N ATIO'NA 1, COOPERATIV,.

1E HIGHWAY RESEARCH PROGRAM

Roadway Widths for

Low-Traffic-Volume Roads

Transportation. Research Board

National Research Council
TRANSPORTATION RESEARCH BOARD EXECUTIVE COMMITTEE 1994
OFFICERS
Chair: Joseph M. Sussman, JR East Professor and Professor of Civil and Environmental Engineering, Massachusetts Institute of Technology
Vice Chair: Lillian C. Liburdi, Director, Pori Authority, The Port Authority of New York and New Jersey
Executive Director: 77tomas B. Deen, Transportation Research Board

MEMBERS
BRIAN J. L. BERRY, Lloyd Viel Berkner Regental Professor & Chair, Britton Centerfor Development Studies, University of Texas at Dallas
JOHN E. BREEN, The Nasser 1. Al-Rashid Chair in Civil Engineering, )'7te University of Texas at Austin
KIRK BROWN, Secretary, Illinois Department of Transportation
DAVID BURWELL, President, Rails-ro-Trails Conservancy
L. GARY BYRD, Consulting Engineer, Alexandria, Virginia
A. RAY CHAMBERLAIN, Executive Director, Colorado Department of Transporiation (Past Chair, 1993)
RAY W. CLOUGH, Nishkian Professor of Structural Engineering, Emeritus, University of California, Berkeley
RICHARD K. DAVIDSON, Chairman and CEO, Union Pacific Railroad
JAMES C. DELONG, Director ofAviation, Stapleton International Airport, Denver, Colorado
JERRY L. DEPOY, former Vice President, Properties & Facilities, USAir
DELON HAMPTON, Chairman & CEO, Delon Hampton & Associates
DON C. KELLY, Secretary and Commissioner of Highways, Transportation Cabinet, Kentucky
ROBERT KOCHANOWSKI, Executive Director, Southwestern Pennsylvania Regional Planning Commission
JAMES L. LAMMIE, President & CEO, Parsons Brinckerhoff. Inc.
WELLLA,M W. MILLAR, Executive Director, Port Authority ofAllegheny County. Pennsylvania (Past Chair, 1992)
CHARLES P. O'LEARY, JR., Commissioner, New Hampshire Department of Transportation
JUDE W. P. PATIN, Secretary, Louisiana Department of Transportation and Development
NEIL PETERSON, former Executive Director, Los Angeles County Transportation Commission
DARREL RENSINK, Director, Iowa Department of Transportation
JAMES W. vAN LOBEN SELS, Director, California Department of Transportation
C. MICHAEL WALTON, Ernest H. Cockrell Centennial Chair in Engineering and Chairman, Department of Civil Engineering, The University of Texas at Austin
DAVID N. WORMLEY, Dean of Engineering, Pennsylvania State University
HOWARD YERUSALIM, Secretary of Transporiation, Pennsylvania Department of Transportation
ROBERT A. YOUNG III, President, ABF Freight Systems, Inc.

MICHAEL ACOTT, President, National Asphalt.Pavement Association (ex officio)

ROY A. ALLEN, Vice President, Research and Test Department. Association ofAmerican Railroads (ex officio)
ANDREW H. CARD, JR., President and CEO, American Automobile Manufacturers Association
THOMAS J. DONOHUE, President and CEO, American Trucking Associations (ex officio)
FRANCIS B. FRANCOIS, Executive Director, American Association of State Highway and Transportation Officials (ex officio)
JACK R. GILSTRAP, Executive Vice President, American Public Transit Association (ex officio)
ALBERT J. HERBERGER, Maritime Administrator, U.S. Department of Transportation (ex officio)
DAVID R. HINSON, Federal Aviation Administrator, US. Department of Transportation (ex officio)
GORDON J. LINTON, Federal Transit Administrator, US. Department of Transportation (ex officio)
ROSE A. McMURRAY, Research and Special Programs Administrator, U.S. Department of Transportation (ex officio)
JOLENE M. MOLITORIS, Federal Railroad Administrator, U.S. Department of Transportation (ex officio)
RODNEY E. SLATER, Federal Highway Administrator, US. Department of Transporlation (ex officio) ,
HOWARD M. SMOLKIN, National Highway Traffic Safety Administrator, US. Department of Transporiation (ex officio)
ARTHUR E. WILLIAMS, Chief of Engineers and Commander, US. Army Corps of Engineers (ex officio)

NATIONAL COOPERATIVE HIGHWAY RESEARCH PROGRAM

Transportation Research Board Executive Committee Subcommittee for NCHR
JOSEPH M. SUSSMAN, Massachusetts Institute of Technology (Chair) LILLIAN C. LIBURDI, Port Authority of New York and New Jersey
A. RAY CHAMBERLAIN, Colorado Department of Transportation RODNEY E. SLATER, Federal Highway Administration
FRANCIS B. FRANCOIS, American Association of State Highway and L. GARY BYRD, Consulting Engineer
Transportation Officials THOMAS B. DEEN, Transportation Research Board

Field of Design Area of General Design Project Panel C15-12

NORMAN H. ROUSH, West Virginia Department of Highways (Chair) LARRY F. SUTHERLAND, Ohio Department of Transportation
RONALD W. ECK, West Virginia University W. STEPHEN YOUNG, Allegany County, Maryland
JOHN R. HOLT, USDA-Forest Service JUSTIN TRUE, FHWA Liaison Representative
JAMES E. SHAMBLIN, Oklahoma State University G. P. JAYAPRAKASH, TRB Liaison Representative
RICHARD SORENSON, Idaho Transportation Department

Program Sta
ROBERT J. REILLY, Director, Cooperative Research Programs FRANK R. McCULLAGH, Senior Program Officer
CRAWFORD F. JENCKS, Manager, NCHRP KENNETH S. OPIELA, Senior Program Officer
LOUIS M. MAcGREGOR, Administrative Officer DAN A. ROSEN, Senior Program Officer
STEPHEN E. BLAKE, Senior Program Officer SCOTT A. SABOL, Program Officer
LLOYD R. CROWTHER, Senior Program Officer EILEEN P. DELANEY, Editor
AMIR N. HANNA, Senior Program Officer KAM] CABRAL, Editorial Assistant
NATIONA 'L COOPERATIVE HIGHWAY RESEARCH PROGRAM

Report 362

Roadway Widths for

Low-Traffic-Volume Roads

C.V. ZEGEER, R. STEWART, and F. COUNCIL

University of North Carolina
Highway Safety Research Center
Chapel Hill, NC
and
T.R. NEUMAN
CH2M HILL
Evanston, IL

Subject Areas

Highway and Facility Design

Highway Operations, Capacity, and Traffic Control
Safety and Human Performance

Research Sponsored by the American Association of State

Highway and Transportation Officials in Cooperation with the
Federal Highway Administration

TRANSPORTATION RESEARcH BOARD

NATIONAL RESEARCH COUNCIL

NATIONAL ACADEMY PRESS

Washington, D.C. 1994
NATIONAL COOPERATIVE HIGHWAY RESEARCH NCHRP REPORT 362
PROGRAM

Systematic, well-designed research provides the most effective Project C15-12 FY'88
approach to the solution of many problems facing highway ad-
ministrators and engineers. Often, highway problems are of lo- ISSN 0077-5614

cal interest and can best be studied by highway departments in- ISBN 0-309-05360-9
dividually or in cooperation with their state universities and
L C. Catalog Card No. 93-061751
others~ However, the accelerating growth of highway transpor-
tation develops increasingly complex problems of wide interest
to highway authorities. These problems are best studied through Price $25.00
a coordinated program of cooperative research.
In recognition of these needs, the highway administrators of
the American Association of State Highway and Transportation
Officials initiated in 1962 an objective national highway research
program employing modem scientific techniques. This program is
supported on a continuing basis by funds from participating member
states of the Association and it receives the full cooperation and
support of the Federal Highway Administration, United States De- NOTICE
partment of Transportation. The project that is the subject of this report was a part of the National Cooperative
The Transportation Research Board of the National Research Highway Research Program conducted by the Transportation Research Board with
Council was requested by the Association to administer the research the approval ofthe Governing Board ofthe National Research Council. Such approval
reflects the Governing Board's judgment that the program concerned is of national
program because of the Board's recognized objectivity and under- importance and appropriate with respect to both the purposes and resources of the
standing of modem research practices. The Board is uniquely suited National Research Council.
for this purpose as it maintains an extensive committee structure
The members of the technical conunittee selected to monitor this project and to
from which authorities on any highway transportation subject may review this report were chosen for recognized scholarly competence and with due
be drawn; it possesses avenues of communications and cooperation consideration for the balance of disciplines appropriate to the project. The opinions
and conclusions expressed or implied are those of the research agency that performed
with federal, state and local governmental agencies, universities the research, and, while they have been accepted as appropriate by the technical
and industry; its relationship to the National Research Councii committee, they are not necessarily those of the Transportation Research Board,
is an insurance of objectivity; it maintains a full-time research the National Research Council, the American Association of State Highway and
Transportation officials, or the Federal Highway Administration, U.S. Department of
correlation staff of specialists in highway transporiation matters to Transportation.
bring the findings of research directly to those who are in a position
Each report is reviewed and accepted for publication by the technical committee
to use them.
according to procedures established and monitored by the Transportation Research
The program is developed on the basis of research needs identi- Board Executive Committee and the Governing Board of the National Research
fied by chief administrators of the highway and transportation Council.
departments and by committees of AASHTO. Each year, specific
areas of research needs to be included in the program are proposed
to the National Research Council and the Board by the American
Association of State Highway and Transportation Officials. Re-
search projects to ftilfill these needs are defined by the Board,
and qualified research agencies are selected from those that have
submitted proposals. Administration and surveillance of research
contracts are the responsibilities of the National Research Council
and the Transportation Research Board.
The needs for highway research are many, and the National
Cooperative Highway Research Program can make significant con-
tributions to the solution of highway transportation problems of
'Published reporls of the
mutual concern to many responsible groups. The program, how-
ever, is intended to complement rather than to substitute for or NATIONAL COOPERATIVE HIGHWAY RESEARCH PROGRAM
duplicate other highway research programs.
are available from:

Transportation Research Board

National Research Council
Note: The Transportation Research Board, the National Research Council, the 2101 Constitution Avenue, N.W.
Federal Highway Administration, the American Association of State Highway and
Washington, D.C. 20418
Transportation Officials, and the individual states participating in the National
Cooperative Highway Research Program do not endorse products or manufac-
turers. Trade or manufacturers names appear herein solely because they are
considered essential to the object of this .reporl. Printed in the United States of America
FOREWORD This report is recommended to persons involved in highway design, safety, mainte-
nance, and administration. It describes the results of comprehensive safety, operational,
BY Staff and economic analyses regarding minimum roadway widths for low-volume roads. It is
Transportation Research believed that this report provides useful information, documents a detailed research ap-
Board proach, describes a useful model for investigating many facets of low-volume roads, and
presents recommendations for revisions to the design standards presented in the American
Association of State Highway and Transportation Officials' (AASHTO) Policy on Geomet-
ric Design for Highways and Streets, 1990 Edition (referred to in this report as Policy or
Green Book).

The AASHTO Green Book established minimum roadway widths for new construc-
tion and major reconstruction of roadways according to road classification, traffic volume,
and design speed. These minimum roadway widths were based on limited research findings
augmented with engineering judgment. The impact of using these minimum widths was
highlighted in the results of a recent research effort conducted by the Transportation
Research Board that was published in TRB Special Report 214, "Designing Safer Roads:
Practices for Resurfacing, Restoration, and Rehabilitation" (1987). That report indicated
that acceptable operational and safety experience can be found on low-traffic-volume
(less than 2000 ADT) roadways having widths narrower than the minimum roadway
widths specified in the Green Book. This implies that designs for new construction and
major reconstruction of low-traffic-volume roadways may be excessive. Because about
1.3 million miles of rural roadways (or one-third of the total rural mileage) in the United
States fall within the low-traffic-volume category, a reduction in the roadway-width
requirements could have a major impact on the state programs.
The objectives of this research were to develop an engineering analysis procedure
for determining optimum roadway width for the construction and reconstruction of low-
volume roadways. This procedure was then to be used to develop "minimum width of
traveled way and shoulder" recommendations for consideration by the Geometric Design
Task Force of the AASHTO Highway Subcommittee on Design.
This project consisted of (1) a critical review of all pertinent literature dealing with
safety, operations, and geometrics of low-volume roads as they pertain to roadway width;
(2) development of a plan for acquiring and analyzing data on traffic volume, vehicle
speed, percent trucks, roadway functional classification (arterial, collector, and local),
level of service, accidents, and other information; (3) extensive data collection and analyses
to assess safety, operational, and economic aspects; (4) development of an engineering
analysis procedure for optimizing roadway width, considering the costs and safety benefits;
(5) comparison of the- recommended roadway widths to current width criteria for low-
traffic-volume roadways to determine the safety, operational, and economic impacts that
would be associated with adopting recommended standards; and (6) preparation of a final
report documenting the efforts and findings of the research.
This project was highlighted by (1) the use of the Highway Safety Information
System (HSIS), which allowed the analyses of thousands of miles of low-volume, two-
lane roadways; (2) extensive field investigations to obtain or verify critical information
on the widths, cross sections, and roadside conditions for these roadways; (3) thorough
analyses of the data leading to the formulation of a model for the analysis of traffic and
roadway design conditions that affect the safety and operation of low-volume roads; and
(4) an extensive economic analysis using data gathered from the states and case studies
of actual low-volume roadway improvement projects that had been completed over the
last 5 years. These analyses revealed that narrower roadways could function safely and
effectively in some situations, allowing highway improvement funds to be stretched over
a greater proportion of the two-lane road network. The results cover most low-volume
roadways because the sample included paved and unpaved roadways, roadways in various
terrains, a range of traffic volumes and rnix, and roadways from different regions of the
nation. Recommendations for changes to the AASHTO Green Book are presented in the
report.
 CONTENTS

I SUMMARY

5 CHAPTER ONE Introduction and Research Approach

Problem Statement, 5
Current AASHTO Width Criteria, 6
Background Safety Research, 9
Research Objective and Scope, 10
Research Approach, 10
11 CHAPTER TWO Study Findings
Analysis Issues, I I
Analysis of Highway Safety Issues, 11
Development of the Primary Accident and Roadway Database, 11
Statistical Testing and Results, 14
Highway Operational Considerations Related to Design, 21
Functional Classification, 21
Design Speed, 22
Roadway Design Requirements, 26
Cost-Effectiveness and Width-Related Sensitivities, 26
Comparison of 10-Ft Lanes with 5-Ft Shoulders to 10-Ft Lanes with 2-Ft
Shoulders, 31
Comparison of 11 -Ft Lanes with 4-Ft Shoulders to 11 -Ft Lanes with 2-Ft
Shoulders, 31
Comparison of 12-Ft Lanes with 4-Ft Shoulders to 12-Ft Lanes with 2-Ft
Shoulders, 32
Comparison of 11-Ft Lanes with 2-Ft Shoulders to 10-Ft Lanes with 2-Ft
Shoulders, 32
Comparison of 10-Ft Lanes with 6-Ft Shoulders to 11-Ft Lanes with 2,-Ft
Shoulders, 33
Comparison of 11 -Ft Lanes with 4-Ft Shoulders to 10-Ft Lanes with 5-Ft
Shoulders, 33
Comparison of 12-Ft Lanes with 3-Ft Shoulders to 1 1-Ft Lanes with 4-Ft
Shoulders, 34
39 CHAPTER THREE Results, Interpretation, Appraisal, and Application
Comparison of Recommended Widths to Current AASHTO Policy, 39
Format, 39
Local Road Design Values, 39
Collector Design Values, 43
Arterial Design Values, 46
Application to 3R Design Problems, 46
Guidelines for 3R Design Problems, 46
Procedures for Evaluafion of Low-Volume Road Design Options, 46
Evaluation of Case Studies, 47
Case Study No. 1: US 62 and 281 —Widening and Reconstruction; Caddo
County, Oklahoma, 47
Case Study No. 2: State Highway 46—Reconstruction;.'Gooding and Camas
Counties, Idaho, 47
Case Study No. 3: Oak Harbor Southeast Road—Resurfacing; Ottawa
County, Ohio, 48
Case Study No. 4: Grande Ronde River Road 51 —Reconstruction; La
Grande, Oregon, 48
Summary of Case Studies, 48
50 CHAPTER FOUR Conclusions
Accidents on Low-Volume Roads, 50
Accident Effects of Lane and Shoulder Width, 50
Accident Effects of Paved vs. Unpaved Surfaces, 50
Accident Effects of Large Trucks, 50
Accident Benefits of Widening Low-Volume Roads, 50
Capacity and Level-of-Service Effects on Width of Low-Volume Roads, 50
Speed-Related Widths, 51
Shoulder Functions Related to Width, 51
Truck Operations Related to Width, 51
Costs of Construction of Low-Volume Roads, 51
Cost-Effective Width Combinations, 51
Recommended Standards, 51
52 REFERENCES

54 APPENDIX A Cost-Effectiveness Guide for Design Improvements to Roadway Cross

Sections on Low-Volume Rural Roads
64 APPENDIX B Critical Review of Literature
72 APPENDIX c Analysis of Low-Volume Roads Data
90 APPENDIX D Survey of Construction Costs—Two-Lane Rural Highways
100 APPENDIX E Derivation of Consistency Between Lane Width, Shoulder Width,
Functional Class, and Volume
105 APPENDIX F Guidelines for Derivation of Consistency Between Design Speed and
Lane Width
 ROADWAY WIDTHS FOR
LOW-TRAFFIC-VOLUME ROADS

SUMMARY in recent years, maintaining and reconstructing the two-lane highway system has
emerged as a serious problem. A primary problem results from the extensive size of the
system. A 1983 study estimated that there are approximately 3.1 million miles of two-
lane highways in the United States, which represent 97 percent of the rural mileage and
80 percent of the total U.S. highway mileage. Further, much of thii mileage has relatively
low traffic volumes. For example, of the 3.1 million miles of two-lane rural roads,
approximately 90 percent (2.8 million miles) have an average daily traffic (ADT) of less
than 1,000 vehicles per day (vpd). About 80 percent have ADTs of less than 400 vpd,
and 38 percent carry less than 50 vpd.
A great proportion of the two-lane rural road system is 30 years old or more, necessitating
investment to replace pavement, repair shoulders, and address other problems. The age
of the system adds another dimension to the problem. Significant mileage of two-lane
highways was designed and built to standards that are outdated compared with current
design policy. For example, over one-quarter of the mileage of such roads have lane
widths of 9 ft or less, and two-thirds have shoulder widths of 4 ft or less. In addition,
11.5 percent of two-lane highway mileage (356,500 miles) have no shoulders. These
statistics contrast with current design values shown in the 1990 American Association of
State Highway and Transportation Officials (AASHTO) Policy on the Geometric Design
of Highways and Streets (henceforth referred to as Policy or Green Book). For all but
extremely low-volume and low-speed highways, the current Policy calls for 22- to 24-ft
roadways (11 - to 12-ft lane widths) regardless of terrain or other conditions (AASHTO,
1990). 1

There are insufficient funds to reconstruct the entire two-lane system to design values
implied by the 1990 AASHTO Policy. At the same time, abandoning roads or allowing
continued, long-term deterioration are not viable options. Thus, state and local highway
agencies find their engineers confronted with difficult choices related to repaving or
reconstructing (to AASHTO design standards). The impacts can be dramatic if, for exam-
ple, an agency's budget allows for either 20 miles of roadway repavement or 12 miles
of roadway reconstruction.
Of course, both repairs and reconstruction would be expected to produce different levels
of safety and operational quality. Yet, as currently written, the AASHTO Policy does not
offer the flexibility to consider the safety, operations, and cost trade-offs on either a
statewide programming or site-specific basis. Indeed, many design engineers perceive the
Policy as a hindrance to arriving at the most cost-effective design solution. This is because
there are many miles of two-lane rural highways considered "substandard" (with respect
to the current AASHTO Policy values) that have acceptable safety experience. Further-
more, in many cases, accident problems that do exist may be attributable to variables
other than those cited in the AASHTO Policy (i.e., the width of traveled way and shoulder),
such as roadside conditions and horizontal or vertical alignment. However, once a decision
is reached to reconstruct rather than resurface or maintain a highway, the current AASHTO
Policy applies, which calls for certain minimum cross-sectional dimensions. There is no
current evidence that such dimensions produce significant safety benefits, but they do
generally have major cost implications. Revision to the values shown in the AASHTO
Policy, reflecting greater sensitivity to safety and construction cost trade-offs, could offer
the opportunity for agencies to increase the safety return on their reconstruction investment.
The objectives of this study were to (1) develop an engineering analysis procedure that
would allow for the determination of optimum roadway widths for construction and
reconstruction of low-volume roadways (ADT less than 2,000 vpd) and (2) use the
analysis procedure to develop recommendations for "minimum width of traveled way
plus shoulder." These recommendations would be considered by the design task force of
the AASHTO highway subcommittee on design, for inclusion in future editions of the
Green Book. The efforts in this study relate to paved or unpaved two-lane rural roads
with ADT volumes of 2,000 vpd or less. Functional classifications studied included arterial,
collector, and local roadways.
The basic approach to this research involved first reviewing current AASHTO Green
Book values on roadway width, as well as other proposed guidelines, and the roadway
factors considered to be important in the formulation of such standards. Next, a detailed
review of safety literature was conducted to determine what is currently known about
accident effects of roadway width and other traffic and roadway features on two-lane
rural roads. Safety effects of numerous features, in addition to roadway width, were
considered important because no single data item alone is the sole factor in resulting
crashes. Instead, the interaction among several roadway elements, and perhaps driver and
vehicle factors, can affect the frequency and/or severity of accidents.
The literature review revealed that in addition to lane and shoulder width, other factors
found to affect safety on two-lane rural roads include shoulder type, roadside slope,
roadside obstacles (placement, number, and rigidity), horizontal alignment, vertical align-
ment, traffic volume (and mix of traffic), narrow bridges, and location of intersections
and driveways, among others. The identification of such safety relationships was useful
in formulating a data collection and analysis plan. This plan identified the important
variables for the analysis of the interaction of the factors that affect roadway design. For
example, increasing lane width can reduce shoulder width and recovery area, if limited
right of way is available. If shoulders are widened, then roadside slopes may -need to be
steepened, possibly leading to more severe crashes. Clearly, the trade-offs that exist from
what seems to be a simple and logical increase in lane width must be studied together to
determine optimal lane widths from a safety perspective.
To maximize the extent of the safety, operational, and economic analyses in this project,
existing databases that offered adequate detail were used to the extent possible. The
primary data source selected was a database of approximately 2,400 miles of two-lane
road from seven states (with ADT of 2,000 vpd or below) from two previous FHWA
studies (see References 5 and 7). This database was supplemented with data from approxi-
mately 1,700 miles of paved and unpaved two-lane roadway (mostly local and collector
streets) from North Carolina, Utah, and Michigan, for a total of approximately 4,100
miles. Field data were collected on these supplemental sections to provide information
on roadside conditions, intersections, driveways, and terrain to match the other database.
 3

A detailed statistical analysis was conducted on the primary database, and accident rates
were determined for various lane and shoulder widths. To validate the findings and to
investigate these relationships further, three independent databases totaling more than
54,000 miles of low-volume, two-lane roads were obtained from three states (Illinois,
Minnesota, and North Carohna) and analyzed.
A project cost analysis methodology was developed for roadway widening projects for
existing roadways and also for comparing roadway width effects on new road construction
costs. An economic analysis analyzed the cost and benefit relationships for the different
lane and shoulder width combinations in various situations. Economic benefits of accident
reductions were estimated using research by others that established societal costs of
motor vehicle injuries and fatalities. The research approach used FHWA's recommended
methodology for tabulating such costs, resulting in a recommended value of $60,000 per
low-volume road accident. This value was used to estimate benefits of accident reductions.
A construction cost model, based on a comprehensive survey of state design engineers
and practices, was used to investigate the sensitivities of variable lane and shoulder width
designs, as well as other relevant factors such as terrain on total construction costs.. The
accident and construction cost models were used to evaluate the cost-effectiveness of
various combinations of lane and shoulder width.
Design standards for highways are based on more than safety performance. The research
investigated the operational requirements and effects of different roadway and shoulder
widths. Issues investigated included capacity; speed; highway level of service; consider-
ation of wider, larger vehicles; and compatibility with alignment and other design standards
and criteria published in the Policy. A central theme of the study was that width-related
design standards should be compatible with primary design policy controls such as design
speed and functional classification.
The safety, operational, and economic analyses covering all facets of low-volume roads
resulted in a number of conclusions. Specific conclusions from the safety analyses included
the following:

- Low-volume roads were found to experience a slightly higher percentage of injury

accidents than the full sample of rural roads. Single-vehicle accidents (fixed object,
rollover, run-off-road) are greater proportionately and multivehicle accidents (rear-end,
angle, turning) lesser proportionately on low-volume roads than on all two-lane rural
roads. The studies found that low-volume road accidents are affected primarily by roadway
width, roadside hazard, terrain, and driveways per mile. Accident rates are significantly
associated with varying lane and shoulder widths for single-vehicle and opposite-direction
accidents.
- The presence of a shoulder is asso' ciated with significant accident reductions for lane
widths of at least 10 ft. For lane widths of 11 and 12 ft, shoulder widths of at least 3 ft
have significant effects. For a combination of reasons, some perhaps relating to lower
vehicle speeds on narrow roadways, there is no apparent accident benefit of widening
lanes from 9 ft to 10 ft. Indeed, the study produced evidence that 9-ft lanes with wide
shoulders may be preferable to 10-ft lanes with narrow shoulders. Roadways with lane
widths of 11 or 12 ft have significantly lower accident rates than roadways with 10-ft
lane widths.
- Accident experience does not appear significantly different for unpaved versus paved
roadway surfaces at traffic volume levels of 250 vpd or less. Above this volume, accident
rates are significantly greater for unpaved roads, which provides evidence for paving
roadways with traffic volumes above 250 ADT for safety reasons.

The study produced a linear model that was used to estimate expected accident effects
of various combinations of lane width and shoulder width. For 10-ft lanes, accident rates
were lower by 0.98 per million vehicle miles (MVM) when accompanied by shoulders
of greater than 4 ft versus shoulders of 4 ft or less. For 11 - and 12-ft lanes, shoulder
widths of 3 ft or greater produced accident rates 0.56 per MVM lower than with shoulder
widths less than 3 ft. The accident model produced identical accident benefits for 11 - and
12-ft lanes.
Specific conclusions derived from the analysis of operational issues included the fol-
lowing:

* Based on known operational effects of lane and shoulder widths, capacity requirements
should ir~fluence design standards for certain terrain conditions and traffic composition.
For designing roadways with ADT of 1,500 vpd or more, combinations of lane and
shoulder width in excess of those required merely for safety are necessary to produce
levels of service compatible with AASHTO design policy guidelines.
* Lane widths should produce operating speeds compatible with the selected design
speed. Wider lane widths (say, I I - or 12-ft) on roadways designed with lower design
speeds (say, 40- or 50-mph) may be undesirable. Such widths may promote operating
speeds above those for which the alignment was intended to accommodate.
- It was concluded that for certain combinations of functional classification, traffic
volume, and composition, shoulder widths in excess of those required solely for safety
are necessary. For the upper range of traffic volumes (greater than 1,500 vpd), this study
recommended minimum shoulder widths of 6 ft to 8 ft depending on the functional class
of the facility.
- It was concluded that, where longer, wider vehicles can be expected to operate,
providing incrementally greater lane and/or shoulder widths is desirable to accommodate
the off-tracking and encroachment problems associated with these vehicles.

The economic analysis led to the following specific conclusions:

- Cost per mile of construction was found to be greatly influenced by terrain, total
roadway width, and lane width.
* "Breakeven" ADT levels were computed to deten-nine when widening appeared
justified.

Finally, based on the accident, operational, and economic analyses results, revised width
guidelines were developed for new or reconstructed low-volume roads. Guidelines also
were developed for evaluation of 3R projects. This study concluded that there are opportuni-
ties to refine current lane and shoulder width standards for low-volume roads. A major
conclusion of this exercise was that width combinations narrower than those called for
by current AASHTO Policy appear appropriate. In particular, where total roadway widths.
(i.e., lanes plus shoulders) currently exceed 30 ft, there are many cases where narrower
widths could be employed at lower cost with no apparent degradation in safety. Also, the
cost-effectiveness analysis demonstrated that less variation in widths for the range in
traffic volumes is appropriate compared to current AASHTO Policy. Revised design
values based on the safety studies, operational analyses, and cost-effectiveness evaluations
would produce meaningful savings in construction and reconstruction costs associated
' rural system.
with the low-volume
The major objective of the research—to develop recommended design values for
lane and shoulder width for low-volume roads—was accomplished. This research also
recommended revisions to the format of design standards as well as to the values for the
range of relevant condition. Overall, the research supports conclusions of earlier work
from TRB Special Report 214; therefore, narrower width combinations for lanes and
shoulders on low-volume roads should be considered in subsequent editions of the
AASHTO Policy.
CHAPTER 1
INTRODUCTION AND RESEARCH APPROACH

PROBLEM STATEMENT ft to 24-ft traveled way (i.e., I I -ft to 12-ft lanes) for all terrain
and other conditions.
In recent years, maintaining and reconstructing the two-lane Much of the existing rural mileage was designed to criteria that
highway system has emerged as a serious problem. The problem prevailed more than 30 years ago. Indeed, there is undoubtedly
primarily results from the extensive size of the system and the significant mileage of low-volume roads that may not ever have
fact that significant mileage of two-lane highways was designed been "designed" in the conventional sense. Such roads evolved
and built to outdated standards not reflective of current design over the years, beginning as horse paths and eventually being
policy. Also, a large portion of low-volume roads is unpaved, upgraded over time to a paved surface. What evolved was func-
which presents maintenance problems in addition to safety con- tional for the type and level of traffic that prevailed locally, yet
cerns. n-fight not have met any formal criteria for highway design.
A 1983 study estimated that there are approximately 3.1 mil- There are insufficient funds to reconstruct the entire two-lane
lion miles of two-lane highways in the United States, which system to design values implied by the 1990 AASHTO Policy.
represents 97 percent of the rural mileage and 80 percent of the At the same time, abandoning roads or allowing continued, long-
total U.S. highway mileage. Further, much of this mileage has term deterioration is not a viable option. Thus, state and local
relatively low traffic volumes. For example, of the 3.1 million highway agencies find their engineers confronted with difficult
miles of two-lane rural roads, approximately 90 percent (2.79 choices. Should a highway merely be repaved within its existing
million miles) have an average daily traffic (ADT) of less than cross section, or should reconstruction to values shown by
1,000 vehicles per day (vpd). About 80 percent have an ADT AASHTO be considered? For example, assume that a given
of less than 400 vpd, and 38 percent carry less than 50 vpd. Of highway budget may enable repaving of 20 miles of facility in
all local and minor collector roads on the two-lane system, 90 need of repair; if the latter approach— reconstruction —is taken,
percent have an ADT of 2,000 vpd or less; more than 60 percent only 12 miles may be built for the same money.
of minor rural arterials have an ADT of 2,000 vpd or less (1). Of course, both repairs and reconstruction would be expected
Thus, in terms of their extensive mileage, low-volume roads are to produce different levels of safety and operational quality. Yet,
clearly an important component of the highway transportation as currently written, the AASHTO Policy does not offer the
system. flexibility to consider the safety, operations, and cost trade-offs
A great proportion of the two-lane rural road system is 30 on either a statewide programming or site-specific basis. Indeed,
years old or more, necessitating investment to replace pavement, many design engineers perceive the Policy as a hindrance to
repair shoulders, and address other problems. The age of the arriving at the most cost-effective design solution. This is because
system adds another dimension to the problem. Significant mile- there are many miles of Iwo-lane rural highways considered
age of two-lane highways was designed and built to standards "substandard" (with respect to the current AASHTO Policy val-
that are outdated compared with current design policy. For exam- ues) that demonstrate acceptable safety records. Furthermore, in
ple, over one-quarter of the mileage of such roads have lane many cases, accident problems that do exist may not be attribut-
widths of 9 ft or less, and two-thirds have shoulder widths of 4 able necessarily to the variables that describe the cross section
ft or less. In addition, 11.5 percent of two-lane highway mileage in the AASHTO Policy (i.e., the width of traveled way and
(356,500 miles) has no shoulders. These statistics contrast with shoulder), but to other factors such as the roadside or horizontal
current design values shown in the 1990 American Association or vertical alignment. However, once a decision is reached to
of State ffighway and Transportation Officials A Policy on the reconstruct rather than resurface or maintain a highway, the
Geometric Design of Highways and Streets (henceforth referred current AASHTO Policy applies, which calls for certain mini-
to as Policy or Green Book) (2). For all but extremely low- mum cross-sectional dimensions. There is no current evidence
volume and low-speed highways, the 1990 AASHTO Policy that such dimensions produce significant safety benefits; gener-
calls for 22-ft to 24-ft roadways (11-ft to 12-ft lane widths) ally, they do have major cost implications.
regardless of terrain or other conditions (2). Recent research concerning resurfacing, restoration, and reha-
Over the last 50 years, design criteria for all highways have bilitation (3R) practices [TRB Special Report 214, (3)] has sug-
evolved to reflect changes in vehicle width and performance, gested that acceptable safety and operational experience could
greater understanding of driver behavior, and advances in mate- be expected on low-volume roadways with lane widths that are
rial and construction techniques. Cross-section values in the somewhat narrower than those proposed in the AASHTO Policy.
AASHTO policies from 1940 to the present have gradually If this result could be demonstrated through carefully conducted
changed to the point where, except for extremely low-volume research, the implications would be significant. Revision to the
and low-speed highways, AASHTO policy criteria specify a 22- values shown in the Policy, reflecting greater sensitivity to safety
 Sideslope Center!!ne Ditch bottom
Original (alse QrIgInaI
ground foreslope) Shoulder Lane Lane Shoulder Foreslope Backslope ground

Shoulder break --y

Hinge
point Depth

K1
of cut
tHelght
0 I Shoulder
I
1 fill width
4 - -Lane wlctth_- _L
Paved
S
Ditch bottom'
shoulder width (variable)
Slope S: I vAdth

Clear zone width

(to obstacle or
non traversable slope) Roadway width

Figure 1. Elements of rural highway cross sections.

and construction cost trade-offs, could offer the opportunity for roads with ADT of less than 250 vpd with design speeds of 20
agencies to increase the safety return on their reconstruction or 30 mph.
investment. Shoulder width requirements for rural arterials range from 4
Of course, the findings implied by TRB Special Report 214 ft to 8 ft of usable shoulders (each side of pavement), where
are somewhat tentative in nature at this point, because there such usable shoulders should be paved. For rural collector and
have been very few well-conducted studies of lane-width-related local roadways, grav' el shoulders of 2 ft to 8 ft are specified,
crashes for low-volume roadways. Instead, much of the informa- depending on design traffic volume.
tion currently available is based on data collected on higher The 1990 AASHTO width values incorporate several revisions
volume roadways where accidents per mile are greater. There is since the 1984 Green Book. For example, the 1984 Policy did
also a need to establish the effects of lane width on crash rates not include a 40-mph design speed category for rural arterials.
for low-volume roads and, more specifically, to relate this infor- Also, the 1984 criteria for rural arterials specified design traffic
mation to the issue of widening existing narrow pavements on volume categories with DHV of 200 to 400 vpd and DHV above
low-volume roads when reconstruction is done. No optimum set 400 vpd, which now are combined into one category (DHV 200
of lane and shoulder width values has yet been established for vpd or above) in the 1990 Policy. Further, the 1990 Policy allows
various traffic or geometric conditions on low-volume roads. 22-ft traveled-way widths on reconstructed rural arterials in cases
where alignment and safety records are satisfactory. Finally, on
rural local roads with DHV of 200 to 400 vpd, the 1990 Green
CURRENT AASHTO WIDTH CRITERIA
Book specifies a 6-ft shoulder width, compared to an 8-ft width
in the 1984 version.
Width-related elements of rural highway include roadway fea-
tures (lanes and shoulders) as well as roadside features (fore- In recent years, standards less stringent than implied by Green
slopes, ditches, backslopes). Figure I describes those individual Book values have been proposed for lane and shoulder widths.
elements of interest in rural cross-section design. For example, AASHTO 3R design values have minimum lane
The 1990 AASHTO Policy (Green Book) design values for widths of 10 ft or I I ft and 2-ft minimum shoulder widths (see
width of traveled way for new construction or roadway recon- Table 4). These values are independent of ADT but do consider
struction are given in Tables 1, 2, and 3 for arterial, collector, percent trucks for average running speeds over 40 mph (TRB
and rural local roads, respectively. The values in these tables Special Report 214, 1987).
generally are quite stringent for many roadway situations. For The roadway width standards proposed in 1978 by the Federal
example, a 24-ft traveled way width (i.e., two 121ft lanes) is Highway Administration (FHWA) generally fall somewhere be-
suggested for nearly all design traffic volumes and design speeds tween the AASHTO 1990 Policy and AASHTO 3R values dis-
on rural arterial streets, except for some combinations of lower cussed previously in terms of minimum width requirements (see
design speed (i.e., 40 or 50 miles per hour (rnph)) and design Table 5). These standards generally call for minimum lane widths
traffic volumes (i.e., design hourly volumes of less than 200 of 10 ft to 12 ft. However, for minor roads, 9-ft lanes are allowed
vpd). On rural collector streets, 24-ft widths are specified for on routes with ADT below 400 vpd and less than 10 percent
design hourly volumes (DHV) over 400 vpd, with 20-ft to 22- trucks (TRB Special Report 214, 1987). Minimum width shoul-
ft widths specified for other situations. On rural local roads, ders of 2 ft to 3 ft are specified based on various combinations
traveled-way-width requirements range from 18 ft to 24 ft, with of ADT, design speed, and design trucks.
24-ft widths specified again for DHV of 400 vpd and above, In TRB Special Report 214, the authors modified the 1978
and 18-ft traveled-way widths (9-ft lanes) permitted only for FHWA values based on cost-effectiveness and practicality of
TABLE 1. Minimum width of traveled way and usable shoulder for rural arterials

Width of Traveled Way (feet)'

Design Speed Current ADT Current ADT DHV DHV
(mph) Under 400. vpd 400 vpd & 100-200 Over 200
Over

40 22 22 22 24
50 22 24 24 24
60 24 24 24 24
70 24 24 24 24

Width of Usable Shoulder (feet) on Each Side of Pavementb

All Speeds 4 1 6 , . 1 6 1 8
.Width of traveled way may remain at 22 feet on reconstructed highways where alignment and
safety records are satisfactory.
b Usable shoulders on arterials should be paved.

Source: AASHTO. 1990. Geometric Design of Highways and Streets.

TABLE 2. Minimum width of traveled way and graded shoulder for rural collector roads

Width of Traveled Way (feet)

Current ADT
Design Speed Current ADT 400 vpd and DHV DHV DHV
(mph) Under 400 vpd Over 100-200 200400 Over 400

20 20 20 20 22 24
30 20 20 20 22 24
40 20 22 22 22 24
50 20. 22 22 24 24
60 22 22 22 24 24
70 22 22 22 24 24

Width of Graded Shoulder—Each Side of Pavement

All Speeds 28 1 4 1 6 8 8

'Minimum width is 4 feet if roadside barrier is utilized.

Source: AASHTO. 1990. Geometric Design of Highways and Streets.

TABLE 3. . Minimum width of traveled way and graded shoulder for rural local roads

Width of Traveled Way (feet)

Design Speed Current ADT Current ADT Current ADT DHV DHV DHV
(mph) Under 250 vpd 250-400 vpd I Over 400 vpd 100-200 1 200400 Over 400

20 18 20 20 20 22 24
30 18 20 20 20 22 24
40 20 20 22 22 22 24
50 20 20 22 22 24 24
60 20 22 22 22 24 24
Width of Grade Shoulder—Each Side of Pavement
All Speeds 1 2 2 1 - 4 1 6 6 8
Source: AASHTO. 1990. Geometric Design offfighways and Streets.
8

TABLE 4. AASHTO roadway width standards for 3R projects

Minimum Widths (ft)

Average Running Speeds (mph) Percent Trucks Lane Shoulder
40 or less All 10 2
Over 40 15 or less 10 2
Over 15 11 2

NOTE: The standards specify a range of widths for all cases. Only the minimums are reported here.

TABLE 5. FHWA roadway width standards (1978)

Width (ft)
10 Percent Less Than
Current
or More Trucks 10 Percent Trucks
Traffic Design Speed
(ADT) (mph) Lanes Shoulders Lanes Shoulders
1-400 50 or less 10 2 9a 2
Over 50 10 2 10 2
401-4,000 50 or less 11 2 10 2
Over 50 12 3 11 3
Over 4,000 All 12 4 Il 4

NOTES: The 1978 FHWA proposed standards were actually defined in terms of lane width and total
roadway(lane plusshoulder) width. The standardsare shown here in termsoflane and shoulder width so
thattheycan be moreeasily compared with AASHTO standards. In the actual standard, I I -ft lanesand a
I-ft shoulder are permitted where 10-ft lanes and 2-ft shoulders are specified in the table.
a"Minor roads!' only; otherwise 10-11 lanes.

Source: Transportation Research Board. Special Report 214. 1987.

TABLE 6. FHWA standards revised by TRB

Width (ft)
10 Percent Less Than
Design Year Average
or More Trucks 10 Percent Trucks
Volume Running
(ADT) Speed (mph) Lanes Shoulders Lanes Shoulders
1-750 Under 50 10 2 9 2
50 or more 10 2 10 2
751-2,000 Under 50 11 2 10 2
50 or more 12 3 11 3
Over 2,000 All 12 6 11, 6

NOTES: Shoulders may be I ft less for highways in mountainous terrain. The standards were actually
defined in termsof lane width and total roadway(lane plusshoulder) width. They aregiven here in terms
of lane and shoulder width for easier comparison with AASHTO standards. For the purpose of simplicity,
weighted average design speed was used for average running speed in the analysis.

