from Dubuque to SCPL is 20% longer than the average typical rail
distance from Waterloo to SCPL, and
3. The combined truck-barge distance from Waterloo to SCPL is 27% longer than the direct
rail shipment from the same origin to the same destination.
The longer barge distance is caused by the meandering of the Mississippi River. Figure 1
illustrates the impact of this meandering on river distances.
All of the NOLA grain export elevators are located within the Baton Rouge-Myrtle Grove
section of the river. The river distance between these two points is 167 miles (Blue Water Shipping
Company). The MapQuest driving distance between these two points is 107 miles. Thus, the
meandering of the river increases the river distance between these two points 56% above the
driving distance. The entire Mississippi River meanders in a similar fashion up to its source near
Minneapolis.
The total miles to an importing country is even more important for calculating total fuel
consumption for grain destined for export. For example, corn exports to Japan typically move in
two directions. One is by barge, rail, or truck to Gulf of Mexico ports (including NOLA) for ocean
vessel movements through the Panama Canal. The second is by rail to the West Coast and ocean
vessel to Japan. The rail movement from Iowa to the West Coast is longer than to NOLA and it is
over the Rocky Mountains. This suggests that fuel consumption would decrease if Iowa corn was
shipped to NOLA ports for export. However, the ocean distance from NOLA to Japan is more than
double the distance from Seattle–almost 6,000 miles longer (Baumel et al. 1985). The net result is
that corn shipped by rail from western Iowa to Tacoma and ocean vessel to Japan uses less total fuel
than any modal combination through NOLA (Gervais and Baumel 1999).
Barge NTMG are typically calculated by dividing total net-ton miles of freight hauled by
all barges on all navigable rivers, by the total number of gallons of fuel consumed. The Lower
Mississippi River–that portion of the river south of the confluence of the Ohio and Mississippi
Rivers–is the mostfuel efficient river on the Mississippi River system (Gervais and Baumel 1999).
NTMG increase sharply as the number of tons increases in barge tows. Barge tows on the Lower
Mississippi River can have 50 or more barges. That compares with a maximum of 15 barges on the
Upper Mississippi River and as few as two on the upper Missouri River. Second, the river current
on the Lower Mississippi River is swift because it is not impeded by dams. The swift current pushes
84
Mode of transport Miles NTMG d
Gallons
per ton
Total
gallons
per ton
Rail direct to St. Charles Parish 1,180 a 413 2.86 2.86
Truck to Dubuque 91b 155 0.59
Barge (Dubuque to St. Charles Parish) 1,413c 576 2.45
Total truck-barge 1,504 3.04
Sources:
a. Association of American Railroads (2010)
b. MapQuest Driving Directions North America
c. Iowa Department of Transportation and Blue Water Shipping Company
d. Texas Transportation Institute (2009)

JTRF Volume 50 No. 1, Spring 2011
Figure 1: Blue Water Shipping Company Mississippi River Deep Water Corridor Map
the loaded barges downstream, further reducing fuel consumption. Third, since there are no locks on
the Lower Mississippi River, barge tows move nonstop on this river segment. Barge tows on most
other rivers must stop to transit the locks at each dam. This suggests that barge NTMG should be
calculated for individual rivers to generate more accurate NTMG estimates.
Similar issues exist for railroads. The large tonnages and direct shipments of unit-trains make
them more fuel efficient than trains consisting of a mix of different commodities. Trains that cross
mountains consume more fuel than trains that essentially follow, but don’t meander, along the
Mississippi River (Gervais and Baumel 1999). Finally, new technology locomotives are highly fuel
efficient (ICF International 2009). This suggests that NTMG should be estimated for different types
of rail service.
Burton (1990), using the TVA’s Barge Costing Model, estimated that barges operating on the
Lower Mississippi River obtained 917 NTMG. Yet, TVA estimates, based on 1995-1997 fuel taxes
collected from barge companies, ranged from 604 to 646 NTMG (Gervais and Baumel 1999). If the
TVA’s Barge Costing Model estimates are correct, barge companies paid taxes on fuel that they did
not consume. This is highly unlikely. Fuel meters and physical measurements may be more accurate
than some computer models not designed specifically to estimate NTMG. However, if fuel meters or
physical measurements are used for barges, the fuel consumed by switch boats in switching barges
into and out of tows must be added to the total fuel consumption. Moreover, fuel used to generate
electricity on towboats must also be added to total fuel consumption. Only fuel used for propulsion
is counted in waterway fuel tax receipts (IRS 2009).
85

Transportation Fuel Efficiency
Finally, backhaul estimates are needed to improve the accuracy of fuel consumption estimates.
The backhaul rate for most unit-trains is typically zero. However, barge tows and ocean vessels
typically have some level of backhaul, which increases the NTMG. Trucks also have backhauls but
typically at a lower rate than barges and ocean vessels.
How to Measure NTMG
The most common, easiest, and least costly method to estimate NTMG is to divide aggregate data on
ton miles of product hauled by a mode of transport, by the total gallons of fuel consumed to move
those ton miles. It is also probably the least accurate method of estimation because it averages
NTMG over thousands of products and movements and over many different types of equipment.
Previous studies indicate that NTMG estimates vary substantially among products hauled (ICF
International 2009), type of equipment used (Gervais and Baumel 1999), terrain and river segment
(Baumel et al 1985), speed (ICF International 2009), and distance hauled.
A second method to measure fuel consumption is to install fuel meters on trucks, rail locomotives
and barge tow boats. Fuel meters work well on trucks and railroad locomotives but not on barge
tow boats. An alternative to fuel meters on tow boats is to use Internal Revenue Service excise
fuel tax collections on barge fuel consumption to calculate barge fuel consumption (IRS 2009).
The excise tax is collected on each gallon of fuel used to propel inland waterway and intra coastal
waterway commercial vessels for the transport of commercial property. Therefore, fuel used to
generate electricity and heat must be added to the estimated fuel used to propel the vessel. These
excise tax collection data are available by river segment.
A third method is to develop computer simulation models that incorporate all of the major
characteristics of the movement being simulated. On railroads, these characteristics include type and
size of train, number and type of locomotive, product mix, distance, grade severity, curvatures, and
speed. Truck characteristics include type and size of truck, road grades, road congestion, tire types,
and speed. Barge characteristics include number and type of barges, size and type of towboat, river
segments, locks traversed by size and congestion, and water levels. Properly constructed simulation
models appear to have the potential to estimate total fuel consumption for a larger number of
movements and types of equipment more accurately and cheaper than other methods that have been
used in past studies.
CONCLUSIONS
1. NTMG, when used alone, is frequently an incomplete and misleading measure for modal
fuel efficiency comparisons. It is an accurate measure of comparative fuel efficiency only if
the comparative mode shipments are from the same origin to the same destination, the same
distance from the origin to the destination, and there are no intermodal movements in each
shipment.
2. A more accurate measure of comparative modal fuel efficiency is the total fuel consumed by
each mode over the entire movement in the transfer of the product from the origin to the final
destination. This measure can be calculated by dividing the total miles traveled by each mode
by the appropriate NTMG for each mode. Fuel consumption by each mode in the transfer
should be added to obtain the total fuel consumption for the entire multimodal shipment.
3. Appropriate NTMG should be estimated for different products, different rivers, and for different
sizes and types of trains traveling over different terrains and with different types of locomotives.
4. There is no one “greenest” mode of freight transport. The greenest mode or group of modes
depends on several variables. These include the origin, final destination, product, type of
shipment, level of backhaul, accuracy of the NTMG, and the miles traveled by all modes
involved in the transfer.
86

JTRF Volume 50 No. 1, Spring 2011
5. The comparisons should be made over actual shipment routes rather than over routes that may
possibly be used sometime in the future.
6. Government officials should carefully examine proposals for public infrastructure investments
that use NTMG alone to justify the proposal’s fuel efficiency. These proposals may have the
unintended consequence of increasing total fuel consumption. Accurate proposals should
provide estimates of total fuel consumption from the beginning origin to the final destination.
Total fuel consumption should be based on appropriate NTMG measures, routes and total miles
traveled by each mode involved in the transfer. These estimates should be compared with the
next best alternative movement. Table 1 is an example of that type of comparison.
7. The large differences in methodologies and results of previous fuel efficiency studies provide
opportunities for university researchers, in cooperation with transportation organizations
and agencies, to develop simulation models and data to generate unbiased and reliable fuel
efficiency estimates for alternative modes, products, routes, and distances. These models
would be very useful to public and private decision makers in allocating investment funds
among transportation investment alternatives.
References
Abacus Technology Corporation. Rail vs Truck Fuel Efficiency: The Relative Fuel Efficiency of
Truck Competition Rail Freight and Truck Operations Compared in a Range of Corridors. DOT/
FRA-92/2, Federal Railroad Administration, U.S. Department of Transportation, Washington D.C.,
1991.
Association of American Railroads. Personal Communication. Washington D.C. July, 2010.
Baumel, C. Phillip. “Is the Era of Cheap Barge Rates Over?” Feedstuffs 81 (40), September 28,
2009.
Baumel, C. Phillip, Charles R. Hurburgh, and Tenpao Lee. “Estimates of Total Fuel Consumption
in Transporting Grain from Iowa to Major Grain-Importing Countries by Alternative Modes and
Routes.” Special Report 90, Agriculture and Home Economics Experiment Station, Iowa State
University, Ames, Iowa, August 1985.
Baumel, C. Phillip, Robert J. Hauser and Jeffrey Beaulieu. “Impact of Inland Waterway User
Charges on Corn, Wheat and Soybean Flows: Study of Inland Waterway User Taxes and Charges.”
National Technical Information Service, U.S. Department of Commerce, PB82-196023, Springfield,
Virginia, March, 1982.
Bernert Barge Lines, Inc. “Save Our Dams.” www.gorge.net/digitaldesigner/bbl/dams.htm.
Accessed July 27, 2010.
Blue Water Shipping Company. “Blue Water Mississippi River Deep Water Corridor Map.” Metairie,
Louisiana.
Burton, Mark L. “Missouri River Navigation Benefits: Incorporating the Effects of Air Quality
Improvements.” Prepared for the U.S. Army Corps of Engineers, Missouri Division, Attachment to
Affidavit of Mark L Burton, Exhibit 28, MO-ARK Association, Motion to Alter Judgment, Case No.
96 4086-CV-C-66BA, Knox County, State of Tennessee, September 5, 1997.
Cambridge Systematics, Inc. “National Rail Freight Infrastructure Capacity and Investment Study.”
Cambridge, Massachusetts, September, 2007.
87

Transportation Fuel Efficiency
Eastman, S. E. “Fuel Efficiency in Freight Transportation.” American Waterway Operators, Inc.,
Arlington, Virginia, 1980.
Gervais, Jean-Philippe and C. Phillip Baumel. “Fuel Consumption for Shipping Grain Varies by
Origin and Destination.” Feedstuffs 71 (36), August 30, 1999.
ICF International. Comparative Evaluation of Rail and Truck Fuel Efficiency on Competitive
Corridors. Federal Railroad Administration, U.S. Department of Transportation, Washington D.C.,
November, 2009.
Iowa Department of Transportation. River Barge Terminal Directory, Revised 2005. www.iowadot.
gov.
IRS. Excise Taxes, Publication 510 (04/2009). www.irs.gov/publications/p510.html.
MapQuest Driving Directions North America, www.mapquest.com/directions.
National Waterways Foundation. “New National Study Compares Freight Transportation by Barge,
Truck and Train.” News Release, Arlington, Virginia, July 8, 2008.
Quigley, Leo. “Grain Transportation’s Green Alternative.” World Grain, October, 2009.
Texas Transportation Institute. “A Modal Comparison of Freight Transportation Effects on the
General Public.” Center for Ports and Waterways, Texas A&M University, Houston, Texas,
December, 2007, Amended July, 2009.
The Journal of Commerce and Commercial. “Ship Fixture Breakdown.” 110 Wall Street, New York,
N.Y. 10005 (February 1-July 31 1983 and January 1-July 31, 1998).
U.S. Department of Transportation. The Replacement of Alton Locks and Dam 26: An Advisory
Report of the U.S. Department of Transportation to the Senate Commerce Committee. Washington,
D.C. September, 1975.
U.S. Department of Transportation. Environmental Advantages of Inland Barge Transportation.
Maritime Administration, Washington D.C., August, 1994.
Van Der Kamp, Jerry. Personal Communication. President, AGRI Industries, Ankeny, Iowa,
December, 2010.
Waterways Council, Inc. “Keep America Moving.” Video presentation, Arlington, Virginia,
February 25, 2010.
C. Phillip Baumel is a Charles F. Curtiss Distinguished Professor Emeritus of Economics at Iowa
State University. He received the Transportation Research Forum’s Distinguished Transportation
Researcher Award in 1993, the Transportation Research Forum’s Outstanding Research Paper
Awards in 1982, 1989, and 1999, and the Transportation Research Forum’s Best Agricultural and
Rural Transportation Paper Awards in 1989, 1991, and 1995. He is a Fellow in the American
Agricultural Economics Association. Baumel pioneered the use of economic network models for
analyzing issues of agricultural product transportation by rail, truck, and barge. He retired in 2003
and currently lives in Iowa (summer) and Arizona (winter).
88