Source: Transportation Research Board. Special Report 214. 1987.

use. As given in Table 6, the authors also used ADT, average in place of design speed, perhaps because the term "n-tinor road"
running speed, and percent trucks as criteria in their suggested is vague and design speed was considered difficult to detem-dne
guidehnes. They further provided for shoulders to be I ft less for an existing highway. TRB Special Report 214 also discusses
in mountainous terrain than those width values given in Table various roadway width standards.
5. The revised table (Table 6) eliminated "minor road" as a In 1979, Glennon developed minimum roadway width require-
criterion for 9-ft lanes, and substituted "average running speed" ments for roads with ADT of 400 vpd and below (4). As given
 TABLE 7. Glennon roadway width requirements for low-volume roads

Total Road Width Requirements (ft) ~~/ b/

Lower % Busses & Trucks Higher % Busses & Trucks

(as specified below) (as specified below)
Design > 28% for 0- 50 ADT
< 28% for G- 50 ADT
Speed < 12% for 51-100 ADT > 12% for 51-100 ADT
7% for 101-200 ADT > 7% for 101-200 ADT
(.ph)'/ —All% for 201-400 ADT
NA for 201-400 ADT

Infrequent Trips by Frequent Trips by Infrequent Trips by Frequent Trips by

Farm Machineryi
d d/
Farm Machinery— Farm Machinery Farm Machinery

20 mph 18 ft. 22 ft. 20 ft. 24 ft.

25 20 24 22 26

30 20 24 22 26

35 22 24 24 26

40 22 26 24 28

45 26 26 26 28

50 30 30 30 30

1 ft - .305 m
Widths above 24 ft. (7.3m) include appropriate shoulder widths.
I mph - 1.61 kph
The determination of "frequent" and "infrequent" are at the discretion of the designer.

in Table 7, criteria for minimum widths included design speed, incidence of these related accident types (and included in the
percent buses and trucks, ADT groups below 400 vpd, and fre- predictive model) were wider lanes, wider shoulders (paved
quency of use of farm machinery. Recommended minimum shoulders were found to be slightly safer than unpaved shoul-
widths range from 18 ft to 30 ft. For design speeds of 40 mph ders), better roadside condition, flatter terrain, and lower traffic
or less, higher percentages of buses and trucks require 2 ft of volume. For lane widths between 8 and 12 ft, the predictive
additional width. accident model showed that related accidents were reduced by
A review of such other proposed width standards provides 12 percent for I ft of lane widening (for example, 10-ft lanes to
insight into the roadway criteria (e.g., design speed, running I I -ft lanes), 23 percent for 2 ft of lane widening, 32 percent for
speed, ADT, design hourly volume, percent trucks) considered 3 ft of lane widening, and 40 percent for 4 ft of lane widening.
in these standards, as well as the roadway width requirements The study also modeled the effects of shoulder widening on
that are proposed. To date, much of the development of current related accidents for paved and unpaved shoulders. For shoulder
or proposed width standards has been based on "reasonableness" widths between 0 and 12 ft, the percent reduction in related
and, in some cases, partly on expected safety and operational accidents resulting from widening paved shoulders was predicted
relationships. to be 16 percent for 2 ft of widening, 29 percent for 4 ft of
This report builds on previous work, focusing on the determi- widening, and 40 percent for 6 ft of widening. Predicted accident
nation of accident relationships with roadway width on low- reductions resulting from adding unpaved shoulders were slightly
volume roads, as well as on other highway operational require- less than reductions for adding paved shoulders (5).
ments considered essential to width standards. Another study that addressed low-volume rural roads was
one by Griffin and Mak (6), which attempted to quantify the
relationship between accident rate and roadway surface width
BACKGROUND SAFETY RESEARCH
on two-lane rural roads in Texas with ADT of 1,500 vpd or less.
Log-linear accident prediction models were developed for 36,215
Over the past 20 years, many studies have been conducted
miles of roadway within several ADT categories. Multivehicle
concerning the relative safety of various roadway widths. For
accident rates (accidents per mile per year) were not found to
example, one of the most comprehensive and more recent studies
be related to surface width for any of the ADT groups tested.
conducted to date on safety effects of roadway width was a 1987
Single-vehicle accident rates were found to increase as roadway
study by Zegeer, et al., for the FHWA (5), which involved an
analysis of 4,951 miles of two-lane roadway in seven states. width decreased for ADT groups between 401 and 1,500 vpd.
Statistical testing was used along with an accident prediction Accident reduction factors were developed for various widening
model to determine the expected accident reductions related to projects within these ADT ranges. Based on an economic analy-
various geometric improvements. sis, widening was not found to be cost-beneficial for ADT below
Accident types found in the Zegeer, et al. study to be most 1,000 vpd (6).
related to cross-section features included run-off-road, head-on, It is interesting to note that accident reduction factors devel-
and sideswipe (same direction and opposite direction) accidents oped by these two studies for lane widening generally were
(5). The roadway variables found to be associated with a reduced quite similar. The studies were reviewed in detail along with
to

approximately 70 other accident studies to better understand such were considered important, because no single data item alone is
accident relationships (see Appendix B for a complete literature the sole factor in resulting crashes. Instead, the interaction among
review). Considerable variation in results was observed, how- several roadway elements, and perhaps driver and vehicle factors,
ever, for some of the older studies. While past research has laid can affect the frequency and severity of accidents.
the groundwork for what is currently known on the subject, there The literature review revealed that in addition to lane and
is a need to look more closely at accident relationships for low- shoulder width, other factors found to affect safety on two-
volume roads (i.e., ADT of 2,000 vpd or less) including paved lane rural roads include shoulder type, roadside slope, roadside
and unpaved roads, and for roads in a variety of functional obstacles (placement, number, and rigidity), horizontal align-
classifications (arterial, collector, and local) with varying road- ment, vertical alignment, traffic volume (and mix of traffic),
way conditions. Also, there is a need to determine what specific narrow bridges, and location of intersections and driveways,
traveled way and shoulder width combinations provide reason- among others.
able levels of safety for various conditions. The identification of such safety relationships was useful in
developing an accident study plan. For this study plan, it is
important to select appropriate variables for data collection. For
RESEARCH OBJECTIVE AND SCOPE example, increasing lane width can decrease shoulder area and
recovery area if limited right of way is available. If shoulders
The objective of this research was first to develop an engi- are widened, roadside slopes may be steepened, possibly leading
neering analysis procedure that would allow for the determination to more severe crashes. Clearly, the trade-offs that exist from
of optimum roadway widths for construction and reconstruction what seems to be a simple and logical increase in lane 'width
of low-volume roadways (ADT less than 2,000 vpd). Based on must be studied together to determine optimal lane widths from
the analysis performed in this effort, recommendations should a safety perspective.
be developed for "minimum width of traveled way plus shoulder" Analysis issues identified for this study relate primarily to
(for consideration, by the design task force of the AASHTO safety, operational, and cost issues of various roadway widths.
highway subcorim-littee on design, for inclusion in future editions To address these issues, existing databases were reviewed to
select those that would enable the development of accident rela-
of the Green Book).
tionships with important roadway variables, so that expected
Thus, this study involved producing two separate, but related,
accident benefits could be computed for various roadway width
products. One product was a set of design tables specifying
alternatives. The primary data source selected was a database of
minimum width values for lanes and shoulders based on certain
approximately 2,400 miles of two-lane road from seven states
roadway factors (e.g., design speed, ADT, roadway classifica-
(with ADT of 2,000 vpd or below) from previous FHWA studies
tion) to modify and expand the tables now found in the AASHTO
(5, 7). This database was supplemented with data from approxi-
Green Book for ADT below 2,000 vpd. These tables should
mately 1,700 miles of paved and unpaved two-lane roadway
provide suggested minimum lane and shoulder widths for at least
(mostly local and collector streets) from North Carolina, Utah,
three ADT classes below 2,000 vpd.
and Michigan, for a total of approximately 4,100 miles. Field
To produce these numbers, another study output included tech-
data were collected on these supplemental sections, including
nical background and a framework for estimating the effects of
information on roadside safety, intersections, and driveways, as
various roadway widths on safety and operations. This should
described later. The resulting database subsequently is termed
enable informed judgments by design engineers in applying the
and referred to as the "primary database."
research findings to 3R-type problems or in assessing situations
A detailed statistical analysis was conducted on the primary
where unusual conditions may warrant design exceptions.
database, and accident rates were determined for various lane and
The efforts in this study relate to paved or unpaved two-lane
shoulder widths. To validate and to investigate these relationships
rural roads with ADT volumes of 2,000 vpd or less. Functional further, three independent databases totalling more than 54,000
classifications studied include arterial, collector, and local miles of low-volume, two-lane roads were obtained from three
roadways. states (Illinois, Minnesota, and North Carolina) and analyzed. A
project cost methodology was developed for roadway widening
projects for existing roadways and also for comparing roadway
RESEARCH APPROACH width alternatives on new road construction.
Finally, based on the accident analyses results, cost estimates,
The basic approach to this research involved first reviewing and other operational considerations, revised width guidelines
current AASHTO Green Book values on roadway width, as well were developed for new or reconstructed low-volume roads.
as other proposed guidelines, and the roadway factors considered Guideline also were developed for an evaluation of 3R design
to be important in the formulation of such standards. Next, a problems. Additional evaluations of the operational effects of
detailed review of safety literature was conducted to determine variable widths were performed. Issues investigated include
what is currently known about accident effects of roadway width speed related to lane width; width effects on highway level of
and other traffic and roadway features on two-lane rural roads. service; consideration of wider, larger vehicles; and compatibility
Safety effects of numerous features, in addition to roadway width, with alignment and other design standards and criteria.
 II

CHAPTER 2
STUDY FINDINGS

ANALYSIS ISSUES ANALYSIS OF HIGHWAY SAFETY ISSUES

The data analysis was structured to address the safety-related

The analysis in this study was aimed at determining appro- issues listed previously, and an appropriate database was devel-
priate design values for roadway width on rural roads carrying oped for use in statistical testing. Then, appropriate analyses
'
less than 2,000 vpd. Accomplishing this objective required quan- were conducted to quantify safety consequences of roadway
tifying the safety effects of various widths on such roads, and width and other roadway features. Details of these topics are
identifying and controlling for other important traffic and road- discussed next.
way variables that, with roadway width, affect accident experi-
ence. The results of this accident analysis should be used along
Development of the Primary Accident and
with project cost information (for roadway widening), vehicle
Roadway Database
operations, capacity of roadway designs, and other considerations
in the development of appropriate roadway width design values.,
Crash experience on rural highways is a complex function of
Thus, the specific issues addressed in this study include the
many factors, including those associated with physical aspects
following: Issue I —What are the characteristics of accidents of the roadway, and many other factors related to driver, vehicle,
on low-volume, two-lane roads? Issue 2—What accident types traffic, and environmental conditions. One 1978 study, for exam-
are related to roadway width? Issue 3—What traffic and road- ple, estimated that at least 50 roadway-related features could
way variables have a significant effect on accidents on low- have an effect on accidents (8). Note, however, in typical accident
volume, two-lane roads? Issue 4—What are the relationships analyses, there are often relatively few important traffic and
between accidents and various combinations of lane and shoul- roadway variables that show individually significant relation-
der width for paved, low-volume roads (i.e., roads with ADT ships with accidents.
below 2,000 vpd)? Issue 5—How are accident frequencies and
severities affected by having a paved road surface versus an
unpaved road surface? Issue 6—How does the presence of Needed Roadway and Traffic Variables
trucks influence operations and safety for various roadway
widths on low-volume roads? Issue 7—What are the expected
On the basis of their relationships to accidents developed in
accident benefits (i.e., accident reduction factors) on low-vol- past research, the traffic and roadway variables selected for data
ume roads for various amounts of roadway widening as a func- collection included section information (section identification
tion of ADT groups, percent trucks, vehicle speed, and other and length); pavement type (paved or unpaved); lane width,
roadway features? shoulder width, and type of shoulder (i.e., paved, gravel, or
In addition to safety, other operational and design requirements earth); terrain; type of area and development; design speed;
should play a role in the formulation of width standards. Issues functional roadway class; number of driveways (per mile); num-
8 through 14 address these additional requirements, as well as ber of intersections (per mile); percent trucks; speed limit; aver-
the manner in which all inputs are used to develop appropriate age annual daily traffic (AADT); horizontal alignment (i.e., per-
width-related standards: Issue 8—How do roadway width c-om- cent of the section with curvature greater than 2.5 deg.); vertical
binations influence capacity and level of service on low-volume alignment (i.e., the percent of the section with grade greater than
roads? Issue 9—How does vehicle speed (i.e., design speed 2.5 percent); sideslope ratio (2:1 and steeper, 3:1, 4:1, 5:1, 6:1,
versus running speed) influence accidents and vehicle operations or 7:1 and flatter); and measures of general roadside hazard (see
on different roadway widths for low-volume roads? Issue 10— discussion below).
What are the width-related operational requirements for shoul- The two measures of roadside hazard used in the data collec-
ders on low-volume roads? Issue ]]—What are the width-re- tion and analysis were termed "roadside recovery distance" and
lated operational effects of wide trucks? Issue 12—What are "roadside hazard rating." These measures were used in the 1987
the construction and maintenance costs associated with roadway FHWA study by Zegeer, et al. (5) on two-lane rural roads, and
widening in different terrain and for other conditions? Issue were both found to have a significant relationship to Accidents.
13—What roadway widths are most cost-effective relative to The researchers developed a seven-point roadside hazard scale
expected safety benefits on low-volume roads? Issue 14—What based on a literature review and on the results of a workshop
combinations of lane width and shoulder width are appropriate involving highway and roadside safety professionals. The road-
for design standards for low-volume roads? side hazard rating developed for the study was a subjective
12

V 3~

701 "W
A'—

Mi I

-V

Ro,.idside hazard rating of* 3 Roadside h,

.iz~ird i-ating of' 4

Figure 2(a). Sample photographs of the roadside hazard scale: hazard ratings 1, 2, 3, and 4.

measure of the hazard associated with the roadside environment. utility poles, culvert headwall, bridge rail, steep slope (i.e., 3:1
The rating values indicated the accident likelihood and damage or steeper), and so forth. Thus, like the "roadside rating," C, the
severity expected to be sustained by errant vehicles on a scale roadside recovery distance basically measures the degree of "for-
from one (low likelihood of an off-roadway collision or overturn) ,giveness" of the roadside.
to seven (high likelihood of an accident resultina in a fatality or Note that the foregoing list of variables includes measures of
severe injury). lane and shoulder width and shoulder type, which are of primary
The ratinos for the roadside hazard ratina used in this study concern in the development of roadway width values. Also in-
are based on a seven-point pictorial scale for rural highways (see cluded are other variables currently included in AASHTO width
Figures 2(a) and 2(b)). The data collectors chose the rating value values, such as functional class, design speed, and traffic volume
(one through seven) that most closely matched the roadside (ADT). The list also contains other important variables that re-
hazard level for the roadway section in question. In many cases. search has shown to affect accidents on two-lane roads, such as
the roadside hazard along a section varied considerably, so that horizontal and vertical alignment, roadside conditions, number
an "average" value was selected to represent the general roadside of intersections and driveways. and others. These other important
hazard rating of that section. For example, if ratings generally variables must be included in the accident analysis and predictive
range from four to six along a section, a rating of five was chosen models to obtain a more accurate assessment of the true effects
to represent the roadside hazard rating of the section. of roadway width on accidents.
In addition to the subjective roadside hazard rating, a measure
termed "roadside recovery distance" also was determined for
each section along with detailed data on the number of roadside Needed Accident Variables
obstacles by type and distance from the roadway. Although this
measure is relatively similar to a "clear zone" definition, it is Although many accident variables could have been chosen for
the lateral distance from the edgeline (i.e., outer edge of the analysis purposes, only those necessary for the analysis were
traffic lane) to the closest object that would cause a fixed object selected. For each roadway section, accident information in-
or rollover collision, which is the closest lateral distance to trees, cluded years of crash data (5 years in each case); total number
 13
- - __ . ~_ 1, ;%]
, -

1Y~l e

Roadside hazard rating of 7

Figure 2(b). Sample photographs of the roadside hazard scale: hazard ratings 5, 6, and 7.

of accidents on the section, number of accidents by severity unstable accident rates. Note that even with these longer section
(property-darnage only, A-injury, B-injury, C-injury. and fatal)~ lengths, many of the low-volume sections would have zero acci-
number of people killed; number of crashes by light condition dents in a given 5-year time period.
(daylight or dirkness)~ number of accidents by pavement condi- Sample size requirements were computed to enable detection
tion (dry, wet, or icy); and number of crashes by type (fixed of at least a 10-percent difference in accident rate between road-
object, rollover, other run-off-road, head-on, opposite-direction way width groupings at a significance level of 0.05 (i.e., a 95-
sideswipe, same-direction sideswipe, rear-end, backing or park- percent confidence level). The analysis revealed that a sample
ing, pedestrian or bike or moped, angle or turning, train related, of approximately 2,500 miles to 3,000 miles would be adequate.
aninial related, and other or unknown). Ultimately, a sample of 4,137 n-Wes was chosen for use in the
primary analysis. Independent samples of roadway sections also
were used to validate these accident relationships, as discussed
Sample Si7e Requirements later.

The data sample selected for analysis was a computer file

consisting of sections of two-lane roads. each with its correspond- Selection of the Database
ing roadway, traffic, and accident characteristics. This type of
database allows a comparison of the accident experience associ- The criteria used for selecting and developing a database for
ated with different roadway widths, paved versus unpaved road- the analysis were as follows: Criterion I —The database should
way surfaces, and so on. Ideally, each roadway section should consist of roadway sections that can be merged with correspond-
be of sufficient length to allow for calculation of accident rates ing traffic and accident data. Criterion 2—The database should
in ternas of accidents per million vehicle miles (accidents per have a substantial sample of unpaved as well as paved roads,
MVM). Section lengths of one mile or greater were chosen to and all sections should have ADT below 2,000 vpd; also, a wide
help ensure data stability because very short sections can yield range of roadway widths was needed to fulfill the sampling
14

requirements. Criterion 3—The accident data should be of good tions with a total of 4,137 miles, including 556 miles of unpaved
quality in terms of location accuracy, with a reasonably low and 3,581 miles of paved roads. The average section length was
reporting threshold. Criterion 4—Accurate computerized data 3.24 miles.
should be available for lane width, shoulder width, ADT, ter-rain, The selected roadway sections from the four databases con-
and other needed roadway variables; also, data on percent trucks tained many, but not all, of the needed data variables. Therefore,
and design speed should be available or be easy to obtain. field surveys were conducted on all of the supplemental sections
A detailed review was made of several existing databases on in Michigan, Utah, and North Carolina to obtain data on roadside
rural two-lane roads, including those developed for national condition (i.e., roadside hazard rating and roadside recovery
safety research studies (i.e., FHWA safety research), as well as distance), the number of driveways and intersections, and the
available computer databases from many state highway agencies width and type of pavement and shoulder. Also, for these sec-
and transportation departments. No single database was sufficient tions, state and local files and records were used to obtain data
by itself for the purposes of this analysis. Thus, several databases on ADT, percent trucks (where available), speed limit, and acci-
were obtained and were supplemented with field and other data dent data. For the seven-state cross-section database, state files
collection for this analysis. also were used to obtain functional class and percent trucks and
The bulk of the data came from the database on two-lane rural speed limit. However, design speed data were only available for
roads developed for the Transportation Research Board (TRB) sections in Montana.
and FHWA in the study, "Safety Effects of Cross-Section Design In summary, a computer data file consisting of 4,137 miles,
for Two-Lane Roads" (5). The database developed for this earlier which was created as described in the foregoing, represented the
effort is perhaps the most complete multistate database on two- "primary database" used for developing accident relationships
lane roads, in terms of roadway section representation, the with roadway width and other variables. Three other independent
amount of data sampled (4,137 miles of two-lane roadway sec- state databases also were selected to validate the accident rela-
tions in seven states), and the wide variety of accident, traffic, tionships found with the primary database. These included com-
roadway, and roadside variables for which data were collected. puterized databases from Illinois, Minnesota, and North Carolina
The FHWA database consists of a large sample of mileage (mileage in North Carolina excluded sections used for the pri-
on paved, two-lane roads from Alabama, Michigan, Montana, mary database for analysis and model building). These three
North Carolina, Utah, Washington, and West Virginia. Perhaps, validation databases did not contain roadside data, number of
the most pertinent data variables collected in that study, which driveways and intersections, and some other variables that were
are not available from standard state accident or roadway inven- contained in the primary database; however, they did contain
tory files, were those related to sideslope and roadside hazard. ADT, accident data, width and type of pavement and shoulder,
Such variables not only are important in assessing safety on two- and many other valuable variables.
lane roads, but they also are expensive and time-consuming to The validation databases were analyzed separately in detail,
collect. and their accident relationships were compared to those of the
While the cross-section data file provided a valuable source primary database as discussed later.
for the required analysis file, it alone was not sufficient. The Sample sizes of the validation databases were 3,791 miles in
analysis in the current study is concerned only with two-lane Illinois, 36,247 miles in Minnesota, and 14,134 miles in North
roads with 2,000 vpd or less, and the FHWA cross-section data- Carolina. A substantial sample of unpaved roads was available
base provided only approximately 2,700 miles of the 4,137-mile from Minnesota (11,948) with a lesser amount from North Caro-
sample in this lower ADT range. Also, it had no samples of lina (455 miles). Thus, a total of 58,309 miles of roadway was
unpaved roads, and inadequate samples of roads with "local" analyzed, which included 21,827 total sections. The primary
functional class and within a very low ADT range (particularly database contained numerous detailed roadway variables (i.e.,
ADT below 750 vpd). Thus, other data sources were needed to roadside data and number of intersections and driveways) not
fill these gaps. available on the validation databases.
Three state/local databases were selected to supplement the
cross-section database: State of North Carolina, State of Utah,
and Oakland County (Michigan) database. Statistical Testing and Results
Each of these databases contained a substantial sample of
unpaved and low-volume sections with "mergeable" data files The data analysis focused on addressing each of the analysis
and good-quality data. They also contained many of the needed issues listed earlier. The results are summarized in the following
roadway, accident, and traffic data variables. These three states discussion. More details of analysis activities are given in Appen-
also had provided data to the FHWA cross-section database. dix C.
Selection of additional sections in three of the seven "cross-
section" states prevented additional state biases to be introduced
from different state reporting thresholds, state coding practices, Issue .1—Characteristics of Accidents on Low-
or other factors. The wide variety of climate, driver characteris- Volume Roads
tics, roadway design practices, and other factors contained within
the seven states were considered to ensure a diverse sample of The total sample of accidents that occurred on the primary
roadway and traffic conditions. sections in a 5-year period is given in Table 9, along with accident
Within the three supplemental computerized databases, road- summaries by type, severity, light conditions, and pavement
way sections were selected as needed to fill the data "gaps." conditions. A total of 14,888 accidents was noted on the 4,137.19-
This resulted in sample sizes for the primary database, as shown mile primary database, which included 8,973 property-damage-
in Table 8. The primary database contained 1,277 roadway sec- only accidents (60.3 percent of the total); 5,632 nonfatal injury
 15

TABLE 8. Summary of sample sizes for primary and validation databases

Number of Miles Number of Sections

Database Paved Unpaved Totals Pav;; Unpaved I Totals

Primary 3,581.05 556.14 4,137.19 1,113 164 1,277

Illinois
(Validation) 3,790.86 0 3,790.86 1,795 0 1,795
Minnesota
(Validation) 24,298.92 1 11,948.58 36,247.50 8,911 4,900 13,811 1

North Carolina
(Validation) 13,678.30 455.40 14,133.70 4,708 236 4,944

Totals 45,349.13 12,960.12 58,309.25 16,527 5,300 21,827

TABLE 9. Summary of accident types and characteristics for low-volume road sites

Primary Database on Cross-Section

Low-Volume Roads Database
Percent Percent
Number of of Total Number of of Total
Accident Type Accidents Accidents Accidents Accidents

Total 14,888 100.0 62,676 100.0

Property Damage Only 8,973 60.3 38,857 62.0
Injury 5,632 37.8 22,944 36.6
Fatal 283 1.9 875 1.4

Injuries' 8,768 N/A 37,321 N/A

Fatalitiee 328 N/A 1,068 N/A

Daylight 8,050 54.1 37,402 59.7

Dawn/Dusk 820 5.5 2,888 4.6
Dark with Lights 160 1.1 2,770 4.4
Dark without Lights 5,809 39.0 19,496 31.1
Light Unknown 49 0.3 120 0.2

Dry 10,306 69.2 41,957 66.9

Wet 2,442 16.4 13,487 21.5
Snow/lee 1,952 13.1 6,657 10.6
Unknown Pavement 188 1.3 575 0.9

ROR - Fixed Object 4,017 27.0 12,091 19.3

ROR - Rollover 1,999 13.4 4,245 6.8
ROR - Other 2,287 15.4 2,840 4.S
Head-On 475 3.2 2,113 3.4
Opposite Direction Sideswipe 642 4.3 2,997 4.8
Same Direction Sideswipe 330 2.2 2,288 3.7
Rear-End 893 6.0 12,420 19.8
Parking/Backing 264 1.8 1,155 1.8
Ped/Bike/Moped 117 0.8 655 1.0
Angle & Turning 1,773 11.9 14,730 23.5
Train 20 0.1 47 0.1
Animal 1,404 9.4 5,212 8.3
Other or Unknown 1 667 1 4.5 1,883 1 3.0

'rbe data for these variables represent the number of People injured or killed, and not the number of
accidents.
N/A Not Applicable.
ROR Run-off-Road.

accidents (37.8 percent); and 283 fatal accidents (1.9 percent). volume roads versus the full sample of 4,785 miles of rural two-
A total of 8,768 people were injured and 328 people were killed lane roads in the database from the earlier FHWA study (with
in these 14,888 accidents. A comparison was made of the percent- a full range of ADT). The sample of low-volume roads had a
age of crashes by severity between the 4,137 miles of low- slightly higher percentage of injury accidents (37.8 versus 36.6
16

percent) and fatal accidents (1.9 versus 1.4 percent) than the full accident types that are associated or related significantly to road-
sample. way width. The independent roadway variables used in the analy-
Similar comparisons also were made between low-volume sis were lane width, shoulder width, terrain, and roadside haz-
roads and the full cross-section rural sample for other accident ard rating.
variables, as shown in Table 8. Both samples showed high per- Accident rates were found to be associated significantly with
centages of accidents when no roadway lighting was present; varying lane and shoulder widths for single-vehicle accidents
low-volume roads had a higher percent of nighttime, no-lighting and opposite-direction accidents. Rates of other accident types
accidents than the higher volume, cross-section sample (39.0 were found to not be significantly related to lane or shoulder
versus 31.1 percent). The greater difficulty in driving on low- width. These findings agree closely with the 1987 study by
volume roads at night, because there are fewer opportunities for Zegeer et al. of rural, two-lane roads—that is, all ADT ranges
"caravaning," could have contributed to this result. Also, there in that study, not just for ADT of 2,000 vpd or below (5).
may typically be a reduced level of nighttime visibility on low- However, that study not only related single-vehicle and oppo-
volume roads associated with less delineation and less-frequent site-direction accidents to roadway width, but also found that
overhead lighting compared to higher volume roadway sections. "same-direction sideswipe" accidents were marginally signif-
The low-volume sample also showed a slightly higher proportion icant.
of snow and ice accidents than the full rural sample (13.1 versus In virtually all of the remaining analyses, single-vehicle and
10.6 percent), possibly because of better snow removal opera- opposite-direction accidents were combined and were referred to
tions on higher volume routes. as "related accidents." To investigate accident severity questions,
In terms of accident types, low-volume roads have a greater certain types of injury accident rates (for all injury accidents on
percentage of fixed-object crashes (27.0 versus 19.3 percent), a section) also were analyzed.
rollover crashes (13.4 versus 6.8 percent), and other run-off-road
crashes (15.4 versus 4.5 percent) than the full sample of rural
roads. This may be expected because there are fewer other vehi- Issue 3—Important Traffic and Roadway Variables
cles to strike on low-volume roads than on higher volume routes.
Conversely, the data showed a lower percentage of crashes in- The next issue involved identifying what traffic and roadway
volving rear-end collisions and angle and turning collisions for variables have a significant effect on accidents for low-volume,
low-volume roads (6.0 versus 19.8 percent and 11.9 versus 23.5 two-lane roads. This question was important for two major
percent, respectively). reasons.
The average accident rate for the total primary database was First, correctly quantifying the accident effect of roadway
3.47 accidents per MVM; the rates for paved sections and un- width requires that all other important traffic and roadway vari-
paved sections were 2.86 and 7.59 accidents per MVM, respec- ables (confounding variables) also be accounted for or controlled.
tively. The average number of accidents per mile per year was For example, assume that wider roadways and safer roadsides
0.79 for the total sample, and was somewhat higher for paved are both associated with lower accident rates. Also assume that
sections (0.84) than for unpaved sections (0.69), due to higher wider roadways (higher class roads) typically have safer road-
traffic volumes on paved roads than on unpaved roads (and, sides as well. If one simply tries to find the relationship between
consequently, a greater number of opportunities for crashes). roadway width and accidents without controlling for roadside
condition, the results will show an exaggerated, erroneous effect.
In other words, the computed accident benefit of wider roadways
Issue 2—Detennining Related Accident Types would, in fact, be the combined accident benefit of wider roads
plus better roadsides. Thus, the effect of roadside hazard (or
The next analysis issue concerned the identification of specific other important traffic and roadway variables) must be accounted
accident types that are related to roadway width. This issue is for in the analysis to provide an accurate indication of roadway
important in that wider roadways may logically be expected to width effects on accidents.
reduce some accident types (head-on or run-off-road crashes) The second reason for determining effects of other roadway
but not necessarily others (rear-end or angle). For example, if variables (besides roadway width) on accidents is to gain a better
widening lanes on low-volume, two-lane roads does not affect understanding of how safety can be improved on low-volume
rear-end collisions, the rear-end accident type should be excluded rural roads. For example, it would be important to know that
before developing relationships between accidents and road- roadside hazard, horizontal alignment, or other roadway features
way width. affect accidents, in addition to roadway width. This would allow
The following accident types were considered in the investiga- engineers and designers to have more justification for considering
tion of what types may be related to roadway width: single- improvements to roadsides (i.e., cutting trees near the road, flat-
vehicle accidents (fixed-object, rollover, and other run-off-road); tening steep slopes), improving horizontal curves (i.e., improving
opposite-direction accidents (head-on and opposite-direction superelevation), and other improving roadway features on low-
sideswipe); angle accidents (vehicles on the main roadway strik- volume, two-lane roads in addition to (or instead of) roadway
ing at an angle with turning vehicles or vehicles from driveways widening (i.e., as a part of 3R or major reconstruction improve-
or intersections); same-direction accidents (sideswipe and rear- ments).
end); and other accident types (vehicle strikes pedestrian, bicycle, The traffic and roadway variables found to be significantly
animal, or train; backing or parking collisions; and others). All related to the rate of "related" accidents (i.e., rate of single-
accident rates for each accident type were expressed in terms of vehicle plus opposite-direction accidents) included lane and
accidents per MVM. shoulder width (or total roadway width); roadside hazard rating
Analysis of covariance accident models was used to identify and roadside recovery distance (which were essentially inter-
 17

changeable, with only one being used in a particular analysis); and accidents. An analysis of covariance models was used to
number of driveways per mile; terrain, with values of I for flat, estimate rates of related accidents as a function of lane and
2 for rolling, and 3 for mountainous terrain; and state—which shoulder width, while adjusting for roadside hazard rating, ter-
consists of three discrete values representing the following three rain, state, and driveways per mile. The following discussion of
groups of states (states within a group were similar with respect lane and shoulder effects pertains to paved roads only, where
to related accident rate): (1) Alabama, Montana, and Washing- shoulders are either paved or unpaved. The lane and shoulder
ton; (2) North Carolina and Michigan; and (3) Utah and West width refer to average width on one side of the road. For example,
Virginia. a shoulder width of 6 ft refers to a 6-ft shoulder on each side
Variables for percent grade and curvature were not considered of the road. Because shoulder type was not found to affect
for further analysis because they were available for only about accident rate significantly on low-volume roads, shoulder width
half of the study sections. The terrain variable was significant as used in these analyses corresponds to the total width of each
and served as a general measure of alignment for use as a control shoulder, regardless of the shoulder type.
variable. The results revealed that lane and shoulder width each has a
Variables that were not found to be associated significantly significant effect on related accident rate. The lane width catego-
with accidents on low-volume roads were intersections per mile ries used were 8 ft or less, 9 ft, 10 ft, I I ft, 12 ft, and 13 ft or
(i.e., most sections had no major intersections), speed limit (i.e., more. Shoulder widths were combined into five categories: 0 ft,
most sections had 55-mph speed limits whether posted or not, I to 2 ft, 3 to 4 ft, 5 to 6 ft, and greater than 6 ft. Some analyses
regardless of the alignment or design speed), and percent trucks were conducted for various combinations of lane and shoulder
(i.e., very few of the sections had a substantial volume of heavy width, termed "total roadway width."
trucks). The functional class variable was found to relate highly Two separate models were developed for related accident rate
to roadway width (i.e., higher functional classes generally have by total roadway width, as shown in Figure 3. Model I represents
wider roads) and state (i.e., some states tended to assign the the estimated rate of related accidents for various widths of
same one or two functional class categories to all their low- roadway (lanes plus shoulders) while controlling for state and
volume roads, but such designations differed from state to state). terrain as class variables, and roadside recovery distance and
Thus, the accident effect of functional class actually resulted number of driveways per mile as continuous variables. For Model
from state and width effects. 11, state and two levels of functional class (local versus all others)
It is also interesting to note that shoulder type (paved versus were taken as class variables, while terrain, roadside hazard
unpaved shoulders) was not found to affect accidents signifi- rating, and driveways per mile were included as continuous
cantly on low-volume roads. The 1987 study by Zegeer, et al., variables. Note that both models have the same general shape
did find a small (generally less than 10 percent) but significant for related accident rate by roadway width. Also, the rate of
accident reduction from paved shoulders compared to unpaved related accidents tends to decrease as roadway widths increase
shoulders on higher volume, two-lane roads. These two seem- from 20 ft to 32 ft for both models. However, the rate for the
ingly different findings may actually be explainable for several most narrow roadway widths of 18 ft or less was much lower
reasons. First, because higher volume roads (i.e., typically higher than that for most wider roadways. Also, no clear accident reduc-
class) tend to carry more of the heavy truck traffic than lower tion was found for roadway widths greater than 32 ft.
volume roads, paved shoulders would logically be more benefi- The adjusted rate of related accidents was next determined
cial on higher volume roads to handle truck encroachments over for various categories of roadway width, as shown in Figure 4.
the edgeline. Also, because higher volume roads typically have The rate for roadways of 18 ft or less was 1.43 compared to
better alignment and, subsequently, higher speeds than lower 1.90 for 19-ft to 24-ft roads, 1.41 for widths of 29 ft to 39 ft,
volume roads, paved shoulders may be particularly beneficial to and 1. 10 for greater than 40-ft roadway widths.
assist higher speed vehicles that run over the roadway edge. It Next, accident relationships (rates of related accidents) were
is also interesting that AASHTO criteria for roadway width determined for various categories of lane and shoulder widths.
specify paved shoulders for higher class roadways (typically Lane and shoulder width groupings were determined on the basis
higher volume), but allow unpaved shoulders on lower class of available sample sizes and by considering where significant
roads (typically lower volume). accident differences exist. The resulting rates are shown in Figure
In summary, the results of these analyses indicate that acci- 5. Lane widths of 8 ft and 9 ft were grouped together for all
dents on low-volume roads are affected primarily by roadway shoulder widths, and the resulting accident rate was 1.66. This
width, roadside hazard, roadway terrain (i.e., alignment), and the was lower than the rate for 10-ft lanes with narrow (zero to 4
number of driveways per mile. ft) shoulders, which had a rate of 2.41. Related accidents per
All of these roadways features should be considered for possi- MVM for wider shoulders (greater than 5 ft) with 10-ft lanes
ble improvement, as needed, in conjunction with 3R or major resulted in a rate of 1.43. Note that the sample size of 8-ft lane
reconstruction projects. State differences also can affect accident widths was very small in the primary database. Further review
rates, perhaps because of accident reporting differences, driver of accident rates from several validation databases was helpful,
differences, weather differences, roadway maintenance practices, as discussed next.
or other differences. No significant difference was found between I 1-ft and 12-ft
lane widths; therefore, these lane widths were grouped together.
Issue 4—Accident Effects of Lane and Shoulder Rates of related accidents on 11 -ft or 12-ft lanes were 1.87 for
Width on Paved Roads narrow (2 ft or less) shoulders and 1.31 for wide (3 ft or greater)
shoulders. Lane widths of 13 ft resulted in rates of 1.57 for
A detailed analysis was conducted on the primary database narrow (0 to 4 ft) shoulders and 0.76 for wider (5 ft or greater)
to quantify the relationships between lane and shoulder width shoulders.
18

Model 1

3.0 Model 11
2.83
2.64
2.52
2.5 2.43 2.37 2.36
> ED 2.41 2.28
2.23
2 2.14
2-0 2.17 1.87 1.92
1% Z 2.03 1.85
1.87 1.83 1.91 1.49
12
0 1.5 1.45
iz U 1~ 1.59 \
< 1.44 1-0 1.13
'Z 1.37
< 1.0 1.19
1.05
0.5

0.0
< 18 20 22 24 26 28 30 32 34 36 38 40 > 42

ROADWAY WIDTH (FT.)

(WIDTH OF LANES PLUS SHOULDERS)
Figure 3. Accident rates by roadway width from two different models.

2.0 I on than they are for 10-ft lanes, with only a slight accident reduction
1.8 for lanes of 13 ft or wider.
The lower accident rate for roadways with 8-ft and 9-ft lanes
1.6
was unexpected, open to question, and thus warranted further
1.4
Z investigation using other independent databases. These investiga-
1.2
tions included a similar analysis of related accident rates by lane
Z U oei Oemider width caragoTies using "validation" daiabases of
8 '! o,8 paved, low-vnInmr, mads frnm three, states- Illinois ( ' 3.791
0.6 miles), Minnesota (24,299 miles), and North Carolina (13,678
0.4 miles).
0.2
Illinois data showed trends in accident rates similar to the
primary database, by lane and shoulder width, for lane widths
0.0
of 9 ft to 13 ft or greater. Lower accident rates were found for
<18 . 19 to 28 29 to 39 > 40
9-ft lane widths (all shoulder widths) compared to 10-ft lanes
ROADWAY WIDTH (FT.) with narrow shoulders (rates of 1. 13 versus 2.03 accidents per
(WIDTH OF LANES PLUS SHOULDERS) MVM, respectively). However, 8-ft lane widths in the Illinois
Figure 4. Least-squares mean accident rates by roadway database had a much higher rate (3.57 accidents per MVM) than
widths. other lane width categories. Accident rates then decreased for
roads with lane widths (narrow shoulders) greater than 10 ft.
The results of the Minnesota analysis revealed trends similar
to those noted for the Illinois data. Again, rates of related acci-
. Note that different width values were used for narrow and dents were highest(2.32 accidents per MVM) for 8-ft lane widths
wide shoulders, depending on specific lane width categories. (for all shoulder widths), dropped to 0.85 for 9-ft lanes, and
This is because shoulder width categories were determined based increased to 1.03 for 10-ft lanes with narrow shoulders. The
on actual accident rate differences. Selecting uniform shoulder accident rate again dropped for lane widths of 11 ft (0.67 acci-
width categories (e.g., 0 to 4 ft = narrow, 5 ft or greater = wide) dents per MVM) with narrow shoulders, with no further decrease
for all lane width categories would be subjective in nature and in rate for lane widths of 12 ft to 13 ft. Again, for a given lane
would not reflect where actual differences occur. width, wider shoulders were associated with reduced accident
experience.
The North Carolina data showed rates of related accidents to
Validation ofAnalyses Results. In terms of lane width effects, be constant for lane widths of 8 ft or less (1.95 accidents per
the analysis revealed that 10-ft lane widths with narrow or no MVM) and 9 ft(1.94 accidents per MVM). The rate then dropped
shoulders have higher accident rates than 9-ft lane widths on to 1.73 for 10-ft lane widths, and to 1.69 for 11 -ft and 12-ft lane
low-volume roads with any shoulder width. The 8-ft lane widths widths. Shoulder widths of 5 ft or greater were associated with
showed similar low rates as 9-ft lane widths, although the sample reduced accident rates.
sizes were smaller. Further, for sections with 0- to 4-ft shoulders,
accident rates are significantly lower for I 1-ft and 12-ft lanes Discussion of Results. The results from the analysis of the
 19

Database: Low Volume Roads

0 All shoulder widths
Shoulder Width Categories:
Varied
W Narrow (N)

2.41 ADT Catetorv: < 2,000 1E Wide (W)

2.5

2.0
cn
Z
Z
1.5
a
U
U
1.0

0.5

0.0
SHOULDER 0-2 > 3 >5
All 0-4 >5 0-4
WIDTH (FT.)

LANE 8 and 9 10 11 and 12 13

WIDTH (FT.)

Figure5. Rates of related accidents by lane and shoulder width from the low-volume roads
database.

primary and validation databases have several important implica- the important "break points" (or categories of shoulder width)
tions concerning safety effects of various lane and shoulder varied for different lane widths and databases. However, only
widths. First, based on the primary database, the presence of a the North Carolina database did not show an accident rate lower
shoulder is associated with a significant accident reduction for for 9-ft lane widths than for 10-ft lane widths, including adjust-
various lane width categories, particularly for shoulder widths ments for shoulder width.
of at least 3 ft to 4 ft. This was found to be true for lane widths The effects of lane width on accident experience using the four
of between 10 ft and 13 ft or greater. For 10-ft lanes, a shoulder databases (the primary database plus the three state validation
of 5 ft or greater appears to be needed to affect accident rate databases) were compared with the findings from the 1987
significantly. For 11-ft and 12-ft lane widths, shoulders of 3 ft FHWA study by Zegeer, et al., of two-lane rural roads (all ADT
or greater have significantly beneficial effects. For lane widths ranges). The results in that 1987 study used a mathematical model
of 8 ft and 9 ft, the shoulder width relationship is less clear. An that indicated a "smoothed" relationship of reduced accident
inadequate sample of 8-ft lane widths was available within the experience for wider lanes (for lane widths between 8 ft and 12
primary database. Eight-foot lane widths were, therefore, com- ft) and shoulders. This model resulted in an assumed reduction
bined with those of 9 ft for analysis purposes. Because the overall of 12 percent in related accidents for 1 ft of widening between
sample of 9-ft lanes and narrow shoulders had a lower accident 8 ft and 12 ft. However, a more in-depth analysis of the four
rate than the 10-ft lanes, it may be concluded that roads with 9- low-volume roads databases indicates that high accident rates
ft lane widths also may contain narrow shoulders. exist for roads with 8-ft lanes from the Illinois and Minnesota
With respect to lane width, two of the three validation data- databases. Note that the sample size of paved 8-ft lanes is. small
bases (Illinois and Minnesota) support the finding of a reduced (43 miles) for these two States. The low-volume databases also
accident rate for 9-ft lane widths compared to 10-ft lanes with show that accident rates level off at a lane width of 11 ft, with
narrow shoulders. However, these two databases differ from the little additional reduction for lanes of 12 ft or greater (primary
primary database analysis in the findings for 8-ft and 9-ft and and validation databases). Also, the low-volume databases show
wider lane widths, for which the 8-ft lane widths had substantially that 10-ft lanes with narrow or no shoulders produce accident
higher rates. Inasmuch as the primary datitbase had a very low rates that are higher than rates for 9-ft lanes with the range of
sample of 8-ft lanes (83 miles), the low rate is not considered shoulder widths. Although the 1987 study did not show this
reliable. Also, the primary database and the same two validation finding, the study did confirm the safety benefits of wider shoul-
databases both show that 11 -ft lane widths have substanlially ders (with regard to accidents).
lower accident rates compared to 10-ft lane widths, particularly The analysis results of these four low-volume databases gener-
where narrow shoulders exist. It is also clear from those databases ally agree with engineering intuition. Wider shoulders logically
that little, if any, real accident benefit can be gained from having result in reduced accident rates because drivers have more room
lane widths wider than 11 ft on low-volume roads. to recover after encroaching over the edgeline. Lanes of I I ft
The North Carolina database revealed some similar trends, or wider produce lower accident rates than 10-ft lanes, which is
but also some conflicting trends. For example, North Carolina again intuitively expected. The fact that - 12-ft and 13-ft lanes
data supported the finding of all other databases that increases appear to offer minimal accident reduction compared to I I -ft
in shoulder width reduced rates of related accidents, even though lanes on low-volume roads agrees with results of a 1979 study
20

by Zegeer et al.(9)of over 10,000 miles of rural, two-lane roads Paved

in Kentucky. Unpaved

The main issue in question concerns the lower calculated

3.95
3.88
accident rates for roadways with 9-ft lanes compared to those 4.0
with 10-ft lanes. There are two possible explanations for this 3.5
counter-intuitive finding. First, speeds on these narrower road-
3.0
ways may be lower, reflecting not only the width effect of speed,
2.5
but also other variables such as functional class and terrain. The
majority of roads with such narrow lanes may be more local in 2.0
U
character, carrying lower speed, local traffic. (Note that no speed 1.5
U
data were collected as part of this study.) Roadways with 10-ft 1.0
lanes are commonly found on higher class facilities, which tend
0.5
to operate at higher speeds.
A second possible explanation also is related to functional 0.0
~~ 9 10to 11 > 12
class. It may be that many of the roadways with 9-ft lanes are
of the local, very low-volume nature (i.e., with ADT of 200 vpd LANE WIDTH (Fr.)
or less). Nominal traffic volume data for such roads could be
Figure 6 Rate of related accidents for paved versus unpaved
more variable and less reliable. than for the rest of the study
roads by lane width.
sample.
Accident rate trends for roadways with 8-ft lanes are not
entirely clear. There is an indication from two databases (Illinois
and Minnesota) that 8-ft lane widths are associated with consider-
ably higher rates than lane widths of 9 ft or 10 ft. This finding, rates were higher for roads with less roadside recovery distance,
however, is based on a small sample of 8-ft lane widths in those and with more narrow roadway widths.
States. Two other databases (primary and North Carolina) show For each of three lane width categories (less than 9 ft, 10 ft
the rate on 8-ft lanes to be similar to the rate for 9-ft lanes. The to I I ft, and 12 ft or greater), unpaved roads had higher rates
primary database, however, contained only a small sample of 8- of related accidents than paved roads (see Figure 6). This was
ft lane widths and, therefore, was not considered very reliable. also true using rate of injury accidents.
The true effect of these "very narrow" lane widths (8-ft or less) A separate comparison of various lane and shoulder widths
probably relates to many factors, such as volume of nonlocal and also was made using the rate of total injury accidents (where all
truck traffic, roadway alignment, roadside conditions, presence of accident types were included involving an injury or death, not
roadway delineation, and so on. These factors could not be just "related" accident types). Models were fit to three types of
explored any further with the primary database. injury rates, including total injury accidents per MVM (i.e., injury
In any case, the analysis results would tend to support the rate), accidents resulting in K, A, or B injuries per MVM (i.e.,
continued use of 9-ft lanes on some roadways that have accept- KAB rate), and accidents involving K or A injuries per MVM
able accident rates. It appears that widening an existing roadway (i.e., KA rate). The original 29 lane and shoulder width categories
with 9-ft lanes to provide 10-ft lanes can not be expected to were used in the model along with state, terrain, clear recovery
improve its safety, unless such widening is accompanied by a distance, and driveways per mile.
shoulder width of at least 5 ft. Widening lanes from 10 ft or less The pattern found in the variation of injury rates across lane
(particularly those that have little or no shoulders) to 11 ft plus and shoulder width categories was similar to the pattern noted
greater than 3-ft shoulders appears effective in terms of expected for rate of related accidents. This similarity suggests that lane
reduced accident rates. and shoulder width do not affect the distribution of accident
severity. This result may be because narrow low-volume roads
may have basically the same type of accidents as wider low-
Issue 5—Paved Versus Unpaved Road Surface volume roads (primarily fixed-object, rollover, head-on), even
though a higher rate of these accident types occurs for some of
The next question involved comparing accident experience on the more narrow width categories (particularly 10-ft lane widths
low-volume roads for paved versus unpaved roadway surfaces. with narrow shoulders). This would explain the fact that roadway
From the primary database, rates of related accidents were com- width has a similar effect on rate of related accidents and rate
pared between paved and unpaved sections from the same three of injury accidents.
states where both types of sections were available (Michigan, Next, a comparison was made between paved and unpaved
Utah, and North Carolina). Three different accident rate models roadways for various ADT categories to determine the levels of
were used to compare the safety of paved versus unpaved roads. traffic at which paved surfaces were most beneficial. Rates of
Again, each analysis controlled for important traffic and roadway related accidents were computed for paved and unpaved road-
variables such as state, terrain, roadside recovery distance, and ways for three ADT groups: ADT less than 250 vpd, ADT
roadway width. between 250 and 400 vpd, and ADT greater than 400 vpd. Again,
Accident rates were found to be greater for more severe terrain. adjustments were made for important roadway variables. On
Roads in mountainous terrain had greater rates than those in roadways with ADT less than 250 vpd, accident rates did not
rolling terrain, all other variables being held constant or con- differ significantly between paved and unpaved roads. However,
trolled. Similarly, roads in rolling terrain had greater accident for ADT above 250 vpd, rates for unpaved roads were signifi-
rates than those in flat terrain. The analysis also showed that candy higher than for unpaved roads. Note that the top ADT
 21

0.98 ADT above 250 vpd, no effect of pavement surfacing was found
1.0-
from a larger database (Minnesota). Thus, it appears that the
0.82 decision to surface a roadway should be made on a case-by-
0.81 0.78
0.8- 0.74 case basis and should include consideration of surfacing and
> El 0.69
0.64 maintenance costs, ADT,accident experience on the roads, road-
~ZO
72
way alignment (in terms of whether it can handle increased
0.6-
0.63 speeds safely), and other considerations.
0.57 0.58
There is some indication that the width of unpaved roads also
Z -It 0.4- 0.46 can affect accident rates. It appears that while accident rates can
fluctuate considerably for narrow roadways, rates for roadway
0.2- widths of 20 ft or less are low. This may occur as a result of
reduced vehicle speeds on very narrow, unpaved roads. As widths
0.0
increase to about 30 ft, accident rates can increase, perhaps with
S18 20 22 24 26 28 30 32 34 36 38 2~ 40
increases in vehicle speeds. As widths further increase above 30
ft, rates seem to decrease again.
ROADWAY WIDTH (FT.)

Figure 7. Relationship between accident rate and total road- Issue 6—Safety Effects of Large Trucks
way width for unpaved roads in Minnesota. The variable indicating the percent of truck traffic was not
found to be significantly associated with related accident rate in
any of the analyses of the primary database. Although this finding
may, in part, result from the limited variation in the variable and
groups (between 250 and 400 vpd, and greater than 400 vpd) the lack of detailed information concerning oversized trucks or
had to be combined because of small sample sizes in the greater truck type, it is also suspected that the quality of such data on
than 400 vpd ADT group. lower volume roads may be less than for higher volume roads.
The results of this analysis from the primary database provide It is possible and, indeed, intuitive that roadway width is an
some indication that roadways with ADT above 250 vpd should important variable influencing truck accident experience in some
be paved to provide reduced accident experience. However, to situations. Discerning these effects or situations was beyond the
confirm this result, data from Minnesota also were used to com- capabilities of the database assembled for this study.
pare the effect of pavement type (paved versus unpaved). This
analysis made use of a large sample of paved (24,300 miles) Issue 7—Expected Accident Benefits on Low-
and unpaved (11,900 miles) roads. The results, however, showed Volume Roads
no significant effects of pavement on accident rate (p-value of The best linear accident model forms the basis for estimating
0.3980). expected accident effects of variable roadway widths on low-
Another question concerned how total roadway width on un- volume roads, as shown in Figure 5. Table 10 gives the accident
paved roads affects accidents. Using the unpaved road samples rates from this model, compiled by lane width and shoulder
from only the primary database, rates of related accidents were width, that were used in the cost-effectiveness analysis. Annual
much lower on roadways with a total width of less than 18 ft accident benefits are calculated by taking the difference in rates
(1.72 accidents per MVM) compared to roadways of 20 ft to 22 implied by any two cells of Table 3, multiplied by the annualized
ft (3.95 accidents per MVM) or 24 ft or greater (3.88 accidents traffic volume level being analyzed.
per MVM). Similar trends were found using rates of injury The rates reflect a sensitivity to terrain, which is considered
accidents. Thus, increased roadway width seems to have a nega- an indicator of alignment quality. In other words, in more difficult
tive effect, the reverse of the finding for paved roads. terrain, horizontal and vertical alignment tend to be more severe,
As with the previous discussion of very narrow lane widths, producing higher overall levels of accidents.
speed may be an explanation for what appears to be a counter
intuitive finding. Unpaved roads that are very narrow are proba- HIGHWAY OPERATIONAL CONSIDERATIONS
RELATED TO DESIGN
bly driven at very low speeds. Wider, unpaved roads may encour-
age higher speeds, thereby increasing the potential for accidents. Appropriate width values for design are related not only to
A comparison also was made of the effect of width on unpaved safety, but also to other traffic operational considerations. These
Minnesota roadway using similar modeling techniques. As operational considerations include width effects on vehicle
shown in Figure 7, the adjusted rates fluctuate considerably for speeds, capacity and level-of-service implications of widths, ve-
roadway widths of less than 18 ft to 30 ft. This could be due to hicle composition and lane width effects, and shoulder functional
some extent to speed differences associated with various widths requirements.
and, perhaps, to the condition of the surface, percent of local Current design policy established the framework within which
(familiar drivers), and other factors. The rate of related accidents both operational and safety concerns must fit. A highway's design
tends to decrease as total roadway width (i.e., width of lanes is affected significantly by its functional classification and de-
plus shoulders) increases from 30 ft to greater than 40 ft (lane sign speed.
widths of 15 ft to 20 ft).
Functional Classification
In surnmary, the effects of paved versus unpaved surface are
somewhat unclear based on an analysis of the various databases. The concept of functional classification is central to the devel-
While there are indications from one database (the primary data- opment of all highway design standards. As stated in the 1990
base) that paved surface can reduce accidents for roads with AASHTO Policy:
22

TABLE 10. Accident rates used for analysis of cost-effective widths on low-volume roadsa

Accident Rates (Related Accidents per MIM

Terrain
Lane Width Shoulder Width
(feet) (feet) Level Rolling Mountainous

0-2 2.28 2.41 2.86

10 3-4 2.28 2.41 2:86
>4 1.30 1.43 1.88
0-2 1.74 1.87 2.32
11 3-4 1.18 1.31 1.76
1
>4 1 1.18 1.31 1.76
0-2 1.74 1.87 2.32
12 3-4 1.18 1.31 1.76
>4 1.18 1.31 1.76
'Values in this table are based on the accident model produced in the analysis of the primary database.

Under a functional classification system, standards and level speed, however, should not be assumed for a secondary road where
of service vary according to the function of the highway facility. the topography is such that drivers are likely to travel at high speeds.
Volumes serve to further refine the standards for each class. Drivers do not adjust their speeds to the importance of the highway,
Arterials are expected to provide a high degree of mobility for but to their perception of the physical limitations and traffic thereon
the longer trip length. Therefore, they should provide a high op- (AASHTO, 1990, pp. 63-64).
erating speed and level of service. Since access to abutting property
is not their major function, some degree of access control is desirable Table 11 provides general design speed guidelines for various
to enhance mobility. The collectors serve a dual function in accom-
modating the shorter trip and feeding the arterials. They must provide combinations of highway type and terrain. Note that such speeds
some degree of mobility and also serve abutting property. Thus, an are iffespective of traffic volume levels.
intermediate design speed and level of service is appropriate. Local Analysis issues 8 through I I concern the relationships between
roads and streets have relative short trip lengths, and because prop- roadway widths and highway operations. These issues are ad-
erty access is their main function, there is little need for mobility
or high operating speeds. This function is reflected by use of a
dressed in the foRowing sections within the framework estab-
lower design speed and level of service. lished earlier, in which functional classification and design speed
The functional concept is important to the designer. Even though serve as primary design controls.
many of the geometric standards could be determined without refer-
ence to the functional classification, the designer must keep in mind
the overall purpose that the street or highway is intended to serve.
This concept is consistent with a systematic approach to highway Issue 8— Width Effects on Capacity and Level of
planning and design (AASHTO, 1990, p. 16). Service

Both lane and shoulder widths affect the capacity of a two-

Deskgn Speed
lane roadway. Table 12, from the 1985 Highway Capacity Man-
ual, shows that a roadway with 9-ft lanes and no shoulders has
Design speed is on' e of the basic controls in establishing the
only two-thirds the capacity of a two-lane roadway with 12-ft
design of any highway. According to the 1990 AASHTO Policy: lanes and 6-ft shoulders.
Design speed is the maximum safe speed that can be maintained
There is an additional relevant consideration regarding design
over a specified section of highway when conditions are so favorable standards and level of service. Consistent with the framework
that the design features of the highway govern. The assumed design of functional classification, the current AASHTO Policy advises
speed should be a logical one with respect to the topography, the the use of guidelines given in Table 13 for design levels of
adjacent land use, and the functional classification of highway. service. Lower levels of service (i.e., "D") are considered accept-
Except for local streets where speed controls are frequently included
intentionally, every effort should be made to use as high a design able on lower class highways and in more difficult terrain. For
speed as practicable to attain a desired degree of safety, mobility, rural highways, however, in most cases the design should produce
and efficiency while under the constraints of environmental quality, level-of-service B or C.
economics, aesthetics, and social or political impacts. Once selected, Recommended n-dnimum standard widths for design of low-
all of the pertinent features of the highway should be related to the
volume roads should be consistent with minimum AASHTO
design speed to obtain a balanced design.
The design speed chosen should be consistent with the speed a Policy requirements for level of service and with the known
driver is likely to expect. Where a difficult condition is obvious, operational relationships between highway level of service and
drivers are more apt to accept lower speed operation than where lane and shoulder width. Appendix E presents an evaluation of
there is no apparent reason for it. A highway of higher functional these relationships. Tables 14(a) and 14(b) summarize the results.
classification may justify a higher design speed than a less important
facility in similar topography, particularly where the savings in The tables present the minimum required values forf, (the capac-
vehicle operation and other operating costs are sufficient to offset ity factor expressing the contribution of lane width and shoulder
the increased costs of right-of-way and construction. A low design width) to produce the minimum level of service called for by
 23

TABLE 11. Guidelines for selecting appropriate design speed

Design Speed (mph)

Rural
Highway Type Level Rolling Mountainous

Freeway 80 70 60a
Primary Highway or Street 60-70 50-60 40-50
Local or Collector Highway or Street 30-60 30-50 2040
3Special case, SO mph

Source: AASHTO 1990 Geometric Design of Highways and Streets, (Derived from
Tables V-1, p. 4, VI-I, p. 469, and p. 494).

TABLE 12. Adjustment factors for the combined effect of narrow lanes and restricted shoulder width

12-foot Lanes 11 -foot Lanes 10-foot Lanes 9-foot Lanes

Usable' Shoulder LOS I LOSb LOS I LOSb LOS LOSb LOS I LOSb
Width (feet) A-D E A-D E A-D I E A-D E

6 1.00 1.00 0.93 0.94 0.84 0.87 0.70 0.76

4 0.92 0.97 0.85 0.92 0.77 0.85 0.65 0.74
2 0.81 0.93 0.75 0.88 0.68 0.81 0.57 0.70
0 0.70 0.88 0.65 0.82 0.58 0.75 0.49 0.66
ere should r width is different on each side of the road, use the average shoulder width.
bFactor applies for all speeds less than 45 mph.
Source: Transportation Research Board. Special Repon 209. 1985. Highway Capacity Manual. pp.
8-9. -
I

TABLE 13. Guidelines for selecting design levels of service

Type of Area and Appropriate Level of Service

lEghway.Type t Rural Level Rural Rolling I Rural Mountainous I Urban and Suburban
Freeway B B C C
Arterial B B C C
Collector C C D D
Local D D D D
Note: General operating conditions for levels of service:
A-free flow, with low volumes and high speeds.
B-reasonably free flow, but speeds beginning to be restricted by traffic conditions.
C-in stable flow zone, but most drivers restricted in freedom to select their own speed.
D-approaching unstable flow, drivers have little freedom to maneuver.
E---unstable flow, may be short stoppages.

Source: AASHTO. 1990. Geometric Design of Highways and Streets. p. 92.

AASHTO (see Table 12), in typical design conditions. The fol- not a factor in determining minimum values for lane and shoulder
lowing is concluded from the analysis: width for typical combinations of functional classification, ter-
rain, and traffic composition.
1. At traffic volumes less than 1,500 vpd, capacity effects are 2. At traffic volumes approaching 2,000 vpd, the known width
24

TABLE 14(a). Minimum required lane and shoulder width capacity factor (Q for low-volume two-lane
high WaySa (trucks = 10% of ADT)

Average Daily Traffic Volume

Rural ffighway Type/Terrain 400 vpd 750 vpd 1,500 vpd 2,000 vpd
Arterial
- Level 0.12 0.22 0.43 0.58
- Rolling 0.18 0.34 0.68 0.91
- Mountainous 0.20 0.38 0.75 1.00

Collector
- Level 0.07 0.13 0.26 0.35
- Rolling 0.10 0.18 0.36 0.49
- Mountainous 0.13 0.24 0.47 0.63

Local
- Level 0.04 0.08 0.16 0.21
- Rolling 0.06 0.12 0.24 0.32
- Mountainous 0.13 0.24 0.47 0.63
'As determined in Appendix E. Values for f, are representative of typical traffic and geometric conditions for a
given highway type and terrain. See Table 12 for the relationship between f. and lane widths and shoulder
widths.

TABLE 14(b). Minimum required lane and shoulder width capacity factor (Q for low-volume two-lane
highways' (trucks = 15% of ADT)

Average Daily Traffic Volume

Rural lEghway Type/Terrain 400 vpd 750 vpd 1,500 vpd 2,000 vpd
Arterial
- Level 0.12 0.23 0.46 0.61
- Rolling 0.21 0.39 0.78 1.04
- Mountainous 0.25 0.46 0.93 1.24

Collector
- Level 0.07 0.14 0.28 0.37
- Rolling o. I 1 0.21 0.42 0.55
- Mountainous 0.16 0.30 0.60 0.79

Local
- Level 0.04 0.08 0.17 0.22
- Rolling - 0.07 0.14 0.28 0.37
- Mountainous 0.16 0.30 0.60 0.79

'As determined in Appendix E. Values for f. are representative of typical traffic and geometric conditions for a ,
given highway type and terrain. See Table 12 for the relationship between f. and lane widths and shoulder
widths.

effects on level of service should begin to influence minimum of drivers negotiating curves and found that a 2-mph to 3-mph
lane and shoulder width design values, particularly for highways difference in mean speeds occurs for the same alignment for
with significant truck volumes (greater than 10 percent), in rolling narrow roadways (i.e., with lane widths less than 10 ft) versus
or mountainous terrain and on primary arterials. Justification of wide roadways (lane widths of 12 ft or greater).
full lane and shoulder widths on the basis of their influence on Lamm and Choueiri (12) developed models from speed studies
capacity is evident for primary arterials with traffic volumes in New York to quantify the effect of lane width on 85th-percen-
approaching 2,000 vpd. For collector and local roadways, justifi- tile speeds (which are or should be close to the design speed).
cation based on capacity needs of total width dimensions of 22 This research looked at speeds on curves, as well as speeds
ft for roads with low truck volumes, and dimensions of 28 ft to through varying continuous alignment. It was found that, for
30 ft for roads with higher truck volumes, is evident at 2,000 alignment typically occurring in level terrain, 85-percentile
vpd. Refer to Appendix E. speeds on roads with 10-ft lanes are about 6 mph lower than
speeds on roads with 12-ft lanes. The difference in speeds associ-
ated with lane width increases to almost 8 mph for alignment
Issue 9-Effect of Width on Vehicle Speeds associated with mountainous terrain.
Appendix F presents an analysis of the relationship among
A number of studies have documented a lane width effect on vehicle operating speeds, lane width, and design speed based on
free vehicle speeds. Glennon et al. (11) studied the free speeds the foregoing cited research. Its application to development of
 25

TABLE 15. Acceptable shoulder width values for shoulder functions (values in feet)

Functional Classification
Shoulder Function Arterial Collector & Local
Shoulder Widths of I to 2 fed
Roadway and Shoulder Drainage I I
Lateral Support of Pavement 1.5 1
Off-tracking of Wide Vehicles 2 2
Encroachment of Wide Vehicles 2 2
Errant Vehicles (Run off road) — 2
Mail and Other Deliveries 2
Shoulder Widths of 3 to 4 feet
Pedestrians 4 4
Bicycles 4 4
Errant Vehicles (Run off road) 3
Shoulder Widths of 6 feet
Emergency Stopping 6 6
Mail and Other Deliveries 6 —
Garbage Pickup 6 6
Emergency Vehicle Travel 6 6
Routine Maintenance — 6
Law Enforcement 6
Emergency Call Box Services 6

Shoulder Widths of 7 to 10 feet

Parking
Trucks 10 N/A
Residential 8 7
Commercial 10 8
Turning and Passing at Intersections 10 9
Routine Maintenance 8
Major Reconstruction and Maintenance 9 9
Slow-Moving Vehicles 10 9
Law Enforcement 8
Emergency Call box Services 8

Adopted From: Downs, Jr. H.G. and Wallace, D.W. "Shoulder Geometrics and Use Guidelines."
NCHRP Report 254. Transportation Research Board, Washington, D.C. December
1982.

design standards focuses on identifying compatible (hence, ap- search, shows the range in functions and widths. Table 15 indi-
propriate) lane widths for given design speeds. It is concluded cates that there is a stepwise utility of shoulder widths. Minimum
that I I -ft and 12-ft lane widths generally are appropriate (i.e., widths of 2 ft serve to drain the pavement, supply lateral support,
produce speeds compatible with the established design speed) and provide for off-tracking on sharp curves and encroachment
for higher speed highways (those with design speeds of 60-mph of wide vehicles. An additional width of 4 ft offers little additional
and greater). Lane widths of 10 ft and I I ft are appropriate for functional utility, other than the minimum space for pedestrians
intermediate design speeds (50 mph), and lane widths of 10 ft and bicycles. On most low-volume rural roads, providing addi-
are appropriate for lower design speeds (40 mph or less). tional shoulder width just for pedestrian and bicycle use is not
The analysis and its application to lane width design values normally justified by the added expense. At a width of 6 ft,
presents an important research recommendation. Lane widths the utility of shoulders improves significantly. A 6-ft width is
should be compatible with the terrain and design speed. Lane acceptable for emergency stopping, mail and package deliveries,
widths that are too wide for a given set of design conditions are emergency vehicle travel, and routine maintenance (on lower
undesirable. They may encourage higher speeds than those for class facilities). Shoulder widths of 8 ft or more generally have
which the alignment and roadside were designed. utility only for higher class, higher volume roads, where the
effects of slowed vehicles, parked vehicles, and routine mainte-
nance are more significant.
Issue 10—Width-Related Shoulder Functional
Requirements
Issue 11—Operational Effects of Wider Trucks
Shoulders serve many functions, each of which requires a
minimum acceptable width. An earlier NCHRP study by Downs As discussed earlier, the specific accident effects of large
and Wallace (13) catalogued the various shoulder functions and trucks on low-volume roads could not be properly quantified in
noted the required widths. Table 15, summarized from that re- this study because of data limitations. However, a previous re-
26

search study for FHWA has investigated the operational impacts factors, or default values (derived from the survey). Figure 8
of large trucks on rural two-lane roads. Specifically, tractor semi- summarizes the construction cost model. It is based on the rela-
trailers with long trailers (e.g., 48 ft or longer) have been found tively common practice in preliminary or planning-level analyses
to have problems with "off-tracking." As these long trucks travel of estimating quantities for the major construction items of pave-
around a left-hand curve, the trailers may encroach across the ment, earthwork, and shoulders, and applying factors or percent-
centerline. On a right-hand curve, the trailers may encroach off ages for all other items. The result is a "cost per mile" that is
the shoulder. This off-tracking problem increases for narrow representative of the typical conditions and unit values cited in
lanes and shoulders, and for sharper horizontal curves. the survey responses.
This recent study conducted for FHWA by Harkey et al.(10) The following methodology and assumptions apply to the
recommended paved shoulders of 3 ft or more and minimum derivation of the cost model:
lane widths of 12 ft. This design handles the larger trucks allowed
Item ]—Pavement. Quantities for 18-, 20-, 22-, and 24-ft
by the Surface Transportation Assistance Act (STAA) of 1982
traveled-way widths were calculated directly to produce square
(i.e., trucks 102 in. wide, twin 28-ft trailers, or semitrailers of
feet of pavement per mile. Unit costs of pavement were devel-
48 ft or greater), particularly, on roads with severe horizontal
oped for two traffic-volume and two heavy-vehicle percentage
and vertical alignment.
ranges. Tables 16(a) and 16(b) summarize unit quantities and
To best accommodate trucks, it may be a useful design policy
costs.
to differentiate between local, low-volume roads and higher facil-
Item 2—Shoulders. Quantities for unpaved and paved shoul-
ity types such as low-volume arterials. Such differentiation would
ders were calculated to produce square feet of shoulder per mile
involve knowing the percentage and volume of larger trucks. On
(see Table 17(a)). The survey produced the recommended unit
local-type facilities with more local delivery trucks, single-unit
costs given in Table 17(b).
trucks, and smaller trucks, lesser widths may in fact be sufficient.
Item 3—Earthwork. Quantities of cubic yards of earthwork
On the higher class, low-volume roads, the presence of the longer
per mile were computed for typical cross-section designs in
trucks may be sufficient justification for incrementally greater
level, rolling, and mountainous terrain. For level and rolling
lane widths and paved shoulder widths.
terrain, 3:1 sideslopes were assumed. For mountainous terrain,
2:1 sideslopes were assumed. An evaluation of project cost and
quantity data provided by the Colorado Department of High-
ROADWAY DESIGN REOUIREIVIENTS ways was performed to establish the average depth of earthwork
as a function of terrain. This was determined to be 2.5 ft for
Cost-Effectiveness and Width-Related Sensitivities
level terrain, 3.0 ft for rolling terrain, and 4.5 ft for mountainous
terrain. Table 18(a) shows per mile quantities of earthwork, with
Cost-effectiveness is the primary concern in establishing the foregoing assumptions, as a function of the full range of the
width-related design standards. Greater widths may be associated traveled way and shoulder width. Unit costs suggested for use
with safety and operational benefits, but they may also cost are given in Table 18(b). These costs were derived from the
significantly more to build and to maintain. An important re- survey responses for average projects.
search task was to investigate the width-related sensitivities of Item 4—Clearing and Grubbing. Survey responses were lim-
construction and maintenance costs. ited for this item. The Colorado cost information was used to
establish unit costs (see Table 19(a)) for typical right-of-way
values (see Table 19(b)).
Issue 12—Construction and Maintenance Costs Item 5—Utility Relocation. A unit cost of $40,000 per mile
was derived from the survey responses.
The research team performed a comprehensive evaluation of Item 6—Drainage. Costs of drainage structures, ditches, and
construction cost values and cost estimating methodologies to the like are related generally to the earthwork and pavement
uncover width-related sensitivities. The evaluation included a costs. From the survey, a unit value of 15 percent of the costs
nationwide survey of state design agencies, development of a of Items I through 5 was derived. The net effect of this cost
cost model, and calculation of typical per mile construction costs was tested against project data and found to be reasonable~.
for the full range of reasonable lane and shoulder widths. The Other Items. Other minor, but necessary, cost items include
survey and analysis focused on identifying costs of new construc- signing, striping, and permanent traffic control (Item 7); appurte-
tion or reconstruction (4R). nances, such as guardrail (Item 8), and erosion control, environ-
mental mitigation, and landscaping (Item 9). The survey re-
sponses resulted in the following recommended typical unit costs
for these items:(1) permanent traffic control— 1 percent of Items
ConstructionCost Survey. All 50 state transportation agencies
I through 5,(2) appurtenances-9 percent of Items I through
were surveyed to develop a methodology for estimating construc-
5, and (3) erosion control-4 percent of Items I through
tion costs for rural highways. Appendix D, Tables D-I to D-7,
5. Other costs (Items 10 through 14) typically are computed or
surrimarizes the survey findings resulting from a response rate
estimated as a percentage of total cost. The results of the survey
of about 50 percent. The purpose of the survey was (1) data
were used to develop typical percentages for costs of traffic
were sought for 1991 construction unit costs around the country;
control during construction; mobilization and contingencies;
and (2) cost estimating assumptions or models were investigated
preliminary, final, and construction engineering; and other non-
for use or adaptation to the research.
construction costs.
The cost estimating model developed from the survey was
structured to enable users to apply their own unit values or The model was developed to enable the use of substitute cost
 27

Item Cost per Mile

I . Pavement [see Tables 16(a) and 16(b)] $

Shoulders [see Tables 17(a) and 17(b))

Earthwork (see Tables 18(a) and 18(b)]

Clearing and Grubbing [see Tables 19(a) and 19(b)] $

S. Utility Relocation—$40,000 per mile (suggested)

Subtotal (Items 1-5)

Drainage (15 Percent of Items 1-5*) $

Signing, Striping, and Permanent Traffic Control (I Percent of Items 1-5 $_

Appurtenances (9 Percent of Items 1-5*) $_

Erosion Control, Other Environmental Mitigation, $

Landscaping, Seeding Sodding, etc. (4 Percent of Items 1-5*)

e Subtotal (Items 1-9) $

Traffic Control During Construction (4 Percent of Items 1-9*) $

Mobilization, Miscellaneous Costs, and Contingencies (15 Percent of Items 1-9*) $

0 Construction Subtotal (Items 14 1) $

Preliminary and Final Engineering (9 Percent of Items 1-11) $

Construction Engineering (10 Percent of Items 1-11) $

Legal, Administrative, and Other Non-Construction Costs (5 Percent of Items 1-11*)

Total Estimated Cost per Mile for Project (Items 1-14)

Total Estimated Cost for Project

Project Length mi.

Project Cost per Mile x $ /nii.

TOTAL PROJECT COST $

*Percentages other than those suggested may be used if desired.

Figure 8. Summary of construction cost model—cost estimating procedure for two-lane

rural highways (new construction).

factors or unit values for the basic items. The cost-effectiveness Tables 20 and 21 reveal the following: (1) roadways in moun-
analysis results presented here were based on the unit costs and tainous terrain are significantly more costly than roadways.in
percentages previously described. rolling terrain, and the cost difference between roadways in
Tables 20 and 21 present the resulting unit costs per mile for rolling and level terrain is less pronounced; (2) the incremental
constructing two-lane highways by terr ain, traffic volume, and cost of an additional 4 ft of shoulder width (2 ft on each side
roadway cross section, for ADT less than 400 vpd and greater of the roadway) varies from about $50,000 to $70,000 per mile,
than 400 vpd. The.values were checked against typical project depending on terrain; and (3) the difference in costs between
costs from various states and from the research team's informa- paved and unpaved shoulders varies from about $25,000 to
tion on other projects. I
$80,000 per mile, depending on width.
28

TABLE 16(a). Quantities for item I — TABLE 17(a). Quantities for item 2—shoulders
pavement
SHOULDER SHOULDER
TRAVELED PAVEMENT
WIDTH, EACH AREA
WAY WIDTH AREA
(ft.) (sq. ft./mi)
(ft.) I (sq. ft./mi)
2 21,120
18 95,040 4 42,240
20 105,600 6 63,360
22 116,160 8 84,480
24 126,720 10 105,600
H

TABLE 16(b). Suggested unit costs for item 1—pavement*

TABLE 17(b). Suggested unit costs for
item 2—shoulders*
% HEAVY AVERAGE DAILY TRAFFIC
TRUCKS < 400 400 vpd SHOULDER UNIT
TYPE COST
<10 $ 1.50 / sq.ft. $ 1.90 sq. ft.
>10 $ 1.70 / sq.ft. $ 2.20 sq. ft. PAVED $1.45 / sq. ft.
UNPAVED $0.85 / sq. ft.
* Other unit costs may be substituted if desired.
* Other unit costs may be substituted if desire(

Calculation for Item 1: Pavement Calculation for Item 2: Shoulders

sq. ft./mi. X $ —/sq. ft.—= $ —/mi sq. ft./mi. X $ — / sq. ft. = $ _/ mi.

Maintenance Costs Survey. A review of the literature uncov- 214 includes an analysis of the net operational benefits of widen-
ered no identifiable "width-sensitive" costs associated with main- ing roads as a function of ADT. The findings show that user
tenance. TRB Special Report 214 notes that increased pavement savings are less than 2 percent of the cost of widening for roads
maintenance costs, when considering a typical widening project with ADT less than 1,000 vpd, and about 7 percent for roads with
for a two-lane highway, represent about 6 percent of annualized ADT less than 2,000 vpd (3). Such net benefits are considered
construction cost. negligible.

Maintenance Costs. For analysis purposes, incremental main-

Issue 13—Cost-Effective Roadway and Shoulder
tenance costs are considered inconsequential and are ignored.
Width Combinations Relative
For analysis purposes, then, the comparison of economic bene-
to Safety and Operational Benefits
fits and costs is restricted to a comparison of annualized construc-
tion costs and annual accident cost savings. Ignoring incremental
A primary input to determination of cost-effective widths is
traffic operational benefits and maintenance costs would, in total,
an evaluation of the econonfic trade-offs involving costs versus
have little or no effect on the analysis because the incremental
safety and operational benefits. Costs include construction cost
benefits and costs (1) are minor in terms of their individual
and annual maintenance costs. Theoretical benefits include
effect, and (2) tend to cancel each other out. (For example, wider
travel-time savings and accident ' reductions. For low-volume
cross sections have marginally greater operational benefits, but
roads (ADT less than 2,000 vpd), a number of simplifying as-
marginally higher maintenance costs.)
sumptions are appropriate. These include treatment of traffic
operational benefits and incremental maintenance costs.
Cost Assumptions. Assumptions are required for the discount
rate, traffic growth, useful life of improvements, and average
Traffic Operational Benefits. Wider roadways promote higher accident costs.
speeds, which have a dual effect in terms of operating cost.
Travel times are lowered (producing a benefit), but vehicle op-
erating costs are increased (producing a cost). TRB Special Report Discount Rate. The 1977 AASHTO "Manual on User Benefit
 29

TABLE 18(a). - Quantities for item 3-earthwork (cu yd/mi) TABLE 19(a). Quantities for item 4-clearing and
grubbing
ITRAVELED SHOULDER WIDTH, EACH (ft.)
TERRA114 WAY (ft.) 0 1 2 1 4 1 6 1 WIDTH OF
18 28,113 30,069 32,025 33,981 35,937 37,893 TERRAIN CLEARING- (ft.)
LEVEL 20 29,091 31,047 33,003 34,959 36,915 38,871
22 30,069 32,025 33,981 35,937 37,893 39,849 LEVEL 80
24 31,047 33,003 34,959 36,915 38,871 40,827
ROLLING 8
is 34,619 36,967 39,315 41,663 44,011 46,359
MOUNTAINOUS 66
ROLLING 20 35,793 38,141 40,489 42,837 45,185 47,533
22 36,967 39,315 41,663 44,011 46,359 48,707
24 38,141 40,489 42,837 45,185 47,533 49,881 *Width of clearing may be changed if desired.
18 51,920 55,440 58,960 62,480 66,000 69,520
MOUNTAINOUS 20 53,680 57,200 60,720 64,240 67,760 71,280
22 55,440 58,960 62,480 66,000 69,520 73,040
24 57,200 60,720 64,240 67,760 71,280 74.800 TABLE 19(b). Suggested unit costs for item 4-
clearing and grubbing*

TERRAIN UNIT COST

TABLE 18(b). Suggested unit LEVEL $ 170 / ft. width / Mi.
costs for item 3-earthwork* ROLLrNG _L170 /ft. width/ mi.
TERRAIN UNIT COST MOUNTArNOUS J$ 520 / ft. width / mi.
LEVEL $ 3.50 /cu. d.
ROLLING $ 3.50 /cu. d.
MouNrAiNous $ 5.50 /cu. d.
* Other unit costs may be substituted if desired.

* Other unit costs may be substituted if desired.

Calculation for Item 3: Earthwork

Calculation for Item 4: Clearing and Grubbing
cu..yd./mi. X S /cu. yd. = $ /mi. - ft. width X $_ /ft. width/mi. = $ - /mi.