Demand Analysis for Coal on the United
States Inland Waterway System: Fully-Modified
Cointegration (FM-OLS) Approach
by Junwook Chi and Jungho Baek
The Phillip-Hansen fully-modified cointegration (FM-OLS) approach is applied to examine the
dynamic relationship between demand for coal barge transportation, and explanatory variables such
as barge and rail rates, domestic coal consumption and production, and coal exports. The results
provide strong evidence that there exists a long-run equilibrium relationship between demand for coal
barge transportation and the selected variables. It is also found that, in the long-run, the domestic
coal consumption and coal exports are more important than other variables in determining the
demand for coal barge transportation. In the short run, on the other hand, domestic coal production
is found to be the only significant determinant of coal demand. This dynamic analysis will shed new
light on the dynamic interrelationships between the demand for coal barge transportation and its
major determinants, and contribute to the empirical literature on transportation economics.
INTRODUCTION
The United States inland waterways consist of over 12,000 miles of navigable waters, including the
Mississippi, the Ohio River Basin, the Gulf Intercoastal Waterway, and the Pacific Coast systems
(Clark et al. 2005). The inland waterway system plays a crucial role in supporting the nation’s
movements of bulk commodities and raw materials, and barge carriers provide economically sound
and efficient services. In 2007, for example, the inland waterway system transported 403 million
tons and 157 billion ton-miles of commodities (Bureau of Transportation Statistics 2007). Based on
low transportation costs, such large amounts of freight have long served to increase business and
economic activities, thereby contributing to regional economic development.
In the literature on transportation economics, a number of studies have been conducted to
examine demand for freight transportation on the inland waterway system and to evaluate the
determinants of barge traffic systems. One branch of literature relates to analyses of barge demand
and grain movements forecasts (Tang 2001, Babcock and Lu 2002, Thoma and Wilson 2004a, Yu
and Fuller 2005, DeVuyst et al. 2009). For example, Tang (2001) uses an autoregressive integrated
moving average (ARIMA) model to forecast soybean and wheat tonnage on the McClellan-Kerr
Arkansas River Navigation System; the results show small differences between the actual and
forecasted soybean and wheat tonnage (less than 10%). Thoma and Wilson (2004a) adopt a vector
error-correction (VEC) model to forecast annual tonnages on the Mississippi River; they find that
the annual growth rates for the upper, mid, and lower Mississippi River segments range from 1.45%
to 3.33% (1.68% as a whole). DeVuyst et al. (2009) employ a spatial optimization model to forecast
grain flows on the Mississippi River; they provide evidence that longer-term projections have much
larger errors, due mainly to uncertainty.
Another group of studies focuses on transportation rates on the inland waterway system
(Harnish and Dunn 1996, Miljkovic et al. 2000, Thoma and Wilson 2004b, Yu et al. 2007, Babcock
and Fuller 2007). For example, Harnish and Dunn (1996) use a reduced-form equation to examine
the factors determining grain barge rates on the Mississippi River; they conclude that in the short
run, grain exports, coal barge rates, input costs, and distance are the key determinants of barge
rates on the Mississippi River. Thoma and Wilson (2004b) employ a vector autoregressive (VAR)
model to characterize short-run grain movements on the Mississippi and Illinois Rivers; they show
89

Demand Analysis for Coal
that barge rates are significantly determined by lockages and rail rates. Babcock and Fuller (2007)
estimate demand for corn and soybean shipment on the Ohio River using OLS; they show that corn
and soybean shipments in the region are mainly determined by lagged shipments, corn and soybean
stocks, and exports.
Until recently, however, empirical studies have mostly concentrated on assessment of the
demand for grain barge transportation and barge rates to identify factors affecting grain movements
by barge, as well as barge demand and grain movement forecasts. Accordingly, little attention has
been paid to the factors affecting coal shipments and the substitution effect between water and
rail carriers for coal shipments. 1 Given that coal is one of the primary commodities on the inland
waterway system, it is important to fully understand the determinants of demand for coal barge
transportation. In 2009, for example, coal accounts for approximately 24% of total commodities
shipped on the inland waterway system in the United States (U.S. Army Corps of Engineers 2010).
Particularly, coal is the highest traffic volume moving within and through the Ohio River Basin
because of the substantial amount of reserves in the region (Clark et al. 2005); in 2008, barge
delivered 123 of the 744 coal movements, representing 16.5% of the total movements (U.S. Energy
Information Administration 2010). Another important point frequently overlooked in the literature
is that although some analysts use time-series data/methods (Harnish and Dunn 1996, Miljkovic et
al. 2000, Yu and Fuller 2005), they tend to take little cognizance of the unit root problems associated
with level variables. In other words, those studies use the level of each variable in their regression
analysis without taking into account the non-stationarity in the data. When data are not stationary,
standard critical values used in determining the significance of estimated coefficients are not valid
(Wooldridge 2006). 2 Finally, the earlier studies have concentrated on the short-run movements on
the inland waterway system (Tang 2001, Babcock and Lu 2002, Thoma and Wilson 2004b). Since
the short-run effects of barge transportation and barge rates could be different from the long-run
effects, it is important to include the long-run dynamics in a model.
This paper has attempted to expand the scope of previous work by assessing determinants of
demand for coal barge transportation in a dynamic framework of cointegration. The empirical focus
is on identifying the dynamic effects of transport rates–such as barge and rail rates, coal production,
coal consumption, and coal exports–on the quantity of coal shipped by barge. For this purpose, a
fully-modified cointegration (FM-OLS) technique developed by Phillips and Hansen (1990) is used.
The FM-OLS method is a convenient tool to examine dynamic interactions when variables used in
the model are non-stationary I(1) processes. In addition, the FM-OLS is less sensitive to changes in
lag structure and performs better for small or finite sample sizes than other cointegration techniques
(Engle and Granger 1987, Johansen 1988). This dynamic analysis could enhance the understanding
of demand for coal barge transportation on the U.S. inland waterway system and provide vitally
important information for policymakers and shippers. From a policymaker’s perspective, it is
essential to understand the determinants of coal movements to build adequate regional planning and
port capacity. Appropriate planning and investment for the inland waterway system can increase the
capacity of the waterway system and efficiency of barge movements, and can reduce the time and
costs of coal shipments. From a shipper’s perspective, on the other hand, it is necessary to assess
transportation equipment needs and to develop business plans. Shippers can make labor and capital
investment plans based on the determining factors for coal shipments.
The remainder of the paper is organized as follows. The next section develops an empirical
model, with a discussion of the equation being estimated, the FM-OLS approach, and data
description. The third section presents the empirical results, focusing on key determinants for coal
barge movements and the implications of the results. The last section summarizes the paper with
concluding remarks.
90

EMPIRICAL METHODOLOGY
Equation to be Estimated
JTRF Volume 50 No. 1, Spring 2011
The demand for barge transportation is typically derived from the demand for commodities in
origin and destination regions, known as a derived demand. The demand for coal barge service, for
example, is determined by factors shifting the supply curve in coal production regions and demand
curve in coal consumption markets (Boyer 1997). If the coal production in origin regions or the coal
consumption at destination markets increases, the demand for coal barge service tends to increase.
In analyzing factors affecting demand for barge transportation in an empirical work, this paper
relies on a theoretical framework developed by Tang (2001) and Yu and Fuller (2005). In its general
form this model specifies the demand for barge service as a function of the supply (production) of
a commodity at origin, the demand (consumption) for a commodity at destination, and price and
service characteristics of barge and competing transportation mode (e.g., barge and rail rates) (see
Tang [2001] for an extensive discussion of the demand for barge transportation). In the empirical
model used here, the standard demand model of barge service is modified in order to examine factors
affecting coal barge demand. The reduced-form equation for the quantity of coal transportation
service (TV t ) is specified as follows:
(1) TV t = f(BR t , EX t , DD t , DS t , OT t )
where BR t is the coal barge rate; EX t is the coal export level; DD t is the domestic consumption of
coal; DS t is the domestic production of coal; OT t is the rate proxy of other transportation modes.
As noted earlier, for the competing transportation mode for coal shipment, rail transportation is
considered as a possible substitute of barge service for coal. 3
To illustrate the FM-OLS modeling approach, Equation (1) is then expressed in a log linear
form as follows:
(2) ln(TV t ) = α + β 1 ln(BR t ) + β 2 ln(EX t ) + β 3 ln(DD t ) + β 4 ln(DS t ) + β 5 ln(OT t ) + ɛ t
Since an increase in the barge transport rate tends to reduce the quantity of coal shipped by barge, it
is expected that β 1 < 0. An increase in coal exports and domestic consumption leads to an increase
in demand for coal; hence, it is expected that β 2 > 0 and β 3 > 0, respectively. Because an increase
in the production of coal from coal mines causes an increase in coal demand for barge service, it
is expected that β 4 > 0. Unlike the four variables, the expected sign of rail rates is ambiguous and
uncertain. In general, water and rail services are considered as substitutes for freight shipments in
long hauls, suggesting that an increase in rail rates can be positively associated with barge freight
volume. However, the substitutability between water and rail services depends on many factors, such
as the type of commodities, the location of origin and destination, and the availability of loading/
unloading and switching facilities. If a rail service is not physically available or economically viable
for the coal barge movements, a shift in rail rates may not lead to a change in the quantity of coal for
barge service (i.e., independent service). As such, the expected sign of the rail rate in Equation (2) is
not predetermined, due mainly to insufficient information on the substitutability between these two
modes for coal shipments. 4
The FM-OLS Approach
The fully-modified cointegration (FM-OLS) method developed by Phillips and Hansen (1990) is
used to examine factors affecting coal barge demand. The FM-OLS model is an alternative dynamic
model to obtain an unbiased estimate of the long-run relationship when the underlying regressors
are nonstationary I(1) processes. 5 Note that, unlike OLS, the FM-OLS approach does not require
91

Demand Analysis for Coal
differencing of nonstationary variables to make them stationary; hence, this method keeps valuable
information concerning long-run properties inherent in the levels of time-series data (Perman 1991).
The FM-OLS uses an econometric model as follows:
(3)
where y t is I(1) variable (in this paper, y t = TV t , where TV t is the quantity of coal transportation
service); and x t is (k×1) vector of I(1) regressors (i.e., x t = BR t , EX t , DD t , DS t , OT t where BR t is
the coal barge rate; EX t is the coal export level; DD t is the domestic consumption of coal; DS t is
the domestic production of coal; OT t is the rate proxy of other transportation modes), which are
assumed not to be cointegrated among themselves. Additionally, x t is assumed to have the following
first-difference stationary process:
(4) Δ x t = µ + v t
where µ is (k×1) vector of drift parameters; 6 v is (k×1) vector of I(0), or stationary variables. It is
t
also assumed that is strictly stationary with zero mean and a finite positive-definite
covariance matrix, Ʃ.
'
The OLS estimators of α and β 1 in Equation (3) are consistent even if x and ɛ (equivalently v and
t t t
ɛ ) are contemporaneously correlated (Engle and Granger 1987; Stock 1987). t 7 In general, however,
the OLS regression involving non-stationary variables no longer provides the valid interpretations
of the standard statistics such as t- and F-statistics in Equation (3). Further, unless non-stationary
variables combine with other non-stationary variables to form stationary cointegration relationships,
the estimation can falsely represent the existence of a meaningful economic relationship (i.e.,
spurious regression) (Harris and Sollis 2003). To address these problems adequately, it is necessary
to correct the possible correlation between v and ɛ , and their lagged values. The Phillips-Hansen
t t
FM-OLS estimator takes account of these correlations in a semi-parametric manner. As a result, the
FM-OLS is an optimal single-equation technique for estimating with I(1) variables (Phillips and
Loretan 1991).
It is worth mentioning that the non-stationarity of time-series data can be first-differenced rather
than in levels in a framework of OLS. First-difference, however, may lose valuable information
concerning long-run properties inherent in the levels of time-series data (Perman 1991). Therefore,
with the long-run information embedded in the levels data the cointegration approach (i.e., FM-
OLS) offers a solution to this dilemma.
Data
The U.S. Department of Energy (DOE) is the primary data source for this paper. The coal domestic
tonnages by barge (measured in million tons) are obtained from Coal Transportation: Rates
and Trends in the United States (U.S. Energy Information Administration 2004). Domestic coal
consumption, production, and exports (measured in million tons) are taken from the EIA. Average
barge and rail rates (measured in U.S. $ per ton) are also collected from the EIA, which are originally
complied by the Federal Energy Regulatory Commission (FERC). 8 The GDP deflator (2000=100) is
used to derive real barge and rail rates. The data set contains 23 annual observations for the period
1979 to 2001. All variables are in natural logarithms.
It should be pointed out that the 1979-2001 period is currently the best data available for the
analysis as the EIA has not yet updated the relevant dataset. 9 For this reason, the sample size could be
a concern for validation of the dynamic relationships estimated by our empirical model; the findings
should thus be viewed with caution. As Hargreaves (1994) notes, however, the FM-OLS procedure
has been proven to have superior small sample properties, which makes it a good choice for our
92

JTRF Volume 50 No. 1, Spring 2011
sample than other cointegration techniques (e.g., Engle and Granger 1987, Johansen 1988); this
should mitigate the concern with the relatively short period of data coverage and provide credibility
of our findings.
EMPIRICAL RESULTS
The first requirement for application of the FM-OLS cointegration procedure is that the variables
in Equation (2) must be non-stationary with I(1) processes. The presence of a unit root in the six
variables is determined using the Dickey-Fuller generalized least squares (DF-GLS) test (Elliot et
al. 1996). The DF-GLS test optimizes the power of the ADF test using a form of detrending. As
Elliott et al. (1996, p. 813) note: “Monte Carlo experiments indicate that the DF-GLS works well in
small samples and has substantially improved power when an unknown mean or trend is present.”
Recently, Ng and Perron (2001) produced a testing procedure which incorporates both the new
information criterion for setting the lag length and GLS detrending. The results show that, with the
level series, the null hypothesis of non-stationarity cannot be rejected for all six variables at the 5%
level (Table 1). With the first-differenced series, on the other hand, all the variables are found to be
stationary; hence, this analysis concludes that all the variables are non-stationary I(1) processes. The
DF-GLS test statistics are estimated from a model that includes a constant and a trend variable. The
lag lengths are selected using Schwarz criterion (SC).
Table 1: Results of Unit Root Testa Variable Level First difference Decision
ln (TV ) t
-1.910
(1)
-5.701*
(0)
I (1)
ln (BR ) t
-2.516
(0)
-4.960*
(0)
I (1)
ln (EX ) t
-2.405
(0)
-3.941**
(0)
I (1)
ln (DD ) t
-2.777
(0)
-5.913**
(1)
I (1)
ln (DS ) t
-2.604
(1)
-7.455*
(0)
I (1)
ln (OT ) t
-2.666
(1)
-4.603**
(0)
I (1)
a TVt , BR EX , DD , DS , and OT represent the total volume of coal, barge rates, quantity of coal exports,
t t t t t
quantity of domestic demand of coal, quantity of domestic supply of coal and rail rates, respectively. * and
** denote the rejection of the null hypothesis of non-stationarity at the 10% and 5% significance levels,
respectively. The 10% and 5% critical values for the DF-GLS, including a constant and a trend, are -2.890 and
-3.190, respectively. Parentheses are lag lengths, which are chosen by the Schwarz criterion (SC).
With evidence that each of the data series is a non-stationary I(1) process, the FM-OLS is
applied to estimate the long-run relationship in Equation (2). First of all, the result of the DF-GLS
test performed on the residual from the estimated Equation (2) shows that the null hypothesis can
be rejected at the 5% significance level (Table 2), suggesting the existence of long-run relationships
between TV t and the set of explanatory variables (BR t , EX t , DD t , DS t , and OT t ) in Equation (2). In
other words, even though individual series may have trends or cyclical or seasonal variations, the
movements in one variable are matched (at least approximately) by movements in other variables
(Perman 1991).
93