National Safety Council. Other recent research by the authors

Analysis and Bus-Transit Improvements" recommends a 4-per- of this study, using the same FHWA approach, derived values
cent discount rate for low-risk investments (14). More recent of $53,700 to $59,000 per average accident on rural highways
research suggests a rate of 6 percent as appropriate for evaluation and rural highway curves, respectively.
of highway improvements (3). For this research, a 6-percent Accident severity data used in this study do not provide the
discount rate is assumed. details to enable a rigorous derivation of a typical accident cost.
Valuations in the $50,000 to $85,000 range appear appropriate.
For analysis purposes, thus, it was decided that a value of $60,000
Traffic Growth. No consideration of traffic growth is assumed per accident is appropriate. This value is slightly higher than
for analysis purposes. For most low-volume roads, particularly unit values derived in recent past studies (5, 35). Recent revisions
with ADT less than 1,000 vpd, traffic growth is negligible, and to unit costs reflecting inflation are noted in Ref. 15. A * ccident
existing or projected traffic is a rough estimate at best. Thus, severity on low-volume roads appears less, however, than in the
economic analysis of width values for the purpose of testing earlier research. Note that a direct relationship exists between
standards assumes that the design traffic remains constant over the selected accident valuation and breakeven traffic volume in
the life of the improvements. cost-effectiveness analysis levels. For higher accident valuations,
breakeven ADT levels that justify wider roadways will be lower.
Average accident rates from the best linear model were used
Useful Life of Improvements. A 20-year useful life of con- in the economic analysis. These rates were summarized in Table
struction items is assumed. 12. Annualized capital costs for comparison of two alternative
designs were taken as the difference in construction costs per
mile (from Tables 20 and 21) multiplied by a capital recovery
Average Accident Costs. The determi nation of an appropriate factor of 0.08718 (for a 6-percent discount rate and 20-year life).
dollar value for an average accident is, perhaps, the most signifi- For example, for a roadway with a design ADT greater than
cant step in the economic evaluation. Accident costs are a func- 400 vpd, in rolling terrain with less than 10-percent trucks, the
tion of the severity distribution of accidents and costs that are following are typical costs per mile (see Table 21): for a 20-ft
considered to be reasonable for a fatal injury and property- traveled way (i.e., two 10-ft lanes) with 4-ft paved shoulders
damage-only accident. (producing a total roadway width of 28 ft), $870,290; and for a
A recent FHWA Technical Advisory suggests that the value 22-ft traveled way (i.e., two 11 -ft lanes) with 4-ft paved shoulders
of a typical accident on a rural road is $85,600 (15). This is (producing a total roadway width of 30 ft), $916,304.
based on the FHWA "willingness to pay" valuation of accidents, The incremental construction cost of the wider design is
which is substantially higher than the values derived by the $46,014-the difference between the 30-ft roadway and the
30

TABLE 20. Estimated construction costs per mile-two-lane rural highways (ADT less than 400 vpd)

ESTIMATED CONSTRUCTION COSTS -- TWO-LANE RURAL GHWAYSMOLLARSI

ADT < 400 vpd SHOULDER WI I H (it.)
0 1 2 1 4 8 10
PERCENT TRAVELED NO SHOU DERTYPE I SHOULDER TYPE SHOULUER 6 TYPE 9HOULDER TYPE I SHOULDER TYPE
TRUCKS TERRAIN WAY (ft.) SHLDR I PAVED JUNPAVEDI PAVED JUNPAVED PAVED JUNPAVEDI PAVED JUNPAVED] PAVED JUNPAVE
18 560,693 632,019 607,8971 703,344 655,101 774,669 702,304 i8~
111:;
92
6~1 1 1~2
786:~2
~73
11 917,319 796,711
LEVEL 20 597,361 668,686 644,5651 740,011 691,768 81 1,L36 738,972 953,986 833,379
22 634,029 705,354 _681,2321 776,679 728,436 848,004 225,~291
215 Ini 919,3291 822,8431 990,654 870,047
24 670,696 742,021 717,9001 813,346 765,103 884,671 955,9961_ 859.5111 1.027321 906,714
18 604,039 677,975 653,8541 751,912 703,669 825,849 753,484 899,785 803,300 973,722 853,115
<10% ROLLING 20 642,012 715,949 691,8271 789,885 741,642 863,822 791,458 937,759 841,273 1,011,695 891,088
22 679,985 753,922 729,8011 827,859 779,616, 901,795 829,431 975,732 879,246 1,049,669 929,092
24 717,959 791,895 767,774 865,832 817,589 939,769 867,404 1,013,706 917,220 1,087,642 967,035
18 956,407 1,051,552 1,027,431 1,146,698 1,098,455 1,241,844 1,169,479 1,336,990 1,240,504 1,432,135 1,311,528
MOUNTAINOUS 20 1 ~004t%5 1,100,130 1,076,009 1,195,276 1,147,033 1,290,422 1,218,057 1,385,568 1,289,082 1,480,713 1,360,106
22 1 ~215
1 . 55 '~21 1!I.9~
3:~66 ~8:ZO8
2ig 1,124,587 1,243,854 1,195,611 1,339,000 .31 1:
1,266:~62~ 1 482:~
~2~ 1~2~
31 1:j8;
327:~6~01
3i i:2
1 57~
~9:~2§
861 1 1 ~~
4'~
28,684
24 86 1,173,165 1,2(
)24-49 1,244,189 1,387,5781 1,315 262
18 596,876 668,2011 ___ -T--
644079 739.526 691,2831 810,8511 738,486IN2,1161 785,69of 953,501 832,894
LEVEL 20 637,563 708,889 684,7671 780,214 731,971 8§
85~1:5~3;
91 8.9:i17
112 '63:35
2~~ 8~11
1 06g 1,034:i77
867:2
~ 18 2211!2 873,581
22 678,251 749,576 725,4551 8 901 772,658 I 914,2691
24 718,939 790,264 766,1431 861,589 813,346 932,9141 860,5501 1 954,957
18 640,221 714,157 690,0361 788,094 739,851 862,031 789,666 935,968 839,482 1,009,904 889,297
> 10% ROLLING 20 682,214 756,151 732,0301 830,088 781,845 904,024 831,660 977,961 881,475 1,051,898 931,291
22 724,208 798,145 774,0231 872,081 823,839 946,018 873,654 1,019,955 923,469 1,093,891 973,284
24 766,202 840,138 816,017 914,075 865,832 988,012 915,647 1,061,948 965t463 1,135,885 1,015,278
18 992,589 1,087,735 1,063,613 1,182,880 --1-
,134,637 1,278,026 1,205,662 1,373,172 1,276,686 1,468,318 1,347,715
MOUNTAINOUS 2U 1,045,187 1,140,3 3 1,116,211 1,235,478 1,187,236 1,330,624 1,258,260 1,425,770 1,329,284 1,520,916 1,400,3 §28
22 1,097,785, 1,192,931 1,168,809 1,288,077 1,239,834 1,383222 1 310858 i:@81 :§
3 81 1::384 8o 1 573 514 1452 0 7
1:~Ii
24 9 1,221,408 1,340,675 1,292,432 1,4j 3 6 f- ,62 '
l- 0;505
112 l,5' 1

TABLE 21. Estimated construction costs per mile-two-lane rural highways (ADT greater than 400 vpd)
ESTIMATED CONSTRUCTION
COSTS -- TWO-LANE RURAL FUGHWAYS (DOLLARS)
AU I ,4vu VP0 SHUULDER WIDTR-(ff.T-
1 0 2 :j- 4 6
T__ 8 10
PERCENT TRAVELEE; NO
I SHOULDERTYPE I SHOULDERTYPE SHOULDER TYPE I SHOULDER TYPE I SHOULDER TYPE
TRUCKS TERRAIN :~~ SHLDR I PAVED JUNPAVEDI PAVED JUNPAVE PAVED JUNPAVEDI PAVED JUNPAVEDI PAVED JUNPAVE5
18 633,0581 704,383 _696,26i V-
775,708 727,465 847,0331 774,6691 918_,35_8T__8_
21_,8_72T 989,683 869,076
LEVEL 20 71 ~7~7 4
~7~7: 21 724,969 820,416 772,173 891, 911,288 1,034,391 913,784
22 41 749"
793,799 8.4
2~9§81
38; §.q:iN
8~~ 124 816,881. 936,i4§
7~'! 89 'Z
ig~:685
Z! 1,90S~
3'066
,774
007,7741 911,2881 1,079,099 958,492
24 767,1821 838,507 861,5891 981,1571 908,7931 1,052,4821 955,9961 1,123,807 1,003,200
18 676,403 750,340 726,218 824,276 776,033 898,213 825,849 972,150 875,664 1,046,086 925,479
< 10 % ROLLING 20 722,417 796,354 772,232 870,290 822,047 944,227 871,863 1,018,164 921,678 1,092,100 971,493
22 768,431 842,367 818,246 916,304 868,061 990,241 917,876 1,064,177 967,692 1,138,114 1,017,507
24 814,445 888,381 864,260 962,318 914,075 1,036,255 963,890 1,110,191 1,013,706 1,184,128 1,063,521
18 1,028,771 1,123,917 1,099,795 1,219,062 1,170,820 1,314,208 1,241,844 1,409,354 1,312,868 1,504,500 1,383,892
MOUNTAINOUS 20 1,085,389 1,180,535 1,156,414 1,275,681 1,227,438 1,370,827 1,298,462 1,465,972 1,369,487 1,561,118 1,440,511
22
~4 438 0g
1!,1§ 8 j 1!~227
9~ 1:~2;
1§
3 48~O!gO~53
61 7 132 2 .§ 1!~46~::1E
05
11 11:;
6~14
7!2
35'!
21 1 533:129
9Z 9
q.;1 §:~1211
:2
6 '21
:~711 !953
0211 1:2
38
22
8 1I!3~ 1 1i8
2:N231 21~2
6§§j
" ':~
i 5~7 38 748
18 687,331 758,656 734,5351 829,981 781,738 901,306 828,942 972,631 876,145 1,043,956 923,349
LEVEL 20 738,070 809,395 7 §5
882720 832,477 952,045 879,680 1,023,3701 926,884 1,094,695 974,0871
-
22 788,808 860,133 E;
85:6
21
.5
31 58 883,215 1,002,783 930,419 1,074,1081 977,622 1,145,433 1,024,82~
24 839,546 910,871 886,7501 982,196 933,954 1,053,522 981,157 1,124,8471 1,028,361 1,196,172 1,075,564
18 730,6761 804,613 780,491 878,550 830,307 952,486 880,122 1,026,423 929,937 1,100,360 979,752
> 10% ROLLING 20 782,7121 856,657 832,536 930,594 882,351 1,004,531 932,166 1,078,467 981,981 1,152,404 1,031,797
22 834,765 908,701 884,580 982,638 934,395. 1,056,575 984,210 1,130,511 1,034,026 1,204,448 1,083,841
24 ~0
886,809 960,74~ 936,624 1,034,682 986,4391 1,108,619 1,036,255 1,182,556 1,086,070 1,256,492 1,135,885
18 1,083,044 1,178,190 1,154,069 1,273,,3
: :3::6
:: 1,225,093 1,368,481 1,296,117 1,463,627 1,367,141 1,558,773 1,438,1661
MOUNTAINOUS 2031,145,693 1,240,839 1,216,717 1,335,_985 1,287,742 1,431,130 1,358,766 1,526,276 1,429,790 _11- fi-
21,422 1,500,8
22 1,208,342 1,303,4 '18,633 1,350,390 1,49,3 779 1 421415 I!g5" ~~Ell2 ~!,~21~22 1,684,071
24 1,270,9911 1,366,Ij9I 7N3~15~1 i 9F2
88 i1:3N -82 1,413,039 1,55 1 i

28-ft roadway unit costs per mile ($916,304 minus $870,290). 10-ft lanes and 4-ft shoulders in rolling terrain, 2.41 accidents per
Annualized, this results in a cost difference of $4,012 per mile MVM; and for a roadway with 11 -ft lanes and 4-ft shoulders in
per year (i.e., $46,014 x 0.08718). rolling terrain, 1.31 accidents per MVM.
The incremental benefits of the 30-ft roadway are based on For the example case, with traffic volumes of 400 vpd and
the difference in accident rates (see Table 10): for a roadway with accidents valued at $60,000, the following is the calculated an-
 31

nual benefit of the wider cross section per mile per year: Bene- levels is it justified to design for 4-ft shoulders (versus the
fits = (difference in accident rates)(365)(400)(10-1)($60,000) = minimum 2-ft shoulders) on highways with 11 -ft lanes?
(2.41 — 1.31)(0.146)($60,000) = $9,636 per mile per year. In For the range of terrain and other conditions, at what ADT
this example above, because the accident benefits exceed the levels is it justified to design for 4-ft shoulders (versus minimum
annualized difference in construction costs, the wider cross sec- 2-ft shoulders) on highways with 12-ft lanes?
tion is considered cost-effective. In some cases, either an 11 -ft or 1 O-ft lane width may be
Although there are many possible cost-effectiveness compari- appropriate for design purposes. For the range of terrain and
sons, inspection of the accident rates and construction cost impli- other conditions, at what ADT levels is a cross section employing
cations greatly simplifies the analysis. Figure 5 and Table 10 10-ft lanes and 6-ft shoulders more cost-effective than one with
summarize the research findings regarding safety-effectiveness 1 1-ft lanes and 2-ft shoulders?
of variable widths. The following points are emphasized. In some cases, either an I I -ft or I O-ft lane width may be
There is no direct, measurable benefit of 10-ft lanes with no appropriate for design purposes. For the range of terrain and
shoulders or 9-ft lanes with no shoulders compared to any nar- other conditions, at what ADT levels is a cross section of I 1-ft
rower roadway design. For roadways with 10-ft lanes, there is no lanes and 2-ft shoulders more cost-effective than one with 10-
incremental safety benefit of shoulders of 4 ft or less. Increasing ft lanes and 2-ft shoulders?
shoulder widths to 5 ft does produce additional expected benefits, In some cases, either an 11-ft or 10-ft lane width may be
but shoulder width increases beyond 5 ft do not produce addi- appropriate for design purposes. Accompanying shoulder width
tional safety benefits. values appropriate for safety cost-effectiveness vary depending
For low-volume roads with shoulders less than 4 ft wide, there on lane width. For the range of terrain and other conditions, at
is a measurable safety benefit of 11 -ft lanes compared to I O-ft what ADT levels is a cross section with 11 -ft lanes and 4-ft
lanes. No safety benefit is evident for roadways with 11 -ft or shoulders more cost-effective than one with 10-ft lanes and 5-
12-ft lanes beyond providing 4-ft shoulders. ft shoulders?
Eleven-ft lanes with shoulders of 4 ft or more are marginally In some cases, either a 12-ft or an I I -ft lane width may
safer than 12-ft lanes with narrow or no shoulders. Also, for be appropriate. For the range of terrain and other conditions, at
lane widths of I I ft or 12 ft, having shoulders 3 ft in width is what ADT levels is a cross section with 12-ft lanes and 3-ft
associated with a lower accident rate than shoulder widths of 2 shoulders more cost-effective than one with I I -ft lanes and 4-
ft or less. foot shoulders?
The following discussion addresses the foregoing questions.

Issue 14—Cost-Effective Roadway and Shoulder

Width Combinations for Design Standards
Comparison of 10-Ft Lanes with 5-Ft Shoulders to
10-Ft Lanes with 2-Ft Shoulders
The following points are relevant regarding cross-section de-
sign standards.
In many cases (lower design speed, lower classification high- Some highways may have minimum lane widths of only 10
ways in rolling to mountainous terrain), the evidence suggests ft. For low volumes, a 2-ft shoulder is operationally sufficient.
that lane widths as low as 9 ft are reasonable (i.e., produce a At what volume level and under what design conditions are
cost-effective design). wider shoulders appropriate for roads with 10-ft lanes?
Design values for lane width should reflect not only known The analysis surnmarized in Table 22 indicates that a "break-
safety relationships but also quantifiable operational require- even" traffic volume level for wider shoulders varies with the
ments, such as width effects on speed and highway capacity, terrain (and, hence, typical construction costs per mile). Note
and large truck requirements. For highways with higher design that, based on the accident analysis, there is no incremental safety
speeds, with design year traffic volumes greater than 1,500 vpd, benefit of shoulders with widths of 4 ft or less with 10-ft lanes.
significant truck volumes, and higher functional classification, For level to rolling terrain, about 400 to 450 vpd is an estimated,
lane widths of up to 12 ft may be appropriate regardless of breakeven ADT for roads with paved shoulders, and 300 vpd is
limited safety cost-effectiveness. calculated for roads with unpaved shoulders. In mountainous
Shoulder width design values should be based on functional terrain, breakeven ADT volumes are 600 vpd and about 450 vpd
requirements associated with drainage, structural support and for paved shoulders and unpaved shoulders, respectively.
traffic operations, as well as on safety cost-effectiveness. Two
feet is considered a minimum functional shoulder width value.
Widths as great as 8 ft are appropriate for functional requirements
associated with maintenance and storage of vehicles, particularly Comparison of 11-Ft Lanes with 4-Ft Shoulders to
11-Ft Lanes with 2-Ft Shoulders
on facilities with higher functional classification.
Thus, cost-effectiveness analysis as it relates to derivation of
design standards can be reduced to the following questions: For highways where an I I -ft lane width is most appropriate,
even for the lowest traffic volume levels, there is a question of
For the range of terrain and other conditions, at what ADT when a more than minimum 2-ft shoulder is justified. The bottom
levels is it justified to design for 5-ft shoulders (versus the of Table 23 summarizes analyses of the ADT levels that justify
minimum 2-ft shoulders) on highways with 10-ft lanes? the addition of 2 ft to each shoulder. From the accident studies,
For the range of terrain and other conditions, at what ADT there is a significant safety benefit of a design with 11 -ft widths
32

TABLE 22. Comparison of 10-ft lanes with 5-ft shoulders to 10-ft lanes with 2-ft shoulders

Typical Construction Costs per Mile ($000s) from Table 21

Level Terrain Rolling Terrain Mountainous Terrain

Lane/Shoulder Width Percent Shoulders Shoulders Shoulders
(reet) Trucks Paved I Unpaved npaved Paved I Unpaved

* 10% $856.1 $795.8 $907.3 $847.0 $1,323.3 $1,263.0

10/5
* 10% $916.4 $856.1 $967.6 $907.3 $1,383.6 $1,323.3

* 10% $749.1 $725.0 $796.4 $772.2 $1,180.5 $1,156.4

10/2
* 10% $809.4 $785.3 $856.7 $832.5 $1,240.8 $1,216.7

Difference in Annualized Construction Costs per Mile

(20-Year Project Life, 6-Percent Discount Rate, and Capital Recovery Factor of 0.08718)

EM
Level Terrain Rolling Terrain Mountain ous Terrain
Percent Sh ulders Shoulders Shoulders
Trucks I I I
1
Paved Unpaved Paved Unpaved Paved Unpaved

< 10% $9,328 $6,172 $9,668 $6,521 $12,449

> 10% $9,328 $6,172 $9,668 $6,521 $12,449 $9,293

Accidents and Accident Costs

Related Accident Rate Annual Related Accidents Annual Related Accident

(per MVM) per Nfile per 100 vpd Costs per Mile per 100 vpd_
Lane Width/Shoulder Width Terrain Terrain Terrain
(feet)
L_F_R FM L M L I R M

1015 1.30 1.43 1.88 0.047 0.052 0.069 $2,820 $3,120

10/2 2.28 2.41 2.86 0.083 __LO88 OA04 $4,980 $5,280 $2
$4'1 201
$6,240
60
Difference — — — $2,160 $2,160

Breakeven ADT for 10-Foot Lane Width and 5-Foot Shoulder Width (vpd)

Level Terrain Rolling Terrain Mo ntainous errain

Percent Shoulders Shoulders Shoulders
Trucks Paved Unpaved Paved Unpaved Paved Unpaved

< 10% 430 286 448 302 576 430

> 10% 430 286 448 302 576 430

and 4-ft shoulders versus I I -ft lane widths and 2-ft shoulders. minimum 2-ft shoulder isjustified. Table 24 summarizes analyses
The difference in costs between the two alternatives varies with of the ADT levels that justify the addition of 2 ft to each shoulder.
terrain, because much of the cost difference is associated with The results of the accident studies show a significant safety
earthwork. benefit of a design with 12-ft lane widths and 4-ft shoulders
Using the analysis results as a general guideline, the following versus 12-ft lane widths and 2-ft shoulders. The difference in
is concluded: for highways in level and rolling terrain, widening costs between the two alternatives varies with terrain, because
to a 4-ft shoulder with 11 -ft lanes is justified for highways with much of the cost is associated with earthwork.
paved shoulders at ADT levels of 500 vpd or greater, and at Using the analysis results as a general guideline, the following
ADT of 400 vpd or greater for highways with unpaved shoulders; is concluded: for highways in level and rolling terr ain, widening
for highways in mountainous terrain, widening to a 4-ft shoulder to a 4-ft shoulder with 12-ft lanes is justified for highways with
with I I -ft lanes is justified at ADT levels of 700 vpd or greater paved shoulder's at ADT levels of 500 vpd or greater, and at
for paved shoulders and 500 vpd or greater for unpaved shoulders. ADT of 400 vpd or greater for highways with unpaved shoulders;
for highways in mountainous terrain, widening to a 4-ft shoulder
with 12-ft lanes is justified at ADT levels of 700 vpd or greater
Comparison of 12-Ft Lanes with 4-Ft Shoulders to for paved shoulders and 500 vpd or greater for unpaved shoulders.
12-Ft Lanes with 2-Ft Shoulders
Comparison of 11-Ft Lanes with 2-Ft Shoulders to
Certain highways should be designed with a minimum 12-ft 10-Ft Lanes with 2-Ft Shoulders
lane width because of the design speed and functional classifica- For certain functional classes and design speeds, either 10-ft
tion. For such highways, there is a question of when a more than or 11 -ft lanes may be appropriate. The analysis summarized in
 33

TABLE 23. Comparison of 11-ft lanes with 4-ft shoulders to 11-ft lanes with 2-ft shoulders

Typical Construction Costs per Mile ($000s) from Table 21

Level Terrain Rolling Terrain Mountainous T

Lane/Shoulder Width Percent Shoulders Shoulders Shoulders

(feet) Trucks Paved I Unpaved Paved I Unpaved Paved I UnpJ

aved

< 10% $865.1 $816.9 1 $916.3 $868.1 $1,332.3 $1,284.1

1114
> 10% $931.5 $883.2 $982.6 $934.4 $1,398.6 $1,350.4

< 10% $793.8 $769.7 $842.4 $818.2 $1,237.2 $1,213.0

11/2
> 10-% $860.1 $836.0 $908.7 $884.6 $1,303.5 $1,279.4

Difference in Annualized Construction Costs per Mile

(20-Year Project Life, 6-Percent Discount Rate, and Capital Recovery Factor of 0.08718)

Level Terrain Rolling Terrain Mountainous Terrain

Percent Shoulders Shoulders Shoulders

Trucks ft~edTlJnpaved Paved Unpaved Paved Unpaved

< 10% $6,216 4,1

15 ~~$6,"3 $4,350 $8,291 $6,198

,443 $4,342 $8,291 $6,198

1 1

Accidents and Accident Costs

Related Accident Rate Annual Accidents per Annual Accident Costs per
(per MVM) Mile per 100 vpd Mile per 100 vpd

Lane Width/Shoulder Width Terrain Terrain Terrain

(feet) L I R - M L R - M L R M

11/4 1.18 1.31 1.76 .0.043 0.048 0.064 $2,584 $2,880 $3,840

11/2 1.74 1.87 2.32 0.064 0.068 0.085 $3,840 $4,080 $5,100

Difference — — — — — $1,256 $1,200 $1,260

Breakeven ADT for 11-Foot Lane Width and 4-Foot Shoulder Width (vpd)

Level Ter in Rolling Terrain Mountainous Terrai

Percent Shoulders Shoulders Shoulders

Trucks Paved Unpaved Paved Unpaved Paved Unpav d

< 10% 1: 495 328-1 ' 537 362 658 491

> 10% 495 328 1 537 362 658 491

1

Table 25 compares an I I -ft lane and 2-ft shoulder design with vpd for mountainous terraIn. Where shoulders are unpaved and
a 10-ft lane and 2-ft shoulder design. The former design has a shoulder costs are lower, the breakeven ADT levels for the 10-
better expected safety performance, at a higher construction cost. ft lane with a 6-ft shoulder design are somewhat lower-400
It is concluded that the wider lane width becomes more cost- vpd in level terrain, about 500 vpd in rolling terrain, and 750
effective at ADT levels of 350 to 400 vpd for level to rolling vpd in mountainous terrain.
terrain, and at about 500 vpd for mountainous terrain.

Comparison of 11-Ft Lanes with 4-Ft Shoulders to

Comparison of 10-Ft Lanes with 6-Ft Shoulders to 10-Ft Lanes with 5-Ft Shoulders
11-Ft Lanes with 2-Ft Shoulders
For certain functional - classes and design speeds, either 10-
Table 26 summarizes a comparison of two distinctly different or 11-foot lanes may be appropriate. Table 27 summarizes the
designs. A cross section with 10-ft lanes and wide, 6-ft shoulders analysis of a comparison between an I I -ft lane width and a 4-
has a lower expected accident rate than one with I 1-ft lanes and ft shoulder width, design with a 10-ft lane width and a 5-foot
narrow, 2-ft shoulders. The wider total cross section has a greater shoulder width design. The wider lane'width design offers a
construction cost as well. slight accident rate advantage at higher typical unit costs of
Using the analysis results as a general guideline, it is concluded construction. Note that the cost differences are greater for cases
that, where paved shoulders are used, the 10-ft lane with a 6-ft in which the shoulder is unpaved and where the pavement is
shoulder design is cost-effective only at relatively high traffic designed to accommodate high truck volumes.
volumes-800 to 900 vpd for level and rolling terrain and 1,200 Using the analysis results as a general guideline, it is concluded
34

TABLE 24. Comparison of 12-ft lanes with 4-ft shoulders to 12-ft lanes with 2-ft shoulders

Typical Construction Costs per Mile ($000s) from Table 21

Level Terrain Rolling Terrain Mountainous Terrain

Lane/Shoulder Width Percent Shoulders Shoulders Shoulders

(Feet) Trucks Paved I Unpaved Paved _T

Unpaved - Paved
____ TUnpaved

* 10% $909.8 $861.6 $962.3 $914.1 $1,388.9 $1,340.7

12/4
* 10% $982.2 $934.0 $1,034.7 $986.4 $1,461.3 E
$"
$1,4

<10% $838.5 $814.4 $888.4 $864.3 $1,293.8 $" 269 .7

' 3 .0
*0
12/2
> 10% $910.9 $886.7 $960.7 $936.6 $1,366.1 342J

Difference in Annualized Construction Costs per Mile

(20-Year Project Life, 6-Percent Discount Rate, and Capital Recovery Factor of 0.08718)

Level Terrain Rolling Terrain Mountainous Terrain

Percent Shoulders Shoulders Shoulders

Trucks Paved Unpaved Paved Unpaved Paved Unpaved

< 10% $6,216 $4,115 $6,442 $4,342 $8,291 $6,190

> 10% $6,216 $4,124 $6,451 $4,342 $8,299 $6,190

Accidents and Accident Costs

Related Accident Rate Annual Accidents per Annual Accident Costs per
(per MVM[) Mile per 100 vpd Mile per 100 vpd

Lane Width/Shoulder Width Terrain Terrain Terrain

(feet) L R M L I R R M

12/4 1.18 1.31 1.76 0.043 0.048 0.064 $2,584 $2,880 $3,840

12/2 1.74 1.87 2=3

2 Oi4:: 0.068
: 0.085 $3,840 $4,080 $5,100

Difference
— $1,256 $1,200 $1,260

Breakeven ADT for 12-Foot Lane Width and 4-Foot Shoulder Width (vpd)

Level Terrain Rolling Terrain MO ntainous errain

Percent Shoulders Shoulders Shoulders

Trucks Paved Unpaved Paved Unpaved Paved Unpaved

< 10% 495 328 537 362 658 491 1

> 10% 495 328 537 362 658 491
1~

that the I I -ft lane width with 4-ft shoulder width design is cost- The accident analyses do not show a significant difference in
effective for highways with paved shoulders at ADT greater than the safety performance of the two alternatives. The question of
400 to 500 vpd. The 11 -ft lane width with 4-ft shoulder width choice, thus, is reduced to an assessment of the difference in
design is cost-effective where unpaved shoulders are used for costs. As shown in Table 28, typical construction costs are greater
highways with low truck volumes (less than 10 percent) and for the design with 12-ft lanes in all cases. Note, however, that
ADT greater than 750 vpd. Where truck volumes are greater the actual cost differential is very small, particularly for roads
than 10 percent, this design is cost-effective when ADT exceeds with paved shoulders and for roads designed for lower truck
1,000 vpd. volumes.
It is concluded that, in general, a cross section with I I -ft lane
widths and 4-ft shoulders is preferred for all volume levels for
Comparison of 12-Ft Lanes with 3-Ft Shoulders to highways with unpaved shoulders and with pavements designed
11-Ft Lanes with 4-Ft Shoulders for truck volumes greater than 10 percent of ADT. For highways
with paved shoulders that are designed to accommodate lower
For certain functiQnal classes and design speeds, either an I I- truck volumes (less than 10 percent of ADT), the 11-ft lane
ft or 12-ft lane width may be appropriate. Table 28 suffunarizes width and 4-ft shoulder width design is only marginally preferred.
a comparison between a 12-ft lane width and 3-ft shoulder width Within the limitations of cost modeling and the assumptions, no
design with an I I -ft lane width and 4-ft shoulder width design. meaningful difference exists.
 35

TABLE 25. Comparison of 11-ft lanes with 2-ft shoulders to 10-ft lanes with 2-ft shoulders

Typical Construction Costs per Mile ($000s) from Table 21

Level Terrain Rolling Terrain Mountainous Terrain
Lane/Shoulder Width Percent Shoulders Shoulders Shoulders
(feet) Trucks Paved I Unpaved Paved aved

11/2
< 10% $793.8 $769.7 1 $842.4 $818.2 $1,237.2 $1,213.0

> 10% $860.1 $836.0 $908.7 $884.6 $1,303.5 $1,279.4

< 10% $749.1 $725.0 $796.4 $772.2 $1,180.5 $1,156.4
10/2
> 10% $809.4 $785.3 $856.7 $832.5 $1,240.8 $1,216.7

Difference in Annualized Construction Costs per Mile

(20-Year Project Lire, 6-Percent Discount Rate, and Capital Recovery Factor of 0.08718)
Level Terrain Rolling Terrain Mountainous Terrain
Percent Shoulders Shoulders Shoulders
Trucks Paled T Unpaved Paved F-
Unpaved Paved Unpaved

< 10% $3,897 $3,897 $4,010 $4,010 $4,943 $4,943

> 10% $4,420 $4,420 $4,542 $4,542 $5,466 $5,466

Accidents and Accident Costs

Related Accident Rate Annual Accidents per Annual Accident Costs per
(per MM Mile per 100 vpd Mile per 100 vpd
Lane Width/Shoulder Width Terrain Terrain Terrain
(feet) L I R I M L I R I M L R M

11/2 1.74 1.87 2.32 0.064 0.068 0.08S $3,840 $4,080 $5,100

10/2 2.28 2.41 2.86 0.083 0.088 0.104 $4,980 $5,280 $6,240

Difference — — — — — — $1,140 $1,200 $1,140

Breakeven ADT for 11-Foot Lane Width and 2-Foot Shoulder Width (vpd)
Level Terrain Rolling Terrain Mountainous Terrain
Percent Shoulders I Shoulders Shoulders
Trucks

< 10%
> 10%
Paved

342
388
Unpaved

342
388
Paved

334
377
Unpaved
334
377
Paved

433
479
Unpaved
.

433
479
J
36

TABLE 26. Comparison of 10-11 lanes with 6-ft shoulders to 11-ft lanes with 2-ft shoulders

Typical Construction Costs per Mile ($000s) from Table 21

Level Terrain, Rolling Terrain Mountain us Terrain

Lane/Shoulder Width Percent Shoulders Shoulders Shoulders

(feet) Trucks ------

Payed T Un aved Paved I Unpaved Paved Unpaved

< 10% $891.7 $819.4 $944.2 $871.9 $1,370.8 $1,298.5

10/6
* 10% $952.0 $879.7 $1,004.5 $932.2 $1,431.1 $1,358.8

* 10% $793.8 $769.7 $842.4 $818.2 $1,237.2 $1,213.0

11/2
> 10% $860.1 $836.0 $908.7 $884.6 $1,303.5 $1,279.4

Difference in Annualized Construction Costs per Mile

(20-Year Project Life, 6-Percent Discount Rate, and Capital Recovery Factor of 0.08718)

Level Terrain Rolling Terrain Mountainous Terrain

Percent Shoulders Shoulders Shoulders

Trucks Paved Unpaved Paved Unpaved Paved Unpaved

< 10% $8,535 $4,333 $8,875 $4,682 $11,647 $7,454

> 10% $8,012 $3,809 $8,352 $4,150 $11,124 $6,922

Accidents and Accident Costs

Related Accident Rate Annual Accidents per Annual Accident Costs per
(per MVM) Mile per 100 vpd Mile per 100 vpd

Lane Width/Shoulder Width Terrain Terrain Terrain

(feet) L R M R M L R I M

10/6 1.30 1.43 1.88 0.047 0.052 0.069 $2,820 $3,120 $4,140

11/2 1.74 1.87 2.32 0.064 0.068 0.085 $3,840 $4,080 $5,100

Difference $1,020 $960 $960

Breakeven ADT for 10-Foot Lane Width and 6-Foot Shoulder Width (vpd)

Level Terrain Rolling Terrain Mountainous Terrain

Percent Shoulders Shoulders Shoulders -

Trucks Paved Unpaved Paved Unpaved Paved ~~

Unpaved

< 10% 837 425 924 488 !21

l' 776

> 10% 785 373 870 432 1 159 721

 37

TABLE 27. Comparison of 11-ft lanes with 4-ft shoulders to 10-ft lanes with 5-ft shoulders

Typical Construction Costs per Nfile ($000s) from Table 21

Level Terrain Rolling Terrain Mountainous Terrain

Lane/Shoulder Width Percent Shoulders Shoulders Shoulders

(feet) Trucks Paved Unpaved Paved I Unpaved Paved I Unp ved

< 10% $865.1 $816.9 $916.3 $968.1 1 $1,332.3 $1,284.1

11/4
> 10% $931.5 $883.2 $982.6 $934.4 $1,398.6 $1,350.4

< 10% $856.1 $795.8 $907.3 $847.0 $1,323.3 $1,263.0

10/5
> 10% $916.4 $856.1 $967.6 $907.3 $1,383.6 $1,323.3

Difference in Annualized Construction Costs per Mile

(20-Year Project Life, 6-Percent Discount Rate, and Capital Recovery Factor of 0.08718)

Level Terrain Rolling Terrain Mountainous Terrain

Percent Shoulde Shoulders Shoulders

Trucks Paved Unpaved Paved Unpaved Paved Unpaved

:< 10% $785 $1,839 $785 $1,839 $785 $1,839'

> 10% $1,316 $2,363 $1,308 $2,363 $1,308 $2,363

Accidents and Accident Costs

Related Accident Rate Annual Accidents per Annual Accident Costs per
(per MM Mile per 100 vpd Mile per 100 vpd

Lane Width/Shoulder Width Terrain Terrain Terrain

(feet) L -F—R :F M L I R I M . L I R M

11/4 1.18 1.31 1.76 0.043 0.048 0.064 $2,580 $2,880 $3,880

10/5 1.30 1.43 1.88 0.047 0.052 0.069 $2,820 $3,120 $4,140

Difference — — — — — — $240 $240 $260

Breakeven ADT for 11-Foot Lane Width and 4-Foot Shoulder Width (vpd)

Level Terrain Rolling Terrain Mountainous Terrain

Percent Shoulders Shoulders Shoulders

Trucks Paved Unpaved Paved 17~;aved Paved 1 Unpaved

< 10% 327 766 327 766 302 707

> 10% 548 984 545 984 503 909

38

TABLE 28. Comparison of 12-ft lanes with 3-ft shoulders to 11-ft lanes with 5-ft shoulders

Typical Construction Costs per Mile ($000s) from Table 21

Level Terrain Rolling Terrain Mountainous Terrain

Lane/Shoulder Width Percent Shoulders Shoulders Shoulders

(feet) Trucks Paved I Unpaved Paved I Unpaved Paved T_Unpaved

* 10% $874.2 $838.0 $925.3 $889.2 $1,341.3 $1,305.2

12/3
* 10% $946.5 $910.4 $997.7 $961.5 $1,413.7 $1,377.5
* 10% 865.1 816.9 916.3 868.1 1,332.3 1,284.1
11/4
* 10% 931.5 883.2 982.6 934.4 1,398.6 1,350.4

Difference in Annualiwd Construction Costs per Mile

(20-Year Project Life, 6-Percent Discount Rate, and Capital Recovery Fac(or of 0.08718)

Lev I Terrain Rolling Terrain Mountainous Terrain

Percent -Shoulders Shoulders Shoulders

Trucks Paved Unpaved ----------

Paved Unpaved Paved I Unpaved

< 10% $793 $1,839 $785 $1,839 $785 F$1,839

> 10% $1,308 $2,371 $1,316 $2,363 $1,316 1 $2,363

Accidents and Accident Costs

Related Accident Role Annual Accidents per Annual Accident Costs per
(per MVM) Mile per 100 vpd Mile per 100 vpd

Lane Width/Shoulder Width Terrain Terrain Terrain

—
(feet) L I R I M M L
I
12/3 1.18 1.31 1.76 0.043 0.048 0.064 $2,584 $2,880 $3,840

11/4 1.18 1.31 1.7 0.043 0.048 0.064 $2,584 $2,880 $3,840

Difference — — — — — — $0 ;$o4~;
$0

Breakeven ADT for 12-Foot Lane Width and 3-Foot Shoulder Width (vpd)

Level Terrain Rolling Terrain Mountainous Terrain

Shoulders Shoulders Shoulders

Percent
Trucks Paved Unpaved Paved Unpaved Paved Unpaved

< 10% N/A N/A N/A N/A N/A N/A

> 10% N/A N/A N/A N/A N/A N/A
 39

CHAPTER 3
RESULTS, INTERPRETATION, APPRAISAL, AND APPLICATION

The primary research objective is to refine current width- lane width considered the absolute minimum acceptable design.
related design values for low-volume roads. The AASHTO Pol- Greater lane widths of up to 12 ft are considered appropriate for
icy suggests the framework for these standards. Design standards even the lowest volumes where terrain and design speed dictate
are appropriately based on the following considerations: func- (refer to Appendix F).
tional classification, design speed, highway operational require- A minimum shoulder width value of 2 ft for the lowest
ments, safety -effectiveness, and construction and maintenance ADT class was established. A maximum shoulder width value
costs. of 8 ft was used for higher volume arterials and 6 ft for other
It is noted that design standards are not based strictly on classes.
"cost-effectiveness." There are certain minimum width values Safety cost-effectiveness was the basis for recommending
necessary to accommodate the functional requirements of the total roadway width values for each ADT range. This total width,
drivers and vehicles that use the highway, regardless of the then, was allocated in each case, given the relationships discussed
volume of traffic. These functional requirements include con- above between design speed and lane width.
struction and design features, shoulder use, and highway opera- Where applicable (see Appendix E), the total lane and
tions. shoulder width values were increased to provide a compatible
Another central point regarding roadway width design stan- design with respect to capacity and level of service.
dards is the need to confirm or establish compatibility with other
design standards and criteria. The researchers consider a set of COMPARISON OF RECOMMENDED WIDTHS TO
design standards flawed if, by their very nature, they produce CURRENT AASHTO POLICY
operational or design conflicts when used with other established
standards. The recommended values for minimum lane and shoulder
Specific to low-volume roads, derivation of appropriate width widths on low-volume roads differ, significantly in some cases,
design standards is based on the following approach. For road- from current AASHTO Policy values.
ways with extremely low volumes, appropriate widths are based
on minimum functional dimensions as well as compatibility with Format
the design speed of the facility. Incrementally, greater widths
for the total roadway (lanes plus shoulders) become cost-effective In keeping with the cost and level-of-service sensitivities asso-
as the design volume increases, depending on the terrain and ciated with terrain, width values are described in terms of level,
other cost sensitivities noted earlier. Consistency with speeds rolling, and mountainous terrain. The cost-effectiveness studies
and sensitivity to shoulder functions controls the allocation of in this research indicate that, in some instances, narrower widths
total width to maintain design standard compatibility. are appropriate in mountainous terrain versus slightly wider val-
Although the accident studies provide little justification for ues in rolling or level terrain. Also, width values are shown
total roadway widths greater than 30 ft, other factors become relative to design year ADT volume rather than to the somewhat
important as design traffic volumes increase to 1,500 vpd or confusing practice in AASHTO of mixing existing traffic, design
more. Lane and shoulder width design values should promote traffic, ADT, and DHV. Finally, values for total width are shown
reasonable truck operations and enable achieving reasonable lev- in addition to lane width and shoulder width.
els of service for the given terrain, functional classification, and
design traffic conditions. Local Road Design Values
Tables 29(a) through 29(f) show recommended minimum de-
sign values for lane width, shoulder width, and total roadway The recommended design widths for local roads represent
width. To summarize, derivation of these widths was based on significant revisions to current AASHTO Policy values. Current
the following general approach: policy specifies a range in lane widths from 9 ft to 12 ft, in
shoulder widths from 2 ft to 6 ft, and in total width from 22 ft
The AASHTO framework for design values was used. to 36 ft. Moreover, the minimum widths (both by element as
Widths are characterized as a function of design speed, terrain, well as total) increase steadily as traffic volumes increase.
functional classification, and traffic characteristics (volume and The research performed here supports a different approach to
percentage of trucks). design standards. First, for local roads, design speeds are gener-
Minimum lane width values are first established for a given ally in the range of 30 mph to 50 mph, with 20 mph possible
design speed and functional classification. The basis of these in mountainous terrain. Under such operating conditions, given
minimum widths is consistericy in operating speed, with a 9-ft the classification and general volume levels, 10-ft lanes are ap-
40

TABLE 29(a). Local roads, design year ADT 0-2000 vpd (!~ 10% trucks)

Design Year Traffic Volumes (ADT) in Vehicles per Day

Design Speed Cross Section

(mph) Elements* (feet) 0-250 250400 1 400-750 1 750-1500 1500-2000 Comments

LEVEL TERAIN

TW (22) (22) (30) (30) (32) 9- or 10-foot lanes are

30 I-W 9 9 10 to 10 appropriate for
SW 2 2 5 5 6 30 mph
TW (22) (22) (30) (30) (32) 9- or 10-foot lanes are
40 LW 9 9 10 10 10 appropriate for
SW 2 2 5 5 6 40 mph

TV (24) (24) (30) (30) (34) *Minimum 10-foot

50 LW 10 10 11 11 11 lanes are appropriate
SW 2 2 4 4 6 for 50 mph

T'W
LW - - - -
SW
ROLLING T ER A I N
TW (22) (22) (30). (30) (32) 9- or 10-foot lanes are
30 I-W 9 9 10 10 10 appropriate for
SW 2 2 5 5 6 30 mph
TW (22) (22) (36) (30) (32) 9- or 10-foot lanes are
40 LW 9 9 10 10 10 appropriate for
SW 2 2 5 5 6 40 mph

rw (24) (24) (30) (30) (34) 10-foot lanes are

50 I-W 10 10 11 11 11 appropriate for
SW 2 2 4 4 6 50 mph
MOUNTAINOUS TER A I N
TV (22) (22) (22)* (30) (32) 9- or 10-foot lanes are
20 LW 9 9 9 10 10 appropriate for
SW 2 2 2 5 6 20 mph

TW (22) (22) (22) (30) (32) 9- or 10-foot lanes are

30 I-W 9 9 9* 10 10 appropriate for
SW 2 2 2 5 -6 30 mph
TW (22) (22) (30) (30) (32) 9- or 10-foot lanes are
40 LW 9 9 10 10 10 appropriate for
SW 2 2 5 5 6 40 mph

TW
50 I-W - - - - -
SW
ota roadway width (lanes plus shoulders) *10-foot lanes and 5-foot shoulders may be
LW—Lane width considered as reasonable minimums for ADT
SW—Shoulder width (paved or unpaved) greater than 600 vpd.

propriate in most cases. Indeed, for lower speed, lower volume Twelve-ft lanes not only are more costly, but may promote
conditions the research supports 9-ft minimum lane widths. For higher speeds (speeds incompatible with horizontal and vertical
50-mph design speed local roads, 11-ft lane widths may be alignment) than desired for these facilities. For these reasons,
appropriate, particularly when ADT is greater than 400 vpd. 12-ft lane widths are not reconunended as a basic standard.
 41

TABLE 29(b). Local roads, design year ADT 0-2000 vpd (> 10% trucks)

Design Year Traffic Volumes (ADT) in Vehicles per Day

Design Speed Cross Section

(mph) Elements* (feet) 0-25OT25040 400-750 1 750-1500 1500-2000 Comments

F (22)
LEVEL TERRAIN

(22) (30) (30) (32) 9- or 10-foot lanes are

TW
30 LW 9 9 10 10 10 appropriate for
SW 2 2 5 5 6 30 mph
TW (24) (22) (30) (30) (32) 9- or 10-foot lanes are
40 LW 9 9 10 10 10 appropriate for
SW 2 2 5 5 6 40 mph

TW (24) (24) (30) (30) (34) 10-foot minimum

50 LW 10 10 11 11 11 lanes are appropriate
SW 2 2 4 4 6 for 50 mph

TW
LW - - - - -
SW

F (22)
ROLLING TERRAIN

(22) (30) (30) (32) 9- or 10-foot lanes are

TW
30 LW 9 9 10 10 10 appropriate for
SW 2 2 5 5. 6 30 mph
TW (22) (22) (30) (30) (32) 9- or 10-foot lanes are
40 LW 9 9 10 10 10 appropriate for
SW 2 2 5 5 6 40 mph

TV (24) (24) (30) (30) (34) 10-foot minimum

50 LW 10 10 11 11 11 lanes are appropriate
SW 2 2 4 4 6 for 50 mph

MOUNTAINOUS TERRAIN

TW (22) (22) (22)* (30) (32) 9- or 10-foot lanes are

20 LW 9 9 9 10 10 appropriate for
SW 2 2 2 5 6 30 mph
T'W (22) (22) (30) (30) (32) 9- or 10-foot lanes are
30 LW 9 9 10 10 10 appropriate for
SW 2 2 5 5 6 40 mph

TW (22) (22) ~30) (30) (32) 10-foot minimum

40 LW 9 9 10 10 10 lanes are appropriate
SW 2 2 5 5 6 for 50 mph

TW
50 LW - - - - -
SW
— otal roadway width (lanes plus shoulders) *10-foot lanes and 5-foot shoulders may be
LW—Lane width considered as reasonable minimums for ADT
SW--Shoulder width (paved or unpaved) greater than 600 vpd.

In terms of combinations of lane and shoulder width, safety is no safety justification for a total width greater than 30 ft. Note
cost-effectiveness identifies traffic volume levels of about 300 that where 11 -ft lanes are considered appropriate, 4-ft shoulders,
vpd as the point at which a 5-ft shoulder is cost-effective, com- producing a total width of 30 ft, result in significant safety
bined with a 10-ft lane width. Beyond this point, however, there benefits. The cost-effectiveness analysis showed that ADT levels
42

TABLE 29(c). Collector roads, design year ADT 0-2000 vpd (!~ 10% trucks)

Design Year Traffic Voltunes (ADT) in Vehides per Day

Design Speed Cross Section

(mph) Elements* (feet) 0-250 250400 400-750 1 ' 750-1500 1500-2000 Comments

LEVEL TERRAIN

TW (22) (22) (30) (30) (32) 9- or 10-foot lanes

40 LW 9 9 10 10 10 appropriate for
SW 2 2 5 5 6 40 mph

TW (24) (26) (30) (30) (34) 10- or 11-foot lanes

50 LW 10 11 11 11 11 are appropriate for
SW 2 2 4 4 6 50 mph
TW (26) (26) (30) (30) '(34) 11-foot lanes are
60 LW 11 11 11 11 11 appropriate for
SW 2 2 4 4 6 60 mph
ROLLING TERRAIN

TV (22) (22) (30) (30) (32) 9- or 10-foot lanes are

30 LW 9 9 10 10 10 appropriate for
SW 2 2 5 5 6 30 mph
TW (22) (22) (30) (30) (32)' 9- or 10-foot lanes
40 LW 9 9 10 10 10 appropriate for
SW 2 2 5 5 6 40 mph

TW (24) (24) (30) (30) (34) 10- or I I -foot lanes

50 LW 10 10 11 11 11 are appropriate for
SW 2 2 4 4 6 50 mph
TW (26) (26) (30) (30) (34) 11 -foot lanes are
60 LW II II II 11 11 appropriate for
SW 2 2 4 4 6 60 mph
MOUNTAINOUS TERRAIN

TW (22) (22) (22)* (30) (32) 9-foot lanes are

20 LW 9 9 9 10 10 appropriate for
SW 2 2 2 5 6 20 mph

TW (22) (22) (22)* (30) (32) 9- or 10-foot lanes are

30 LW 9 9 9 10 10 appropriate for
SW 2 2 2 5 6 30 mph
TW (24) (24) (30) (30) (32) 9- or 10-foot lanes
40 LW to 10 10 10 10 appropriate for
SW 2 2 5 5 6 40 mph

TW (24) (24) (36) (30) (34) 10- or I I -foot lanes

50 LW 10 10 11 11 11 are appropriate for
SW 2 2 4 4 6 50 mph
ota roadway width (lanes plus shoulders) *10-foot lanes and 5-foot shoulders may be
LW—Lane width considered as reasonable minimums for ADT
SW---Shoulder width (paved or unpaved) greater than 600 vpd.

that justify this allocation are around 250 vpd for level and The net effect of the recorrimendations is that the total width
rolling terrain. Finally, for the highest volume range within the of local roads should vary much less with traffic volume than
study(1,500 to 2,000 vpd), a minimum 6-ft shoulder is considered AASHTO indicates. Above an ADT of 250 vpd, (400 vpd in
appropriate for shoulder functions. This leads to a minimum total mountainous terrain),30-ft total widths arejustified. Total widths
width of 32 ft to 34 ft for the higher volume ranges. significantly above this value are notjustified in terms of safety,
 43

TABLE 29(d). Collector roads, design year ADT 0-2000 vpd (> 10% trucks)

Design Year Traffic Volumes (ADT) in Vehicles per Day

Design Speed Cross Section

(mph) Elements* (feet) 0-250 250400 400-750 1 750-1500 1500-2000 Comments

F LEVEL TERRAIN

TW (22) (22) (30) (30) (32) 9- or 10-foot lanes

40 LW 9 9 10 10 10 appropriate for
SW 2 2 5 5 6 40 mph

TW (24) (26) (30) (30) (34) 10- or 11 -foot lanes

50 LW 10 11 11 11 11 are appropriate for
SW 2 2 4 4 6 50 mph

TW (26) (26) (30) (30) (34) 11-foot lanes are

60 LW 11 11 11 11 11 appropriate for
SW 2 2 4 4 6 60 mph

ROLLING TERRAIN

TV (22) (22) (30) (30) (32) 9- or 10-foot lanes are

30 LW 9 9 10 10 10 appropriate for
SW 2 2 5 5 6 30 mph

TW (22) (22) (30) (30) (32) 9- or 10-foot lanes

40 LW 9 9 10 10 10 appropriate for
SW 2 2 5 5 6 40 mph

TW (24) (24) (30) (30) (34) 10- or I I -foot lanes

50 LW 10 10 11 11 11 are appropriate for
SW 2 2 4 4 6 50 mph

TW (26) (26) (30) (30) (34) 11 -foot lanes are

60 LW 11 11 11 11 11 appropriate for
SW 2 2 4 4 6 60 mph

MOUNTAINOUS TERRAIN

TW (24) (24) (30) (30) (32) 9-foot lanes are

20 LW 9 9 10 10 10 appropriate for
SW 3 3 5 5 6 20 mph

TW (24) (24) (30) (30) (32) 9- or 10-foot lanes are

30 LW 9 9 10 10 10 appropriate for
SW 3 3 5 5 6 30 mph

TW (24) (24) (30) (30) (32) 9- or 10-foot lanes

40 LW 10 10 10 10 10 appropriate for
SW 2 2 5 5 6 40 mph

TW (24) (24) (30) (30) (34) 10- or I I -foot lanes

50 LW 10 10 11 11 11 are appropriate for
SW 2 2 4 4 6 50 mph

— ota roadway width (lanes plus shoulders)

LW—Lane width
SW Shoulder width (paved or unpaved)

capacity, or design speed compatibility and, therefore, are not AASHTO minimum width values is in lane width. Ten-ft lanes
recommended as the basis for design standards. are judged sufficient for lower design-speed collectors (30 to 40
mph). Indeed, this width is more compatible with the alignment
Collector Design Values quality expected in rolling and mountainous terrain. Moreover,
The most significant difference between recormnended and wider lane widths are not required to provide sufficient capacity.
44

TABLE 29(e). Arterial roads, design year ADT 0-2000 vpd (!~ 10% trucks)

Design Year Traffic Volumes (ADT) in Vehicles per Day

Design Speed Cross Section

(mph) Elements* (feet) 0-250 -F 2501~1 400-750 1 750-1500 1500-2000 Comments

LEVEL TERRAIN

TW
40 LW
SW
TW
50 LW
SW
TW (26) (26) (30) (36) (40) 11 - or 12-foot lanes
60 LW 11 11 11 12 12 are appropriate for
SW 2 2 4 6 8 60 mph
TW (28) (28) (30) (36) (40) 12-foot lanes are
70 LW 12 12 12 12 12 appropriate for
SW 2 2 4 6 8 70 mph
ROLLING TERRAIN

TW (24) (24) (30) (34) (38) 9-foot lanes are

40 LW 9 9 11 11 11 appropriate for
SW 3a 3a 4 6 8 40 mph

TW (26) (26) (30) (34) (38) 10- or 11-foot lanes

50 LW 10 10 11 or 10 11 11 are appropriate for
SW 3a 3a 4 or 5b 6 8 50 mph
TW (26) (26) (30) (30) (40) 11 - or 12-foot lanes
60 LW I1 11 1 11 12 are appropriate for
SW 2 2 4 4 8c 60 mph
MOUNTAINOUS TERRAIN
TW
20 LW -
SW
TW
30 LW
SW
TW (26) (26) (26) (34) (34) 10- or I I -foot lanes
40 LW . 10 10 11 11 11 are appropriate for
SW 3a 3a 2 6 6 40 mph

TW (26) (26) (26) (34) (34) 10- or 11-foot lanes

50 LW 10 10 11 11 11 are appropriate for
SW 3a 3a 2 6 6 50 mph
TW—Total width (lanes plus shoulders) '346ot shoulders recommended as minimums for arterials with 9- or 10-foot lanes.
LW—Lane width b10/5 design is minimum for unpaved shoulders; 11/4 design is minimum for paved shoulders.
SW----Shoulder width (paved or unpaved) 'Minimum lane and shoulder widths to provide sufficient capacity.

The accident studies confirm that 10-ft lanes, when coupled with cases, note that the minimum shoulder width need only be 4 ft
at least 5-ft shoulders, are cost-effective. For speeds of 50 mph to provide the most cost-effective design. Finally, in no case is
and 60 mph, a greater lane width of'I'1 ft may be used, particu- a recommended minimum 12-ft lane width required, given that
larly, on roads with traffic volumes above 750 vpd. In such collectors are generally designed with, at most, a 60-mph design
 45

TABLE 29(f). Arterial roads, design year ADT 0-2000 vpd (2~ 10% trucks)

Design Year TmWic Volumes (ADT) in Vehicles per Day

Design Speed Cross Section

(mph) Elements* (feet) 0-250 250400 400-750 1 750-1500 1500-2000 Comments

L TW
LEVEL TERRAIN

40 LW
SW
TW
50 LW
SW
TW (26) (26) (30) (30) (38) 11- or 12-foot lanes
60 LW 11 11 11 11 11 are appropriate for
SW 2 2 4 4 8 60 mph

TW (28) (28) (32) (32) (40) 12-foot lanes are

70 LW 12 12 12 12 12 appropriate for
SW 2 2 4 4 8 70 mph

ROLLING TERRAIN

TW (26) (26) (30) (30) (38) 10-foot minimum

11 11 11 11b lanes are appropriate
40 LW 10
SW 3a 2 4 4 8 for high truck percent

TW (26) (26) (30) (30) (38) 10- or 11-foot lanes

11 11 11 llb are appropriate for
50 LW 10
SW 3a 2 4 4 8 50 mph

TW (26) (26) (30) (31a) (40) 11 - or 12 foot lanes

60 LW II II II II 12b are appropriate for
SW 2 2 4 4 8 60 mph

MOUNTAINOUS TERRAIN

TW
20 LW
SW
TW
30 LW
SW
TW (26) (26) (26) (34) (36)
40 LW 10 10 11 11 12b
SW 3a 3a 2 6b 6
TW (26) (26) (26) (34) (36)
50 LW 10 11 11 11 12b
SW 3a 2 2 6 6
TW—Total roadway width (lanes plus shoulders) 'Moot minimum shoulders recommended in rolling and mountainous terrain with
LW—Lane width 10-foot lanes.
SW---Shoulder width (paved or unpaved) bMinimurn lane and shoulder widths to provide required capacity.

speed. For the highest volume design class(1,500 to 2,000 vpd), coupled with wide shoulders. Finally, 11-ft lane widths are
I I -ft lanes and 6-ft shoulders provide sufficient highway capacity reasonably compatible with even 60-mph speeds.
and reasonable shoulder functions. There is no identifiable safety Note, again, that the sensitivity of total roadway width to ADT
benefit of a 12-ft versus I 1-ft lane width, in either case, when is much less than that shown by current AASHTO design values.
 46

The most cost-effective total width of 30 ft is justified at low of widening from 24-ft or 26-ft total roadway widths to 30
traffic volumes-250 vpd in level and rolling terrain and 400 ft. For existing narrow roadways, minimal widening may be
vpd in mountainous terrain. significantly effective at very low volumes.
Conversely, where the existing total width is 30 ft or greater,
width-related 3R safety improvements may be minimal. In such
Arterial Design Values cases, the focus should probably be shifted to roadside improve-
ments, spot geometric improvements, and traffic control.
The recommended lane width values are less than those called Routine 3R widening of very narrow (less than 10-ft lanes)
for by AASHTO for lower volume arterials (less than 400 vpd). roadways should not necessarily be performed. Rather, decisions
For very low-volume and lower speed arterials (40 to 50 mph to widen should be based in large pail on analysis of existing
design speed), there is no evidence to suggest that I I -ft or 12- accident experience. Related accident rates contained in this
ft lanes are necessary as minimums. Also, for such traffic vol- report (also see Appendixes A and B) can be referenced to
umes, 2-ft shoulders rather than 4-ft shoulders generally appear deten-nine whether or not a location is unusual or a "high-acci-
sufficient. Minimum design values for shoulder width (and total dent" location.
width) are not increased substantially until design traffic exceeds For 31Z projects on highways with design volumes of 1,500
750 vpd. For the highest volume design class, full 8-ft shoulders vpd or greater, decisions regarding widening may involve high-
are recommended on arterials in rolling and level terrain. Also, way operational requirements. An appropriate approach would
minimum 12-ft lanes are recommended, in some cases, to provide be to conduct speed and volume studies to determine actual
compatibility with highway capacity and level-of-service require- operations and level of service. Increasing lane widths for capac-
ments. Again, note that the total width requirements vary much ity or truck traffic should be done on a site-specific basis, al-
less by volume than current AASHTO Policy. though the design values given in Tables 29(a) through 29(f)
can be referenced.
Any 3R design project should be undertaken with a sensitiv-
APPLICATION TO 3R DESIGN PROBLEMS ity to the classification and design speed issues discussed herein.
Care should be taken not to widen a narrow road until it can be
The accident relationships developed in this research provide determined that any resulting speed increases are appropriate
guidance toward appropriate, cost-effective treatments for ex- given the existing horizontal and vertical alignment. (Note: The
isting low-volume road cross sections. The following general nature of 31Z projects is that they involve minimal alignment
findings provide a basis for such guidelines: revisions or reconstruction.)
Widening shoulders rather than lanes is less costly and will
Safety-effectiveness is primarily sensitive to total roadway be more cost-effective. For 3R projects where existing total width
width, and less sensitive to the allocation of width between lanes is 26 ft or less, the optimal approach may be to focus on shoulder
and shoulders. (This is true for normal combinations of lane widening. The discussion, earlier, regarding lane width and speed
width and shoulder width.) compatibility can be useful in deciding the need for lane widening
Within the range of typical total roadway widths, the safety- on such narrow roads.
effectiveness of incremental widths is distinctly nonlinear. For Designers should investigate spot widening alternatives be-
total roadway widths of 24 ft or less, related accident rates are cause these options may be more cost-effective than continuous
in the range of 1.7 to 2.3 accidents per MVM. Related accidents widening. Widening through curves, implementation of truck
decrease significantly as total widths increase to 28 to 30 ft. climbing lanes, or spot shoulder widening may all be effective,
Above 30 ft, little or no safety-effectiveness is expected with site-specific alternatives to continuous widening.
incremental widening.
Specifically, the research did not uncover high accident
rates for roadways with 8-ft or 9-ft lane widths. The accident Procedures for Evaluation of Low-Volume Road
effects of 8-ft lane widths were mixed and unclear from the Design Options
available low-volume data because of the small sample of mile-
age. Based on the available evidence, the use of lane widths Appendix A, "Cost-Effectiveness Guide for Design Improve-
below 9 ft is not considered to be desirable. ments to Roadway Cross Sections for Low-Volume Rural
The effect of terrain on accident rates does not appear to Roads," supplements the roadway width guidelines developed
be width-sensitive. ("Terrain" is considered a surrogate variable in this study. The Guide provides a procedure for computing
describing quality of horizontal and vertical alignment, and also the accident benefits related to various design improvements to
roadside quality). The effect of terrain is the same, for example, roadway cross section (i.e., widening lanes, widening shoulders,
on 26-ft roadways as on 38-ft roadways. paving shoulders, or improving roadsides) on sections of rural
two-lane roads that are low volume (i.e., have ADTs of less
than 2,000). Using this procedure, agencies can estimate their
Guidelines for 3R Design Problems expected project costs for one or more design alternatives on a
given roadway section along with expected accident benefits to
The research findings and general approach to design standards determine the most cost-effective project alternative. A project
outlined here provide the following guidelines for use in 3R alternative may consist of upgrading only one design feature (e.g.,
projects. widening 9-ft lanes to I 1-ft lanes) or upgrading combinations of
features (e.g., widening lanes from 9-ft to 11-ft, adding a 4-ft
1. Designers should note the incremental safety-effectiveness paved shoulder, plus increasing the roadside lane zone from I I
 47

ft to 15 ft). This guide is not intended to replace the use of the profile. Key project data, summarized in the three major sections
minimum width guidelines, but can provide useful input when of work, include:
deciding between two or more "acceptable" design options. Spe-
cifically, using the guide can assist a highway agency in getting
the most accident savings per dollar spent on design improve-
ments.
Improvement
Section I Section 2 Section 3
EVALUATION OF CASE STUDIES Preconstruction (0.6 mi) (2.1 mi) (1.7 mi)
Functional Classifi-
Case studies for recent construction and reconstruction proj- cation Collector* Collector Collector Collector
ects were solicited from the state agencies that responded to the Design Speed 35 mph 60 mph 60 mph
survey of construction costs. NCHRP project panel members Terrain Rolling Rolling Rolling Rolling
also provided case studies from their agency files. Design Year Traffic:
The following four case studies were evaluated to compare ADT 770-2,400 vpd 2,880 vpd 1,620 vpd 930 vpd
design decisions reached in the actual project with the proposed Percent Trucks - — 3.6 6.4 11.1
design values derived from this research. Typical Cross Section:
Combined Lane and
Shoulder Width 14 ft 14 ft 14 ft 14 ft
(one direction)
Case Study No. 1: US 62 and 281 —Widening and
Reconstruction; Caddo County, Oklahoma *Federal-Aid Secondary Road

The Oklahoma Department of Transportation recently recon-

structed approximately 2.7 miles of US 62 and 281 in Caddo
Note that a 14-ft paved roadway (28 ft both directions) was the
County. The project included a combination of widening and
existing as well as proposed, reconstructed cross section. The
some realignment both horizontally and vertically. Key project
reconstruction cost of $1.9 million translates to approximately
data are summarized as follows:
$430,000 per mile.
Data provided by the Idaho Department of Transportation
Designed
showed that the existing accident rate for State Highway 46 was
Preconstruction Improvement
2.5 accidents per MVM. It was noted that this rate was only
Functional Classification Arterial slightly higher than the state average. Thus, resolution of an
Design Speed 55 mph existing safety problem did not appear to be a primary objective
Terrain Rolling Rolling of the improvement project.
Design Year Traffic: From inspection of Tables 29(c) and 29(d), the design values
ADT 1,300 vpd 2,000 vpd proposed for the reconstruction are less than would be called for
(1990) by the revised standards. For section 1 of the improvement, a
Percent Trucks — 22 total width of 32 ft (10-ft lanes and 6-ft shoulders) is specified.
For section 2, 1 1-ft lanes and 6-ft shoulders (34 ft total roadway
Typical Cross Section: width) would meet the revised standard. Finally, for section 3,
Lane Width 11 ft 12 ft
1 I-ft lanes and 4-ft shoulders (a 30-ft total roadway width) appear
Shoulder Width — 8 ft
appropriate.
There are a number of interesting facets to Case Study No.
The cost of reconstruction was estimated at $2.7 million, or $1.0 2. First, the great variation in traffic volume and design speed
million per mile. No accident or other operational data were over the 4.4-mile project creates an unusual design problem. The
provided. theoretical design standards would call for three different designs,
Referring to Table 29(f), under the proposed design values, varying both lane and shoulder width. Such variation over a
the lane and shoulder widths to meet full standards would be short distance may create the appearance of a discontinuous
identical to those actually constructed. A 60-mph design speed facility and be unnecessarily costly to design and construct.
is assumed to apply from Table 29(f). In this case, therefore, the While no indication was given regarding how the lanes would
revised standards would not reflect a change in design. be striped, the proposed design offers the opportunity to use 10-
ft lanes and 4-ft shoulders (fully paved), or I I -ft lanes and 3-ft
shoulders. As shown below, the theoretical accident rate for these
designs does not differ significantly from the accident rate for
Case Study No. 2: State Highway 46—
Reconstruction; Gooding and Camas Counties, the full standard. In addition, it is clear that the incremental costs
Idaho of widening from the existing 28-ft roadway to a 30- to 34-ft
roadway would be significant. In this case, therefore, where an
State Highway 46 was recently reconstructed through Camas existing total roadway width of 28 ft occurs, and where there is
and Gooding Counties in Idaho. The 4.4-mile project involved no existing serious safety problem, maintenance of the existing
reconstruction along existing alignments with adjustments to the width appears to be a reasonable design approach.
48

Fufl As Designed Case Study No. 4: Grande Ronde River Road 51—
Standard (28-ft Roadway) Reconstruction; La Grande, Oregon

Section 1: The U.S. Forest Service reconstructed 6.1 miles of Grande

Lane/Shoulder 10/6 10/4 or 11/3 Ronde River Road 51 in La Grande, Oregon, in 1988. The
Width (ft) road, open to public travel, serves primarily logging traffic and
Predicted Related 1.43 2.41 or 1.31 recreational traffic. The $1.7-million project (approximately
Accident Rate $280,000 per mile) included minor horizontal alignment 'r-evi-
sions. The major work involved revision to the cross section, as
Section 2: noted below-

Lane/Shoulder 11/6 11/3

Preconstr~ction Plan
Width (ft)
Predicted Related 1.31 1.31 Functional Classification Minor arterial Minor arterial
Accident Rate Design Speed 30 mph 40 mph
Terrain Rolling Rolling
Section 3: Design Year Traffic:
ADT NA NA
Lane/Shoulder 11/4 11/3 —
Percent Trucks 39
Width (ft)
Typical Cross Section:
Predicted Related 1.31 1.31
Lane Width 10 ft 12 ft
Accident Rate
Shoulder Width 2 ft 1 ft

Case Study No. 3: Oak Harbor Southeast Road— A spot accident problem was noted on one curve that was related
Resurfacing, Ottawa County, Ohio to logging trucks encroaching on the centerline.
While no traffic data were available, it is assumed that the
The Ohio Department of Transportation recently resurfaced road carries less than 400 vpd. Based on the standards developed
3.1 miles of Oak Harbor Southeast Road in Ottowa County, by this study, a design incorporating 11 -ft lanes and 2-ft shoulders
Ohio. The $600,000 project included about $280,000 for structure (see Table 29f) could have been constructed, versus the 12-ft
repair. The project was essentially a 3R project, with no signifi- lane and I-ft shoulder that was constructed. The total width is
cant adjustments to horizontal or vertical geometry. Key project the same in both cases. However, the narrower lane width and
data are summarized as follows: wider shoulder would offer the following advantages over the
12-ft lane width design: construction cost would be lower, the
Functional 2-ft shoulder would meet minimal functional requirements, and
Classification Collector the 11-ft lane width would be more compatible with the low
design speed of 40 mph.
Design Speed 55 mph
The expected safety performance of the various alternatives
Terrain Rolling
is noted below:
Design Year Traffic:
ADT 1,350 vpd
Lane/Shoulder Related Accident Rate
Percent Trucks 10
Width (ft) (per MVM) per Table 10
Typical Cross Section:
Lane Width 12 ft Preconstruction 10/2 2.41
Shoulder Width 4 ft Plan As Constructed 12/1 1.87
Proposed Standards 11/2 1.87
The contract plans indicate that the existing and resurfaced cross
sections are the same (12-ft lanes and 4-ft shoulders). Also, the Note that the previously outlined advantages come with no worse
plans note that a design exception for shoulder width was ob- expected accident experience than the design that was' con-
tained for the project. structed.
Based on the proposed design standards from the research
here, not only would a design exception not be required, but
the proposed cross section would actually exceed the minimum Summary of Case Studies
standard of I 1-ft lanes and 4-ft shoulders.
In the case of existing roads with 12-ft lanes, maintenance of The four case studies are not necessarily representative of
this full width (as opposed to a narrower width per the proposed national experience in low-volume road construction. Nonethe-
standards) is a rational 3R decision. It should be made, however, less, they do illustrate the expected impacts of revised width
with full understanding of the operational and safety performance standards as proposed by this study.
of total widths, and of various lane and shoulder width combina- First, as in Case Study No. 1, there are many miles of roadway
tions. As noted in Table 10, either 11 -ft or 12-ft widths, when for which no changes in standards are proposed. Current
paired with at least 3-ft shoulders (total roadway width of 28 ft AASHTO Policy is confirmed as reasonable in some cases. Sec-
or 30 ft), would optimize safety. ond, there are design benefits of the proposed standards that do
 49

not necessarily translate to construction cost savings, but are had the proposed standards been in place, the design agency
noteworthy nonetheless. The lower overall standards produce would not have had to evaluate, request, and process a "design
greater design flexibility. In Case Study No. 2, application of a exception."
minimum standard everywhere for the three sections would be Third, and perhaps most important, there are clearly examples
,possible. Alternatively, using one design that is consistent of low volume roads that could be reconstructed at lower cost
.throughout (as was actually done) is also possible. This is an without any degradation in expected safety performance. As
important facet of standards that is often overlooked. The pres- shown by Case Study No. 4, construction and reconstruction of
ence.of a minimum standard should not preclude the use of many miles of low-volume road using standards implied by the
greater dimensions, particularly where there is good reason to current AASHTO Policy may not represent a cost-effective use
do so. Another example of design flexibility is offered by Case of highway funds.
Study No. 3. While the design itself would not have changed,
 50

CHAPTER 4
CONCLUSIONS

This study demonstrates that there are opportunities to refine rates are significantly greater for unpaved roads than for paved
current lane and shoulder width standards for low-volume roads. roads, all else being equal.
Revised design values based on the safety studies, operational
analyses, and cost-effectiveness evaluations would produce
meaningful savings in construction and reconstruction costs asso- ACCIDENT EFFECTS OF LARGE TRUCKS
ciated with the low-volume rural system.
The major research conclusions stem from a series of studies
and analyses of all facets of low-volume roads. The study did not uncover a relationship between the composi-
tion of traffic volume (i.e., percent trucks in the traffic stream)
and accident rate for low-volume roads.
ACCIDENTS ON LOW-VOLUME ROADS

Low-volume roads were found to experience a slightly higher ACCIDENT BENEFITS OF WIDENING LOW-
VOLUME ROADS
percentage of injury accidents than the full sample of rural roads.
Single-vehicle accidents (fixed object, rollover, run-off-road) are
greater proportionately and multivehicle accidents (rear-end, The study produced a linear model that was used to estimate
angle, turning) lesser proportionately than all two-lane rural expected accident effects of various combinations of lane width
roads. and shoulder width. For 10-ft lanes, accident rates were lower
The studies found that low-volume road accidents are affected by 0.98 per MVM when accompanied by shoulders of greater
primarily by roadway width, roadside hazard, terrain, and drive- than 4 ft versus shoulders of 4 ft or less. For 11 -ft and 12-ft
ways per mile. Accident rates are significantly associated with lanes, shoulder widths of 3 ft or greater produced accident rates
varying lane and shoulder widths for single-vehicle and opposite 0.56 per MVM lower than the accident rates produced with
direction accidents. shoulder widths less than 3 ft. The accident model produced
identical accident benefits for I I -ft and 12-ft lanes.
Economic benefits of accident reductions were estimated using
ACCIDENT EFFECTS OF LANE AND SHOULDER research by others that establish societal costs of motor vehicle
WIDTH injuries and fatalities. The research approach used FHWA's rec-
ommended methodology for such costs. A recommended value,
The study determined that the presence of a shoulder is associ- considered conservative, of $60,000 per low-volume road acci-
ated with significant accident reductions for lane widths of'at dent was used to calculate benefits of accident reductions.
least 10 ft. For lane widths of 11 ft and 12 ft, shoulder widths Design standards for highways are based on more than safety
of at least 3 ft have significant effects. With respect to all combi- performance. The research also investigated traffic operational
nations of lane width and shoulder width (for lane widths of 10 requirements related to width elements, and width-related effects
ft or more), there is no apparent accident reduction above a total of construction costs. A central theme of the study was that
roadway width (i.e., lanes plus shoulders) of 30 ft. width-related design standards should be compatible with pri-
The study also addressed very narrow roads (i.e., with lane mary design policy controls. These include design speed and
widths of 8 ft or 9 ft). For a combination of reasons, some functional classification.
relating to sample size and others less clear, there is no apparent
accident benefit of widening such lanes to 10 ft. Indeed, the
study produced evidence that 9-ft lanes with wide shoulders may Capacity and Level-of-Service Effects on Width of
be preferable to 10-ft lanes with narrow shoulders. Low-Volume Roads

The study demonstrated that, based on known operational

ACCIDENT EFFECTS OF PAVED VS. UNPAVED effects of lane and shoulder widths, for certain terrain conditions
SURFACES and traffic composition, capacity requirements should influence
design standards. For a design ADT of 1,500 vpd or more,
The study determined that accident experience does not appear combinations of lane and shoulder width in excess of those
significantly different for unpaved versus paved surfaces at traffic required merely for safety are necessary to produce levels of
volume levels of 250 vpd or less. Above this volume, accident service compatible with AASHTO design policy guidelines.
 51

Speed-Related Widths investigate the sensitivities of variable lane and shoulder width
designs, as well as other relevant factors such as terrain. Cost
The study referenced other research on the operating speed per mile of construction was found to be greatly influenced by
effects of various lane widths. It was concluded that lane widths terrain, total roadway width, and lane width.
should produce operating speeds compatible with the selected
design speed. Wider lane widths (say, 11 -ft or 12-ft) on roadways
designed with lower design speeds (say, 40-mph or 50-mph)
Cost-Effective Width Combinations
may be undesirable. Such widths may promote operating speeds
above those for which the alignment was intended to accom-
modate. The accident model and construction cost model led to an
evaluation of the cost-effectiveness of various combinations of
lane and shoulder width. "Breakeven" ADT levels were com-
Shoulder Functions Related to Width puted to determine when widening appeared justified.
A major conclusion of this exercise was that width combina-
The study referenced other research on the function of shoul- tions less than those called for by current AASHTO Policy appear
ders and widths required to accommodate those functions. It was appropriate. In particular, where total roadway widths (i.e., lanes
concluded that for certain combinations of functional classifica- plus shoulders) currently exceed 30 ft, there are many cases
tion, traffic volume and composition, shoulder widths in excess where lesser widths could be employed at a lower cost with
of those required solely for safety are necessary. For the upper no apparent degradation in safety. Also, the cost-effectiveness
range of traffic volumes (greater than 1,500 vpd), this study analysis demonstrated that less variation in widths for the range
recommends minimum shoulder widths of 6 ft to 8 ft, depending in traffic volumes is appropriate compared to current AASHTO
on the functional class of the facility. Policy.

Truck Operations Related to Width

Recommended Standards
The study also referenced other research on the operational
effects of longer, wider vehicles. It was concluded that, where The primary objective of the research was development of
such vehicles can be expected to operate, providing incrementally recommended design values for lane and shoulder width for low-
greater lane and shoulder widths is desirable to accommodate the volume roads. These values were developed from all of the
off-tracking and encroachment problems associated with these foregoing analyses, and were given in Tables 29(a) through 29(f).
vehicles. This research recommended revisions to the format of design
standards as well as values for the range of relevant condition.
Costs of Construction of Low-Volume Roads Overall, the research supports conclusions of earlier work from
TRB Special Report 214. Lesser width combinations for lanes
A cost model based on a comprehensive survey of state design and shoulders on low-volume roads should be considered in
engineers and practices was produced. The model was used to subsequent editions of the AASHTO Policy.
52

REFERENCES

SMITH, S.A., PURDY, J., McGEE, H.W., HARWOOD, D.W., W., DOUGLASS, J., DILLINGHAM, A., and BLOMQUIST, G.,
ST. JOHN, A.D., and GLENNON, J.C., "Identification, Quanti- "The Cost of Highway Crashes." Report No. PB92-163-625,
fication, and Structuring of Two-Lane Rural Highway Safety Federal Highway Administration, Washington, D.C. (Oct.
Problems and Solutions, Volumes I and If." Report Nos. 1991).
FHWAIRD-831021 and 831022, Federal Highway Adminis- DART, O.K. and MANN, L., "Relationship of Rural Highway
tration, Washington, D.C. (June 1983, Unpublished). Geometry to Accident Rates in Louisiana." Highway Re-
AMERICAN ASSOCIATION OF STATE HIGHWAY AND TRANS- search Record 312, Highway Research Board, Washington,
PORTATION OFFICIALS, A Policy on the Geometric Design D.C. (1970).
of Highways and Streets, AASHTO, Washington, D.C. FOODY, T.J. and LONG, M.D., "The Specification of Rela-
(1984 and 1990). tionships Between Safety and Roadway Obstructions." Re-
"Designing Safer Roads: Practices for Resurfacing, Restora- port No. OHIO-DOT-06-74, Ohio Department of Transpor-
tion, and Rehabilitation." TRB Special Report 214. Transpor- tation (1974).
tation Research Board, Washington, D.C. (1987). BARBARESSO, J.C. and BAIR, B.O., "Accident Implications of
GLENNON, J.C. "Design and Traffic Control Guidelines for Shoulder Width on Two-Lane Roadways." Transportation
Low-Volume Rural Roads." NCHRP Report 214, Transpor- Research Record 932. Transportation Research Board,
tation Research Board, Washington, D.C. (Oct. 1979). Washington, D.C. (1983).
ZEGEER, C.V., HUMMER, J., REINFURT, D., HERF, L., and HEIMBACH, C.L., HUNTER, W.W., and CHAO, C.C., "Paved
HUNTER, W., "Safety Effects of Cross-Section Design for Highway Shoulder and Accident Experience." Transporta-
Two-Lane Roads." Report No. FHWAIRD-871008. Federal
tion Engineering J. Proc., American Society of Civil Engi-
Highway Administration and Transportation Research
neers, New York, New York, 100 (TE4): 889-908 (Nov.
Board, Washington, D.C. (Oct. 1987).
1974).
GRIFFIN, L.I. and MAK, K.K., "Benefits to be Achieved
SHANNON, P. and STANLEY, A., "Pavement Width Standards
from Widening Rural, Two-Lane, Farm-to-Market Roads in
for Rural Two-Lane Highways." Idaho Department of Trans-
Texas." Paper Presented at the 1988 Annual Transportation
portation, Boise, Idaho (1976).
Research Board Meeting.
TURNER, D.S., FAMBRO, D.B., and ROGNESS, R.O., "Effects
ZEGEER, C.V., HUMMER, J., HERF, L., REINFURT, D., and
of Paved Shoulders on Accident Rates for Rural Texas High-
HUNTER, W., "Safety Cost-Effectiveness of Incremental
ways. Transportation Research Record 819, Transportation
Changes in Cross-Section Design -Informational Guide. "
Research Board, Washington, D.C. (1981).
Federal Highway Administration, Washington, D.C. (Dec.
1987). ROGNESS, R.O., FAMBRO, D.B., and TURNER, D.S., "Before-
JORGENSON, Roy & ASSOCIATES, "Cost and Safety Effective- After Accident Analysis for Two Shoulder Upgrading Alter-
ness of Highway Design Elements." NCHRP Report 197, natives." Transportation Research Record 855, Transporta-
Transportation Research Board, Washington, D.C. (1978). tion Research Board, Washington, D.C. (1982).
ZEGEER, C.V., MAYES, J.C., and DEEN, R., "Cost-Effective- RINDF, E.A., "Accident Rates Versus Shoulder Widths."
ness of Lane and Shoulder Widening of Rural, Two-Lane California Department of Transportation (Sept. 1977).
Roads in Kentucky." Kentucky Department of Transporta- "Relationship Between Safety and Key Highway Features-
tion (July 1979). A Synthesis of Prior Research." State of the Art Report
HARKEY, D., ZEGEER, C., REINFURT, D., DAVIS, S., STEW- 6 Transportation Research Board, Washington, D.C.
ART, R., and COUNCIL, F., "Operational Impacts of Wider (1987).
Trucks." Report No. FHWAIRD-901103, Federal Highway ZEGEER, C.V. and DEACON, J., "Effect of Lane Width, Shoul-
Administration, Washington, D.C. (June 1991). der Width, and Shoulder Type on Highway Safety: A Synthe-
GLENNON, J.C., NEUMAN, T.R., and LEISCH, J.E., "Safety sis of Prior Literature." Transportation Research Board
and Operational Considerations for Design of Rural High- (Mar. 1986).
way Curves." Report No. FHWAIRD-861035, Federal High- STOHNER, W.R., "Relation of Highway Accidents to Shoul-
way Administration, Washington, D.C. (1985). der Width on Two-Lane Rural ffighways in New York
LAMM, R. and CHOUEIRI, E.M., "Recommendations for Eval- State." Highway Research Board Proc., Vol. 35 (1956).
uating Horizontal Design Consistency Based on Investiga- GUPTA, R.C. and Jain, R.P., "Effect of Certain Geometric
tions in the State of New York." (July 1986, Unpublished). Design Characteristics of Highways on Accident Rates for
DOWNS JR., H.G. and WALLACE, D.W., "Shoulder Geomet- Two-Lane, Two-Way Roads in Connecticut." University of
rics and Use Guidelines." NCHRP Report 254, Transporta- Connecticut (Aug. 1973).
tion Research Board, Washington, D.C. (Dec. 1982). SPARKS, JW., "The Influence of Highway Characteristics
AMERICAN ASSOCIATION OF STATE HIGHWAY AND TRANS- on Accident Rates." Public Works, Vol. 99 (Mar. 1968).
PORTATION OFFICIALS, "A Manual on User Benefit Analysis RAFF, M.S., "Interstate Highway Accident Study." Highway
of Highway and Bus-Transit Improvements." AASHTO, Research Board Bulletin 74, Highway Research Board,
Washington, D.C. (1977). Washington, D.C. (1953).
MILLER, T., VINER, J., ROSSMAN, S., PINDUS, N., GELLERT, BELMONT, D.M., "Effect of Shoulder Width on Accidents
 53

on Two-Lane Tangents." Highway Research Board Bulletin ZEGEER, C.V., STEWART, J.R., COUNCIL, F.M., REINFURT,
91, Highway Research Board, Washington, D.C.(1954). D.W., MITLER, T., HUNTER, W.W." and HAMILTON, E.,
31 BLENSLY, R.C. and HEAD, J.A., "Statistical Determination "Cost-Effective Geometric Improvements for Safety Up-
of the Effect of Paved Shoulder Width on Traffic Accident grading of Horizontal Curves." Report No. FHWA-RD490-
frequency." Highway Research Board Bulletin 240, High- 021, Federal Highway Administration, Washington, D.C.
way Research Board, Washington, D.C. (1960). (May 1990).
HAGENAUER, G.F., UPCHURCH, J., WARREN,D., and ROGEN-
BELMONT, D.M.; "Accidents Versus Width of Paved Shoul- '
BAUM, J.J., "Synthesis of Safety Research Related to Traffic
ders; on California Two-Lane Tangents— 1951 and 1952."
Control and Roadway Elements." Volume 1, Report No. T5-
Highway Research Board Bulletin 117, Highway Research
82-232, Federal Highway Administration, Washington, D.C.
Board, Washington, D.C. (1956). (Dec. 1982).
PERKINS, E.T., "Relationship of Accident Rate to Highway CLEVELAND, W.S., DEVLIN, S.J., and GROSSE, t., "Regres-
Shoulder Width." Highway Research Board Bulletin 151, sion by Local Fitting." J. Econometrics, 37: 87-114 (1988).
Highway Research Board, Washington, D.C. (1956). TOUTENBURG, H.,Prior Information in Linear kodels. John
HEAD, J.A. and Kaestner, N.F. "The Relationship Between Wiley, New York, New York (1982). ~
Accident Data and the Width of Gravel Shoulders in Ore- AMERICAN ASSOCIATION OF STATE HIGHWAY AND TRANS-
gon." Highway Research Board Proc., Vol. 35, Highway PORTATION OFFICIALS, Roadside Design Guide, AASHTO,
Research Board, Washington, D.C. (1956). Washington, D.C. (1989).
APPENDIX A LA
4~

COST-EFFECTIVENESS GUIDE FOR DESIGN

IMPROVEMENTS TO ROADWAY CROSS SECTIONS ON
LOW-VOLUME RURAL ROADS
The purpose of this guide is to assist state and local agencies in determining the accident This is similar to the concept of a clear zone, except that the roadside recovery distance
benefits and costs related to various design improvements to roadway cross sections (i.e., includes only recoverable slopes. According to the AASHTO "Roadside Design Guide," the
widening lanes, widening shoulders, or improving roadsides). The procedure applies only to clear zone includes nonrecoverable slopes.
sections of rural, two-lane roads with low traffic volumes, that is, with average daily traffic
(ADTs) of 2,000 or less.
Procedure Assumptions
This cost-effectiveness procedure is based on the results of the accident analyses conducted
as part of the NCHRP Project 15-12 study, "Roadway Widths for Low Traffic Volume Roads." The procedures used in this guide are based on information compiled and analyzed only
It also makes use of an FHWA 1987 study by Zegeer, et al., "Safety Cost-Effectiveness of for highway sections (i.e., preferably 0.5 mile or longer) under the following conditions:
Incremental Changes in Cross-Section Design—Informational Guide" (7 ). The procedures in two-lane rural roads with an ADT between 50 and 2,000; lane widths of 8 to 12 ft; and
that Guide pertained to rural, two-lane roads of all ADT levels. That procedure has been shoulders (0 to 12 ft wide) that are paved, unpaved, or partly paved and partly unpaved.
modified, herein, based on the accident experience on low-volume roads only (ADT of 2,000
or less). The procedures in this guide assume that no net change in maintenance cost will occur
because of the project alternatives. If an analyst expects maintenance costs to change because
of an alternative, he or she should adjust the project annual cost (line 34 of Form A, provided
GEOMETRIC DESIGN TERMINOLOGY at the end of this appendix).