Additionally, the results shows that the barge rate is not statistically significant even at the 10%
level, indicating that in the long run a change in barge rates has little effect on the quantity of coal
shipped by barge (Table 2). One plausible explanation for the finding is that, since barge rates are
substantially lower than the rates of alternative transportation modes (i.e., rail rates), barge is the
only economically viable transportation option for coal movements in the current system of the
U.S. inland waterways, regardless of barge rate fluctuations. 10 Similarly, the rail rate is found to be
statistically insignificant at the 10% level, suggesting that rail rates play little role in influencing the
quantity of coal shipped by barge transportation in the long run. One explanation for this finding is
that, while water transportation generally has a service disadvantage (i.e., slow transit time), its better
accessibility from coal mines to coal-fueled power plants tends to reduce the substitution effect to
rail service. 11 This may allow the demand for coal barge transportation to be highly insensitive to a
change in rail rates. On the other hand, the quantity of coal exports is found to have a significantly
positive long-run relationship with the volume of coal barge movements; for example, a 1% increase
in exports causes barge shipments of coal to increase by approximately 0.38%. The quantity of
domestic demand for coal is also found to have a significantly positive long-run relationship with
the quantity of coal barge service; for example, coal barge shipments increase by approximately
2.14%, given a 1% increase in the quantity of domestic demand of coal. However, the quantity of
domestic supply of coal has little effect on the demand for coal barge service. Hence, the findings
show that the demand for coal barge transportation is mostly determined by demand for coal at
domestic and international destinations, rather than the supply of coal or barge and rail rates.
Table 2: Result of the Fully-Modified OLS (FM-OLS) Estimationa Variable
Coefficient
ln (TV ) t
-statistic
ln (BR ) t 0.084 0.536
ln (EX ) t 0.381 2.137**
ln (DD ) t 2.137 2.175**
ln (DS ) t -0.037 -0.045
ln (OT ) t 0.001 0.003
Constant -12.573 -2.796**
DF-GLS statistic -4.699 [0]**
a TVt , BR t EX t , DD t , DS t , and OT t represent the total volume of coal, barge rates, quantity of coal exports,
quantity of domestic demand of coal, quantity of domestic supply of coal and rail rates, respectively. * and
** denote significance at the 10% and 5% levels, respectively. A bracket in the DF-GLS statistic is lag length.
The 10% and 5% critical values for the DF-GLS, including a constant and a trend, are -2.890 and -3.190,
respectively.
For completeness, the error-correction model (ECM) is also estimated using the residual
obtained from Equation (2) in order to examine the short-run adjustment to the long-run steady
state, as well as to confirm the existence of the cointegration relationship (Table 3). The results show
that the coefficient of the error-correction term (ec t‒1 ) is negative and statistically significant at the
5% level. The negatively significant coefficient of ec t‒1 implies that the equilibrium relationship will
hold in the long run, even with shocks to the system. Additionally, the statistically significant ec t‒1
further supports the validity of cointegrating relationship in Equation (2). The multivariate diagnostic
tests on the estimated model as a system indicate no serious problems with serial correlation,
heteroskedasticity, and normality; hence, the model is well defined. Notice that the domestic supply
of coal is only found to have a significantly positive effect on the demand, indicating that the supply
of coal is a crucial determinant of the demand for coal barge transportation in the short run.
94

Table 3: Result of Error-Correction Model (ECM) a
Variable
JTRF Volume 50 No. 1, Spring 2011
Δln (TV t )
Coefficient -statistic
Δln (BR t ) 0.312 1.46
Δln (EX t ) -0.071 -0.29
Δln (DD t ) -1.286 -0.78
Δln (DS t ) 1.831 2.31**
Δln (OT t ) 0.861 1.13
Constant
ec t-1
Serial correlation
Heteroskedasticity
Normality
RESET
0.033
-0.989
0.444 [0.652]
0.003 [0.954]
0.741 [0.954]
0.894 [0.362]
0.586
-3.31**
a TVt , BR t EX t , DD t , DS t , and OT t represent the total volume of coal, barge rates, quantity of coal exports,
quantity of domestic demand of coal, quantity of domestic supply of coal and rail rates, respectively. * and **
denote significance at the 10% and 5% levels, respectively. A bracket in the DF-GLS statistic is lag length. The
10% and 5% critical values for the DF-GLS, including a constant and a trend, are -2.890 and -3.190, respectively.
Serial correlation of the residuals of a whole system is examined using the F -form of the Lagrange-Multiplier
(LM) test. Heteroskedasticity is tested using the F -form of the LM test. Normality of the residuals is tested
with the Doornik-Hansen test (Doornik and Hendry 1994). Statistics in brackets are p -values.
CONCLUDING REMARKS
This paper explores the demand for coal barge transportation on the U.S. inland waterway system
in a cointegration framework. Previous studies have mostly examined the demand for grain barge
transportation and barge rates, but little attention has been given to factors determining coal barge
shipments and the substitution effect between water and rail services for coal shipments. Using a
FM-OLS approach, this paper examines the dynamic interrelationships between the demand for
coal barge transportation and barge rates, rail rates, domestic coal consumption, domestic coal
production, and coal exports. The results of the FM-OLS suggest that there is one stable long-run
equilibrium relationship between the demand for coal barge transportation and the selected variables.
The negatively significant coefficient of the error-correction term in the vector error-correction
model further validates the existence of an equilibrium relationship among the variables. The results
also show that, in the long run, the demand for coal barge transportation is mostly responsive to the
domestic coal exports and domestic coal consumption, rather than barge and rail rates. In the short
run, on the other hand, domestic coal production is the only significant determinant of the demand
for coal barge transportation.
Currently, the U.S. Army Corps of Engineers (USACE) builds, maintains, and plans all
infrastructure needs for the inland waterway system, yet insufficient information hinders its ability
to develop the comprehensive regional investment plans to support freight movements and energize
the economy. The information derived from this paper may help policymakers develop better
infrastructure strategies and improve investment decisions for the inland waterway system. The
U.S. inland waterway system has 230 lock sites, which include 275 lock chambers that support coal
shipments. The crucial factors identified in this paper can be used to help predict coal traffic at lock
sites in the short and long run and assess the lock investment projects. For barge carriers, on the
other hand, the information obtained from this paper will be valuable in implementing their labor
and capital investment plans.
95

Demand Analysis for Coal
Finally, it should be pointed out that, although this study conjectures that more disaggregated
demand structures (data) is both desirable and necessary to provide more useful information about
investment decisions for the inland waterway system, limited data availability does not allow
making future analysis to examine the validity of this conjecture. In addition, this paper does not
consider major policy and market shocks in the model that may result in a significant change in
the demand for coal barge transportation. For example, does the partial deregulation (or mergers)
of railroad companies in the United States affect demand for barge transportation? How does the
Clean Air Act Amendment of 1990 influence shipments of coal volumes transported via the inland
waterway system? All these issues should be addressed in future research.
Endnotes
1. Water carriers are known to compete with rail carriers for the shipments of bulk commodities
for long hauls. While water and rail transportation modes are empirically found to be partial
substitutes for grain movements (e.g., Miljkovic et al. 2000), the substitution effects between
these two modes have not been investigated for coal barge shipments.
2. Some of the previous literature has indeed investigated the presence of a unit root in their timeseries
variables using the augmented Dickey-Fuller (ADF) test (Thoma and Wilson 2004a, Yu
et al. 2007). However, when dealing with finite samples, the power of the standard ADF test is
known to be notoriously low (Maddala and Kim 1998, Harris and Sollis 2003). In other words,
the ADF test has high probability of accepting the null hypothesis of non-stationarity when the
true data-generating process is, in fact, stationary. Therefore, to overcome the shortcoming of
the ADF, this paper adopts a more powerful test known as the Dickey-Fuller generalized least
squares (DF-GLS) detrended test.
3. In this paper, truck transportation is not considered as a substitute of barge service for coal
shipments, because truck transportation is not an economical transportation option for bulk
commodities and raw materials for long hauls.
4. It is worth mentioning that the FM-OLS uses a cointegration framework to take into account the
non-stationarity in the series as well as potential endogeneity of the explanatory variables and
serial correlation of the error term. More specifically, in general, TV t and BR t are endogenously
determined in equation (2); by definition, BR t in this case is likely to be correlated with the
error term ɛ t , which causes the OLS estimators to be biased. Additionally, when using the
traditional OLS, the Durbin-Watson (D-W) test shows that equation (2) is suffering from AR(1)
serial correlation at the 5% level. To solve these problems, this paper uses the FM-OLS to
estimate equation (2). Hargreaves (1994) demonstrates that, particularly in small samples, the
FM-OLS has been proven to be a fully efficient method of estimating the long-run equilibrium
(cointegration) relationship.
5. A stationary series, or integrated of order zero (I(0)) is defined as a series that tends to return
to its mean value and fluctuate around it within a more or less constant range. A non-stationary
series (unit roots), on the other hand, is defined as a series that has a different mean at different
points in time and its variance increases with the sample size (Harris and Sollis 2003). Since
OLS regression with non-stationary series no longer provides the valid interpretations of the
standard statistics (i.e., t - and F -statistics), non-stationary variables should be differentiated
to make them stationary. Integrated of order one, I(1), is a time-series process that needs to
be first-differenced to produce a stationary series, or I(0), which is often said to be a (first)
difference-stationary process (Wooldridge 2006). The first difference of a time series means the
series of changes from one period to the next. However, Engle and Granger (1987) show that,
96

JTRF Volume 50 No. 1, Spring 2011
even in the case that all the variables in a model are non-stationary (i.e., I(1)), it is possible for a
linear combination of integrated variables to be stationary (i.e., I(0)); in this case, the variables
are said to be cointegrated.
6. A drift parameter is usually said to be a random walk with drift. A random walk means a
time series process where next period’s value is obtained as current period’s value, plus an
independent error term. A random walk with drift means a random walk that has a constant (or
trend) added in each period (Wooldridge 2006).
7. The explanatory (independent) variables and error terms are correlated in the same time period
(Woodridge 2006). An explanatory variable is a variable that is used to explain variation in the
dependent variable. An error term means a variable that contains unobserved factors that affect
the dependent variable.
8. EIA collected coal transportation rates primarily from coal burning utilities and they are not
necessarily available to industrial customers that tend to purchase coal under smaller supply
contracts. According to EIA (2004), the weighted average rates are actual averages as individual
rates and mine price outliers are not included. The rates represent coal delivered under contract
and excludes coal deliveries scheduled for 12 or fewer months, commonly referred to as spot
coal purchases.
9. River Transport News (1993-current) could be used as an alternative rate data source but it is
difficult to estimate the annual average rates from the insufficient number of the samples.
10. In 2000, for example, the average domestic barge rates for coal and rail per ton are $2.26 and
$10.30, respectively (U.S. Energy Information Administration 2004).
11. U.S. Army Corps of Engineers (2005) reports that in the case of lock closure at McAlpine on
the Ohio River in Louisville, Kentucky, the most frequent response from survey respondents
is to wait until lock operation resumes and do not switch to rail service. High rail rates and
inaccessibility to loading/unloading rail facilities restrain the shippers’ ability to switch their
transportation modes.
References
Babcock, M.W. and X. Lu. “Forecasting Inland Waterway Grain Traffic.” Transportation Research
Part E 38 (1), (2002): 65-74.
Babcock, M.W. and S. Fuller. “A Model of Corn and Soybean Shipments on the Ohio River.”
Journal of the Transportation Research Forum 46 (2), (2007): 21-34.
Boyer, K.D. Principles of Transportation Economics. Addison Wesley, Reading, MA, 1997.
Bureau of Transportation Statistics. Shipment Characteristics by Mode of Transportation for the
United States: 2007. U.S. Department of Transportation. www.bts.gov. 2007.
Clark, C., K.E. Henrickson, and P. Thoma. An Overview of the U.S. Inland Waterway System. U.S.
Army Corp of Engineers, Alexandria, VA, 2005.
DeVuyst, E., W.W. Wilson, and B. Dahl. “Longer-term Forecasting and Risks in Spatial Optimization
Models: the World Grain Trade.” Transportation Research Part E 45, (2009): 472-485.
97

Demand Analysis for Coal
Doornik, J. and D. Hendry. Interactive Econometric Modeling of Dynamic System (PcFiml 8.0).
International Thomson Publishing, London, UK, 1994.
Elliott, G., T. Rothenberg, and J. Stock, “Efficient Tests for an Autoregressive Unit Root.”
Econometrica 64, (1996): 813–836.
Engle, R.F. and C.W.J. Granger. “Co-Integration and Error Correction: Representation, Estimation,
and Testing.” Econometrica 55, (1987): 251-276.
Hargreaves, C.P. Non-stationarity Time-Series Analysis and Cointegration. Oxford University
Press, Oxford, UK, 1994.
Harnish, G.L. and J.W. Dunn. “A Short-Run Analysis of Grain Barge Rates on the Mississippi
River System.” Proceedings of the 38th Annual Meeting of the Transportation Research Forum. San
Antonio, Texas, 1996.
Harris, R. and R. Sollis. Applied Time Series Modeling and Forecasting. John Wiley and Sons, Inc.,
West Sussex, England, 2003.
Johansen, S. “Statistical Analysis of Cointegration Vector.” Journal of Economic Dynamics and
Control 12, (1988): 231-254.
Maddala, G.S. and I.M. Kim. Unit Roots, Cointegration, and Structural Change. Cambridge
University Press, Cambridge, UK, 1998.
Miljkovic, D., G.K. Price, R.J. Hauser, and K.A. Algozin. “The Barge and Rail Freight Market
for Export-Bound Grain Movement from Midwest to Mexican Gulf: an Econometric Analysis.”
Transportation Research Part E 36 (2), (2000): 127-137.
Ng, S. and P. Perron. “Lag Length Selection and the Construction of Unit Root Tests With Good Size
and Power.” Econometrica 69, (2001): 1519-1554.
Perman, R. “Cointegration: An Introduction to the Literature.” Journal of Economic Studies 18,
(1991): 3-30.
Philips, P.C.B. and B.E. Hansen. “Statistical Inference in Instrumental Variables Regressions with
I(1) Processes.” The Review of Economic Studies 57, (1990): 99-125.
Philips, P.C.B. and M. Loretan. “Estimating Long-run Economic Equilibria.” The Review of
Economic Studies 58, (1991): 407-436.
Stock, J.H. “Asymptotic Properties of Least Squares Estimators of Cointegrating Vectors.”
Econometrica 55, (1987): 1035-1056.
Tang, X. “Time Series Forecasting of Quarterly Barge Grain Tonnage on the McClellan-Kerr
Arkansas River Navigation System.” Transportation Quarterly 40, (2001): 91-108.
Thoma, M. and W.W. Wilson. Long-Run Forecasts of River Traffic on the Inland Waterway System.
U.S. Army Corps of Engineers, Alexandria, VA, 2004a.
Thoma, M. and W.W. Wilson. A Study of Short-Run Grain Movements on the Inland Waterway
System. U.S. Army Corps of Engineers, Alexandria, VA, 2004b.
U.S. Army Corps of Engineers. McAlpine Lock Closure in August 2004: Shipper and Carrier
Response. U.S. Army Corps of Engineers, Alexandria, VA, 2005.
98