The following terminology is used in the procedures for defining geometric and safety The term, related accidents, on the basis of a research study for which this supplement was
relationships. developed, includes the accident types that were found to be related to roadway cross section
(i.e., lane and shoulder widening, and roadside improvements):

Cross-Section Elements 0 Single-vehicle accidents (fixed-object, rollover, and other run-off-road crashes), and

Figure A- I describes the elements of rural highway cross sections. Such elements include 0
. Opposite-direction accidents (head-on and opposite-direction sideswipe crashes).
lane and shoulder width, roadside, and sideslope, while the ADT and terrain are also important
in safety analyses. The only nonstandard terminology in this procedure regards roadway width,
roadside recovery distance, and related accidents. The first two terms are defined as follows; ANALYSIS PROCEDURE FOR DETERMINING PROJECT COST-EFFECTIVENESS
and the third term, related accidents, is identified under "Procedure Assumptions."
The Form A worksheet helps an analyst calculate the accident benefits from cross-section
Roadway width includes the combined width of the lanes and shoulders on a highway improvements (lane widening, shoulder widening, and increasing roadside recovery distance).
section. Total roadway width is one of the most important features in the safety performance Analysts begin the worksheet by entering the "header" identification information at the top of
of a two-lane highway. Roadside recovery distance is defined as a relatively flat, unobstructed the form. The form provides room for analyzing as many as three predefined project
and smooth area adjacent to the edge of travel lane (i.e., edgeline) where there is reasonable alternatives. Analysts interested in considering more than three alternatives should use more
opportunity for safe recovery of an out-of-control vehicle; the roadside recovery distance is the than one form. The form is completed using the following steps.
lateral distance from the outside edge of the travel lane to the nearest of the following:
Step One of the form provides spaces to enter basic data on the current conditions of the
A hinge point where the slope first becomes steeper than 4:1, roadway section and on the alternatives. Analysts need to provide data for lane width, shoulder
width, roadside recovery distance, sideslope, and current ADT that are averages for both sides
A longitudinal element such as a guardrail or bridge rail, of the road along the entire section. If a value of one of these variables changes dramatically
along a section, the section should be broken up and analyzed as several shorter sections. It is
An unyielding and hazardous object, recommended that the procedure be applied to sections of 0.5 mile or longer.

The ditch line of a nontraversable side ditch, and The number of total accidents per year on the section is given on line 14, and the number
of "related" (run-off-road plus opposite-direction) accidents per year is given on line 15. Note
Other features, such as a rough or irregular surface, loose rocks, or a watercourse, that that if the number of "related" accidents is not known, it can be approximated as 63 percent of
pose a threat to errant vehicles. the total number of accidents. For example, assume that 20 total accidents occurred on a 2.5-
mile section over a 3-year period. This would correspond to (20 accidents) - (2.5 miles) =
8.0 total accidents per mile per yeari or (8) x (0.63) = 5.0 related accidents per year From Table A-3, the improved condition should result in an average rate of 1.31. Thus,
the AR factor, AR (i.e., percent reduction) in related accidents between the before (B) and after
In line 16, the rate P, of related accidents [related accidents per million vehicle miles (A) conditions, would be:
MVM)] is computed as:

(No. of related accidents per year) (1,000,000) AR = B - = (2.75) - (1.31) - 1.44 = 0.52
B 2.75 2.75
R, =
(365)(ADT) (Section length in miles)

Lines 17, 18, and 19 involve inputting accident costs for property damage only (PDO) a 52 percent reduction in related accidents. A value of 0.52 would be recorded on line 25.
accidents and per injury or death (i.e., per person injured or killed, not per injury accident or
fatal accident). These inputs are necessary for computing accident benefits (savings) that are Note that if the current (untreated) lane width is 10 ft or greater, it would also be possible
expected because of the improvement. The users may input their values here, but default values to select rate values from Table A-3 for both the before and after conditions to compute the AR
of $3,000 per PDO accident, $15,000 per injury, and $1,500,000 per fatality are suggested, factor. For example, assume a level roadway section with a 10-ft lane and no shoulder to be
based on recent costs recommended by FHWA. Such costs may be updated each year. considered for widening to 12-ft lanes with no shoulder. From Table A-3, a before rate of 2.28
would be used with an after rate of . 1.74. This would result in an accident reduction (AR) of:
Step Two involves updating the accident rates based on future conditions. On the basis of
the expected project life of a given alternative (10 or 20 years) and the projected annual traffic
growth rate (i.e., 0, 1 percent, 2 percent, or 3 percent per year), the traffic growth factor is AR = B - A - (2.28) - (1.75) = 0.54 = 0.237
B 2.28 2.28
obtained from Table A-1. It is multiplied by the current ADT to obtain future ADT over the
project life (line 21).

For line 22, note that if no accidents have occurred on the roadway section within the last or a 24 percent reduction in related accidents. Such rate values for both the before and after
several (e.g., 3 to 5) years, an economic analysis is not appropriate because no accident benefits conditions can then be used to compute the accident reduction factor for roadway widening
would necessarily be expected as a result of a widening improvement. In this case, any when: (1) the lane width in the current (untreated) condition is between 10 and 12 ft, and (2)
proposed widening would need to be justified based on other considerations (capacity, the analyst decides that using this approach will be more reliable than using the actual accident
operational factors for accommodation of large trucks). rate at the site in the before period (e.g., because of a small sample of accidents or other
accident data problems).
In line 23, the number of future related accidents per year is computed. This is the
expected number of related accidents on the section per year if no treatments are made. This In line 26, the accident reduction due to roadside improvements can also be included in
value is used in determining the expected accident benefits from , one or more project the cost-effectiveness analysis using accident reduction factors from Tables A-3 and A-4.
alternatives. In line 24, the annualization factor is recorded from Table A-2 for later use in Assume a low-volume roadway section with a current roadside recovery distance (RRD) of only
computing annual project costs. 5 ft. A roadway widening improvement is being considered that would also increase the RRD
from 5 ft to 20 ft as a result of tree removal and slope flattening. The proposed increase in
Step Three involves determining the expected percent reduction in related accidents (i.e., RRD would be 20 ft - 5 ft = 15 ft. Using Table A-4, a 35 percent reduction in related
accident reduction factor, AR) resulting from roadway widening, roadside improvements, or accidents would be expected. Thus, a value of 0.35 would be entered on line 26.
other improvements. This is determined in line 25 by first identifying the rate of related
accidents in the current (untreated) condition and comparing it with the appropriate rate after If sideslope flattening is proposed (as the only roadside improvement) on a slope relatively
roadway widening, and computing the percent change (if any). These AR factors are determined clear of fixed objects, Table A-5 can be used to estimate reduction in related accidents. For
using Tables A-3, A-4, and A-5. example, assume a 2: 1 slope which may be flattened to 6: 1. This would result in a 21 percent
reduction in related accidents. The reader should note that the AR factor from either Table A-4
(on increasing RRD) or Table A-5 (flattening sideslope) can be used for a given improvement
To illustrate the use of these tables to compute AR factors, assume an existing section of alternative. Also, an AR factor for roadside improvements (from either Table A-4 or A-5) can
low-volume, two-lane road is in a rolling area with a 9-ft lane width and no shoulder. The be used in an analysis even if no roadway widening is being considered.
current rate is 2.75 related accidents per MVM. The proposed improvement will widen the
roadway to I I-ft lanes with 4-ft shoulders. In Step 27, two or more accident reduction factors (Al, A2, A3, ...) may be combined
into one overall accident reduction (AR) based on the following equation:

LA
LA
 AR = I - (1 - ARI)(I - A%)(I - ARO (I - AR4) SIdeslope CenterlIna t Ditch bottom t.A
Original (also Original Or,
ground foreslope) '~'houlder Line Lane :houlder Foreslope B~ck3lOpa ground
For example, assume that a roadway widening and roadside improvements on a roadway section
Shoulder break
for Alternative A would give accident reductions of 52 percent and 21 percent, respectively. Hinge
point
The combined accident reduction (AR) is-not the sum of those values, but should be computed Depth
of cut
as: fII_ ht lope S., Shoulder
le:lll s Lane width IPaved
__J ~_ Ditch bottom
shoulder
AR = I - (I - 0.52) (1 - 0.21) s vAdth
width (variable)

AR = I - (0.48)(0.79)
AR = I - (0.38) = 0.62
Clear zone width
(to ob tacle or
s
nontra~ersable slope) Roaday width
a 62 percent reduction in related accidents because of the combined improvements for
Alternative A. The combined AR is recorded on line 27 and is used in Step 4 to compute
accident benefits resulting from one or more roadway improvements. AR factors for up to three Figure A-1. Elements of rural highway cross sections.
different alternatives may be recorded on each Form A. More than two AR factors could be
included only if additional roadway improvements are..being proposed (e.g., alignment '
changes)
for which the analyst can estimate the effect on related accidents.

Step 4 involves computing annuaI accident benefits,. The annual number of related
accidents reduced is computed in line 29. Lines 29, 30, and 31 are used to determine the
average cost per accident reduced, based on the severity distribution of crashes and the number
of people injured or killed in such crashes, as found in the analysis of low-volume roads
database. Also, other selected values of average costs for PDO only accidents, injuries, and
fatalities may be inserted by the analyst in lines 17, 18, or 19, respectively. Default values of
$3,000 per PDO accident, $15,000 per injury (per person injured, not per injury accident), and
$1.5 million per fatality (i.e., perperson killed) were used, as recommended in a recent FRWA
study (15).

The annual safety benefit is determined in line 33, and the annualized, project cost is
computed in line 34. The analyst may then compute the benefit-cost ratio (line 35) and annual
net benefits Oine 36) of each alternative and select the most desirable project alternative. An
example problem is worked out on the following Form A to illustrate the procedure.
 57
FORM A

Cross-Section Alternative Worksheet

Analyst: Unit Date:

Study Location: Co. Route Begin MP End MP
Alternative A:
Alternative B:
Alternative C:

Alternatives
Current
A B C

Step One - Inputs xxxx xxxx xxxx xxxx

Average lane width (LW) in feet

Average paved shoulder width (PSW) in ft

Average unpaved shoulder width (USW), ft

Total shoulder width (PSW + USW) ft
Average roadside recovery distance
(RRD), ft
Average sideslope ratio (horiz.: vert.)

Terrain (flat, rolling, or mountainous) xxxx xxxx xxxx

Section length, miles xxxx xxxx xxxx

Current average daily traffic (ADT) xxxx xxxx xxxx

Current average annual traffic growth xxxx xxxx xxxx

rate, percent
Expected project life, years xxxx
Project implementation cost xxxx

Discount (interest) rate, percent xxxx

Avg. number of total accidents per year xxxx xxxx xxxx
over the entire section. This line is
optional--can proceed without it.

Avg. number of related accidents per xxxx xxxx xxxx

year (i.e., single-vehicle plus opposite
direction accidents). If unknown, enter
(.63) x (line 14).
Rate of related accidents (current) per xxxx xxxx xxxx
million vehicle miles (R.) (see text) . I I
58
FORM A (Continued)

Alternatives
Current
A B C

Cost per PDO accident. Default - $3,000 xxxx xxxx xxxx

Cost per injury. Default = $15,000 xxxx xxxx xxxx
Cost per fatality. Default = $1,500,000 xxxx xxxx xxxx
Step Two - Update Inputs xxxx xxxx xxxx xxxx
Enter growth factor from Table 1 using xxxx
the average annual traffic growth rate
(line 10) and project useful life (line
11)
Multiply line 20 by line 9. This is the xxxx
updated ADT.

If line 16 (related accidents per year) xxxx

is 0, an economic analysis is not
appropriate, and no further economic
analysis is necessary. If line 15 is a
positive value, enter it here.

Multiply lines 20 by line 22. This is xxxx

the future number of related accidents
per year in the untreated condition.
Enter annualization factor from table 2. xxxx
Step Three - Determine Accident Reduction xxxx xxxx xxxx xxxx
Record the percent accident reduction xxxx
which would result from lane and/or
shoulder widening (see discussion in
text).
Record the percent reduction due to xxxx
roadside improvements from table 2 or 3.
Combine two or more accident reduction xxxx
factors for the same alternative, where
Total AR = I - (1-AR,) (1-AR2 ) ( 1-AR3) . . . .

Step Four - Compute Annual Benefits xxxx xxxx xxxx xxxx

Annual number of accidents reduced. xxxx
Multiply line 27 by line 23.
Mu ltiply line 17 by .603. xxxx xxxx xxxx
9. Multiply line 18 by .588.
30. xxxx xxxx XXXX
 59
FORM A (Continued)

Alternatives
Current

31. Multiply line 19 by .022. xxxx xxxx xxxx

Add lines 29, 30, and 31. xxxx xxxx xxxx

- 32.
Multiply line 28 by line 32. This is xxxx
the annual safety benefit.
Multiply line 12 by line 24. This is xxxx
the annual cost. Compare lines 33 and
34 to assist choosing an alternative.
Compute benefit/cost ratio of each xxxx
alternative. Divide line 33 by line 34.
Compute net benefits of each alterna- xxxx
tive. Subtract line 34 from line 33.
Choose the alternative which is most xxxx
desirable in terms of benefit/cost
ratio and/or net benefits.
60
FORM A

Cross-Section Alternative Worksheet

Analyst: C V Unit DO 7- Date: 11 L73

Study Location: Co. /V14in Route S R /0 Begin MP 0 - 0 End MP 2. 7

. -
Alter native A: - IAJideyl la-e; -ErD m CY t o 11 Fl
Alternative B: VWV I'd r n r) e 3 q to a,d 5~ou)der5 (0 to 4 Fi J
v" r- a V e ol h C r f a js e ' drsf. -From 6 to IS
(oad.s;dt rec. Fl.
Alternatives
Current
A B C
Step One Inputs xxxx xxxx xxxx xxxx
Average lane width (LW) in feet )2-
Average paved shoulder width (PSW) in ft 0 0
Average unpaved shoulder width (USW) , ft 0 0
Total shoulder width (PSW + USW) ft 0 0
Average roadside recovery distance S S j.5
(RRD), ft
6*. Average sideslope ratio (horiz.: vert.) 4:1 4:1 4:1
7. Terrain (flat, rolling, or mountainous) K0)1;A C1 xxxx xxxx xxxx
S. Section length, miles 21-7 xx'xx xxxx xxxx
Current average daily traffic (ADT) 2 00 xxxx XXXX - xxxx
Current average annual traffic growth xxxx xxxx xxxx
rate, percent 2%

Expected project life, years xxxx 10 10

Project implementation cost xxxx Soo,
Discount (interest) rate, percent xxxx 5Vlo 5 '7*
Avg. number of total accidents per year xxxx xxxx xxxx
over the entire section. This line is 12—
optional--can proceed without it.
Avg. number of related accidents per xxxx xxxx xxxx
year (i.e., single-vehicle plus opposite 7
direction accidents). If unknown, enter
x (line 14).
Rate of related accidents (current) per xxxx xxxx xxxx
million vehicle miles (R.) (see text). I

(7)(),000,000) —
'Rc~+e
~: 5.ljl
(St.5)()Z00)(2.7)
 61

FORM A (Continued)

Alternatives
Current
A B C

17. Cost per PDO accident. Default - $3,000 3. 0 0 0 xxxx xxxx xxxx

Cost per injury. Default = $15,000 1'5,o oo xxxx xxxx xxxx

Cost per fatality. Default - $1,500,0100 1 A 560,o0o xxxx xxxx xxxx

Two - Update Inputs xxxx xxxx xxxx xxxx
- Step
Enter growth factor from Table 1 using xxxx
the average annual traffic growth rate 1.12 1.12-
(line 10) and project useful life (line

multiply line 20 by line 9. This is the xxxx 1344 1344

updated ADT.

22.* If line 16 (related accidents per year) xxxx

is 0, an economic analysis is not
appropriate, and no further economic 7 7
analysis is necessary. If line 15 is a
positive value, enter it here.

Multiply lines 20 by line 22. This is xxxx

the future number of related accidents 7,8 7,8
per year in the untreated condition.
Enter annualization factor from table 2. xxxx .13o - 13 P.
Step Three - Determine Accident Reduction xxxx xxxx xxxx xxxx
Record the percent accident reduction xxxx.
which would result from lane and/or
shoulder widening (see discussion in .78
text).
Record the percent reduction due to xxxx
roadside improvements from table 2 or 3.

Combine two or more accident reduction xxxx

factors for the same alternative, where .835
Total AR = I - (I-ARI ) (I-AR?) (I-AR3)....

Step Four - Compute Annual Benefits xxxx xxxx xxxx xxxx

Annual number of accidents reduced. xxxx b, -3 G. -5

Multiply line 27 by line 23.

Multiply line 17 by .603. /.?T 0 xxxx kxxx xxxx

Multiply line 18 by .588. 4 8820 1 xxxx xxxx xxxx

b- A - 1-87 ori)
,Dn 13 :
D A 8 co.b;,, ed
5. 9
7,F)
7 .2.5)
Or +,'O n r:5 AR 3- A
9 7 2 (.2-2)(~- 7 5)~
6 5.9
I
62
FORM A (Continued)

Alternatives
Current
A B C
Multiply line 19 by .022. 3 o0o xxxx xxxx xxxx
Add lines 29, 30, and 31. Low- dfd) — , $ 43,&00 xxxx xxxx xxxx
4f .4
Multiply line 28 by line 32. This is xxxx 2,3j)D80 2.84445
the annual safety benefit.
Multiply line 12 by line 24. This is xxxx 4 4
the annual cost. Compare lines 33 and
34 to assist choosing an alternative.
Compute benefit/cost ratio of each xxxx
alternative. Divide line 33 by line 34. 3. 3,4
Compute net benefits of each alterna- 4
xxxx I(D~Mc 198,10D
tive. Subtract line 34 from line 33.
Choose the alternative which is most xxxx
desirable in terms of benefit/cost
ratio and/or net benefits.
a I i6M5~
e f"he r
P f; 0 r% I'_5
~e s jrob)e -

Table A-1. ADT and accident growth factors.

Growth Factor
Useful Life
(years)
( Years)
Ave.
Annual
Traffic Growth
Rate (Percent)
3c 10 10

0 1.00 1.00
1 1.06 1.11
2 1.12 1.24
3 1 1.18 1 1.38
 63

Table A-2. Annualization factor.

Useful
U seful Life

Discount
Rate 10 20

2 0.111 0.061
4 0.123 0.074
5 0.130 0.080
6 0.136 0.087
8 0.149 0.102

Table A-3. Accident rates used for determining accident reductions

due to roadway widening on low-volume roads.

Accident Rates (Accidents per MVM)

Land Width Shoulder Width
(feet) Terrain
(feet)
Level Rollinq Mountainous

10 0-2 2.28 2.41 2.86

3-4 2.28 2.41 2.86
1 >4 1.30 1.43 1.88

11 0-2 1.74 1.87 2.32

3-4 1.18 1.31 1.76
>4 1.18 1.31 1.76

12 0-2 1.74 1.87 2.32

3-4 1.18 1.31 1.76
>4 1.18 1.31 1.76

Table A-4. Accident reduction factors due to increasing

roadside clear recovery distance.

Amount of Increased
Roadside Recovery Percent Reduction in
Distance (feet) Related Accident Types
5 13
8 21
10 25
12 29
15 35
20 44

Table A-5. Summary of expected percentage reduction in related

, accident types due to sideBlope flattening.

Sideslope Sideslope in After Condition

in Before
Condition 7:1 or
3:1 4:1 5:1 6:1 Flatter-

2:1 2 7 11 is 20
:1 - 6 10 14 1
4:1 - 4 9 14
5:1 4 10
6:1 - 1 6
APPENDIX B
Z
CRITICAL REVIEW OF LITERATURE

A critical review of the literature dealing with the safety and geometrics of low-volume respect to crashes. The model includes ADT, terrain, lane width, paved shoulder width,
roads as they pertain to roadway width was conducted. As found in the literature, the types of unpaved shoulder width, and roadside condition. The roadside was expressed in terms of a
design and traffic elements that affect accidents include ADT, lane width, shoulder width, seven-point roadside hazard scale or a measure of roadway clear recovery distance. The
shoulder type, and other geometric features that combine with roadway width to affect model also was expressed as a nomograph (see Figure B-1) and accident reduction factors then
accidents. Other important variables include roadside condition, horizontal and vertical were developed for lane widening and shoulder widening (paved and unpaved shoulders),
alignment, sight distance, number of intersections and driveways, access control, and pavement roadside improvements, and side-slope flattening. A summary of accident reduction factors for
edge drop-off. The influences of terrain, percent trucks and roadway design speed on safety various combinations of lane and shoulder improvements is given in Table B-2.
and operations for low-volume roads were also considered to be of interest prior to developing
roadway width guidelines. , The relationship between related (AO) accidents (i.e., the combined accident types—run-
off-road, sideswipe and head-on) and lane and shoulder width are illustrated in Figure B-2
The critical reviews focused on two general types of studies: (1) studies that provide based on the accident model. For example, assume an ADT of 1,000 vpd on a rolling terrain,
recommended roadway width guidelines for rural, two-lane roads; and (2) studies that develop and a roadside rating of 5. Widening lanes from 9 to I I ft would be expected to reduce
relationships between accidents and roadway width (and/or lane width, shoulder width, and related accidents from 0.77 to 0.42, a 45-percent decrease. The Zegeer, et al., study also
shoulder type). developed a project cost model to assist in estimating various types of project costs (Zegeer, et
al., October 1987). An informational guide also was written to present step-by-step
Approximately 70 studies related to safety effects of various roadway features were information for estimating costs and benefits due to various improvements on specific sections
critically reviewed. Detailed critical reviews were prepared for 14 studies that pertain most of rural, two-lane roads (Zegeer, et al., December 1987).
directly to roadway width effects and guidelines (see Table B-1). Of those 14 studies, those
involving lane width effects on accidents include Dart and Mann (1970) in a study of 246
highway sections involving 6,000 accidents in Louisiana from 1962-1966; Foody and Long TRB SPECIAL REPORT 214 (TRB, 1987)
(1974) in a study of 1,800 roadway sections in Ohio using 1969-1970 accident data; Roy
Jorgensen & Associates (1978) in an NCHRP study of 12,400 miles of roadway in Washington In 1987, the Transportation Research Board published Special Report 214 titled, "Designing
and Idaho (with preliminary analysis using Maryland data); and Zegeer, Mayes, and Deen Safer Roads: Practices for Resurfacing, Restoration, and Rehabilitation." The study was
(1979) in an analysis of 16,760 accidents (1976 data) on 15,944 I-mile sections in Kentucky. conducted under contract with FHWA in response to a provision in the Surface Transportation
Assistance Act of 1982 to study "the safety cost-effectiveness of highway geometric design
Studies critically reviewed by project team members on shoulder width and type include a standards and recommend minimum standards for resurfacing, restoration, and rehabilitation
1983 study by Barbaresso and Bair in Oakland County, Michigan; a 1974 study by Heimbach, (311) projects on existing federal-aid highways, except freeways." As a part of the overall
Hunter, and Chao in North Carolina; a 1976 study by Shannon and Stanley (Idaho and TRB study, critical reviews of prior research were sponsored on safety effects of numerous
Washington); and studies in 1981 and 1982 by Turner, Fambro, and Rogness in Texas (1981- highway features, including lane width, shoulder width, and shoulder type; alignment; sight
1982). Studies by Rinde (1977), Glennon (1979), and Griffin and Mak (1988) focused on distance; bridge width; pavement surface condition; and pavement and shoulder drop-offs.
overall roadway (i.e., lane plus shoulder) width. Three of the studies (TRB 1987, Glennon
1979, and Griffin and Mak 1988) resulted in proposed guidelines for roadway width, as A summary of the results of these reviews and the most-likely accident relationships for
discussed previously. The 1987 study by Zegeer, et al., using data from seven states and the these topics was published by TRB in the 1987 State-of-the-Art Report 6—"Relationships
October 1987 TRB 3R study (Special Report 214) both involve an analysis of lane width, Between Safety and Key Highway Features—A Synthesis of Prior Research." The critical
shoulder width, and shoulder type. The following is a brief discussion of some of these review on lane width, shoulder width, and shoulder type was conducted by Zegeer and Deacon
studies. (1986). Based on information from four studies (Foody and Long, 1974; Zegeer, Mayes and
Deen, 1979; Turner, et al., 1981; and TRB State-of the-Art Report 6, 1987), a model was
developed by the authors for run-off-road and opposite-direction accident rate as a function of
CROSS-SECTION STUDY lane and shoulder width, as shown in Figure B-3. Accident relationships with other roadway
features (i.e., roadsides, bridge width) also are contained in TRB's Special Report 214 (1987).
One of the major studies of importance is the FHWA/TRB report, "Safety Effects of Cross-
Section Design for Two-Lane Roads," by Zegeer, Hummer, Reinfurt, Herf, and Hunter
(1987). The study quantified the benefits and costs resulting from various cross-sectional NCHRP 214 (GLENNON, 1979)
improvements for two-lane rural roads. Accident types found to be related to cross-sectional
elements included single-vehicle, head-on, and sideswipe (same and opposite direction), and The objective of this 1979 study by Glennon was to evaluate existing highway safety
were termed "related accidents." An accident predictive model, developed for related guidelines, requirements, and criteria for roads with fewer than 400 vpd. This differed from
accidents, shows the interrelationships between various traffic and roadway variables with most other roadway width studies that include rural roads of a wider range of ADT.
Guidelines for roadway width (in addition to other features) were developed, based on a review OTHER STUDIES OF LANE WIDTH, SHOULDER WIDTH, AND TYPE
of the literature and an analysis of safety requirements.
There have been numerous other research studies conducted relative to lane width and
safety. Examples of such studies include Stohner (1956) of 9,299 accidents on 8,746 miles of
On the basis of an analysis for tracking and head-on clearances, shoulders were found to be
needed on low-volume rural roads with design speeds above 45 mph. In terms of conflict rates two-lane highways in New York using 1952 accident data; Gupta and Jain (1973) of crash data
for 1,470 rural, iwo-lane sections in Connecticut; Sparks (1968) of 1964 accident data from
alone, no separate justification was found for shoulders'to accommodate emergency and leisure
Oklahoma; Raff (1953) of 16,421 accidents in 1953 on 32,091 highway sections from 15
stops. Vehicular tracking and lateral clearances to opposing vehicles were reviewed in an
analysis to determine requirements for total roadway width (i.e., width of traveled way plus states; and a synthesis report (1.982) by the Texas Transportation Institute for FHWA.
shoulders). The minimum total roadway width requirements ranged from 18 ft on a 50-vpd
roadway designed for 20 mph to 30 ft on a 400-vpd roadway designed for 50 mph. Specific Studies on accident effects of shoulder width and type on safety include older studies by
Barbaresso and Bair (1983); Belmont (1954) in California; Blensley and Head (1960) in
width guidelines were developed for these low-volume roads, as discussed earlier in this
Oregon; Belmont (1956) in New York State; E. T. Perkins (1956); and Head and Kaestrier
report.
(1956) in Oregon.

While the results of the previous studies varied widely on the relationship between accidents
GRIFFIN AND MAK 11988)
and roadway width, several of the studies did develop accident relationships and expected
effects of roadway widening on accidents. For example, Figure B-4 illustrates the relationship
. Another study that addressed low-volume rural roads was a 1988 study by Griffin and Mak,
between lane width and accident rate developed in five separate studies, as summarized in the
which attempted to quantify the relationship between accident rate and roadway surface width
1982 Texas Transportation Institute FHWA synthesis report. Note that considerable
on two-lane rural roads in Texas with ADT of 1,500 vpd or less. Log-linear accident
differences were found in some of the studies for a variety of reasons (i.e., data for different
prediction models were developed for 36,215 miles of roadway within several ADT categories.
states had a variety of roadway alignment or roadside conditions; weather conditions were not
Multivehicle accident rates (accidents per mile per year) were not found to relate to surface controlled for; etc.).
width for any of the ADT groups tested. Single-vehicle accident rates were noted to increase
Accident reductions associated with roadway widening projects are given in Table B-4
as roadway width decreased for ADT groups between 401 and 1,500 vpd. A6cident reduction
based on studies by Rinde in California (1977) and Rogness, et al. (1982) in Texas for various
factors were developed for various widening projects within these ADT ranges. On the basis
ADT groups. These illustrate only a few of the many studies previously conducted on
of an economic analysis, widening was.not found to be cost-beneficial for ADT below 1,000
roadway width and its relation to safety. Note that these results include roads with traffic
vpd. volumes as high as 7,000 vpd. -
A comparison of accident reduction factors for lane widening from the 1988 Griffin and
Mak study, with the October 1987 Zegeer, et al., study, reveals very similar results, as
indicated in Table B-3. When averaging the accident reductions for the three ADT groups
used by Griffin and Mak, the accident reductions were 11 to 12 percent for 2 ft of lane
widening, 21 to 22 percent for 4 ft of widening, 30 to 31 percent for 6 ft of widening, and 38
percent for 8 ft of widening. These values are within I to 2 percent of the values found by
Zegeer, et al., using data from seven states (Zegeer, 1987).

The similarity of accident reductions may seem surprising, since the seven-state database
moM was able to control for roadside, terrain, arid other roadway factors not analyzed in the
Griffin and Mak study. However, it should be mentioned that while roadside hazard,
alignment, and other such factors are known to affect accidents, they have also been shown to
be related to ADT. In other words, roadsides are typically more hazardous on narrow roads
(compared to wide roads) and on low-volume roads (compared to high-volume roads). Thus,
because Griffin and Mak controlled for ADT by developing models within various ADT
groups, they may have also indirectly controlled for roadside and other roadway features
within each ADT group. Note also that the Griffin and Mak study found accident reduction
factors to be about twice as high for the highest ADT groups (ADT of 1,001 to 1,500 vpd)
than they were for the lowest ADT groups (ADT of 401 to 700 vpd).
 Table B-1
Summary of Selected Studies Critically Reviewed
Cross-Sectional Elements Analyzed Type of Analysis

Comparative Analysis

Roadway
Width
States ShId. Rdwy. ShId. Funct. Before/ No Predict Predict Guidelines
Author [Reference] Date Included Lane Width Width Width Type Analy. After Equation Equation Proposed

Dart and Mann [16) 1970 LA X X X

Heimbach, Hunter and Chao [191 1974 NC X X

Foody, Long [17] 1974 OH X X X X

Shannon, Stanley [201 1976 ID, WA X** X** X

Rinde [23] 1977 CA X X

Jorgensen& Assoc. [81 1978 WA** ID X X X X

Zegeer, Mayes, Deen [9) 1979 KY X X X

Turner, Fambro Rogness [211 1981 TX X X

Rogness, Fambro, Turner [221 1982 TX X

X
Barbaresso and Bair [181 1983 Oakland Co. X X
(MI)

Glennon [41 1979 No state data X X X

TRB 3R Study [31 1987 Based on data X X X X

(KY,OH)

Zegeer, et al. (Cross-Sect. Study) 1987 AL, MI, MT, X X X X

[5, 71 NC, WVA,
UT, WA

Griffin, Mak [61 1988 TX X X

X
*New York State was used for initial analysis, but excluded for development of accident relationships.
**In this study, pavement width was the variable used in the analysis, which included paved width (lanes plus shoulders).
 67

Table B-2
Accident Reduction Factors for Related Accident Types
for Various Combinations of Lane and Shoulder Widening

Existing Shoulder Percent Related Accidents Reduced

Condition
(Before Period) Shoulder Condition in After Period

Amount of Lane 2-ft Shoulder 4-ft Shoulder 6-ft Shoulder 8-ft Shoulder
Widening Shoulder Surface
(in ft) Width (11) Type F— P U P U P U P U

0 N/A 43 41 52 49 59 56 65 62
2 Paved 32 - 43 - 52 - 59 -
2 Unpaved 34 33 44 41 53 49 60 56
4 Paved - - 32 - 43 - 52 -
3 4 Unpaved - - 36 32 46 41 54 49
6 Paved - - - - 32 - 43 -
6 Unpaved - - - - 37 32 47 41
8 Paved - - - - - - 32 -
8 Unpaved - - - - - - 39 32

0 N/A 35 33 45 42 53 50 61 56
- - -
2 Paved 23 35 - 45 53
2 Unpaved 25 23 37 33 46 42 55 50
4 Paved - - 23 - 35 - 45 -
2 4 Unpaved - - 27 23 38 33 48 42
6 Paved - - - - 23 - 35 -
6 Unpaved - - - - 29 23 -40 33
8 Paved - - - - - - 23 -
8 Unpaved - - - - - - 31 23

0 N/A 26 24 37 34 47 43 55 50
2 Paved 12 - 26 - 37 - 47 -
2 Unpaved 14 12 28 24 39 34 48 43
4 Paved - - 12 - 26 - 37 -
4 Unpaved - - 17 12 20 24 41 34
6 Paved - - - - 12 - 26 -
6 Unpaved - - 19 12 31 24
8 Paved - - - - 12 -
8 Unpaved 21 12

Notes:

Blank cells correspond to projects that would decrease shoulder width and/or change paved shoulders to unpaved shoulders.

P = paved, U = unpaved.

ese values are for two-lane rural roads only.

Source: Zegeer, et al. ( October 1987).

68

Table B-3
Comparison of Accident Reductions From Lane Widening

Percent Accident Reduction'

Roadway Width
(rt) Griffin and Mak 1988 Study
Amount of
Lane ADT = ADT = ADT Avg. of Zegeer,
Widening 401-700 701-1,000 1001-1,500 3 ADT et al.
(ft) Before After (vpd) (vpd) (vpd) Groups

18 20 7 12 14 11 12
20 22 7 12 15 11 12
2 22 24 7 13 16 12 12
24 26 7 13 17 12 12

18 22 13 23 27 21 23
4 20 24 13 23 28 21 23
22 26 13 24 30 22 23

6 18 24 19 32 38 30 32
20 26 1 19 1
33 1
40 31 32

8 18 26 1 25 1 41 49 38 40
1

aAccident reduction factors are rounded to the nearest whole number.

bZegeer, et al., October 1987.

Table B-4
Summary of Accident Reduction Found in the Literature for Actual Roadway Widening Projects

Expected Percent Reduction in Accidents

ADT Total Single-Vehicle Head-On

State Type of Project Range(vpd) Accidents Accidents Accidents

California' Widening 20- to 24-ft 0-3,000 16 22 45

roadway to 28 ft
Widening 18- to 24-ft < 5,000 35* 49* 48*
roadway to 32 ft
Widening 18- to 24-ft > 5,000 29* 22* 51*
Roadway to 40 ft

Texasb Adding full-width I.,

aved 000-3,000 27* 55* Unknown
shoulders to two- 3000 13* 21* Unknown
roads 5,000 18* 0 Unknown—

aValues from the Rinde (1977) study in California.

bValues from the Rogness, et al. (1982) study in Texas.
ese pe ent differences were significant at the 95 percent level of confidence for California sites and
90 percent confidence level at the Texas sites.

Source: Turner, et al., 1981; Rogness, et al., 1982; and Rinde, 1977.
70

ADT = 1,000
Terrain = Rolling
1.0 1 Roadside Hazard Rating = 5

0.8
CL
4)

0.6
CL

rW
4)
'0
0.4

0.2

0
8 10 11 -11 12
Lane Width (Feet)

Figure B-2. Plot of related accidents for lane and shoulder widths based on the predictive model.

Source: Zegeer, at al.. October 1987.

2
CL Legend
rA
0 0 40 No Shoulder
1.8- Stabilized Shoulder
— Unstabilized Shoulder

C 2
0 1.6-

NI

1.4 -

U) 4
0
CL
a
0 1.2 -
Shoulder Width (feet)
a

10
0
2

0
10
.8

4) .
M 6
E
3 7 8 9 10 11 12
Z
Lane Width (feet)

Figure B-.3. Rate of related accidents based on most-likely accident relationship.

Source: Zegeer, C.V. and J. Deacon. "Effect of Lane Width, Shoulder Width,
and Shoulder Type on Highway Safety: A Synthesis of Prior Literature."
Transportation Research Board: Washington, D.C. March 1986.
 71

r
0

r
Legend
SOURCE:
Symbol Reference
0 0 pit VI
U 0 7 KY 1980
0 4 LA 1970
5 OH 1974
6 MD&WA 1978

0L
7 8 9 10 11 12 13

Lane Width in Feet

Figure B-4. Accident Rate Based on Lane Width

for Two-Lane Rural Roads

SOURCE: Sp thesis of Sa I aty Research Related to Tra I I ic Cc mro I and Roadmy Elements.
Vo;ume 1, Chapter 1, "Roadway Cross Section and Alinement," Federal Highmv
Administration Report No. FIIWA-TS-82-232, December 1982.
 Variables that were tested and found not significant include number of intersections per
APPENDIX C mile, speed limit, and percent trucks. It should be noted that the posted speed limit varied
very little in the data (i.e., usually 55 mph), because all of the sections were on rural roads.
ANALYSIS OF LOW-VOLUME ROADS DATA Sections with no posted speed limit were assumed to have 55-mph speed limits. A separate
analysis of paved roads was carried out using 0 available data. Data on unpaved roads were
obtained only from Michigan, North Carolina, and Utah. Another analysis was carried out for
PRELIMINARIES both paved and unpaved roads from these three states.

Some basic issues that arose prior to carrying out the main data analysis tasks involved
the choice of models and analysis techniques to be used, the choice of the dependent variables GENERAL ANALYSES OF ACCIDENT RATES ON PAVED ROADS
to analyze, and the choice of independent variables to include in the analysis. The nature of
the research question and the available data seemed to dictate that some form of co-variance Two types of models were developed for paved roads—one using both lane width and
analysis would be appropriate because a study goal is to compare the accident experience on shoulder width separately and one based on total roadway width. Six lane-width categories
roads with specific lane and shoulder widths while taking into account certain other factors, were used: 8 ft or less, 9 ft, 10,ft, I I ft, 12 ft, 13 ft or greater. Shoulder width categories
some of which may be measured as continuous variables. Based on previous work by Zegeer were 0, 1-2, 3-4, and 5-6 ft. All combinations of these categories were present in the data
and Stewart, et al. (1989), a weighted, least-squares analysis of some types of accident rate with the exception of lane widths of 8 ft or less together with shoulder widths of greater than
aJso seemed appropriate for this study. To select the dependent variable, preliminary analyses 6 ft. A basic model for the lane- and shoulder-width analysis relative to related accident rate
were carried out to determine what types of accident rates were correlated with lane width and contains the following variables: lane/shoulder width—consists of 29 discrete values
shoulder width. For these analyses, co-variance models were fit to the following accident representing the combinations of lane and shoulder width; state—consists of three discrete
rates: single-vehicle accidents (fixed-object, rollover, and run-of-road); opposite-direction values representing the three groups of states: Alabama, Montana, and Washington; North
accidents (head-on and opposite-direction sideswipe); angle accidents; same-direction accidents Carolina and Michigan; Utah and West Virginia. It was found in preliminary analyses that the
(sideswipe and rear-end); and other accidents (including pedestrian and animal accidents). states within these groups had similar effects relative to related accident rates; number of
driveways per mile—a continuous variable; hazard rating—a roadway rating scale ranging from
All accident rates were given as accidents per million vehicle miles. The independent variables I to 7 (good to bad) treated as a continuous variable (note that clear recovery distance, a
in these analyses included lane width, shoulder width, terrain, and roadside hazard rating. continuous variable, often was used in place of hazard rating and had essentially the same
effects; and terrain—a value of 1, 2, or 3 representing flat, rolling, or mountainous terrain,
The results of the preliminary analyses showed only single-vehicle accident rates and respectively (this variable was used in the initial model as a continuous variable and later
opposite-direction accident rates to be associated significantly with varying lane and shoulder included as a class variable, with essentially the same effect in either form).
widths. Following these analyses, single-vehicle and opposite-direction accidents were
combined (this combination is referred to as "related accidents"). Thus, related accidents per Results from a weighted, least-squares analysis of these variables using SAS PROC GLM
million vehicle miles was the primary dependent variable in the subsequent analyses. To are given below. The product ADT x section length was used as the weight function as in an
investigate severity questions, certain types of injury accident rates also were analyzed (as earlier study by Zegeer, et al. (1990). Thus, the accident effects on longer and more heavily
discussed in a later section of this appendix). traveled sections were given more weight in the development of the model coefficients.

Independent variables were selected on the basis of being associated significantly with The resulting model contained 33 estimated parameters: 28 for lane and shoulder width,
related accident rates and having valid values on a large portion of the study sections. Thus, 2 for state, and I each for driveways per mile, hazard rating, and terrain. Each of the first
variables related to grade and curvature were not considered because they were available for four factors was significant at p = 0.0001. Terrain was not very significant at p = 0. 1226.
only about half of the study sections. Other independent variables used in the analyses were The model had an R' = 0.234. The estimated model coefficients of driveways per mile,
state, functional class, clear recovery distance, terrain, and number of driveways per mile. hazard rating, and terrain were 0.030, 0.268, and 0. 118, respectively. Thus, an increase in
Lane width, shoulder width, and total roadway width of lanes (plus shoulders) were treated as each of these factors corresponded to an increase in related accident rate.
class variables; that is, each specific value (i.e., width in feet) was considered as a treatment
level. State and functional class also were class variables. Terrain was given the following Of particular interest are the least-squares means of related accident rate by lane and
values: I = flat, 2 = rolling, 3 = mountainous; it was sometimes used as a class variable shoulder width class. These are adjusted mean accident rates, adjusted according to the other
and sometimes as a continuous variable (the effects were quite similar). Hazard rating and factors in the model, or mean accident rates by lane and shoulder width, "all other things
clear recovery zone were continuous and essentially interchangeable, with only one being used being equal" (certain average values depend on the variables in the model and the manner in
in a given model. The state, functional class, and the road-width variables tended to be which they are classified). Thus, while the relative values of the least-squares means resulting
interrelated strongly, and after two similar models were estimated, one containing functional from a given analysis are meaningful, the values themselves have little meaning and cannot be
class as a variable and the other with that variable omitted so that the results could be compared to least-squares mean values resulting from a different analysis using different
compared. variables and classifications.
 The least-squares means of accident rate by lane and shoulder width categories from the considered: 9 ft or less, 10, 11, 12, ..., 20, and 21 ft or more. Along with the independent
basic model described previously are given in Table C-1. In addition to the least-squares variables described earlier, a variable indicating whether or not the shoulder was paved was
means, Table C-1 also shows the sample size within each lane and shoulder width category and included in the initial analyses of total roadway width. This variable was not significant (p
p-values from two sets of significance tests. The first set of tests (Test 1) compares each 0.662), and was not included in later analyses.
shoulder-width category with the preceding smaller category within the same lane width.
Thus, the first value (0. 19) in the Test I column is the p-value for a test comparing accident Results from two different models for related accident rates on the 13 roadway width
rates on roads with lane widths of 8 ft or less and no shoulders, with the rates on roads with categories are presented in Table C-3. Model I, in addition to roadway width, included state
lane widths of 8 ft or less and I- to 2-ft shoulders. (three levels) and terrain as class variables, and clear recovery distance and number of
driveways per mile as continuous variables. In Model II, state and two levels of functional
Very few of these "pairwise" comparisons are statistically significant. For example, none class (local versus all other) were taken as class variables, while terrain, hazard rating, and
of the comparisons are significant for 9-ft lane widths; for 10-ft lanes, the only significant driveways per mile were included as continuous variables. Table C-3 shows least-squares
difference (of those tested) was between 3- to 4-ft shoulders and 5- to 6-ft shoulders, and so means of related accident rate by the road width categories for each of the two models, and
on. For lane widths of 8 ft or less, a significant difference is indicated between 3- or 4-ft test results for comparing adjacent width categories. Use of functional class Model 11 reduces
shoulders and 5- or 6-ft shoulders. However, because only two sections in the data had 5- or the sample size from 1,107 to 982 because functional class is unknown in the Michigan data.
6-11 shoulders and lane widths of 8 ft or less, and because the least-squares mean rate of With all the differences, the pattern of least-squares means is very consistent between the two
related accidents is negative, the foregoing result does not appear meaningful. models. Specifically, 9-ft widths have lower rates than 10-ft widths. The rates are about the
same for 10- to 14-ft widths and, then, they decrease again for widths of 20 ft and over. A
The second set of tests involved a simultaneous comparison of several shoulder width comparison of Models I and II is given in Figure C-1.
categories within a fixed lane width. For example, for 9-ft lanes, a test of the equality of all
five shoulder width categories had a p-value of 0.90 (which does not indicate significant When the width categories were combined in exactly the manner described, and Model I
variation in related accident rates across the five shoulder width categories for roads with 9-ft was reestimated using the reduced number of width categories (four), the results were as given
lanes). In fact, of the Test II results shown in Table C-1, only the results comparing the four in Table C-4. The results of these analyses seem quite consistent with those of the lane width
shoulder width categories for 8-ft lanes was very significant. This result, again, seems to stem and shoulder width analysis, with 10-ft lanes and narrow shoulders generating many of the 10-
from the fourth shoulder width category and is not a meaningful result. For 12-ft shoulder and 11-ft total width sections. The rates of related accidents from Table C-4 are shown in
widths, there were significant differences, but,the rate fluctuated in a nonuniform way. For Figure C-2.
example, the accident rate for 5- or 6-ft shoulders was higher than the rate for 3- or 4-ft
shoulders (1.85 versus 1.40).
FURTHER REFINEMENT OF ACCIDENT RELATIONSHIPS ON PAVED ROADS
For the next step in the analysis, a similar model was run with the shoulder width
categories tested under Test 11 combined. This resulted in 10 lane-width by shoulder-width Using the primary low-volume roads database (Zegeer, et al., October 1987), the
categories (rather than the original 29). Lzast-squares means from this model are presented in information shown in Table C-2 was used to further combine lane and shoulder width
Table C-2 along with least-squares means from a second model in which clear recovery categories when the rates of related accidents did not differ significantly. This process led to
distance was substituted for hazard rating, ar~d terrain was entered as a class variable with the seven lane and shoulder width categories given in Table C-5, which gives least-squares
three levels (rather than as a continuous variable). The two sets of least-squares means are rates of related accidents (i.e., single-vehicle plus opposite-direction accidents) estimated from
quite similar and, more importantly, the pattern down the column is the same. Namely, roads a model with co-variates indicating state, terrain type, clear recovery distance, and number of
with 8- and 9-ft lane widths have about the same accident rates, the rate increases for roads driveways per mile. It should be noted that the data in Table C-2 were derived from the
with I 0-ft lanes and narrow shoulders, and the rates then decrease for I I - and 12-ft lanes with application of a similar (collapsing) procedure applied to Table C-1, which resulted in the
narrow shoulders. Roads with 10- or 12-ft lanes and wider shoulders have rates that are lower unequal shoulder width categories in Table C-2 and, also, in Table C-5. An illustration of the
than those for 12-ft lanes with narrow shoulders. For the relatively few roads in the study data rates from Table C-5 are shown in Figure C-3. Note that for lane widths of 10 ft and above,
with lane widths of 13 ft or greater, those with narrow shoulders have accident rates that fall related accident rates decrease as shoulder widths decrease. No difference in accident rate was
between the 11- and 12-ft lane and narrow shoulder Tate and the 10- to 12-ft lane and wide found for lane widths of I I versus 12 ft. Lane widths of 8 and 9 ft had a lower rate of related
shoulder rate. The roads with 13-ft lanes and wide shoulders have the lowest rates of all. The accidents (1.66) than 10-ft lanes with narrow shoulders (2.41). This could be the result of
li-values from significance tests of these comparisons are shown in the last column of Table C- lower vehicle speeds on roads with 8- and 9-ft lanes.
2.
The accident rates in Table C-6 represent a slight modification to the data in Table C-5,
The second type of model developed was based on categories of lane width and shoulder width where consistent shoulder width categories of zero to 4 ft (narrow) and 5 ft and over (wide)
(i.e., half of the total width of lanes plus shoulders). Initially, 13 total width categories were have been selected. As expected, the only noticeable difference between Tables C-6 and C-5
involves the 11- and 12-ft lane, narrow shoulder category. Figure C-4 illustrates the rates C-6 and C-9 uses similar categories of shoulder widths. While the rates differ for some
from Table C-6. categories of lane and shoulder width, the general rate trends are relatively similar, suggesting
that similar lane and shoulder width effects on accidents is supported by both databases.
The pattern of related accident rates given in Tables C-5 and C-6(see Figures C-3 and C-
4) seems to fit well with expectations for roads with 10-ft and wider lane widths. Note that
very little information was available for roads with 8-ft lanes; thus, more than 90 percent of ANALYSIS OF UNPAVED ROADS
the data for roads with 8- and 9-ft-wide lanes, in fact, results from a survey of 9-ft-wide lanes.
Moreover, the estimated rate of related accidents for 8- and 9-ft lanes in Tables C-5 and C-6 A total of 164 roadway sections of unpaved roads (from Michigan, North Carolina, and
of 1.66 is quite consistent with the rates for other lane widths on roads having wide shoulders. Utah) was available for analysis. These sections were placed into three width categories
The puzzling result, then, is that the estimated rates for roads with 9-ft lanes and narrow (corresponding to lane width plus shoulder width for paved roads) with roughly equal sample
shoulders (see Table C-1), seems to be lower than would be expected relative to rates for 9-ft sizes: 9 ft or less, 10 and 11 ft, and 12 ft and greater. Because of the small number of
lanes and wide shoulders, and rates for 10-ft lanes and narrow shoulders. unpaved roads and the interest to compare paved and unpaved conditions, paved and unpaved
roads from the three states were analyzed together.
In an effort to investigate the unexplained accident rate reduction for 8- and 9-ft lanes
(compared to 10-ft lanes), the primary database was split into two subgroups: data from the A class variable, with six levels corresponding to the three width categories of paved and
cross-section database (Zegeer, et al., December 1987) and data from the supplemental unpaved roads, was analyzed in models containing many of the other variables used in the
databases (e.g., data from roads in Utah and in Oakland County, Michigan; and in North analyses described above. Because functional class was not available for Michigan data, and
Carolina from minor roads). As shown in Figure C-5, rates were higher for 10-ft lanes and was highly confounded with state and roadway width variables in the other two states, no
narrow shoulders in each database subset. This indicated that there was not one portion of the attempt was made to include functional class in any of the analyses. Comparisons of the
data that was causing this trend. In general, all of these databases revealed this same trend in estimated effects were made of width within pavement types, and of paved versus unpaved
reduced accident rate for 8- and 9-ft lanes. roads within width categories. Table C-10 contains results from fitfing three different models
(labeled A, B, and Q to the data. Model A contained only the road class variable, terrain,
To take one further look at this question, models used to generate Tables C-1, C-2, C-5, and clear recovery distance. In Model B, driveways per mile was added, and in Model C,
and C-6 were fit to the cross-section database (developed for FHWA by Zegeer, et al., state was added as a class variable. The R2 values for these three models were 0.147, 0.162,
December 1987). This database originally included approximately 5,000 miles of rural, two- and 0. 170, respectively. Figure C-7 (which illustrates related accident rate by lane width for
lane roads in seven U.S. states with a full range of ADT (i.e., 100 vpd to approximately paved and unpaved roads using Model A) clearly shows that unpaved roads have higher
20,000 vpd). From this database, a subsample was selected for all rural road sections with accident rates than paved roads for given lane widths.
ADT less than or equal to 5,000 vpd. Initially, I I lane and shoulder width categories were
considered (the same as those in Table C-2), but with the 9-ft lane width roads split into two Table C-10 lists three sets of least-squares means and the significance levels from three
categories—one with shoulder widths of zero to 4 ft and the other with shoulder widths of 5 ft sets of pairwise comparison tests, where the symbols P, P2, P31 U11 U2, and U3 are used to
or greater. Estimated rates of related accidents from this analysis are given in Table C-7. denote the three width categories for paved and unpaved roads, respectively. The same pattern
Again, note that the rates for the two 9-ft lane categories are virtually identical and are less of least-squares mean accident rates is seen for all three models. This pattern also ig consistent
than the rates for the 10-ft lane, narrow-shoulder roads. with the previous results—accident rates are lower for narrower roadways. The significance
tests consistently show a significant difference between accident rates on unpaved roads with
On the basis of hypothesis tests of the equality of various sets of the rates in Table C-7, lane plus shoulder width less than or equal to 9 ft wide, and those for unpaved roads 10 and I I
categories were further collapsed into the six categories given in Table C-8. A similar set of ft wide. Accident rates for paved roads differ significantly from accident rates for unpaved
analyses was carried out using only shoulder widths of zero to 4 ft and 5 ft or greater. The roads for width categories of 10 to I I ft and for 12 ft and greater.
resulting estimated rates of related accidents (following collapsing of categories in Table C-8)
are presented in Table C-9. Another type of comparison of paved versus unpaved roads was done by partitioning the
data according to three ranges of ADT rather than by roadway width. The three ADT ranges
The low-volume roads database and the cross-section database overlap—the cross- section used were ADT :5 250 vpd, 250 vpd < ADT :!~ 400 vpd, and ADT > 400 vpd. Again, a
data on rural roads with ADT less than or equal to 2,000 vpd contained in both. Nonetheless, road type variable with six classes is given: paved, ADT 15 250 vpd; paved, 250 vpd <
the results given in Tables C-7, C-8, and C-9 seem to provide further validation of the results ADT :5 400 vpd; ... unpaved, ADT > 400 vpd. Table C- I I gives results from a model for
given in Tables C-5 and C-6. related accident rate as a function of road type, state, terrain, and clear recovery distance.
These analyses show that rates of related accidents on paved roads decrease significantly with
Figure C-6 compares rates of related accidents by lane and shoulder width categories for increases in ADT category. On unpaved roads, rates on roads with ADT less than 400 vpd
the low-volume roads database (ADT less than or equal to 2,000 vpd) to rates for the cross- were shown to be significantly lower than rates on roads with 250 vpd :5 ADT < 400 vpd,
section database (ADT less than or equal to 5,000 vpd). This comparison of rates from Tables though the sample sizes were quite small for these two categories.
 The rates did not differ significantly between paved and unpaved roads with ADT less accident rates of Table C-10, injury rates were significantly higher on 10- to 11-ft unpaved
than 250 vpd. Paved roads had significantly lower rates than unpaved roads in the ADT range roads than on unpaved roads with widths less than or equal to 9 ft. In addition, paved roads
from 250 to 400 vpd. For roads with ADT greater than 400 vpd, again the difference was not differed significantly from unpaved roads with respect to injury rate for the categories of 10-
significant. and I I-ft lanes and greater than or equal to 12-ft lanes. A similar pattern also can be seen for
the KAB rate column. Again, the same comparison tests were statistically significant. The
Because of the very small number of unpaved roads with ADT greater than 400 vpd, the pattern is much less discernable for the KA rate column and, with these rates, none of the
upper two ADT categories were combined and then compared. The last line of Table C-11 comparisons was statisficaJly significant (p < 0.05). Figure C-9 shows a plot of injury rates
indicates that this difference is statistically significant. Accident rates from Table C-11, for by surface type and lane width.
paved versus unpaved roads for various ADT groups, are shown in Figure C-8.
The similar pattern in the variation of injury rates across lane and shoulder width
categories and for related accident rates suggests that the distribution of accident severity
ANALYSES OF INJURY RATES remains relatively constant with respect to lane and shoulder width. In particular, there are
relatively high injury rates on the narrowest roads where there are relatively low accident
The question of how roadway features relate to crash severity can be approached in two rates.
different ways: (1) calculate various types of injury rates in the form of injury accidents (at
some level of severity) per million vehicle miles, and fit models to these rates in the same
manner as was done with related accident rates; and (2) examine how the distribution of VALIDATION AND EXTENSION OF LANE AND SHOULDER WIDTH ANALYSES -
injuries, given an accident rate, varies with roadway features. Because our basic data file
consists of counts of accidents of various types (e.g., injury accidents, fatal accidents), only Data from three state data systems (Illinois, Minnesota, and North Carolina) were used to
the former approach was pursued in this study. validate and extend the results presented in the primary analysis of the safety implications of
varying lane widths and shoulder widths on both paved and unpaved, low-volume roadway
Three types of injury rates were used: (all) injury accidents per million vehicle miles; sections. The North Carolina data represented additional sections not used in the primary
accidents resulting in K, A, or B injuries per million vehicle miles; and accidents r esulting in analysis. Again, all accident rates were in terms of related accidents only (i.e., single-vehicle
K or A injuries per million vehicle miles. These rates are referred to simply as injury rate, plus opposite-direction crashes), unless otherwise stated.
KAB rate, and KA rate, respectively. It should be noted that for these analyses, accidents are
not restricted to related accidents. In the analyses of the seven-state primary database, lane and shoulder width effects on
related accidents per million vehicle miles were estimated while adjusting for differences in
To investigate relationships between injury rate and lane and shoulder width, models were functional class, terrain, clear recovery distance (or hazard rating), and number of driveways
fit to each type of injury rate using the original 29 lane and shoulder width categories, along per mile. Functional class was the only variable available in all of the three secondary analysis
with state, terrain, clear recovery distance, and driveways per mile. Patterns of least-squares files; terrain also was available in the North Carolina data.
mean injury rates clearly resembled those of related accident rates, and a similar set of test
statistics indicated it would be appropriate to group the lane and shoulder width categories into Table C-14 shows the available paved roadway sections for the primary data set and for
10 groups as done previously for related accident rates. Least-squares mean injury rates from the three validation data sets, classified into the 30 lane width by shoulder width categories
the resulting combined categories are given in Table C-12. The pattern down the injury rate used in the primary analysis (and shown earlier in Table C-1). Differences in the distributions
column is similar to that of the related accident rates in Table C-2, with the exception of the of lane and shoulder width from state to state are apparent from Table C-14. The following
relatively high injury rates for roads with 13-ft lanes. As injury severity increases (KAB rate discussion of each state will highlight these differences.
and KA rate) the overall pattern becomes less pronounced. A part of the pattern that is
retained is that the injury rates for roads with 10-ft lanes and narrow shoulders are consistently
higher than the rates for roads with 8- and 9-ft lanes. This could logically be related to lower Analysis of Illinois Data
vehicle speeds on roads with lane widths of 8 and 9 ft, compared to 10 ft. A test of the
difference between rates for roads with 9-ft lanes versus 10-ft lanes with narrow shoulders is The Illinois data contained no information on unpaved roads. With respect to paved
statistically significant for all three injury rates (p 15; 0.0002). The relatively higher injury roads, Table C-14 indicates that Illinois has very few roads with lane widths of 8 ft or less,
rates for roads with 13-ft lanes may result from more angle accidents (which were not included and few roads with shoulders of less than 3 ft in width, making comparisons with the primary
in related accident rates). analysis results very difficult in certain categories. However, even given these difficulties, a
first comparison of Illinois rates with those from tile primary study involved the classification
Table C-13 shows least-squares means of the three injury rates from the three-state, paved of roadways into the same 10 groups as used in Table C-1. Least-squares mean rates,
versus unpaved analysis. As was pointed out with respect to Table C-12, the injury rates in presented in Table C-15 and Figure C-10, represent related accidents per million vehicle miles
Table C-13 are quite similar to the rates for related accidents in Table C-10. Also, as with the from a model that contained only one other factor, functional class (local versus other).
 For comparison purposes, the same model (i.e., a model only involving functional class) shoulders (p = 0.6959). For the other shoulder width categories, there are statistically __J
also was run on the primary data set (see the last column of Table C-15 for the results). 01
significant differences by lane width, but accident rates do not consistently decrease with
Clearly, the two sets of least-square estimates differ in magnitude or scale. A state variable increasing lane width (holding shoulder width constant).
was a significant factor in the primary analysis, therefore, a difference in scale between the
two data sets is not unexpected. However, the state difference between Illinois and the other Rates of total injury accidents were computed as total number of all injury crashes on a
states was much greater than the state differences within most states of the primary database. section (i.e., not just related accidents divided by million vehicle miles) in addition to the rate
For whatever reasons, the Illinois rates are consistently lower than the primary analysis rates. of more serious injury accidents (i.e., fatal, * A-type, plus C-type, termed KAB rate). The
injury rate and KAB rate are also given in Table C-17 by lane and shoulder width
Also, note that the Illinois rate for 8-ft lanes is based on very few observations, and the combinations. Such rates are by far the highest for lane widths of :5 8 ft (i.e., 1.92 injury
9-ft rate is derived primarily from roads with shoulder widths of 5 ft or more. Given these rate and 1.21 KAB rate). For 9-ft lanes and 10-ft lanes with narrow (6 ft wide or less)
limitations, and even with the lower rates, there is an increased rate for the 10-ft-wide lanes shoulders, injury rates drop to between 0.84 and 0.88, while KAB rates drop to between 0.58
with narrow shoulders. Patterns for other lane width groups also indicate increased rates with and 0.64. Injury and KAB rates drop further and fluctuate somewhat for wider lanes and
respect to decreased shoulder width. shoulders.

Because most of the Illinois data are for roads with lanes of 9 ft or more and relatively Least-squares mean rates of related accidents by functional class also were computed for
wide shoulders, rates of related accidents were estimated for five categories of these roads and Minnesota paved roadways. These values were arterials-0.64, collectors-0.75, and
compared with rates from similar roads from the primary database. Results from these local-0.87.
analyses are given in Table C- 16 and shown on Figure C- 11. It should be noted here that the
last entry under the primary data column in Table C-16 is based on relatively few As expected, the arterials had the lowest rates, probably because of generally better geometric
observations. Both sets of estimates show steadily decreasing related accident rates with design. Local roads had the highest rates, as may be expected.
increasing lane width.

Using the Illinois estimates, a decrease of about 10 percent of related accidents is found Analysis of Minnesota Unpaved Roads
for each additional foot of lane width (for roadways with relatively wide shoulders). Using the
primary database, the decrease is roughly the same for lane widths of 9 to 12 ft. The Minnesota sample contained data on 4,900 unpaved road sections with lane widths
ranging from 6 to 16 ft. For 96.7 percent of these sections, the file also indicates a shoulder
of at least I ft in width. Analyses were carried out on both lane and shoulder width categories
Analysis of Minnesota Paved Roads and on total width categories. Table C-18 gives results from a model with 19 lane width by
shoulder width categories and functional class at two levels (local versus collector) as
From Table C-14, the Minnesota data (similar to data from Illinois) contain virtually no independent class variables. Table C-18 gives the least-squares mean rate of related accidents
information for roadways with lane widths of 8 ft or less, and relatively little information for and sample size for each of the 19 categories.
those with 9-ft lanes. Very large samples, however, are available for wider roads and,
especially, for wider roads with wide shoulders. Starting with the basic categories of Table C- The width factor was not statistically significant in the estimated model (p = 0.3717),
14, related accident rates were estimated and categories with similar rates were combined to and the mean accident rates do not display consistent tendencies with respect to lane width or
yield 14 categories for further analysis. The first two categories again consisted of all roads shoulder width. Functional class was significant (p = 0.0057) with mean rates of related
with lane widths of 8 ft or less, and all roads with 9-ft lanes, respectively. For roads with lane accidents of 0.64 and 0.78 for collector and local roads, respectively.
widths of 10 ft or more, three shoulder width categories were considered: narrow, medium,
and wide. These groups are specified explicitly in Table C-17, which lists the least-squares The second analysis of Minnesota unpaved roads involved adding the width of the lanes
mean accident rates from an analysis that included three levels of functional class (arterial, and shoulders to form a single width variable. Twelve width categories, ranging from 18 ft or
collector, and local) as a second factor. less to 40 ft or more, were formed and analyzed in a model with functional class as a second
factor. In this model, the width categories were marginally significant (p = 0.0715), while
For a fixed lane width, reductions in rates of related accidents are generally found as functional class was significant (p = 0.0052). The total width categories, their estimated mean
shoulder width increases, as given in Table C-17 and shown on Figure C-12. For example, rates of related accidents, and sample sizes are given in Table C-19 and shown on Figure C-
for a 10-11 lane width, shoulder widths of zero to 2, 3 to 6, and greater than 7 ft yield accident 13. As with the lane width by shoulder width categories, there does not seem to be a
rates of 1.03, 0.86, and 0.42, respectively. Accident rates also generally decrease for discemable pattern of change in rate of related accidents with increasing roadway width. This
increases in lane width for fixed shoulder width categories in only some cases. This is further finding differs from the findings from two studies of the primary database, where wide
borne out by the hypothesis tests at the bottom of Table C-17, which show no significant unpaved roads had lower accident rates than narrow unpaved roads. It is not clear why the
differences due to lane width (10 ft to more than 13 ft wide) on roads with the widest two databases do not agree.
Combined Analysis of Minnesota Paved and Unpaved Roads Data shoulders 5 ft and wider) in the North Carolina and the primary data analyses showed a lower
accident rate for 8- and 9-ft lanes, a higher rate for 10-ft lanes, and a much reduced rate for
A third analysis was carried out on the Minnesota file, this time on the combined data for 11- and 12-ft lanes. The North Carolina analyses, however, revealed the highest rates of
paved and unpaved roads. This analysis involved estimating a model with 16 lane width by related accidents for lanes of 5 9 ft (1.95), with lower rates for 10-, 11-, and 12-ft lanes (a
shoulder width categories as one factor, functional class (three levels), and a variable rate of approximately 1.7 for each of these three widths). See Table C-21.
indicating paved or unpaved as two additional factors. Both the width categories and
functional class were statistically significant (p = 0.0001 for each). Pavement status was not
significant (p = 0.3980). Analysis of Unpaved North Carolina Roads Data

The width categories and their estimated accident rates are given in Table C-20. As might be The North Carolina data file contained information on 236 unpaved road sections. All
expected from the previous analyses, the results in Table C-20(paved plus unpaved) appear to but nine of these sections were classified as having zero shoulder width, which suggests an
agree with the results for paved roads shown in Table C-17, except for the 8-ft lane width inventory coding decision has been made in North Carolina not to attempt to separate lane
category, where the rate for paved roads was much higher. This is perhaps because of the width from shoulder width on unpaved roads. This may result from the nature of most
higher speeds on narrow paved roads (8-ft lane width) compared to unpaved roads. unpaved North Carolina roads—the gravel cover and grading extend from ditch line to ditch
line. The data were grouped into six lane width categories: 5 to 6, 7 to 8, 9, 10, 11 to 12,
and 13 to 16.
Analysis of North Carolina Data on Paved Roads
In an analysis with lane width as one factor and terrain and functional class (local versus
Unlike the data for Illinois and Minnesota, data on the additional North Carolina sections collector) as other factors, neither lane width nor terrain were statistically significant (p =
not used in the primary analysis contained relatively high frequencies of roads in the less than 0. 1895 and p = 0. 5398, respectively). With terrain deleted from the model, lane width was
or equal to 9-ft lane width range, but virtually none with lane widths greater than 12 ft. The still not significant (p = 0. 1301). Least-squares mean rates of related accidents by lane width
first analysis of these data was done on the first 25 lane width by shoulder width categories and functional class are given in Table C-23.
given in Table C-14. The analysis was formulated as a model with a lane width factor, a
shoulder width factor, a lane width by shoulder width interaction, and two additional . In addition to lane width not being statistically significant, little evidence suggests that
factors—functional class at three levels and terrain at three levels. any logical regrouping by expanded lane width categories would lead to meaningful results.
The rates in Table C-23 reveal an inconsistent relationship between width and related accident
~ The lane width by shoulder width interaction was not statistically significant rate for unpaved roads. Possible reasons for this trend include the relatively small sample
(p = 0.8115); therefore, the model was rerun with the interactions removed. Least-squares sizes for most width categories, possible problems with accident reporting on unpaved roads,
mean rates of related accidents by each of the other factors are given in Table C-21 and shown or reduced speeds on some unpaved road sections that may result in reduced accident rates.
on Figure C-14, which controlled for variables of importance. It would appear from Table C- Regardless of the explanation, it does not appear that changes in lane width on unpaved roads
21 that some of the rates listed.are essentially equal, (i.e., lane width :5 8 and 9 or I I and 12, affect accident rate. This conclusion differs from that of the primary analysis, which indicates
shoulder width of 0 to 4 ft). that wider, unpaved roads have lower accident rates than more narrow, unpaved roads.

Statistical hypotheses tests confirmed these conjectures. As a result, lane widths were A comparison of the estimated rates of related accidents of Table C-23 with those of
reclassified into two categories (:5 9 ft, > 10 11);and shoulder widths into three categories (0- Tables C-21 and C-22 suggests that North Carolina rates are considerably higher on unpaved
4, 5-6, >7 ft). The model was rerun with this reclassification. Again, an interaction term roads than on paved roads. The overall, unadjusted mean rate of related accidents was found
was included because, even though this factor was not significant, including it results in the to be 1.63 for paved roads versus 3.45 for unpaved roads, which supports this conclusion.
calculation of least-squares mean rates of related accidents for each lane width by shoulder However, the fact that most of the paved roads, even with narrow lanes, tended to have some
width combination. These recalculated rates are given in Table C-22 and shown on Figure C- shoulders, while the unpaved roads tended to have narrow lanes and no shoulder, complicated
15. the analysis of paved and unpaved roads together to produce a meaningful measure of road
width. In an analysis with four lane width categories (i.e., :5 8 ft, 9, 10, > 11 ft), together
Table C-22 indicates that rates of related accidents generally decrease as shoulder width with factors for terrain, functional class, and pavement status, all factors were significant
increases for a given lane width category. For example, for lane widths of 10 ft or more, the (p = 0.0001 for each factor). The least-squares mean rates of related accidents for paved road
related accident rate drops from 1.85 (for shoulder widths of 0 to 4 ft) to 1.65 (for shoulder versus unpaved road, after adjusting for the other factors, were 1.82 for paved and 3.26 for
widths of 5 to 6 ft) to 1.43 (for shoulder widths of 7 ft or greater). Also, for a given shoulder unpaved roads.
width category, lane widths of greater than 10 ft had slightly lower rates than lane widths of
less than 9 ft. These results, considered with those in Table C-21, show some similarities and
some differences with the primary analysis. First of all, wider shoulders (particularly for
SUMMARY AND CONCLUSIONS __j
Table C-1 . 00
For. paved roads, aside from differences in magnitude from state to state, the validation I,east-Squares Mean Rate of Related Accidents and Significance Test
analyses seemed, overall, to confirm the findings of the primary analyses. Rates of related for 29 Lane and Shoulder Width Categories
accidents (i.e., single-vehicle plus opposite-direction accidents) decrease with both increasing Lane Width Shoulder Width Number of L.S. Mean Test I Test H
lane widths and increasing shoulder widths. In Illinois, most road sections had relatively'wide (feet) (feet) Sections Accident Rates p-Value p-Value
shoulders, and accident rates decreased consistently with increasing lane width. The Minnesota
data (which contained much information on wider roads with a full range of shoulder widths), .:~.8 0 7 1.38 -
showed rates of related accidents to decrease consistently with increasing shoulder width, but .!L8 1-2 9 2.45 0.19 0.03
* 8 34 6 2.12 0.69
much less consistently with increasing lane width. In fact, for roads with the widest shoulders,
* 8 5-6 2 -0.21 0.01
rates did not differ significantly with differing lane widths. North Carolina's data were shifted
more towards roads with narrower lanes and a fairly full range of shoulder widths. On these 9 0 24 1.56, -
roads, consistent decreases in related accident rates were found with both increases in lane 9 1-2 38 1.83 0.52
width and increases in shoulder width, although the range of the decreases due to shoulder 9 34 97 1.66 0.58 0.90
width was somewhat greater. 9 5-6 90 1.60 0.79
9 6+ 2S 1.43 0.61
Overall, the data seem to show that, with respect to related accident rates on paved roads, 10 0 28 2.31 -
shoulder width is at least as important a factor as lane width. In particular, increasing lane 10 1-2 52 2.23 0.85 0.84
width at the expense of decreasing shoulder width would not appear to be a cost-effective 10 3-4 85 2.40 0.56
approach. This finding is consistent with the results of recent research on higher volume, two- 10 5-6 54 1.45 0.00 0.19
lane roads. 10 6+ 29 1.08 0.19

A large amount of data on unpaved roads was available from Minnesota and a lesser 11 0 14 1.79 -
1-2 48 1.81 0.97 0.97
amount from North Carolina. Neither of these samples showed related accident rates to vary 11
11 3-4 51 1.21 0.04
significantly with roadway width. In Minnesota, the rate of related accidents did not differ
significantly on paved and unpaved roads, while in North Carolina, rates of related accidents 11 5-6 50 1.13 0.74 0.85
on unpaved roads were estimated to be nearly double those on paved roads. 11 6+ 43 1.27 0.57
12 0 42 1.81 - 0.91
12 1-2 89 1.78 0.91
12 3-4 71 40
12 5-6 36 1,
1.85 0.10
0." 0.06
12 6+ 65 1.22 0.02
* 13 0 30 1.20 -
* 13 1-2 8 1.32 0.84 0.96
* 13 34 8 1.36 0.96
> 13 5-6 8 0.80 0.45 0.58
.L13 1 6+ 1 4 1 0.40 1 0.58
 Table C-2 - Table C-3
Results from Models of Lane Plus Shoulder Width
Least-Squares Mean Rates of Related Accidents for
Collapsed Lane and Shoulder Width Categories Width of
Lane Plus Model I Model I Model 11 Model H
Model 2 Model 3 Shoulder Number L.S. Rates of Pairwise Tests L.S. Rates of Pairwise Tests
Lane Width Shoulder Width L.S. Mean L.S Mean Significance Tests (feet) of Sections Related Accidents p-Value Related Accidents p-Value
(feet) (feet) Accident Rates Accident Rates p-Values
< 9 34 1.44 2.17
8 all 1.62 1.55 0.01 0.09
0.7170 10 46 2.41 2.83
9 all 1.63 1.67 0.25 0.57
0.0001 11 69 2.03 2.64
0.49 0.35
10 0-4 2.34 2.41 12 145 1.87 2.43
0.0001 0.82 0.74
I1 0-2 1.80 1.88 13 180 1.83 2.37
0.84 0.35
12 0-2 1.78 1.86 14 194 1.87 2.52
0.0077 0.11 0.11
10 ~t 5 1.30 1.43 15 119 1.59 2.23
0.04 0.11
1.32 16 92 1.19 1.91
11 z- 3 1.21
0.38 0.35
17 60 1.37 2.14
12 ;-> 3 1.41 1.30 0.64 0.42
18 41 1.49 2.36
13+ 0-4 1.24 1.57 0.90 0.82
0.0506 19 39 1.45 2.28
13+ 5 0.65 0.76 0.17 0.30
20 39 1.05 1.92
0.75 0.79
a 21 55 1.13 1.85

Table C-4
Least-Squares Mean Rates for Four Categories
of Lane Plus Shoulder Width

Width of Iane Plus L.S. Mean Rates of

Shoulder (feet) Related Accidents Test p-Value

<9 1.43
10-14 1.90 0.09
15-19 1.41 0.0001
20 & Over 1.10 0.03
 00
Table C-5 0
Table C-7
Mean Rates of Related Accidents on Collapsed Lane and Shoulder Width Estimated Rates of Related Accident for 11 Lane
Categories from Primary Database and Shoulder Width Categories from Cross Section Database
L.S. Mean Rate of Lane Width (feet) Shoulder Width (feet) L.S. Mean Accident Rate
Lane Width (feet) Shoulder Width (feet) Related Accidents
8 all 1.52
8,9 all 1.66
10 0-4 2.41 9 0-4 1.37
10 ~:5 1.43 9 a5 1.34
11,12 0-2 1.87
11,12 23 1.31 10 0-4 1.82
13 0-4 1.57 10 a 5 1.22
13 2t5 0.76
11 0-2 1.64
11 a3 1.05

12 0-2 1.50
Table C-6 12 ~t 3 0.93
Rates of Related Accidents for Modified Lane and Shoulder Width Categories
13 0-4 1.57
L.S. Mean Rate of 13 2t 5 0.46
Lane Width (feet) Shoulder Width (feet) Related Accidents

8,9 all 1.66

10 0-4 2.41
10 ~6 1.43
11,12 0-4 1.65 Table C-8
11,12 z5 1.30 Estimated Accident Rates of Related Accidents
13 0-4 1.54 for Lane and Shoulder Width Categories'
13 ~5 0.74 Lane Width (feet) Shoulder Width (feet) L.S. Mean Accident Rate

8 & 9 all 1.36

10 0-4 1.81

11,12,13 0-2 (11- & 12-foot Lane) 1.57

0-4 (13-foot Lane)

10 5 1.22

11,12 0 0.98

13 5 0.47

-Collapsed from Table C-7

 Table C-9 Table C-11
Estimated Rates of Related Accidents for Lane and Equal Shoulder Width Paved Versus Unpaved Roads Within ADT Range
Categories from Cross Section Database L.S. Mean Rates
of Related
Lane Width (feet)
- FShoulder Width (feet) -FL S.
Mean Accident Rate
Pave ent Type ADT Accidents Sample Size
8 & 9 all 1.36
P, - Paved :5 250 3.71 86
10 0-4 1.82
P2- Paved 250-400 2.61 91
11,12,13 0-4 1.38
P, - Paved > 400 1.96 430
10 5 1.21
U, - Unpaved !~ 250 3.94 128
11,12 5 0.93
U, - Unpaved 250-400 4.16 23
13 2t 5 0.47
U3 - Unpaved > 400 2.29 13

F-Comparisons P-Value

P, = P, 0.046
Table C-10
P, =_P3 0.032
Model Results Comparing Paved and Unpaved Roads

Model A Model B Model C U, = U, 10.751

Lane L.S.Mean Rates of L.S.Mean Rates of L.S.Mean Rates of
Road Type I Width (feet)
Related Accidents Related Accidents Related Accidents U, = U, 0.013
P, - Paved 15 9 1.20 1.13 1.32 P, = U, 0.730
P, - Paved 10-11 2.06 1.94 2.04
P, = U, 0.015
P3 - Paved a 12 1.99 1.99 2.09

1.94 PI = U3 0.519
U, - Unpa T, 9 1.72 1.85

PI + PI = U2 + U3 0.024
U2 - Unpaved 10-11 3.95 3.95 3.85

U3 - Unpaved 2: 12 3.88 3.92 3.83

Model A Model B Model

Pairwise, Comparisons P-Values P-Values P-Valu

PI = P2 0.074 0.089 0.131

P2 = P3 0.802 0.866 0.872

Ul = U2 0.006 0.010 0.019

Ul = U3 0.921 0.958 0.994

P, = U, 0.517 0.353 0.427

P2 = U2 0.001 0.001 0.001

P3 = U3 0.001 0.001 0.001

 00
Table C-12 Table C-14 t')
Estimated Injury Rates as a Function of Lane and Shoulder Width Data Set Sample Sizes for Paved Roadway Sections
Lane Shoulder Primary Illinois Minnesota North Carolina
Lane Shoulder Width (feet) Width (feet) Data Data Data Data
Width (feet) Width (feet) Injury Rate KAB Rate KA Rate
0 4)
7(25.1) 0(0 0(0) 24(45.5)
1-2 9 (33. 0(0) 6(7.7) 46(140.5)
8 or less all 1.08 0.34 0.13 s 8 34 6 (16.5) 1 (6.9) 2(5.6) 234 (580.8)
s8 5-6 2(8.2) 1 (1.2) 0(0) 109(308.6)
~ 8 7+ 0(0) 10(22.1) 0(0) 9(16.8)
9 all 1.04 0.38 0.20
9 0 24 (87.0) 5 (12.7) 0(0) 109(238.1)
9 1-2 38 (134.1) 1 (3.5) 14 (27.9) 89(263.6)
10 0-4 1.56 0.74 0.35 9 3-4 97 (325.6) 13 (26.7) 23(44.8) 770 (2,082.8)
9 5-6 90(264.8) 125 (250.8) 4 (7.3) 1,080 (3,210.9)
9 7+ 25 (78.9) 137 (283.2) 8 (13.1) 158 (459.4)
10 5+ 0.89 0.33 0.20
10 0 28 (83.5) 3 (3.9) 5(14.7) 50(117.0)
11 0-2 0.96 0.72 0.38 : 0
0
10
34
'_2
5-6
52 (2
175.8)
85 47.
1)

54 (151.6)
2(7.9)
10 ( 18.6)
74 (146.5)
70(157.0)
209(471.6)
26(78.0)
46(135.5)
420 (1,217.7)
798 (2,385.6)
11 3+ 0.82 0.36 0.19
10 7+ 29(94.4) 74(154.4) 101 (303.6) 216(773.4)

0 14(48.8) 3 (5.0) 31 (68.7) 3 (5.2)

481 (142 .2)
(171.3) 2(4.3) 723 (2,009.6) 7 (28.6)
12 0-2 0.99 0.90 0.48
3-4
'_2 5 63 (137.1) 1,278 (3,618.2) 87 (244.8)
1 5-6 50(149.9)
(1 74(135.1) 214 (601.5) 160(487.5)
12 3+ 0.76 0.51 0.26 1 7+ 43 45.6) 292 (615.5) 106(299.8) 81 (309.2)

12 0 42(146.8) 9(14.2) 27(64.5) 4 (7.3)

* 13 0-4 12
1 89(302.6) 822 (2,311.2) 2 (4.3)
1.03 0.92 0.41
2 '_2
3-4 71 (244.8) 144(282.0)
- (2-1) 2, 832 (7,872.8) 36(80.4)
2 5-6 36 (114.5) 177 (378.5) 1,086 (2,919.0) 61 (160.4)
* 13 5+ 0.63 0.62 0.30 2 7+ 65 (206.0) 456 (974.6) 849 (2,137.5) 104 (349.7)
13 0 30(96.9) 2(5.0) 28 (53.9) 0(0)
k 13 1-2 8(27.8) 7 (20.7). 120(321.2) 0(0)
z 13 34 8 (25.3) 29 (73.4) 170(495.5) 2 (3.2)
z 13 5-6 8(24.0) 16 (37.4) 120(319.6) 0(0)
z 13 7+ 4 (9.6) 64 (167.2) 37 (81.4) 3 (21.1)

TOTALS 1,113 (3,581.1) 1,795 (3,790.9) 8,911 (24,258.9) 4,708 (13,678.3)

Table C-13
Estimated Injury Rates on Paved and Unpaved Roads Numbers in parentheses indicate mileage within sample.