JTRF Volume 50 No. 1, Spring 2011
U.S. Army Corps of Engineers. The U.S. Waterway System Transportation Facts & Information.
U.S. Army Corps of Engineers, Waterborne Commerce Statistics Center, New Orleans, LA, 2010.
U.S. Energy Information Administration. Coal Transportation Rates and Trends in the United
States. U.S. Department of Energy. www.eia.doe.gov., 2004.
U.S. Energy Information Administration. Annual Coal Distribution. U.S. Department of Energy.
www.eia.doe.gov., 2010.
Woodridge, J.M. Introductory Econometrics: A Modern Approach. South-Western College
Publishing, Mason, OH, 2006.
Yu, T. and S.W. Fuller. “The Measurement of Grain Barge Demand on Inland Waterways: a Study
of the Mississippi River.” Journal of the Transportation Research Forum 44 (1), (2005): 27-39.
Yu, T.H., D.A. Bessler, and S.W. Fuller. “Price Dynamics in U.S. Grain and Freight Markets.”
Canadian Journal of Agricultural Economics 55, (2007): 381-397.
Junwook Chi is a transportation economist in Nick J. Rahall, II Appalachian Transportation
Institute and Center for Business and Economic Research at Marshall University. He received his
Ph.D. in transportation and logistics from North Dakota State University. His specialty area of
research is transportation economics and regional economic development. He currently manages
the multi-year transportation research and economic development projects for the U.S. Army Corps
of Engineers and West Virginia Department of Transportation. He is also an instructor in Lewis
College of Business at Marshall University and teaches an Economics of Transportation course.
He received the awards for “2008 Best Graduate Student Paper” from the Transportation Research
Forum and “2002 Outstanding Master’s Thesis” from the Canadian Agricultural Economics
Association.
Jungho Baek is an assistant professor of economics in School of Management at University of
Alaska Fairbanks (UAF). He received his Ph.D. from Michigan State University in 2004. Prior to
UAF, he has conducted research for the Korea Institute for International Economic Policy (KIEP),
in South Korea, and the Center for Agricultural Policy and Trade Studies (CAPTS), at North Dakota
State University. His principal areas of research are international trade and policies, transportation
economics, environmental economics, and econometric modeling. His work in these areas has been
presented at various professional conferences and published in international refereed journals.
99

100

State-of-the-Art: Centerline Rumble Strips
Usage in the United States
by Daniel E. Karkle , Margaret J. Rys, and Eugene R. Russell
Centerline Rumble Strips (CLRS) are used to avoid cross-over roadway departures, making rural
highways safer. The objectives of this study were to obtain nationwide, updated information about
states’ policies and guidelines for utilization of CLRS and to provide a list of gaps in research
along with good practices. Results indicate that 36 states reported the use of CLRS. The total CLRS
approximate mileage is 11,333 miles. The predominant CLRS pattern is: milled, length 16”, width
7”, depth 0.5”, spacing 12”, continuous. This survey reported that 17 states have written policies
or guidelines. A list of good practices used by the states is presented.
INTRODUCTION
Roadway departure fatalities are a serious problem in the United States. A roadway departure crash
is defined as a non-intersection crash which occurs after a vehicle leaves the traveled way, crossing
the center line of undivided highways, or crossing an edge line (longitudinal pavement marking
located at the edge of the traveled lane and the shoulder) of the roadway. Roadway departures
are usually severe and involve run-off-the-road (ROR), sideswipes, and head-on crashes. There
are many contributing factors for the occurrence of roadway departures, and the principal of them
are driver drowsiness, fatigue, alcohol/drug impairment, and inattention, along with poor visibility
caused by inclement weather. Roadway departure crashes account for the majority of rural highway
fatalities. According to data from the Fatality Analysis Reporting System (FARS), in 2009 there
were 11,185 fatal roadway departure crashes on rural highways, resulting in 23,169 fatalities
(NHTSA 2009). Furthermore, roadway departure crashes correspond to approximately 40% of all
crashes in the United States, and the estimated annual cost of roadway departure crashes is $100
billion (FWHA 2003).
In order to reduce the number of roadway departure crashes, since 1955, several state
departments of transportation have installed rumble strips and other accidents countermeasures on
U.S. highways (Carlson and Miles 2003).
Rumble strips are raised or indented patterns utilized to alert drivers that they are moving out of
the travel lane. When vehicles’ tires pass over the rumble strips, noise and vibration are produced by
this contact, which provides motorists with a warning that they are leaving the travel lane. Rumble
strips are designed to alert drowsy and inattentive motorists and can generally be classified by
their position in relation to the travel lane as: a) shoulder rumble strips (including edgeline rumble
strips), b) centerline rumble strips, c) midlane rumble strips, and d) transverse rumble strips. Figure
1 illustrates the position of each type of rumble strips in relation to the travel lane. The commonly
referred dimensions of rumble strips are: length, normally defined as the dimension perpendicular
to the traffic direction; width, usually defined as the dimension parallel to the traffic direction;
depth of height; and spacing, usually measured from center to center of rumble strip patterns. The
spacing can be continuous, if the rumble strips are placed with constant spacing along the roadway,
or alternatively, if the spacing changes along the roadway (for example: 12 in., followed by 24 in.,
spacing).
101

Centerline Rumble Strips
Figure 1: Placement of Rumble Strips in a Roadway
102
2
Travel Lane
1
Shoulder
Shoulder
Travel Lane
3
4
Center Line
Note: (1) Shoulder Rumble Strips, (2) Centerline Rumble Strips, (3) Midlane Rumble Strips,
(4) Transverse Rumble Strips
Shoulder rumble strips (SRS) are placed on the shoulders or on the edge line (along the
longitudinal pavement marking located at the edge of the travel lane and the shoulder) of the roadway
and are a countermeasure for ROR type crashes. On divided highways, SRS may be installed on
both the outside and median shoulders. When installed along the edge lines, they are commonly
referred to as “rumble stripes” or edgeline rumble strips. The benefits of rumble strips are the
greater free space allowed on the shoulders for motorists to perform corrective maneuvers, for other
users such as bicyclists to use the shoulders, and that they can be installed on roadways with narrow
or nonexistent shoulders. Centerline rumble strips are placed on the center of the roadway and are
designed to mitigate cross-over crashes.
Midlane rumble strips is a concept with no actual installations known. Their theoretical
placement would be in the center of the travel lane, serving to potentially prevent both cross-over
and ROR crashes.
Transverse rumble strips are usually placed across the full width of the travel lanes. They are
designed to alert motorists of approaching roundabouts, intersections, and toll plazas.
According to Elefteriaou et al. (2000), there are four types of rumble strips classified by their
installation process: a) raised, b) milled, c) rolled, and d) formed, as presented in Figure 2. The
milled is the most common type of rumble strip in the United States. They can be installed on
new or existing asphalt and Portland cement concrete (PCC) pavements. This type of rumble strip
is produced by a machine, which cuts a groove in the pavement. Raised rumble strips are made
by adherence of proper material to new or existing pavement surfaces. Formed rumble strips are
installed on PCC surfaces by forming grooves or indentations into the concrete during its finishing
process. Rolled rumble strips are installed only on asphalt surfaces by a roller that presses grooves
into the hot surfaces when the asphalt is being compacted.
Figure 2: Types of Rumble Strips
a) Raised b) Milled c) Rolled d) Formed
Source: Richards and Saito (2005)

Centerline Rumble Strips
JTRF Volume 50 No. 1, Spring 2011
This study focuses on the applications of CLRS. Centerline rumble strips are primarily installed on
the center line of undivided two-lane highways, and their main purpose is reduction of cross-over
crashes, more specifically, head-on and opposite direction sideswipe and front-to-side type crashes,
which are usually caused by driver inattention and drowsiness. The data available on the FARS
database reveal that in 2009, 56% of the fatal crashes occurred on rural roads. Among these, 74%
occurred on undivided two-lane roads, and 20% of these accidents involved two vehicles traveling
in opposite directions, totaling 2,579 cross-over fatal crashes per year (NHTSA 2009). CLRS are
accepted as a countermeasure that reduces approximately 25% of cross-over crashes, making twolane
rural roads safer. Therefore, the use of them in the United States has increased over the years.
Several authors have reported advantages other than crash reduction in installing CLRS, such
as low interference in passing maneuvers, versatile installation conditions, and public approval
(Miles et al. 2005; Richards and Saito 2007). Due to their associated low costs of installation
and maintenance, CLRS provide high benefit-cost ratios. For instance, Carlson and Miles (2003)
reported estimated benefit-cost ratio associated with CLRS in the range of 0.17 to 39.16 (the higher
the roadway traffic volume, the greater the benefit), considering five states and assuming cross-over
crash reduction of 20%. However, some concerns involving CLRS, such as the levels of exterior
noise, potential decreased visibility of the painted strips, potential tendency to speed up pavement
deterioration, possibility of causing driver erratic maneuvers, and ice formation in the grooves,
have been cited in the current literature (Russell and Rys, 2005). The policies and guidelines for
CLRS installation are very distinct among the states using them. A better understanding of good
practices and gaps in research about the use of CLRS would contribute to future enhancement
of their associated advantages and reduction of their potential weaknesses. For these reasons, the
objectives of this study are to obtain nationwide, updated information about states’ policies and
guidelines for utilization of CLRS and to provide a list of gaps in research along with good practices
in the country. It is expected that the information from this study will be useful for planners and
policy makers, providing guidance for future applications of CLRS.
LITERATURE REVIEW
This section presents a review of the pertinent studies that focus on the different effects of CLRS
and the previous national surveys on CLRS policies. Studies of other types of rumble strips, for
example, shoulder rumble strips, are not part of the scope of this work.
Safety Effectiveness of CLRS
There are several published and unpublished studies revealing that CLRS reduces cross-over crashes.
Generally, the methods utilized in these studies are the Naïve before-and-after, which just compares
the before and after numbers with no adjustments, and the Empirical Bayes method, which uses
more sophisticated, state-of-the-art statistics. Some of these studies are summarized in Table 1.
The results of these studies are not uniform. The differences in the crash reduction effects may be
partially attributed to differences of the CLRS applications, since different patterns of rumble strips
have proven to generate different levels of noise and vibration stimuli for drivers. The best pattern
and application of CLRS along the roadway can be considered a gap in research since it remains
unknown. Chen et al. (2003) claim that the performance of rumble strips should be a function of
the difference between noise and vibration stimuli over rumble strips and over smooth pavement
conditions (the best pattern would be the one that produces the largest differences). In addition, an
increase in order of 9 to 10 dBA (dBA corresponds to the unit of the A-scale on a sound-level meter,
which is the scale that best approximates the frequency to which that human ear can respond) in
the level of sound is necessary for a person to be alerted by the presence of that sound (Lipscomb
103

Centerline Rumble Strips
1995, cited by Rys et al. 2008). Therefore, CLRS should raise the levels of sound by at least 10 dBA.
Miles and Finley (2007) stated that the “standard” rumble strips dimensions (milled, length equal or
greater than 12 in., width of 7 in., depth of 0.5 in., and spacing of 12 to 24 in.) in the United States
provide adequate increase in the sound level to alert all drivers, regardless of the speed or the type
of pavement.
Table 1: Safety Effectiveness of CLRS
104
State Study Statistical Method
Arizona
AECOM (2008) Comparison Group
Kar and Weeks (2009) Naïve Before-and-After
Type of Crash
Studied
Fatal and serious
injury cross-over
Fatal and serious
injury cross-over
Crash
Reduction
61.0%
56.0%
California
Fitspatrick et al. (2000)
Persaud et al. (2003)
Naïve Before-and-After
Empirical Bayes
Fatal head-on
Total head-on
Cross-over
All types
90.0%
42.0%
12.0%
14.0%
Colorado
Outcalt (2001)
Persaud et al. (2003)
Naïve Before-and-After
Empirical Bayes
Head-on
Sideswipe
Cross-over
All types
34.0%
36.5%
31.0%
11.0%
Head-on 95.0%
Drove left to the
center
60.0%
Delaware DOT (2003) Naïve Before-and-After
PDO Increase 13%
Delaware
Injury Increase 4%
All Types 8.0%
Persaud et al. (2003) Empirical Bayes
Cross-over
All types
81.0%
23.0%
Fatal head-on 80.0%
Head-on 81.0%
Naïve Before-and-After
Sideswipe 78.0%
Kansas Karkle et. al (2009)
Cross-over 80.0%
Fatal and serious
injury cross-over
59.0%
Empirical Bayes
Cross-over
All types
85.0%
33.0%
Maine Unpublished Maine DOT Naïve Before-and-After
Head-on
ROR
91.7%
28.9%
Maryland Persaud et al. (2003) Empirical Bayes All types 19.0%
Massachusetts Noyce and Elango (2004) Comparison Group Several Inconclusive

Table 1: Safety Effectiveness of CLRS (continued)
State Study Statistical Method
Minnesota
Persaud et al. (2003) Empirical Bayes
Briese (2006)
Knapp and Schmit (2009)
Cross-Sectional
Comparison
Cross-Sectional
Comparison
Torbic et. al (2009) Empirical Bayes
Missouri Unpublished Missouri DOT
Naïve Before-and-After
Empirical Bayes
JTRF Volume 50 No. 1, Spring 2011
Type of Crash
Studied
Crash
Reduction
Cross-over Increase 12%
All types 0.0%
Cross-over 43.0%
All types 42.0%
Cross-over - Fatal and
severe injury
All types - Fatal and
severe injury
Cross-over - Fatal and
severe injury
All types - Fatal and
severe injury
Increase 13%
73.0%
47.0%
40.0%
All Types 11.1%
Fatal and injury 21.8%
Cross-over 48.9%
Fatal and injury crossover
44.7%
Head-on 29.0%
Sideswipe 61.0%
Head-on 53.0%
Sideswipe 62.0%
Nebraska Unpublished Nebraska DOT Naïve Before-and-After Cross-over 64.0%
Oregon
Pennsylvania
Washington
Monsere (2002) cited by
Russell and Rys (2005)
Naïve Before-and-After Cross-over 69.5%
Comparison Group Cross-over 79.6%
Persaud et al. (2003) Empirical Bayes All Types 46.0%
Galenabiewski et al. (2008) Naïve Before-and-After Cross-over 48.0%
Torbic et. al (2009) Empirical Bayes
Persaud et al. (2003) Empirical Bayes
Torbic et. al (2009) Empirical Bayes
All Types 1.6%
Fatal and injury 6.2%
Cross-over 25.8%
Fatal and injury crossover
44.4%
Cross-over 21.0%
All types 25.0%
All Types Increase 2.3%
Fatal and injury Increase 4.1%
Cross-over 35.4
Fatal and injury crossover
35.4
105

Centerline Rumble Strips
Pavement Deterioration Due to Water/Ice Accumulation, and Winter Maintenance Issues
Water and ice accumulation in CLRS grooves may or may not cause accelerated pavement
degradation. Torbic et al. (2009) claimed that several DOTs’ maintenance crews have reported that
heavy traffic would speed pavement deterioration due to the presence of rumble strips and that the
water and ice accumulated in the grooves would crack the pavement. The authors state that these
concerns have not been validated. Moreover, in a survey conducted in 2005, 15 DOTs did not believe
that CLRS cause pavement deterioration due to ice or water accumulation in the grooves (Russell
and Rys 2005). However, a Virginia inspection on the milled CLRS found that approximately 1%
of the strips inspected were deteriorating (Torbic et al. 2009). The reason for the deterioration may
be poor pavement conditions before the installation of CLRS, as found by the following studies.
According to Kirk (2008), the Kentucky Transportation Center (KYTC) held a meeting with
personnel from the Kentucky DOT to investigate if the joint deterioration found on Daniel Boone
Parkway and Mountain Parkway in Kentucky was caused by CLRS. The conclusion was that these
roads had poor pavement performance even before the rumble strip installation. In addition, the
conclusion was that water and ice accumulation in the centerline rumble strip is a non issue. Another
study also suggests that the center joint degradation promoted by CLRS only appears to occur when
the pavement condition is not adequate before the CLRS installation (Knapp and Schmit 2009). The
same authors also conducted a survey about winter maintenance problems caused by CLRS. Seven
of the nine surveyed states indicated that they were not aware of any maintenance problems. Two
states responded that the snow/ice in the CLRS may melt and then refreeze at a time when winter
maintenance activities are no longer occurring. Minnesota DOT engineers anecdotally noted that
more salt appears to be needed along roadway sections with CLRS, which might suggest the need
to reconsider CLRS designs and/or winter maintenance practices.
Regarding the effect of CLRS on winter maintenance and operation activities, additional passes
of snowplow appeared to be needed in Alaska due to the presence of milled CLRS. However, CLRS
may be beneficial because they provide guidance for snowplow drivers (Russell and Rys 2005). In
addition, Hirasawa et al. (2005) claimed that the Japanese CLRS pattern produces sufficient warning
(sound and vibration) for drivers on slushy winter roads, even when the center line was invisible.
The concerns reviewed in this section can be qualified as gaps in research because there is
limited literature about these topics, and a specific scientific investigation is yet to be done in order
to prove or disprove any hypothesis. Results available and presented in this section were obtained
mainly from questionnaires.
Other Users of the Highways
The noise and vibration caused by CLRS may affect bicyclists, motorcyclists, and residents near
highways. The policies on CLRS can play a role to equilibrate the trade-off between safety and
other aspects. Three studies are consistent with the conclusion that CLRS did not appear to be a
safety hazard to motorcyclists (Miller 2008, Hirasawa et al. 2005, Bucko and Khorashadi 2001).
Only one study evaluated the safety effectiveness of CLRS. Miller (2008) investigated 26 of the 29
motorcyclist crashes that occurred in Minnesota after the installation of CLRS and concluded that
those crashes were unrelated to CLRS. An estimate of the safety effectiveness of CLRS regarding
motorcyclists remains a gap in research.
Three studies concluded that the patterns of rumble strips that produce the greatest levels of
noise and vibration for drivers are the least comfortable for bicyclists (Bucko and Khorashadi 2001,
Outcalt 2001, and Elefteriadou et al. 2000). In addition, Torbic (2001) concluded that there is a
linear relationship between bicyclists’ whole-body vibration and comfort. Another study found that
the space that drivers leave between their vehicles and bicyclists is greater along roadway sections
with CLRS as compared with similar situations without CLRS (Zebauers 2005 cited by Knapp and
Schmit 2009).
106

JTRF Volume 50 No. 1, Spring 2011
Several studies have found that rumble strips increase the level of external noise, which may
affect roadside residents. Finley and Miles (2007) concluded that pavement type and rumble strip
dimensions affect the levels of exterior noise. Karkle et al. (in press) concluded that distance, type
of vehicle, and speed of vehicles affect the levels of exterior noise and that at the studied distances
up to 150 ft., the noise caused by a 15-passenger van and a sedan hitting CLRS could disturb
residents. The authors recommended that a minimum distance from houses and businesses should
be considered for installation of CLRS and suggested that 200 ft. of distance from the center of the
roadway should be considered as the minimum. Makarla (2009), based on a survey with a limited
number of roadside residents, suggests that the respondents were willing to accept the levels of noise
generated by the CLRS due to the increase in safety aspects.
The Operational Usage of the Travel Lane by Drivers
CLRS may affect the lateral position, i.e. may cause vehicles to operate closer to the shoulders, the
speed at which the drivers travel and other operational aspects. Several studies found that CLRS
cause drivers to move to the right, farther away from the center line (Torbic et al. 2009). If installed
in conjunction with rumble stripes, drivers appear to position the vehicle closer to the center of lanes
at locations with lane widths as narrow as 11 ft. and shoulder widths of 3 ft. (Finley et al. 2008).
Moreover, the vehicle travel speed does not appear to be changed much by the presence of CLRS
and the passing opportunity maneuvers seems to be unchanged by the presence of CLRS (Miles et
al. 2005).
In addition, CLRS may influence other operational aspects, such as: a) the presence of both CLRS
and shoulder rumble strips on the same roadway may cause drivers to react to the left after hitting
CLRS under drowsiness or inattention condition. (Noyce and Elango [2004]), using a simulated
environment, reported that 27% of the participants initially reacted leftward after encountering
CLRS; and b) CLRS may affect operational aspects of emergency vehicles. This result was not
confirmed in a survey conducted in 2005, which revealed that 17 DOTs had no evidence or opinion
of CLRS causing people to react to the left (Russell and Rys 2005).
The Visibility of Pavement Markings
It is controversial how CLRS affect the visibility of pavement markings. According to Bahar and
Parkhill (2005), there is a debate whether the degradation of the pavement marking visibility occurs
faster if the markings are painted on top of the rumble strips. However, several authors reported that
the visibility of pavement markings placed over rumble strips is higher than over smooth pavement,
especially during wet-night situations (Torbic et al. 2009). The current belief is that CLRS improve
the night visibility of the pavement markings.
METHODOLOGY
A survey was emailed to the 50 state DOTs between April and May 2010, and consisted of 17
questions regarding the following topics: use of CLRS, type of construction and pattern dimensions,
total mileage, placement of CLRS in relation to the longitudinal joint and center line, type of
CLRS application along the longitudinal roadway, type of pavement and policy on depth and age
of pavement, minimum lane and shoulder width requirements for CLRS installation, and concerns
from the public about CLRS.
107

Centerline Rumble Strips
RESULTS AND DISCUSSION
The total response rate of this survey was 60%, or 30 state DOTs. The results are summarized below.
1. Are there any centerline rumble strips installed on your highways (yes or no)?
Among the total of 30 respondents, 90% (n=27) answered “Yes” and 10% (n = three) answered
“No” to this question.
Combining the information from three previous state-of-the-art studies (Russell and Rys 2005,
Richards and Saito 2007, Torbic et al. 2009) with this current survey, the number of state agencies
that have at least once reported the use of CLRS is 36.
2. What is the type of construction used by your agency (milled, rolled, raised, or combination)?
Among the 27 respondents that have reported the use of CLRS, only one state (Florida) does not use
the milled type. Florida has reported the use of only the raised type of CLRS. Two states (Texas and
North Carolina) reported the use of a combination, i.e., both raised and milled types. The other 24
states reported the use of the milled type of CLRS.
3. What are the strip dimensions used by your agency? The length refers as the dimension
perpendicular to the center line and spacing is measured from center to center.
Florida uses a continuous raised pattern with length and width of 2.5 in., height of 0.5 in. and
spacing of 30 in.
Among the states that use the milled CLRS type, the dimensions varied as follows:
• Length: the range was 6 to 24 in., with 16 in. the predominant value used by about 42% (n
= 11) of the respondents.
• Width: the range was 5 to 9 in., with 7 in. the predominant width used by about 85% (n =
22) of the respondents.
• Depth: the range was 0.375 – 0.625 in., with 0.5 in. the predominant depth used by about
73% (n = 19) of the respondents.
• Spacing: the range was 5 to 48 in., with12 in. the predominant spacing used by about 77%
(n = 20) of the respondents.
• Continuous or Alternating: About 65% (n=17) answered continuous, about 19% (n =
five) reported the use of alternating pattern, and about 12% of the respondents use both
continuous and alternating patterns.
• Class of Highway: the answers for this topic varied. Some of the reported classes of
highways were all classes, rural undivided and rural two-lane arterial.
4. How many miles are there installed by type of highway and dimensions?
Responses varied from three miles (Delaware) to 3,200 miles (Pennsylvania) as shown in Table 2.
The total mileage reported was approximately 11,333. This number does not include the states of
Colorado and Texas that did not report the number of CLRS miles installed.
5. Where are the rumble strips installed in relation to the longitudinal joint and centerline (CLRS
completely within pavement markings, CLRS extended into the travel lane, CLRS on either side
of pavement markings)?
Among the 27 states using CLRS, about 67% (n=18) answered that CLRS are installed completely
within pavement markings. About 45% (n=12) answered CLRS extended into the travel lane and
about 15% answered CLRS on either side of pavement markings. Some of the states reported more
than one type of CLRS placement.
108

Table 2: Number of Miles of CLRS per State
State # Miles State # Miles
AK 118 MI 3,000
AR 74 MN 30
AZ 174 MS 400
CO Unknown MO 700
DE 3 NE 300
FL 68 NC 32
HI 10 NH < 100
ID 268.28 OK 9.25
IA 60 OR 93
KS 232 PA 3,200
KY 190 TX Unknown
LA 408 VA 18.5
MD 412 WA 1,425
ME 7 - 8 Total Approx. 11,333
JTRF Volume 50 No. 1, Spring 2011
6. Where are the CLRS installed in relation to longitudinal roadway (continuous or specific
locations)?
Among the states using CLRS, about 89% (n=24) install them in a continuous manner. Only 18.5%
(n = five) of the states reported the use of CLRS at specific locations such as curves and no passing
zones. Some of the states reported both alternatives.
7. In what type of pavement has your agency installed centerline rumble strips (only asphalt, only
concrete, or both)? Do you have any policy regarding depth and age of the pavement?
About 74% (n=20) of the respondents reported the use of CLRS only on asphalt pavements. About
26% (n = seven) reported the use of CLRS on both asphalt and concrete pavements. The guidelines
regarding the age and minimum depth of the pavement for installation of CLRS are summarized in
Table 3. Examples of guidelines are given below.
• Kansas: CLRS are installed in asphalt pavement surfaces 1.5 in. or more in depth. Age of
pavement is not addressed in the policy. However, they are typically installed as part of
resurfacing projects.
• Pennsylvania: CLRS should not be installed on existing concrete pavements with overlay
less than 2 ½ in. depth. New pavements (less than one-year-old) should present a minimum
1½ in. depth and existing concrete pavements should not have overlays less than 2.5 in. in
depth for installation of CLRS. The pavement should be in sufficiently good condition, as
determined by the district, to effectively accept the milling process without deteriorating.
Otherwise the pavement needs to be upgraded prior to milling.
• Washington has no specific policy. However, the policy reads: “Ensure that the pavement
is structurally adequate to support milled rumble strips. Consult the Region Materials
Engineer to verify pavement adequacies.”
109

Centerline Rumble Strips
Table 3: Guidelines Regarding Age and Depth of Pavement
State Min. Pavement Depth (in.) Min. Pavement Age (years)
AK 2 No
DE Requires consultation of pavement management section
IA 2.5 7
KS 1.5 No
KY Pavement in good condition
LA 2 ≥ 10
MD Pavement in good condition
MI Engineering judgment
MN Engineering judgment
MS Considering for new pavement in future
MO 1.75 New overlays
NE No New Pavement
OR Pavement in good condition
PA 1.5 Older than 1 year
TX 2 No
WA Pavement is structurally adequate
A supplementary question was sent to the seven state DOTs that reported the installation of
CLRS on concrete pavement. This question asked the state DOTs about their experience and if they
have any center joint deterioration caused by CLRS on concrete pavements. The answers are given
below.
• Texas: “I have not heard of any reports of pavement deterioration caused by CLRS. Most
of our centerline rumble strips are installed on hot mixed asphaltic surfaces and we have
also not had any negative pavement reports.”
• Nebraska: “We do not place rumble strips on the joint. We place them on the south side of
east-west highways and the east side of north south highways to match our paint striping.”
• Iowa: “We have yet to install any on PCC pavement.”
• Idaho: “I haven’t heard of any deterioration yet, but we are fairly new to the installations. We
may know more in a few years.”
• Missouri: “To date, I am not aware of joint deterioration due to the CLRS with our concrete
pavements. As I indicated previously, we have installed the CLRS more in the last year
or two. This may be an issue more after a few years, but currently we do not seem to be
having issues.”
• Colorado: “I have not seen or heard of any deterioration of the concrete joints, but I have
not inspected them for such an occurrence.”
• Michigan: “I can tell you that we have very little experience with CLRS on concrete, but
what I heard recently from two of our regions is that milling on the CL joint on an old PCC
pavement is a bad idea. We will be changing our specifications to reflect that.”
8. Is there a minimum lane width requirement for the installation of centerline rumble strips (yes
or no, elaborate)?
About 67% (n=18) of the respondents answered “Yes” to this question and about 33% (n = nine)
did not report a lane width requirement. Some states have suggestions or guidelines rather than
requirements. Table 4 shows the lane width values reported by the respondents.
110

Table 4: Minimum Lane Width for Installation of CLRS
State Min. Lane Width (feet)
AK Requires Lane + Shoulder ≥ 14
WA Requires Lane + Shoulder ≥ 12
MI, MO, PA Require Roadway ≥ 20
DE, MD Require 10
HI, KY, LA Require 11
NE Requires 12
MN Proposal to Require 12
NC Suggests 10
IA, TX Suggest 11
AZ Suggests 12
OK Experimented 12
JTRF Volume 50 No. 1, Spring 2011
9. Is there a minimum shoulder width requirement for installation of centerline rumble strips (yes
or no, elaborate)?
About 70% (n=19) answered “No” to this question. About 30% (n = eight) of the respondents
have a minimum shoulder width requirement for the installation of CLRS. Some states have a
suggested value rather than a requirement. Table 5 shows the shoulder width values reported by the
respondents.
Table 5: Minimum Shoulder Width for Installation of CLRS
State Min. Shoulder Width (feet)
WA Requires Lane + Shoulder ≥ 12
AK Requires Lane + Shoulder ≥ 14
KS Requires 3. Less is allowed to provide continuity
MN Proposal to require 2
AZ Suggests 4
MO Suggests 4
IA Eng. Judgment
OK Experimental sites with 8 shoulder
10. Are there both centerline rumble strips and shoulder rumble strips along the same roadway?
(yes or no, number of miles)?
About 74% of the respondents have installed both CLRS and SRS along the same roadway. The total
number of miles reported for this dual application was approximately 1,600. Some states answered
“Yes” to this question, but did not report the number of miles. Seven states answered “No” to this
question.
11. Are there both centerline rumble strips and edge line rumble strips (also referred as rumble
stripes) along the same roadway (yes or no, number of miles)?
About 33% (n = nine) of the respondents answered “Yes” to this question. The total number of miles
for this type of dual application was 722. The other 18 states answered “No” to this question.
111

Centerline Rumble Strips
12. If you have answered yes on the previous question, has your agency installed both centerline
rumble strips and edge line rumble strips in sections of highway with narrow (width less than
3 feet) or no shoulder?
Only three states (MS, OK, and WA) reported that they have installed dual application on sections
of highways with narrow or no shoulder. Only Washington reported the number of miles (less than
one mile for this case).
13. Are there other requirements for installation of centerline rumble strips (traffic volume, crash
rate, traffic volume, etc)?
About 52% (n=14) of the respondents have other requirements such as crash rates, minimum AADT,
and speed limit for installation of CLRS. For instance, Texas has the following requirements:
“Apply CLRS in roadways with high-incidence crash rate with regard to head-on, opposite direction
sideswipe and/or single vehicle cross-over crashes as a result of inattentive drivers or impaired
visibility of pavement markings during adverse weather; CLRS shall not be milled or rolled into
bridge decks; breaks in the CLRS will start at least 50 ft. and no more than 150 ft. prior to each
approach for the following instances: bridges, intersections, and driveways with high usage or large
trucks; CLRS may be installed along the edge line delineating pavement stripes for two-way left
turn lanes (TWLTL). The TWLTL should have at least a 14-ft. width from the outside edges of the
solid edge lines, and the CLRS will be reduced to 12 in. in width for each edge line. Consider noise
impacts when the installation is near residential areas, schools, and churches. A minimum of 3/18 in.
depth of milled CLRS or rolled CLRS may be considered in these areas. Posted speed limit should
be greater or equal to 45 mph.”
14. Does your agency have a written policy or guidelines for the installation of centerline rumble
strips (yes or no)?
About 63% (n = 17) of the respondents reported that they have some type of written policy or
guidelines for the installation of CLRS. About 37% (n = 10) of the respondents answered “No” to
this question.
15. Has your agency performed a before-and-after study to evaluate the effectiveness of centerline
rumble strips and/or edge line rumble strips (yes or no)?
About 52% (n=14) of the respondents reported that they have, at least anecdotally, performed a
before-and-after safety evaluation of CLRS. About 48% (n=13) of the respondents answered “No”
to this question.
16. Has your agency received any concerns from the public about vehicles hydroplaning due to the
contact with rumble strips?
Only one state (Kansas) reported that only one person has presented a concern about vehicles
hydroplaning after hitting CLRS.
17. Has your agency received other type of concerns from the public about centerline rumble strips
(yes or no, elaborate)?
About 70% (n=19) of the respondents have received concerns from the public regarding CLRS. The
causes of concerns cited were: roadside residents about external noise (n=11), motorcyclists (n=11),
bicyclists (n = three), pavement deterioration (n = two), lack of advance signing of treated sections
(n = one), and snow and ice removal maintenance issues (n = one). Other eight states did not report
any kind of concern received from the public.
Based on the results found in this current survey and in the literature review, it is possible to
summarize the gaps in research and good practices involving the use of CLRS. Good practices are
given below.
• For enhancing the safety effectiveness of CLRS: adopt a minimum AADT (DOTs responses
ranged between 1500 and 3000), a minimum speed (DOTs responses ranged between 40
112

JTRF Volume 50 No. 1, Spring 2011
and 55 mph), a minimum crash rate for the installation of CLRS, a minimum lane width
(DOTs responses ranged between 10 and 12 ft.), and a minimum shoulder width (DOTs
reported two to four feet). In addition, install CLRS in roadways continuously in nopassing
and passing zones, but discontinue the use of CLRS at intersections and at bridge
decks, and adopt a pattern that is able to generate approximately 10 dBA above the ambient
in-vehicle sound level. The predominant pattern in the country (length=16 in., width=7 in.,
depth=0.5 in. and spacing=12 in.) has this characteristic (Miles and Finley 2007). Thus,
this pattern is recommended.
• To avoid potential pavement deterioration caused by CLRS, good practices include: install
CLRS only on new construction or overlays; adopt a minimum pavement depth to install
CLRS (DOTs responses ranged between 1.5 and 2.5 in.). Do not install CLRS if the center
joint is not in good condition (use engineering judgment).
• A widely applied practice to reduce the impact of CLRS on winter maintenance activities
is to avoid the raised type of CLRS in areas where snow is frequent.
• Bicyclists are not expected to hit CLRS very often. However, an intermittent gap in the
spacing of CLRS may help bicyclists to cross the travel lane when needed.
• External noise issues may be addressed by the adoption of a minimum distance from houses
or businesses to install CLRS. Karkle et al. (in press) recommended 200 ft. of distance, but
semi-trucks were not considered in the study.
• To reduce the potential impact of CLRS on vehicles’ position on the travel lane, good
practices include: adopt a minimum shoulder and lane width for installation of CLRS
(DOTs reported lane widths ranging from 10 to 12 ft. and shoulder widths ranging from
two to 4 ft.). Utilize CLRS in conjunction with “rumble stripes” when technically feasible,
since one study showed that CLRS in conjunction with “rumble stripes” resulted in drivers
positioning the vehicle closer to the center of lanes (safer condition) at locations with lane
widths as narrow as 11 ft. and shoulder widths of 3 ft. (Finley et al. 2008).
• In order to avoid potential drivers’ mistakes on initial reactions after hitting CLRS, when
CLRS are installed in conjunction with shoulder rumble strips (SRS) on the same roadway,
different patterns of CLRS and SRS should be used.
• Other factors suggested for inclusion in CLRS installation guidance found in the reviewed
literature were: type of roadway, location of roadway, local and regional conditions,
roadway alignment, consistency within a state, and experience of others (Russell and Rys,
2005). Furthermore, Carlson and Miles (2003) recommended that CRLS may be installed
along the edge line delineating pavement stripes for two-way left turn lanes.
The gaps in knowledge associated with CLRS are: to determine the optimum dimensions for
CLRS pattern, to determine the effects of CLRS on the visibility of pavement markings, to estimate
the safety effectiveness of CLRS regarding motorcyclists, and to verify the effects of CLRS on
pavement deterioration rates.
CONCLUSIONS
This paper presented the most recent survey about the DOT policies and practices regarding CLRS.
The use of CLRS has grown over the years. In 2005, the total mileage of CLRS installed in the
United States was 2,403 miles (Richards and Saito 2007). This current survey found a total mileage
of approximately 11,333 miles (not including the states of Texas and Colorado), which represents
an increase of about 372% over five years. The state DOTs are in the process of implementing
written policies or guidelines for installation of CLRS. In 2006 only seven U.S. states had written
policies or guidelines (Torbic et al. 2009). This survey reported that 17 states have written policies
or guidelines. According to survey results, the milled type of CLRS construction is the predominant
type, and the CLRS predominant pattern dimensions are: length 16 in., width 7 in., depth 0.5 in.,
spacing 12 in., continuous. This pattern is recommended since it produces sufficient amount of
113

Centerline Rumble Strips
noise to alert drivers. Moreover, the installation of CLRS on only asphalt pavement is predominant.
Among the states that use CLRS on concrete pavements, the center joint deterioration appears not
to be an important issue. This result is consistent with the literature review. Some previously cited
studies have reported that pavement deterioration after the installation of CLRS seems to occur on
roads that had poor pavement conditions before the CLRS application. Several state DOTs made the
recommendation to investigate the condition of the pavement and to install CLRS only on sections
with pavement in good condition.
The combination of CLRS and rumble strips is rarely used on sections of highways with narrow
or no shoulder, despite the results that drivers appear to position the vehicle closer to the center of
lanes at locations with lane widths as narrow as 11 ft. and shoulder width of 3 ft. (Finley et al. 2008).
The main causes of concerns received from the public regarding CLRS are the external noise
produced by them that may disturb roadside residents and from motorcyclists, although some
published results from the literature state that CLRS do not have a negative effect on motorcyclists.
Centerline rumble strips are an efficient countermeasure to reduce cross-over crashes. The
policies and guidelines for CLRS installation are not very consistent among the states using them.
Therefore, a list of good practices was given in this study. It can be useful in providing guidance for
future applications of CLRS.
Future research may be performed on the gaps in research topics summarized by this study,
which includes: to determine the optimum dimensions for CLRS pattern, to determine the effects
of CLRS on the visibility of pavement markings, to estimate the safety effectiveness of CLRS
regarding motorcyclists, and to verify the effects of CLRS on pavement deterioration rates.
References
AECOM Transportation. Safety Enhancement Evaluation of Ground-in Centerline Rumble Strips.
Report to Arizona Department of Transportation, Phoenix, AZ, 2008.
Briese, M. Safety Effects of Centerline Rumble Strips in Minnesota. Report Number MN/RC-2008-
44, Minnesota Department of Transportation, St. Paul, MN, 2006.
Bahar, G.J. and M. Parkhill. Synthesis of Practices for the Implementation of Centerline Rumble
Strips. Transportation Association of Canada (TAC), 2005.
Bucko, T.R. and A. Khorashadi. Evaluation of Milled-In Rumble Strips, Rolled-In Rumble Strips
and Audible Edge Stripe. Office of Transportation Safety and Research, California Department of
Transportation, 2001.
Carlson, P.J. and J.D. Miles. Effectiveness of Rumble Strips on Texas Highways: First Year Report.
Report FHWA/TX-05/0-4472-1, Texas Department of Transportation, Austin, TX, 2003.
Delaware Department of Transportation. Centerline Rumble Strips: The Delaware Experience.
http://www.deldot.gov; accessed in July 2008.
Chen, C., E.O. Darko, and T.N. Richardson. “Optimal Continuous Shoulder Rumble Strips and the
Effects on Highway Safety and the Economy.” ITE Journal 5 (73), (2003): 30-41. http://www.ite.
org/itejournal/journalsearch.cfm.
Elefteriadou, L., M. El-Gindy, D. Torbic, P. Garvey, A. Homan, Z. Jiang, B. Pecheux, and R. Tallon.
Bicycle-Tolerable Shoulder Rumble Strip. The Pennsylvania State University, The Pennsylvania
Transportation Institute, Report Number: PTI 2K15, 2000.
114

JTRF Volume 50 No. 1, Spring 2011
Federal Highway Administration. Rumble Strips. FHWA Publication No: FHWA-RC-BAL-04-0015,
Washington, D.C., 2003.
Finley, M.D. and J.D. Miles. Exterior Noise Created by Vehicles Traveling over Rumble Strips.
Transportation Research Board 86 th Annual Meeting, Transportation Research Board, National
Research Council, Washington, D.C., 2007.
Finley, M.D., D.S. Funkhouser, and M.A. Brewer. Studies to Determine the Operational Effects of
Shoulder and Centerline Rumble Strips on Two- Lane Undivided Roadways. Report 0-5577-1, Texas
Department of Transportation, Austin, TX, 2008.
Fitzpatrick, K., K. Balke, D.W. Harwood, and I.B. Anderson. Accident Mitigation Guide for
Congested Rural Two-Lane Highways. NCHRP Report 440, Transportation Research Board,
National Research Council, Washington, D.C., 2000.
Golembiewski, G., R. Haas, B. Katz, and T. Bryer. Evaluation of the Effectiveness of PennDOT’s Low
Cost Safety Improvement Program (LCSIP). Publication FHWA-PA-2008-16-070308. Pennsylvania
Department of Transportation, 2008.
Hirasawa, M., K. Saito, and M. Asano. “Study on Development and Practical Use of Rumble Strips
as a New Measure for Highway Safety.” Journal of the Eastern Asia Society for Transportation
Studies (6), (2005): 3697–3712. http://www.easts.info/on-line/journal_06/3697.pdf.
Kar, K. and R.S. Weeks. “The Sound of Safety.” Public Roads 4 (72), (2009): 10–16.
Karkle, D.E., M.J. Rys, and E.R. Russell. Evaluation of Centerline Rumble Strips for Prevention of
Highway Cross-over Accidents in Kansas. Proceedings of the 2009 Mid-Continent Transportation
Research Symposium, Ames, IA, 2009.
Karkle, D.E., M.J. Rys, and E.R. Russell. “Centerline Rumble Strips: A Study of Exterior Noise.”
Journal of Transportation Engineering. (in press). http://scitation.aip.org/getpdf/servlet/GetPDFSer
vlet?filetype=pdf&id=JTPEXX000001000001000162000001&idtype=cvips&prog=normal
Kirk, A. Evaluation of the Effectiveness of Pavement Rumble Strips. Research Report KTC-08-04/
SPR319-06-1F, Kentucky Transportation Center, University of Kentucky, Lexington, KY, 2008.
Knapp, K.K. and M. Schmit. Assessment of Centerline Rumble Strips in Minnesota: Executive
Summary and Project Task Summary Attachments. Center for Excellence in Rural Safety, State and
Local Policy Program, Hubert H. Humphrey Institute of Public Affairs, University of Minnesota,
Minneapolis, MN, 2009.
Lipscomb, D.M. Auditory Perceptual Factors Influencing the Ability of Train Horns. Third
International Symposium on Railroad-Highway Grade Crossing Research and Safety, 1995: 195.
Makarla, R. “Evaluation of External Noise Produced by Vehicles Crossing Over Centerline Rumble
Strips on Undivided Highways in Kansas.” Thesis (M.S.), Kansas State University, 2009.
Miles, J. D., P. J. Carlson, M.P. Pratt, and T.D. Thompson. Traffic Operational Impacts of Transverse,
Centerline, and Edgeline Rumble Strips. Report No. FHWA/TX-05/0-4472-2. Texas Department of
Transportation, 2005.
Miles, J.D. and M.D. Finley. Evaluation of Factors that Impact the Effectiveness of Rumble Strip
Design. Presented at the 86 th Annual Meeting of the Transportation Research Board, Washington
D.C., 2007.
115

Centerline Rumble Strips
Miller, K.W. Effects of Center-Line Rumble Strips on Non-Conventional Vehicles. Minnesota
Department of Transportation, 2008.
Monsere, C.M. Preliminary Evaluation of the Safety Effectiveness of Centerline (Median) Rumble
Strips in Oregon. Presented at the Institute of Transportation Engineers Quad Conference, Seattle,
WA, 2002.
National Highway Administration (NHTSA). “FARS Database.” Accessed January 25, 2011. http://
www-fars.nhtsa.dot.gov, 2009.
Noyce, D.A. and V.V. Elango. “Safety Evaluation of Centerline Rumble Strips: Crash and Driver
Behavior Analysis.” Transportation Research Record: Journal of the Transportation Research
Board 1862, (2004): 44-53.
Outcalt, W. Centerline Rumble Strips. Report No. CDOT-DTD-R-2001-8, Colorado Department of
Transportation, Denver, CO., 2001.
Persaud, B.N., R.A. Retting, and C. Lyon. Crash Reduction Following Installation of Centerline
Rumble Strips on Rural Two-Lane Roads. Insurance Institute for Highway Safety, Arlington, VA,
2003.
Richards, S.J.N. and M. Saito. “State-of-the-Practice and Issues Surrounding Centerline Rumble
Strips.” WIT Transactions on the Built Environment (94), (2007): 36-45.
Russell, E.R. and M.J. Rys. Centerline Rumble Strips. NCHRP Synthesis 339, Transportation
Research Board, National Research Council, Washington, D.C., 2005.
Rys, M.J., L. Gardner and E.R. Russell. “Evaluation of Football Shaped Rumble Strips Versus
Rectangular Rumble Strips.” Journal of the Transportation Research Forum, 47 (2), (2008): 41-54.
Torbic, D. J. “Comfort and Controllability of Bicycles as a Function of Rumble Strip Design.”
Dissertation (Ph.D). The Pennsylvania State University, 2001.
Torbic, D.J., J.M. Hutton, C.D. Bokenkroger, K.M. Bauer, D.W. Harwood, D.K. Gilmore, J.M.
Dunn, J.J. Ronchetto, E.T. Donnell, H.J. Sommer III, P. Garvey, B. Persaud, and C. Lyon. Guidance
for the Design and Application of Shoulder and Centerline Rumble Strips. NCHRP Report 641,
Transportation Research Board, National Research Council, Washington, D.C., 2009.
Zebauers, V. Centerline Rumble Strips Help Cars and Bicyclists. In the 2005 Annual Meeting and
Exhibit Compendium of Technical Papers, Institute of Transportation Engineers, Washington, D.C.,
2005.
116

JTRF Volume 50 No. 1, Spring 2011
Daniel E. Karkle is a Ph.D. student in the Department of Industrial and Manufacturing Systems
Engineering at Kansas State University. He received his B.S. degree in civil production engineering
in Brazil.
Margaret J. Rys is an associate professor in the Department of Industrial and Manufacturing
Systems Engineering at Kansas State University. She obtained her integrated B.S/M.S degree
from the Technical University of Wroclaw, Poland, in 1979 and M.S. (1986) and Ph.D. (1989)
from Kansas State University, all in industrial engineering. She has almost 20 years of experience
conducting research and teaching courses in human factors engineering, quality, engineering
economy, statistics, and safety. During the past 20 years she has been principal or co-principal
investigator on more than 40 projects and authored and co-authored more than 50 journal papers.
Eugene R. Russell, PE, is a professor emeritus of civil engineering at Kansas State University and
still conducts research on a part-time basis. He has 50 years of transportation and traffic engineering
experience, including 42 in academia. He has directed more than 80 research projects covering a
wide range of highway engineering and highway safety issues, authored or co-authored more than
100 technical papers, and more than 100 presentations at U.S. and international conferences.
117

118

Book Review
Wilner, Frank N. Understanding the Railway Labor Act. Omaha: NE: Simmons Boardman
Books, Inc., 2009. ISBN 978-0911382594.
Understanding the Railway Labor Act
by Gordon P. MacDougall
This is the author’s second book on the subject of the Railway Labor Act (1926), 45 U.S.C. 151,
et seq. The earlier book (Wilner 1991), referred to below as RLA-Dilemma, was the subject of
comment by this reviewer in this Journal (MacDougall 1993).
Mr. Wilner’s latest 172-page textual effort, Understanding the Railway Labor Act (RLA-
Understanding), is preceded (after his 5-page preface), by a 10-page introduction from Lawrence H.
Kaufman, a long-time public relations (public affairs) spokesman for the Association of American
Railroads, and several individual major rail carriers. Following the end of the Wilner text are four
invited essays “In Defense of the RLA,” from RLA sympathizers with a recognized background in
railroad labor-management negotiations or litigation under the RLA. 1
The author’s preface and 11 chapters text of RLA-Understanding are accompanied by 58 pages
of endnotes, bearing no less than 1,018 small-type size entries, 2 directing the reader to source
material, chiefly newspaper or other press accounts, but also to some published books and treatises. 3
What does Wilner’s second effort on the Railway Labor Act, RLA-Understanding, add to or
change in his RLA-Dilemma, written some 18 years earlier? The author is a prolific writer and a very
hard worker on the computer. He is a known entity in the transportation field, but perhaps more as a
journalist or publicist rather than as an expert or authority in the field of labor relations. Mr. Wilner’s
approach is not necessarily favored by some practitioners on the labor side, 4 and this reviewer
opined (MacDougall 1993) RLA-Dilemma takes a partisan view of some features of managementlabor
relations in the railroad industry. The author does not appear to have changed his views in the
intervening years, but the current RLA-Understanding, unlike RLA-Dilemma, does not take sides on
sensitive subjects, such as secondary boycotts, lengthening the terms of National Mediation Board
members, or forcing single representation for each craft or class on a merged rail system. Moreover,
Wilner, in his current book, does not advance recommendations for “improving” the RLA. However,
the author does mention some additional developments such as “interest” bargaining in his preface, 5
and includes a more comprehensive discussion of the railroad strike insurance plan history, among
other embellishments not developed in his earlier work.
I suggest RLA-Understanding does have considerable value, with the Lawrence Kaufman
introduction and the four essays, and a thorough compendium of reference citations for those readers
seeking source material on various aspects of the Railway Labor Act and its workings.
The four invited essays, themselves, are of great value to an understanding of the RLA,
particularly in dealing not only with its origins and history, but also with the practical mechanics of
implementation on individual properties and with national handling. The four presenters have wide
and varied backgrounds in rail labor-management relations. The four essays are ably written, short
and precise, and should be easily understandable even for those with limited prior background in rail
labor relations. Mr. Wilner is to be commended for his ability to gather these four authors, each with
substantial background and standing in the rail industry, under one book cover.
119

Railway Labor Act
Endnotes
1. Harry R. Hoglander (Member & Former Chmn., National Mediation Board), pp. 173-78;
Francis X. Quinn (Arbitrator), pp. 179-82 & App. 183-84; Kenneth R. Peifer (Ret. V.P., Labor
Relations, CSX), pp. 185-91; Clinton J. Miller (Gen. Counsel, UTU), pp. 192-94.
2. An additional 15 endnotes cover the Kaufman introduction and the four outside essays.
3. The author's earlier RLA-Dilemma 118-page book is also replete with footnotes–some 436 in
number.
4. See reviews by William G. Mahoney, longtime counsel for Railway Labor Executives Assn.
and other labor organizations (Mahoney, 1991, and Mahoney, 2010).
5. See pp. ix-xi of RLA-Understanding. Interest-bargaining puts into play legislative support for
carrier objectives, which Wilner says UTU has engaged in, in the trade-off of economic benefits
between carriers and unions. It should not be confused with “interest-arbitration.” See Elkouri
and Elkouri, 2002 (p. 1348), 2008 (p. 511), and 2010 (p. 533).
References
Elkouri, Frank and Edna Asper Elkouri. How Arbitration Works. 6th ed. A.M. Rubin, Ed. Bureau of
National Affairs, 2002.
Elkouri, Frank and Edna Asper Elkouri. How Arbitration Works. Supplement. Bureau of National
Affairs, 2008.
Elkouri, Frank and Edna Asper Elkouri. How Arbitration Works. Supplement. Bureau of National
Affairs, 2010.
MacDougall, Gordon. “Book Review.” Journal of the Transportation Research Forum 33, (1993):
135 ff.
Mahoney, William G. “Book Review.” ICC Practitioners Journal 59, (1991): 67.
Mahoney, William G. “Book Review.” Journal of Transportation Law, Logistics, and Policy 77,
(2010): 181.
Wilner, Frank N. The Railway Labor Act and the Dilemma of Labor Relations. Simmons-Boardman,
Omaha, NE, 1991.
Gordon MacDougall is a transportation attorney practicing in Washington, D.C., and a member of
the Wisconsin, D.C., and New York bars. He is a founding member of TRF and serves at present as
Counsel for the TRF Foundation.
120

Transportation Research Forum
Statement of Purpose
The Transportation Research Forum is an independent organization of transportation professionals.
Its purpose is to provide an impartial meeting ground for carriers, shippers, government officials,
consultants, university researchers, suppliers, and others seeking an exchange of information and
ideas related to both passenger and freight transportation. The Forum provides pertinent and timely
information to those who conduct research and those who use and benefit from research.
The exchange of information and ideas is accomplished through international, national, and
local TRF meetings and by publication of professional papers related to numerous transportation
topics.
The TRF encompasses all modes of transport and the entire range of disciplines relevant to
transportation, including:
Economics Urban Transportation and Planning
Marketing and Pricing Government Policy
Financial Controls and Analysis Equipment Supply
Labor and Employee Relations Regulation
Carrier Management Safety
Organization and Planning Environment and Energy
Technology and Engineering Intermodal Transportation
Transportation and Supply Chain Management
History and Organization
A small group of transportation researchers in New York started the Transportation Research Forum
in March 1958. Monthly luncheon meetings were established at that time and still continue. The
first organizing meeting of the American Transportation Research Forum was held in St. Louis,
Missouri, in December 1960. The New York Transportation Research Forum sponsored the meeting
and became the founding chapter of the ATRF. The Lake Erie, Washington D.C., and Chicago
chapters were organized soon after and were later joined by chapters in other cities around the
United States. TRF currently has about 300 members.
With the expansion of the organization in Canada, the name was shortened to Transportation
Research Forum. The Canadian Transportation Forum now has approximately 300 members.
TRF organizations have also been established in Australia and Israel. In addition, an International
Chapter was organized for TRF members interested particularly in international transportation and
transportation in countries other than the United States and Canada.
Interest in specific transportation-related areas has recently encouraged some members of TRF
to form other special interest chapters, which do not have geographical boundaries – Agricultural
and Rural Transportation, High-Speed Ground Transportation, and Aviation. TRF members may
belong to as many geographical and special interest chapters as they wish.
A student membership category is provided for undergraduate and graduate students who are
interested in the field of transportation. Student members receive the same publications and services
as other TRF members.
121

Annual Meetings
In addition to monthly meetings of the local chapters, national meetings have been held every year
since TRF’s first meeting in 1960. Annual meetings generally last three days with 25 to 35 sessions.
They are held in various locations in the United States and Canada, usually in the spring. The
Canadian TRF also holds an annual meeting, usually in the spring.
Each year at its annual meeting the TRF presents an award for the best graduate student paper.
Recognition is also given by TRF annually to an individual for Distinguished Transportation
Research and to the best paper in agriculture and rural transportation.
Annual TRF meetings generally include the following features:
• Members are addressed by prominent speakers from government, industry, and
academia.
• Speakers typically summarize (not read) their papers, then discuss the principal
points with the members.
• Members are encouraged to participate actively in any session; sufficient time is
allotted for discussion of each paper.
• Some sessions are organized as debates or panel discussions.
122

TRF Council
President
Alan R. Bender
Embry-Riddle Aeronautical University
Executive Vice President
Kristen Monaco
California State University (Long Beach)
Vice President – Program
Jack Ventura
Retired, Surface Transportation Board
Vice President-Elect – Program
Art Guzetti
American Public Transportation Association
Vice President – Chapter Relations
Joe Schwieterman
DePaul University
Vice President – Membership
David Ripplinger
North Dakota State University
Vice President – Academic Affairs
Eric Jessup
Washington State University
Vice President – Public Relations
Joshua Schank
Eno Transportation Foundation
Vice President – International Affairs
Paul Bingham
Wilbur Smith Associates
Secretary
Paul K. Gessner
Gessner Transportation Consulting LLC
Treasurer
Carl Scheraga
Fairfield University
Counsel
John A. DeVierno
Attorney at Law, Washington, D.C
Immediate Past President
B. Starr McMullen
Oregon State University
Chicago Chapter President
Joe Schwieterman
DePaul University
New York Chapter President
Robert L. James
Port Authority of NY & NJ
Washington, D.C. Chapter President
Stephen J. Thompson
Michael Babcock
JTRF Co-Editor
Kansas State University
Kofi Obeng
JTRF Co-Editor
North Carolina A&T State University
TRF Foundation Officers
President
Jack S. Ventura
Retired, Surface Transportation Board
Vice President
Aaron J. Gellman
Northwestern University
Secretary
Tillman H. Neuner
Consultant
Treasurer
Marcin Skomial
Surface Transportation Board
General Counsel
Gordon MacDougall
Attorney
TRF Office
www.trforum.org
123

Past Presidents
2010 B. Starr McMullen
2009 Richard Gritta
2008 Kenneth Button
2007 John (Jack) V. Wells
2006 Anthony M. Pagano
2005 Scott E. Tarry
2004 John S. Strong
2003 C. Gregory Bereskin
2002 Martin Dresner
2001 Mark R. Dayton
2000 Richard S. Golaszewski
1999 Aaron J. Gellman
1998 Richard Beilock
1997 Robert Harrison
1996 Clinton V. Oster, Jr.
1995 Paul K. Gessner
1994 Russell B. Capelle, Jr.
1993 Louis A. LeBlanc
1992 Stephen J. Thompson
1991 Joanne F. Casey
1990 Frederick J. Beier
1989 Thomas N. Harvey
1988 Joedy W. Cambridge
1987 Frederick C. Dunbar
1986 Carl D. Martland
1985 Allan D. Schuster
1984 Douglas McKelvey
1983 William B. Tye
1982 Michael S. Bronzini
1981 Jay A. Smith, Jr.
1980 Samual E. Eastman
1979 Lana R. Batts
1978 Carl J. Liba
1977 Gerald Kraft
1976 Edward Morlok
1975 Barry A. Brune
1974 Richard Shackson
1973 Harvey M. Romoff
1972 Arthur Todd
1971 Herbert E. Bixler
1970 Paul H. Banner
1969 George W. Wilson
1968 Donald P. MacKinnon
1967 David L. Glickman
1966 Edward Margolin
1965 Robert A. Bandeen
1964 Gayton Germane
1963 Herbert O. Whitten
1962 Herbert O. Whitten
1961 Herbert O. Whitten
1960 Herbert O. Whitten
1959 John Ingram (TRF of NY)
1958 Herbert O. Whitten (TRF of NY)
124
Recipients of the TRF Distinguished
Transportation Researcher Award
2011 Martin Wachs
2010 Clifford Winston
2009 Daniel McFadden
2008 Steven A. Morrison
2007 José A. Gomez-Ibanez
2006 Tae H. Oum
2005 Kenneth Button
2004 Kenneth Small
2000 Michael E. Levine
1998 Edward K. Morlok
1997 Carl D. Martland
1996 Benjamin J. Allen
1995 D. Philip Locklin
1994 Martin T. Farris
1993 C. Phillip Baumel
1992 Alan A. Altshuler
1990 George W. Wilson, Ph.D.
1989 Sir Alan Arthur Walters, B. Sci., Hon.
D. Soc. Sci.
1988 Karl M. Ruppenthal, Ph.D.
1987 William S. Vickrey, Ph.D.
1986 William J. Harris, D. Sci., Hon. D.
Eng.
1985 John R. Meyer, Ph.D.
1984 Alfred E. Kahn, Ph.D.
1982 W. Edwards Deming, Ph.D.
1979 James C. Nelson, Ph.D.
1978 James R. Nelson, Ph.D.
1977 Lewis K. Sillcox, D. Sci., D. Eng.,
LL.D., D.H.L.
Recipients of the Herbert O. Whitten
TRF Service Award
2011 Anthony Pagano
2010 Captain James R. Carman
2009 John (Jack) Wells
2008 Gordon MacDougall
2007 Robert T. Beard
2006 Gene C. Griffin
2005 Michael W. Babcock
2004 Jack S. Ventura
2000 Paul K. Gessner
1998 Arthur W. Todd
1996 Stephen J. Thompson
1994 Samuel Ewer Eastman
1991 Carl D. Martland, C.E.
1990 Herbert O. Whitten, M.B.A.

Past Editors of JTRF
Wayne K. Talley (1998-2000)
K. Eric Wolfe (1993-1998)
Kevin H. Horn (1993-1998)
Anthony M. Pagano (1987-1993)
Richard Beilock (1987-1993)
Recipients of the TRF Best Paper Award
2011 Alexander Bigazzi and Miguel Figliozzi, A Model and Case Study of the Impacts of
Stochastic Capacity on Freeway Traffic Flow and Benefits Costs
2010 Tony Diana, Predicting Arrival Delays: An Application of Spatial Analysis
2009 Tae H. Oum, Jia Yan, and Chunyan Yu, Ownership Forms Matter for Airport Efficiency:
A Stochastic Frontier Investigation of Worldwide Airports
2008 C. Gregory Bereskin, Railroad Cost Curves Over Thirty Years – What Can They Tell Us?
2007 Rob Konings, Ben-Jaap Pielage, Johan Visser, Bart Wiegmans, “Container Ports and
Their Hinterland: Multimodal Access and Satellite Terminals within the Extended
Gateway Concept”
2006 Ian Savage, Trespassing on the Railroad.
2005 Patrick DeCorla-Souza, A New Public-Private Partnership Model for Road Pricing
Implementation.
2004 Ian Savage and Shannon Mok, Why Has Safety Improved at Rail-Highway Crossings?
2000 Ian Savage, Management Objectives and the Causes of Mass Transportation.
1999 C. Phillip Baumel, Jean-Philippe Gervais, Marty J. McVey, and Takehiro Misawa,
Evaluating the Logistics Economic Impacts of Extending Six Hundred Foot Locks on the
Upper Mississippi River: A Linear Programming Approach.
1998 Carl Scheraga and Patricia M. Poli, Assessing the Relative Efficiency and Quality of
Motor Carrier Maintenance Strategies: An Application of Data Entry Envelopment
Analysis.
1997 Wayne K. Talley and Ann Schwarz-Miller, Motor Bus Deregulation and Racial/Gender
Effects: Earnings and Employment.
1996 Todd Litman, Full Cost Accounting for Transportation Decision Making: Estimates,
Implications and Applications.
1995 Leon N. Moses and Ian Savage, A Cost-Benefit Analysis of United States Motor Carrier
Safety Programs.
1994 Brian Shaffer and James Freeman, Variation in Producer Responses to Automobile Fuel
Economy Mandates.
1993 W. Bruce Allen and Dong Liu, Service Quality and Motor Carrier Costs: An Empirical
Analysis.
1992 Victor E. Eusebio, Stephen J. Rindom, Ali Abderrezak, and John Jay Rosacker, Rail
Branch Lines at Risk: An Application of the Exponential Survival Model on Kansas
Duration Data.
1991 Patrick Little, Joseph M. Sussman, and Carl Martland, Alternative Freight Car
Maintenance Policies with Attractive Reliability/Cost Relationships.
1990 James A. Kling, Curtis M. Grimm, and Thomas M. Corsi, Strategies of Challenging
Airlines at Hub-Dominated Airports.
125

1989 Cathy A. Hamlett, Sherry Brennan, and C. Phillip Baumel, Local Rural Roads: A Policy
Analysis.
1988 John R. Brander, B. A. Cook, and John Rowcroft, Auctioning Runway Slots: Long Term
Implications.
1987 Clinton V. Oster, Jr. and C. Kurt Zorn, Deregulation’s Impact on Airline Safety.
1986 Chulam Sarwar and Dale G. Anderson, Impact of the Staggers Act on Variability and
Uncertainty of Farm Product Prices.
1985 Brian D. Adam and Dale G. Anderson, Implications of the Staggers Rail Act of 1980 for
Level and Variability of Country Elevator Bid Prices.
Donald J. Harmatuck, Back Haul Pricing: Mathematical Programming and Game
Theory Approaches to Allocating Joint Costs.
Jeffrey L. Jordan, Dieudonne Mann, S. E. Pressia, and C. Thai, Managing the
Transportation of Perishable Commodities: Temperature Control and Stacking Patterns.
1984 K. Eric Wolfe, An Examination of Risk Costs Associated with the Movement of Hazardous
Materials.
1983 Denver D. Tolliver, Economics of Density in Railroad Cost-Finding: Applications to Rail
Form A.
1982 Jeffrey Beaulieu, Robert J. Hauser, and C. Phillip Baumel, Inland Waterway User Taxes:
Their Impacts on Corn, Wheat and Soybean Flows and Transport Costs.
126

Journal of the Transportation Research Forum
Guidelines for Manuscript Submission
1. Submit manuscripts by e-mail or hardcopy to either General Editor at:
Michael W. Babcock Kofi Obeng
Co-General Editor, JTRF Co-General Editor, JTRF
Department of Economics Dept. of Economics & Transportation/Logistics
Kansas State University School of Business and Economics
Manhattan, KS 66506 U.S.A. North Carolina A&T University
Phone: (785) 532-4571 Greensboro, NC 27411 U.S.A.
Fax: (785) 532-6919 Phone: (336) 334-7231
Email: mwb@ksu.edu Fax: (336) 334-7093
E-mail: obengk@ncat.edu
2. E-mail submissions must be in MS WORD.
3. Final manuscript and abstract (hard or e-mail copy) must be received as soon as possible.
4. The text of manuscripts is to be double-spaced. Use one side of each page only. Articles are limited
to a maximum length of 30 pages; industry issue papers and comments are limited to a maximum
of 15 pages.
5. The manuscript should have a title page which includes the names, affiliations, address (mailing
and e-mail) and phone numbers of all authors. Brief biographical sketches for all authors should be
included with the manuscript.
6. The abstract should briefly describe the contents, procedures and results of the manuscript, not its
motivation, and should not exceed 100 words.
7. Endnotes are to be used rather than footnotes, used sparingly and placed at the end of the
manuscript. Do NOT use the endnote feature of the word processing software.
8. The Chicago Manual of Style is to be used for endnotes and references. At the end of the
manuscript, complete references are listed alphabetically (not by number) by author surname,
government agency, or association name. Use the following examples.
Book:
Jones, Robert T. Improving Airline Transportation. General Publishing Company, Washington,
D.C., 2002.
Chapter in a Book:
Bresnahan, T.F. “Empirical Studies of Industries with Market Power.” R. Schmalensee and R.
Williq eds. Handbook of Industrial Organization. Amsterdam: North Holland (1989): 1011-
1057.
Journal Article:
Kane, Louis R. “Grain Transport Rate Projections.” Transportation Journal 2, (1999): 20-40.
Journal Article with Multiple Authors:
Chow, G., R. Gritta, and E. Leung. “A Multiple Discriminant Analysis Approach to Gauging
Air Carrier Bankruptcy Propensities: The AIRSCORE Model.” Journal of the Transportation
Research Forum 31 (2), (1991): 371-377.
127

128
Ph.D Dissertation:
Jessup, E.L. “Transportation Optimization Marketing for Commodity Flow, Private Shipper Costs,
and Highway Infrastructure, Impact Analysis.” Dissertation (Ph.D). Washington State University,
1998.
Government Document:
U.S. Senate. Committee on Foreign Relations, Investigations of Mexican Affairs. 2 vols. 66th
Cong., 2nd session, 1919-20.
9. Tables, figures, graphs, and charts should be titled and numbered by an Arabic numeral (e.g.,
Figure 2). All figures and charts should be submitted as separate files. Any graphics or photos must
be 300 dpi and submitted as a separate .tif or .eps file.
10. Headings should adhere to the following style. Place the abstract below the title of the manuscript.
• First level headings should be all caps, left justified.
• Second level headings should be initial caps, left justified.
• Third level headings should be initial caps on the same line as the first paragraph which it heads.
• Do not number the headings.
11. Submit your final copy in “MS WORD” only.
12. For proper preparation of all charts, figures, maps and photos call Beverly Trittin, North Dakota
State University, at (701) 231-7137 or bev.trittin@ndsu.edu prior to submitting the final version of
your paper. The journal is printed in black and white, therefore, all graphics must be submitted in
black and white, or grayscale.
TIPS:
• Photographs to be submitted as 200 dpi or higher resolution (.jpg, .tif , .eps format).
• Images taken from the web are 72 dpi and not print quality. Secure high resolution image
if possible.
• Do not embed graphics in Word – submit as individual files (.jpg, .tif, .eps format).
• Improper graphics or images will be returned to author for correct formatting.
Book Review Information
Books for review should be mailed to:
Dr. Jack Ventura
Book Review Editor
1025 Chiswell Lane
Silver Spring, MD 20901
Phone: (301) 593-1872
Email: jack.ventura@verizon.net
JTRF Information
The Journal of the Transportation Research
Forum is distributed to members of the
Transportation Research Forum.
Copyright 2011
The Transportation Research Forum
All Rights Reserved
ISSN 1046-1469

The Transportation Research Forum gratefully acknowledges the contributions
of sponsoring and sustaining members.
Supporting Members
Michael Babcock, Kansas State University
Marcus Bowman
Paul K. Gessner, Gessner Transportation Consulting LLC
Larry Jenkins, Embry-Riddle Aeronautical University
Jack S. Ventura, Surface Transportation Board
John (Jack) V. Wells, U.S. Department of Transportation