Least-Squares Mean
Lane
Width (feet) Injury Rate
- KAB Rate KA Rate

Paved _.5 9 0.67 0.48 0.23

Paved 10-11 1.23 0.86 0.43

Paved 2~ 12 1.09 0.77 0.37

Unpaved ~g 9 0.86 0.51 0.24

Unpaved 10-11 2.11 1.43 0.52

Unpaved 12 1.97 1.33 0.47

 Table C-15 Table C-17
Illinois Least-Squares Mean Rates of Related Accidents Related Accident Rates and Total Injury Rates for Paved Minnesota Roads

Illinois Primary Data Lane Width Shoulder Width L.S.Mean Injury

Width Group (feet) (feet) Accident Rate Rate KAB Rate
L.S. Mean Rate L.S. Mean Rate
Lane Width Lane Shoulder of Related of Related 1 !~ 8 all 2.32 1.92 1.21
Group' Width (feet) Width (feet) Accidents Accidents
2 9 all 0.85 0.86 0.58
1 !g 8 all 3.57 2.60
3 10 0-2 1.03 0.88 0.64
2 9 all 1.13 2.56
4 10 3-6 0.86 0.84 0.59

3 10 0-4 2.03 3.11 5 10 2: 7 0.42 0.43 0.29

4 10 5 & over 1.02 2.16 6 11 0-2 0.67 0.65 0.44

1.12 2.28 7 11 3-6 0.63 0.63 0.45

5 11 0-2
8 11 ~: 7 0.37 0.43 0.30
6 11 3 & over 0.88 2.11
9 12 0-2 0.72 0.68 0.47
7 12 0-2 0.84 2.47
10 12 3-6 0.55 0.57 0.40
8 12 3 & over 0.85 2.04
11 12 7 0.37 0.41 0.29

9 2: 13 0-4 0.90 1.79 0-2 0.81 0.85 0.57

12 13

10 13 5 & over 0.73 1.00 13 13 3-6 0.55 0.53 0.37

'See Table C-1 14 13 2: 7 0.39 0.41 0.29

Hypothesis Tests
No Difference by Shoulder Width for Lane Width of:

10 feet F,- = 31.23 p = 0.0001

I I feet F,- = 17.35 P = 0.0001
Table C-16
12 feet F2w = 81.02 P = 0.0001
Comparison of Rates of Related Accidents from the Illinois and 13 + feet F2W = 17.75 P = 0.0001
Primary Databases on Roads with Wide Shoulders
Hypothesis Tests
Primary No Difference by Lane Width for Shoulder Width:
Shoulder Illinois-Rate of Data-Rate of
lane Width (feet) Width (feet) Related Accidents Related Accidents Narrow F,- 5.43 p = 0.0010
Medium F3W 15.79 P = 0.0001
9 5 & over 0.87 1.73 Wide F,- 0.48 p = 0.6959
10 5 & over 0.80 1.38
11 3 & over 0.66 1.31
12 3 & over 0.63 1.23
~t 13 3 & over 0.56 0.41
 00
Table C-18 Table C-19 4~
Rates of Related Accidents by Lane and Shoulder Width Total Width Effects for Unpaved Minnesota Roads
on Unpaved Minnesota Roads
L.S. Mean Rate of
Least-Squares Width Group Road Width (fe Accidents Sample Size
Lane Width Shoulder Width Rates of Related
Width Group (feet) (feet) Accidents Sample Size 1 15 18 0.74 36
1 6-7 all 0.76 67
2 20 0.57 102
2 8 0-2 0.59 71
3 22 0.98 321
3 8 2: 3 0.89 170
4 24 0.81 732
4 9 0-2 1.01 152

5 9 2: 3 0.76 389 5 26 0.63 1,167

6 10 0-2 0.76 317 6 28 0.72 1,121

7 10 :t 3 0.73 724 7 30 0.82 616

8 11 0-2 0.62 464
8 32 0.78 364
9 11 > 3 0.75 641
9 34 0.69 202
10 12 0-2 0.66 505
10 36 0.58 119
11 12 k 3 0.85 455

12 13 0-2 0.69 256 11 38 0.64 60

13 13 z3 0.78 194 '12 a 40 0.46 60

14 14 0-2 0.66 137

15 14 ;t 3 0.71 122

16 15 0-2 0.70 68

17 15 2! 3 0.52 52

18 16 0-2 0.80 74

19 16 2t 3 0.21 52
 Table C-20 Table C-22
Lane and Shoulder Width Effects North Carolina Paved Roads Mean Rates of Related Accidents
on Paved and Unpaved Minnesota Roads
for Combined Lane and Shoulder Width Categories
L.S. Mean Rate of Related
Lane Width Shoulder L.S.Mean Rate of Accidents from Minnesota Shoulder Width
Width Group (feet) Width (feet) Sample Size Related Accidents Paved Roads Only*

0.69 2.32
Lane Width (feet) 0-4 feet 1 5-6 feet 7 feet I' All
1 8 0-2 99
2 8 a 3 217 0.87 2.32
2.10 1.83 1.70 1.88
3 9 0-2 166 0.94 0.85 1.85 1.65 1.43 1.64
4 9 2:3 424 0.74 0.85
A1110 1.98 1.74 1.56
5 10 0-2 392 0.86 1.03
6 10 3-6 959 0,78 0.86
7 10 z3 101 0.41 0.42

0-2 1218 O~65 0.67

1
9 .1
11 3-6 2:133 0.63 0.63
10 11 k 7 .'6 0.36 0.37 Table C-23
11 12 0-2 1,354 0.70 0.72 Least-Squares Mean Accident Rates
12 12 3-6 4,362 0.54 0.55 on Unpaved North Carolina Roads
13 12 2: 7 860 0.36 0.37
Accident
4 >:3 0-2 683 0.73 0.81
1 _> 3 3-6 513 0.56 0.55 Lane Width (feet) Rate Number
16 513 z7 224 0.40 1 0.39 (per M`VM) of Sections
'From Table C-17.
5-6 4.40 63
7-8 2.73 44
9 2.43 40
10 3.47 25
11-12 5.84 34
Table C-21 13-16 1.20 30
Least-Squares Mean Rates of Related Accidents on North Carolina Paved Roads
by Lane and Shoulder Width Categories, Functional Class, and Terrain Accident
Rate
Shoulder (per MM
Functional Class
Lane Width (feet) Accident Rate Width (feet) Accident Rate
Collector 2.49
:58 1.95 0 1.89
Local 4.40
9 1.94 1-2 1.89
10 1.73 3-4 1.96
11 1.68 5-6 1.72
12 1.69 _~ 7 1.53

Functional Class Accident Rate Terrain Accident Rate7

Major Collector 1.45 Flat 1.62

Minor Collector 1.83 Rolling 1.81
Total 2.11 Mountainous 1.96
 Model 1 00
Database: Low-Volume Roads
Model 11 E All shoulder widths C7'
3.0 Shoulder Width Categories:
2.83
Varied M Narrow (N)
2.64
2.52 ADT Category: -~ 2,000 vpd 13 Wide (W)
2.5 2.43 2.37 2.41
2.36 2.5
> E) 2.41 2.28
2.23
W :;_ 2.14
2.0 2.1 1.87 1.92
W > 2.0
1% Z 2.03 1.85
1.87 1.91
1.83 1.49
1.5 1.45
1~ 1.59 z
1.44 1.13 W
[-. z 1.5
U 0 1.37
1.0 1.19 U
1.0
1.05

0.5
0.5

0.0
0.0
< 18 20 22 24 26 28 30 32 34 36 38 40
772:42
8 and 9 ft. 10 ft. I I and 12 ft. > 13 ft.
ROADWAY WIDTH (FT.) LANE WIDTH
(WIDTH OF LANES PLUS SHOULDERS)
Figure C-3. Rates of related accidents by lane and shoulder width from

Figure C-1. Accident rates by roadway width from two different models. the Low Volume Roads database Using ori.-inal shoulder width cate.-ories.

Database: Low-Volume Roads

Shoulder Width Categories: All shoulder widths
N 0-4 ft.
Narrow (N)
W > 5 ft.
ADT Category: -~ 2,000 vpd El Wide (W)
2.5 2.41
2.0
1.8
j 2.0
1.6
>
W 1.4
z I
w 1.2 1.5
z
1.0 Z
z Q
0.8 U
U 1.0
U W 0.6
0.4
0.2 0.5

0.0
:5 18 19 to 28 29 to 39 > 40 0.0

ROADWAY WIDTH (FT.) 8 and 9 ft. 10 ft. I I and 12 ft. > 13 ft.
(WIDTH OF LANES PLUS SHOULDERS)
LANE WIDTH

Figure C-4. Accident rate.,by lane and shoulder width

Figure C-2. Least-squares mean accident rates by roadway widths.
f1'010 the Low Volume Roads database.
 0 Paved
4.0 -
3.56 N Cross-Seciion Data Unpaved
3.5- 10 Supplemental Data
> 2.87 3.95 3.88
;E 3.0 - .80 4.0
2.68
gx,
z 2.5 2.30 111
2. 23
23 0 3.5
z W
1.97 1.99 3.0
U 2.0 -
1.56 1.51 IZ z 2.5
1.5- 30 1.34 1.39
W
L08
z 2.0
1.0-
VMS' 0.691 1.5
0.5 1.0
0.5
0.0-h
LANE WIDTH 8ft 9ft [Oft loft I I ft I I ft 12 ft 12 ft 13 ft ;> 13 ft 0.0
SHOULDER 9 10 to I 1 12
WIDTH
all all 04 ft >5ft 0-2 ft k3ft 0-2 ft ->'jft 0-4 f >5ft
LANE PLUS SHOULDER WIDTH (FT.)

Figure C-5. Rates of related accidents by ]are and shoulder width

Figure C-7. Rate of related accidents for paved versus unpaved roads by lane plus shoulder width (Model A).
for the subgroups of the primary database.

Paved
Low Volume Roads database
2.5 2.41 4.5
Cross-Section database
4.0
> 2.0 W > 3.5
1.66 1.65
- 1.54 ce 3.0
z 1.5 1.43
11.38 z z
Z .30 2.5
2.0
tj 1.0-
<m < 1.5
06.74
047 1.0
< 0.5
0.5
0.0
0.0-
250 251 to 400 > 400
LANE WIDTH 8 and 9 ft. 10 fL 10 ft. I I and 12 ft: I I and 12 ft. ;> 13 ft. > 13 ft.
SHOULDER AVERAGE DAILY TRAFFIC
WIDTH all 0-4 ft. 5 ft. 04 ft. 5 ft. 0-4 ft. > 5 ft.
(VEHICLE PER DAY)

FigLfre C-6. Comparison of accident rates by lane and shoulder width Figure C-8. Compafison of related accident rate for
00
between Low Volume Roads and Cross-Section databases. paved versus unpaved roads within ADT range. _j
 M Paved Illinois datab.,qe 00
00
EM Unpaved
4.V
Primary database
2.5

1.6
> 2.0

1.2
z 1.5

z U
0.8
1.0

0.4
0.5

0.0
0.0
LANE WIDTH 9 ft. 10 ft. 11 ft. 12 ft. > 13 ft.
9 10 to I 1 12
SHOULDER WIDTH ? 5 ft. > 5 ft. > 3 ft. a 3 ft. > 3 ft.
LANE WIDTH (FIr.)

Figure C-1 1. Comparison of least-square mean accident rates

for Illinois and primary databases for roads with wide shoulders.
Figure C-9. Injury rates on paved versus unpaved roads by lane width.

M Illinois database

1
4.0 Lane
Primary database 3.0 Width,
3.57
:5,8 ft. :
3.5 : Lane Lane Width Lane Width Lane Width Lane Width
> 3.0
3.11
2.5
:Width
2.3 :9 ft.
10 ft. 11 ft. 12 ft. 13 ft.

60 256
2.56
2.5 2.47
2.16 2.0
z
J 7
ZW 2.04
2.03
2.0
1.79
1.5
1.5

1.0
1.13
1.02 1.12
0.88
1 1/.20 70/01.1
0.84 ; - 0.85
U
U
U

1.0
1.0
VMM/ 0073777 0.9 0.8
?///, VMM/
0.5

0.0
no
ME 05 0.4
0.7 0.6

0.4
A
0.7
0.6
0.4
0.6
0.4

LANE WIDTH 8ft 9ft loft 10 1`1 11 ft I I ft 12 ft 12 fit 13 ft 13 ft

0.0
SHOULDER all all 0-4. ft > 5 ft 0-2 ft ;> 3 ft 0-2 ft >3ft 0-4 I't >5ft
WIDTH SHOULDER all all 0-2 3-6 7 0-2 3-6 7 0-2 3-6 >7 0-2 3-6 7
WIDTH (Fr.)

Figure C-10. Comparison of related accident rates for Illinois and primary databases. Figure C-12. Accident rates by lane andshoulder width for paved roads in Minnesota.
 2.0-
> 1.95 1.94
7- 1.73
W 1.68 1.69

Z
0.98 1.0-
U
0.82
0.81 0.78 0.5
0.8 0.74

0.64
0 70
0.72
2 0.0
Z 0.6 <8 9 10 11 12
0.63
0.57 0.58 LANE WIDTH (FT.)
z U
1 <
~4 0.4 0.46 ~ on I on 1.96

0.2-
1.5

0.0-
18 20 22 24 26 28 30 32 34 36 38 a40 U 1.0

ROADWAY WIDTH (Fr.)

0.5

Figure C-13. Relationship between accident rate and total roadway width for unpaved roads in Minnesota.
0.0
0 1 to 2 3 to4 5 to 6 7
. SHOULDER WIDTH (Fr.)

Figure C-14. Accident rates by lane width and shoulder width

in North Carolina.

<
,9 ft. lanes
~-- 10 ft. lanes
2.5

Z 1.5
Z

1.0

0.5

0.0
0 to 4 ft. 5 to 6 ft. Z7 ft. All Shoulder
Widths
SHOULDER WIDTH
00
Figure C- 15. Comparison of accident rate% by lane and shoulder width in North Carolina. ID
APPENDIX D

SURVEY OF CONSTRUCTION COSTS-TWO-LANE

RURAL HIGHWAYS
4. What is the intended service life of new pavement construction and resurfaced pavement?

Years for new pavement

A survey was sent to state transportation departments to gather information on highway Years for resurfaced pavement
construction costs. The responses were tabulated, and they are presented in Tables D-1
to D-7.
SHOULDER COSTS

What is a typical or average unit cost of new shoulders for highways constructed or reconstructed
in your state under typical soil and climate conditions?
TYPICAL PAVEMENT COSTS
Per square foot paved
What is atypical or average unit cost of new pavement for highways constructed or reconstructed Per square foot gravel/unpaved
in your state under typical soil and climate conditions? Per square foot turflunstabilized

Average Daily Traffic

EARTHWORK COSTS
Heavy Vehicles 0-400 vpd 401-750 vpd 751-2000 vpd
What is a typical unit cost of earthwork for construction of new highways for the following
< 10% $----jsq. ft. $-/sq. ft. $-/sq. ft. conditions (include excavation, transport, placement, and compaction)? If the size of project
> 10% $_/sq. ft. $_/sq. ft. $-/sq. ft. typically influences unit costs, please indicate this below.

Size of Project or Quantity

What is a typical or average unit cost of resurfiacing existing highways in your state under typical
soil and climate conditions? Larger Average Smaller
(- miles) 'C- miles) (- miles)
Average Daily Traffic per cu. yd per cu. yd per cu. yd
Heavy Vehicles 0-400 vpd 401-750 vpd 751-2000 vpd per cu. yd per cu. yd per cu. yd
* 10%
* 10%
$-/sq. ft.
$-/sq. ft.
$-
$-
/sq. ft.
/sq. ft.
T I
$-/sq. ft.
$-/sq. . What range of earthwork unit costs has your agency encountered? Please note the factors that
contribute to higher or lower unit costs (e.g., transport distance, type of material), and indicate
the low and high range of unit costs for such cases?
What is a typical or average unit pavement cost of widening existing highways in your state
under typical soil and climate conditions? (For example, widening a 20-ft highway to a 22- or High unit cost $ Low unit cost $
24-ft highway). Contributing Factors:

Average Daily Traffic

Heavy Vehicles 0-400 vpd 401-750 vpd 751-2000 vpd_
Please provide typical or guideline unit costs of the following other construction items.
* 10% $-/sq. ft. $--/sq. ft. $--/sq. ft. Informition on either costs per unit length or other dimensions, or as percentages of other cost
* 10% $_/sq. ft. $-/sq. ft. items may be provided based on your agency's actual experience or preliminary cost estimating
$--/sq. ft.
procedures.
ESTIMATED COST OF CONSTRUCTION

Cost per mile, station or Cost as percentage of other Our research includes evaluation of case studies of new and reconstructed highways in rural areas.
other unit dimension or construction item* Please provide one or more examples of recently reconstructed, newly constructed, and resurfaced
or 3R-type recent projects in your state. Plan, profile, cross section, and cost estimate or bid
Drainage $ %
documents; design study reports; or other supporting information would be helpful to us.
Appurtenances $ %
Please send whatever information you ran to:
Traffic control
during construction Mr. Kevin Slack
Signing, striping, Jack E. Leisch & Associates
permanent traffic control 1890 Maple Avenue
Suite 200
Evanston, Illinois 60201
Erosion control, other Phone: (708) 866-9490
environmental mitigation $ % Fax: (708) 866-6394
Landscaping, seeding,
sodding, etc. $ %
Your response by March 15 will be greatly appreciated. Please provide the name and, phone number
Utility Relocation % of an individual who can answer questions about your case studies or responses to the.survey.
Miscellaneous costs $ Contact:
Other major items (list)
Phone:

Preliminary and final

engineering %

Construction engineering %_

Legal, administrative,
other non-construction
costs

*For example, drainage as 12% of earthwork cost

 Table D-I Table D-2
Response to Question #I: Response to Question 042:
What Is a Typical or Average Unit Cost of New Pavement What Is a Typical or Average Unit Cost of Resurfacing
for Highways Constructed or Reconstructed in Your State Existing Highways in Your State Under
Under Typical Soil and Climate Conditions? Typical Soil and Climate Conditions?

Avenwe DailN. Trallic and Percent Heave Vehicle, Average DailyTrafilic and Percent I leavy Vehicles
0.400 \ pd 401-750 Td 751-2( )0 ~pd 0-400 vpd 401-7 0 v 751-2000 vpd
(per square foot) (per square foot) (per square tool) (per square foot) (per! (per square loot)
S l,ATF 10'. 101. 10*o 10o. H STATE < 10% > 10% < 10% < 10% > 1051,

ALABANIAI 2.00 N A 4.05 N-A 6.70 N/A ALABAMAI 0.30 N/A 0.00 N/A 0.00 N/A
ALABANIA2 0.61 0.61 0.61 0.89 1.01 1.01 ALABAMA2 0.23 0.23 0.23 0.23 0.23 0.23
ALASKA* ALASKA*
CALTRANS* CALTRANS-
COLORADO 0.71 0.89 0.89 0.98 1.07 1.16 COLORADO 0.08 0.08 0.08 0.27 0.27 0.36
CONNECTICUT 3.50 3.50 3.50 3.50 3.50 3.50 CONNECTICUT 0.55 0.55 0.55 0.55 0.80 0.80
IDAI-10 0.80 0.80 0.80 0.80 0.80 0.80 IDAHO 0.66 0.66 0.66 0.66 0.66 0.66
ILLINOIS 2.60 2.60 2.60 2.60 2.60 2.60 ILLINOIS 0.40 0.40 0.40 0.40 0.40 0.40
IOWA 0.83 0.96 1.39 1.80 2.07 2.12 IOWA 0.20 0.25 0.30 000 0.45 0.50
KANSAS* KANSAS'
KENTUCKY 1.06 1.26 1.22 1.47 1.80 KENTUCKY 0.21 0.31 0.21 0.31 0.21 0.31
LOUISIANA 0.82 0.85 0.82 0.85 0.82 0.85 LOUISIANA 0.51 0.55 0.51 0.55 0.51 0.55
MARYLAND 1.80 2.00 2.00 2.30 2.30 2.55 MARYLAND 0.20 0.25 0.25 0.40 0.40 0.55
I\I ININ E SOTA 0.80 0.80 0.95 1.00 1.15 1.35 MINNESOTA** 0.25 0.25 0.00 0.00 0.40 0.40
NEWHAMPSHIRE 0.67 0.67 0.67 0.67 0.67 0.67 NEW HAMPSHIRE 0.04 0.07 0.07 0.07 0.14 0.32
NEWMENICO 3.38 4.76 4.81 6.33 8.14 9.38 NEW MEXICO 0.48 3.01 4.41 5.74 6.89 7.76
NEW YORK 2.84 2.84 2.84 2.84 2.84 2.84 NEW YORK 1.76 1.76 1.76 1.76 1.76 1.76
NORTH CAROLINA 0.63 0.74 0.74 0.80 0.80 0.92 NORTH CAROLINA 0.10 0.20 0.10 0.20 0.20 0.20
NORTH DAKOTA 0.48 0.61 0.66 NORTH DAKOTA 0.00 -0.38 0.00 0.59 0.32 0.48
OHIO* OHIO*
OKLA140NIA 1.02 1.25 3.50 4.50 4.50 4.75 OKLAHOMA 0.20 0.28 0.39 0.50 0.50 0.66
RHODE ISLAND 0.20 0.20 0.20 0.20 0.20 0.20 RHODE ISLAND 0.17 0.17 0.17 0.17 0.17 0.17
SOUTH CAROLINA 4.35 4.60 4.35 4.60 4.35 4.60 SOUTH CAROLINA 0.35 0.40 0.35 0.40 0.35 0.40
SOUTH DAKOTA- 3.70 3.70 3.70- 5.07 5.07 5.07 SOUTH DAKOTA 0.51 0;51 0.51 0.51 0.51 0.51
TENN
IESSEE 2.00 2.50 2.50 3.17 2.90 4.21 TENNESSEE 0.22 0.22 0.22 0.26 0.26 0.50
TEXAS] (WEST) 0.84 0.87 0.87 0.91 0.89 0.93 TEXAS I (WEST) 0.05 0.05 0.05 0.05 0.27 0.27
TEXAS2 (EAST) 1.14 1.18 1. 18 1.21' 1.19 1.24 TEXAS2 (EAST) 0.05 0.05 0.05 0.05 0.27 0.27
VIRGINIA 0.69 0.93 0.93 1.32 1.32 1.56 VIRGINIA 0.16 0.16 0.16 0.23 0.23 0.23
WASHINGTON* WASHINGTON-
WEST VIRGINIA 1.13 1.70 1.48 2.22 1.96 2.94 WEST VIRGINIA 0.15 0.39 0.24 0.47 0.34 0.70
WISCONSIN 1.25 1 1.40 1.30 1.50 1.60 1.70 1 WISCONSIN 0.20 0.30 0.35 0.40 0.40 0.55

TOTAL 39.85 42.21 47.89 51.56 59.91 59.40 TOTAL 8.03 11.48 12.02 15.17 16.94 19.53
LOW 0.20 0.20 0.20 0.20 0.20 0.20 LOW 0.04 0.05 0.05 0.05 0.14 0.17
HIGH 4.35 4.76 4.81 6.33 8.14 9.38 HIGH 1.76 3.01 4.41 5.74 6.89 7.76
M EA N 1.53 1.69 1.92 2.15 2.40 2.38 MEAN 0.32 0.46 0.52 0.63 0.68 0.78
MEDIAN 1.04 1.18 1.30 1.50 1.60 1.70 MEDIAN 0.21 0.28 0.25 0.40 0.35 0.48

No response No response
Costs based on equivalent single axle loads (ESAL) Costs based on equivalent single axle loads (ESAL)
S3.70 - AC: S5.07 - PCC
 Table D4
Table D-3
Response to Questions #4 and 115:
Response to Question #3:
4. What Is the Intended Service Life of New Pavement
What Is a Typical or Average Unit Pavement Cost
Construction and Resurfaced Pavement?
ofWidening Existing Highways in Your State
5. What Is a Typical or Average Unit Cost ofNew, Shoulders
Under Typical Soil and Climate Conditions?
for Highways Constructed or Reconstructed in Your State
' Under Typical Soil and Climate Conditions?
Average Daily Traffic and Percent Heavy Vehicles
0-400 vpd 401-750 vpd 751-2000 vpd
Question $14 Question $15
(per square foot) (per square foot) (per square foot)
Service Life Shoulders
STATE < 101/6 > 10% < 10% 10% < 10% > 10%
(years) (per square foot)
New Resurfaced Gravel/ sTurrl
ALABAMAI 0.45 N/A 0.55 N/A 0.60 N/A Unpaved Un tabilized
STATE Pavement Pavement Paved
ALABAMA2 0.69 0.69 0.69 0.69 0.69 0.69
ALABAMAI 20.00 12.00
ALASKA* 12.00 8.00 0.33
ALABAMA2
CALTRANS'
0.98 1.07 1.16 ALASKA*
COLORADO 0.71 0.89 0.89
3.50 CALTRANS*
CONNECTICUT 3.50 3.50 3.50 3.50 3.50
COLORADO 20.00 10.00 0.89
IDAHO 0.80 0.80 0.80 0.80 0.80 0.80
CONNECTICUT (AV) 20.00 13.50 3.50
ILLINOIS 2.20 2.20 2.20 2.20 2.20 2.20
18.00 0.80 0.55
2.50 3.00 3.50 IDAHO
IOWA 2.00 2.20 2.25 1.25 0.40 0.15
ILLINOIS (AV) 20.00 8.00
KANSAS- 15.00 1.60 0.37 0.06
IOWA 30.00
KENTUCKY 1.06 1.26 1.22 1.51 1.47 1.80 0.15
20.00 10.00 1.25 0.65
1.67 1.67 1.67 KANSAS*
LOUISIANA 1.67 1.67 1.67 0.77 0.22
KENTUCKY(AV) 20.00 6.00 L15
MARYLAND 2.20 2.50 2.50 2.75 2.75 3.00
LOUISIANA (AV) 13.50 9.00 0.53 0.35
MINNESOTA* 15.00 2.50
MARYLAND 30.00
NEW HAMPSHIRE 0.89 0.89 0.89 0.89 0.89 0.89
MINNESOTA (AV) 18.00 13.50 0.60 0.55
NEW MEXICO 3.38 4.76 4.81 6.33 8.14 9.38 0.12
NEW HAMPSHIRE (AV 13.50 9.00 1.31 0.64
NEW YORK 3.55 3.55 3.55 3.55 3.55 3.55
NEW MEXICO 20.00 10.00 4.76 3.38
NORTH CAROLINA 0.65 0.78 0.78 0.82 0.82 0.94
NEW YORK 14.00 9.00 1.27
NORTH DAKOTA* 1.50 0.15
NORTH CAROLINA 20.00 10.00
OHIO* 20.00 1.44 0.85
NORTH DAKOTA 20.00
OKLAHOMA** 1.43 1.85 1.85 .2.39 2.39 3.19
0.20 0.20 0.20 0.20 0.20 OHIO*
RHODE ISLAND 0.20 1.50 1.00
OKLAHOMA 20.00 8.00
SOUTH CAROLINA 4.30 4.30 4.30 4.30 4.30 4.30
RHODE ISLAND 20.00 10.00 0.20
SOUTH DAKOTA 1.58 1.58 1.58 1.58 L58 1.58
20.00 10.00 1.00 0.10
0.71 0.67 0.89 SOUTH CAROLINA
TENNESSEE 0.52 0.60 0.60 1.75 1.37
SOUTH DAKOTA 33.00 18.00 1.88
TEXASI (WEST) 1.09 1.13 1.13 1.18 1.15 1.21 0.60
TENNESSEE 20.00 10.00 0.65 0.57
TEXAS2 (EAST) 1.48 1.53 1.53 1.58 1.55 1.61
TEXAS(AV) 15.00 10.00
VIRGINIA 1.04 1.40 1.40 1.99 1.99 2.34
VIRGINIA 10.00 10.00 0.67 0.48
WASHINGTON*
0.80 1.60 2.08 WASHINGTON*
WEST VIRGINIA 0.48 0.64 0.64
WEST VIRGINIA 20.00 5.00 0.47 0.10
WISCONSIN 1.40 1.50 1.50 1.75 1.75 2.00
20.00 15.00 1.26 1 1.58
WISCONSIN 1

TOTAL 37.26 40.41 41.02 44.66 48.32 52.47

507.00 274.00 31.98 14.47 2.77
0.20 0.20 0.20 TOTAL
LOW 0.20 0.20 0.20 0.85 0.40
AVERAGE 20.28 11.42 1 1 . 45 1
HIGH 4.30 4.76 4.81 6.33 8.14 9.38
MEAN 1.55 1.76 1.71 1.94 2.01 2.28
1.45 1.58 1.59 1.80 - No Response
MEDIAN 1.25 1.50
(AV) - Average of Range for Service Lives

No response
Values based on less than 20 %and greater than 20 %heavy vehicles
 Table D-5
~0
4~,
Response to Question #6:
What Is a Typical Unit Cost of Earthwork
for Construction of New Highways for the Following Conditions
(Include Excavation, Transport, Placement, and Compaction)?

Quantity Cost .
Larger Average Smaller
(per cubic yard [CYI) (per cubic yard [CY]) (per cubic yard [CY])
STATE Larger Average Smalle Low High Low High Low High
ALABAMAI I mil CY 1045k CY < IkCY 1.45 2.00 6.75
ALABAMA2 6.50
ALASKA*
CALTRANS*
COLORADO 5 miles 2 miles I milt 1.50 3.50 2.50 4.50 3.50 5.50
CONNECTICUT 1. 15 miles 0.04 mile 0.66 mile 4.00 5.45 2.00 8.50 3.50 7.00
IDAHO 1.40 1.69 8.00
ILLINOIS 3.36
IOWA 10 miles 5 miles I milt 0.90 1.29 1.50
KANSAS 10 miles 5 miles I mile 2.30 4.00 7.00
KENTUCKY >1 mil CY Imil-100kCY 100k-lk CY 2.07 2.77 6.19
LOUISIANA 10 miles 5 miles 0-1 mile 5.00 6.00 8.00
MARYLAND 200k CY 75k CY 1.5k CY 3.00 8.00 12.00
MINNESOTA*
NEW HAMPSHIRE 4 miles 1.5 miles 0.25 mile 3.00 3.75 3.75 4.25 4.25 5.00
NEW MEXICO 10 miles 5 miles <1 Mille 1.70 1.80 2.16 2.54 10.00 12.00
NEW YORK*** 8.30 39.50 9.45 48.92 10.57 58.33
NORTH CAROLINA 7 miles 3 miles <0.5 mile 1.00 3.00 2.00 4.00 4.00 7.00
NORTH DAKOTA 20+ miles 10 miles <1 mile 0.75 0.80 1.50
OHIO*
OKLAHOMA** 5+ miles 1-5 miles <1 mil 2.90 3.50 4.50
RHODE ISLAND 2.00 3.50 5.00
SOUTH CAROLINA >75k CY 75k- Ik CY <IkC 2.75 3.50 5.00
SOUTH DAKOTA 10 miles 7 miles 0.73 0.80 2.08
TENNESSEE 9 miles 3 miles I mi1 0.90 2.10 1.15 2.35 3.75 4.50
TEXAS*
VIRGINIA >50k CY 50k-10k CY <10kC 3.00 6.00 3.50 8.50 4.50 10.50
WASHINGTON 1.20 3.00 5.00 10.00 7.00 15.00
WEST VIRGINIA 2.78 mi:esl 1.02 miles 0.28 mi 3.00 02 8.25
WISCONSIN >5 mi es I- <1 mil 1.37 1.43 1 .88 3.10 1.96 3.43
I

No response *** Project locale has greater influence than project size
Values based on less than 20 %, greater than 20 % heavy vehicles CY = cubic yard
 Table D-6
Response to Question #7:
What Range of Earthwork Unit Costs Has Your Agency Encountered?

Earthwork
(Unit Costs)
State Low High Contributing Factors
ALABAMAI 1.00 10.00 Quantity, rock (type material).
ALABAMA2 0.50 30.00
ALASKA*
CALTRANS*
COLORADO 1.00 10.00 Terrain, altitude, rolling, moutain & plain,size, traffic, etc.
CONNECTICUT 2.00 7.50 Construction competition and quantity, haul distance.
IDAHO 0.50 20.00 Low unit cost is due to large quantities.
ILLINOIS 1.96 20.76
IOWA 0.52 18.00 Unit costs of $5418, when contractor furnished borrow and haul.
KANSAS 1.50 10.00 Length of project, location.
KENTUCKY 1.50 20.00 Type of rock, haul dist., cost of waste areas, geo. features, etc.
LOUISIANA 2.00 10.00 Length of project, land use, local condition, etc.
MARYLAND 2.50 25.00 Historic-based on country and quantity volume.
MfNNESOTA*
NEW HAMPSHIRE 2.50 6.25 Location of project, material avail/disposal site, urban vs. rural.
NEW MEXICO 1.80 12.00 Type of material, haul distance.
NEW YORK 2.50 327.80 Costs for hauling and/or disposal in NYC and on L.I. drive costs up.
NORTH CAROLINA 1.00 15.00 Urban sites, soil types and price-quantity relationships.
NORTH DAKOTA 0.73 4.00 Transport distance, wet conditions, small quantity.
OHIO*
OKLAHOMA" 2.00 10.00 Size and location of project, availability and type of material.
RHODE ISLAND 0.10 10.00 Size of project, haul dist., availability of material, etc.
SOUTH CAROLINA 1.85 10.00 Structure excavation, culverts, conc.footings, haul dist., etc.
SOUTH DAKOTA 5.40 Rock excavation.
TENNESSEE 0.85 5.50 Rock vs. Earth, Rural vs. Urban, New Location vs. Renovation,etc.
TEXAS*
VIRGINIA 1.67 12.50 Quantity of excavation, location of work, bidding competition.
WASHINGTON 1.20 15.00 Amount and type of material.
WEST VIRGINIA 0.01 100.00 Quantity, percent of rock, waste site distance, geographic location.
WISCONSIN 0.85 15.00 Location and size of project.

No response
Values based on less than 20 % and greater than 20 % heavy vehicles
 Table D-7 \D
Cr\
Response to Question #8:
Please Provide Typical or Guideline Unit Costs of the Following Other Construction Items

ESTIMATED COST OF CONSTRUCTION PER MILE, STATION, OR OTHER UNIT DIMENSION

TRAFFIC SIGN, STRIPE, ROSION CONTRO LANDSCAPING,

CONTROL PERMANANT OTHER ENVIR. SEEDING, UTILITY
STATE DRAINAGE APPURTENANCE URING CONSTR TRAF.CONTROL MITIGATION SODDING, etc. RELOCATION
ALABAMAI $150,000 $108,000 $28,000 $30,000 $9,500 $40,000
ALASKA*
CALTRANS
COLORADO $17,900 $4,135 $4,000
CONNECTICUT 1.50/s.f. paved $100,000 $100,000
IDAHO $662/11-mile $720/11-mile $700/11-mile $25,000
IOWA $4.00/L.F. $200 $500/acre
KANSAS $40,000 $2,300/acre $4,000 $15,000 $500/acre $1000/acre $18,000
KENTUCKY $200,000
LOUISIANA $80/ft $1700/each $4,200 $3,200 $9.00/lb $10.00/lb
MARYLAND
MfNNESOTA*
NEW HAMPSHIRE $15,000 $3,000
NEW MEXICO $12,000 $9,000 $1,500 $802/acre
NEW YORK $42,000 $44,000 $1,800 $1,800 $5,000 $200
NORTH CAROLINA $120,000 $8,000 $2,000 $14,000 $8,000
NORTH DAKOTA $18,000 $1,500 $1.15/sq.yd. $180/acre
OHIO*
OKLAHOMA $192,290 $109,762 $12,000 $7,626 $71,157
RHODE ISLAND
SOUTH CAROLINA $21,000 $1,500 $9,000
SOUTH DAKOTA $19,300 $400/acre $7,500 $750 $10,850 $3,880 $33,850
TENNESSEE
TEXAS*
VIRGINIA $16,400 $5,000 $34,506 $7,650 $67,500
WASHINGTON 1.50-2.00/L.F.
WEST VIRGINIA $442,380 $220,724 $238,000 $101,000 $30,000 $23,000 $30/L.F.
WISCONSIN $11,000 $3,100 $1,300 $1,885 $17,200
TOTAL $643,490 $474,486 $428,500 $174,711 $195,698 $87,230 $484,5 0
MEAN $58,499 $118,622 $32,962 $13,439 $21,744 $9,692 $60,5 69
MEDIAN $21,000 $104,881 $8,000 $3,300 $14,000 $8,000 $36,925

* No response
 Table D-7 (continued)

OTHER:
GRUB AND BASE, PRELIMINARY LEGAL, ADMIN,
misc. CLEAR, SEAL COAT, AND FINAL ONSTRUCTIO OTHER NON-
COSTS GRADING BORROW STRUCTURE SURFACING ENGINEERING ENGINEERING CONST. COSTS
STATE
ALABAMAI
ALASKA*
CALTRANS
COLORADO $8,500
CONNECTICUT $415,000 $7.00/cy $175/sf Deck $7000/contract
IDAHO $1260/ft-mile $6000/ft-mile $8570/ft-mile $1230/ft-mile $3840/ft-mile
IOWA
KANSAS
KENTUCKY
LOUISIANA $2000/each $2000/acre
MARYLAND
MrNNESOTA*
NEW HAMPSHIRE
NEW MEXICO $4,800 $20,000

NEW YORK $400

NORTH CAROLINA
NORTH DAKOTA
OHIO*
OKLAHOMA $64,618 $415,448 $100,337
RHODE ISLAND -
SOUTH CAROLINA $10,500 $96,000
SOUTH DAKOTA $42.00/s.f.
TENNESSEE
TEXAS*
VIRGINIA
WASHINGTON
WEST VIRGINIA $91,000/cntrct $173,000 $342,000
WISCONSIN 1 $54,000 1 $24,500
TOTAL $480,018 $192,000 $0 $0 $853,448 $154,337 $29,300 so
MEAN $160,006 $64,000 $0 so $284,483 $77,169 $14,650 $0
MEDIAN $64,6 $10,500 $342,000 $77,169 $14,650

* No response
 Table D-7(continued) ~c
00

ESTIMATED COST OF CONSTRUCTION AS A PERCENTAGE OF OTHER CONSTRUCTION ITEMS

TRAFFIC SIGN, STRIPE, I ROSION CONTRO LANDSCAPING,

CONTROL PERMANANT OTHER ENVIR. SEEDING, I REUTILITY MISC
STATE DRAINAGE PPURTENANCE IDURING CONSTR ITRAF. CONTROL MITIGATION I SODDING, etc. LOCATION I COST!
ALABAMAI 1 9.7 % of TC I 7 % of TC 1 1.7 % of TC 17C I I I
ALABAMA2
ALASKA*
CALTRANS
COLORADO 4-7 % of TC I I%ofTC
CONNECTICUT 5 % of CI I %of Cl 0.5 % of CI 0.5 % of Cl
IDAHO I % of TC I %ofTC
ILLINOIS 6.5 % of TC 8.3 % of TC 2.7 % of TC 0.7 % of TC 1.2 % of TC 7.4 % of TC
IOWA 4 % of TC 5%ofTC
KANSAS 10.0 % of TC
KENTUCKY 10%ofEW 5 % of EW 2 % of EW I % of EW I % of EW I % of EW 10.0 % of EW
LOUISIANA 15 % of TC
MARYLAND 20 % of All 2 % of All 2 % of All 8%ofAll 3 % of All 1-3 % of All
MINNESOTA*
NEW HAMPSHIRE 15 % of CI lo%ofcl 2.5 % of Cl I0%ofCI
NEW MEXICO
NEW YORK 25 %of PC
NORTH CAROLINA 15 % of Cl
NORTH DAKOTA 25 % of EX 2 % of EX 4 % of EX 6 % of EX
OHIO*
OKLAHOMA 19 % of TC 11 %ofTC 1.2 % of TC I %ofTC 7 % of TC 6.5 OXo of TC
RHODE ISLAND
SOUTH CAROLIN 14 % of TC I %of TC 6 % of TC
SOUTH DAKOTA
TENNESSEE 2 % of TC 4 % of TC 0.5 % of TC 0.05 % ofTC 0.005 % of TC 0.008 % of TC arying 6/o of T 3 % of TC
TEXAS*
VIRGINIA 22 % of TC
WASHINGTON 2-10 % of TC 1-3 % of TC I %ofTC
WEST VIRGINIA
WISCONSIN

PC = Pavement Cost
EW= Earthwork Cost
EX = Excavation Cost
 Table D-7(continued)

OT14F.R-
GRUBAND BASE, PRELIMINARY LEGAL, ADMIN,
CLEAR, CONCRETE SEAL COAT, AND FINAL ONSTRUCTIO OTHER NON-
GRADING SURFACES STRUCTURES SURFACING OBILIZATIO ENGINEERING ENGINEERING CONST. COSTS
7.5 % of TC 13 % of TC
8.0 % of TC 7 % of TC

14 % of TC 10%ofTC 20 % of Eng
10 % of TC 15.5 % of TC
7.5-20 % of Cl 12 % Bid Amt 2%ofCI

11 %ofTC 5.2 % of TC 20.6 % of TC 21.2 % of TC 5 % of TC 10%ofTC

2-3 % of TC 2-6 % of TC
6%ofyC 7.5 % of TC
10%ofTC 10%ofTC .5 % of TC
2.5 % of TC
5.0 % of TC

7-10 % of CI 10%ofTC 10-12 % of CI 20 % of Pre Eng

10 % of Pvt 12 % of Pvt 23 % of Pvt 3 % of Pvt

10 % of TC 10 % of TC 5 % of TC

41 % of TC 3.5 % of TC 10 % of TC
15 % of TC
7 % of TC 64 % of TC 7.5 % of TC 15 % of TC
1-7.5 % of TC 10 % of TC
10%ofTC 1.5 % of TC

8-10 % of TC 8-10 % of TC 2 % of TC
10-15 % of TC 5 % of TC

~0
110
APPENDIX E
8
CD

DERIVATION OF CONSISTENCY BETWEEN LANE WIDTH,

SHOULDER WIDTH, FUNCTIONAL CLASS, AND VOLUME AASHTO Policy recommends minimum levels of service as a function of terrain and functional
classification. Values for the volume to capacity ratio, v1c, are a function of terrain, percent
ANALYSIS APPLICATION
no-passing zones, and level of service. For analysis purposes, level terrain is assumed to
produce, on average, 20 percent no-passing zones; rolling terrain 40 percent no-passing zones;
Design standards for lane width and shoulder width should be consistent with the basic and mountainous terrain 60 percent no-passing zones. Values for (v1c), can thus be obtained
operational requirements that they affect. The most basic of these is the relationship between from the Highway Capacity Manual (see Table E-1) . A value for fd was obtained from the
lane width and shoulder width, and capacity. Highway Capacity Manual, assuming a 60/40 distributional split. Values for fH, were
calculated using heavy-vehicle passenger car equivalents specified in the Highway Capacity
For two-lane rural highways, capacity effects are as described in the 1985 Highway Manual, assuming two ranges of truck percentages-10 percent and 15 percent trucks.
Capacity Manual published by the Transportation Research Board. The general relationship is:
A final analysis step involves conversion of daily traffic volumes (the basis for width
design standards) to hourly volumes, on which capacity calculations are based. Assumptions
SFi = 2,800 X (vlc)i X fd X f~ X fHV (FI) for "PHF" (peak hour factor) and "K" (design percentage of ADT) were derived from the
literature.
where:
Table E-4 summarizes all traffic and design assumptions used in the analysis.
SFi total service flow rate in both directions for prevailing roadway and
traffic conditions, for level-of-service i, in vehicles per hour;
ANALYSIS RESULTS
(VIC)i ratio of flow rate to ideal capacity for. level-of-service i, obtained from
Table E-1; The analysis produces values for f. as a function of traffic volume for the full range of
design conditions studied. The f. values represent minimums, i.e., values at which the
fd adjustment factor for directional distribution of traffic, obtained from combined effects of lane and shoulder width produce at least the desired level of service.
Table E-2; What actually is produced is a set of linear relationships between f. and traffic volume.

f. adjustment factor for narrow lanes and restricted shoulder width, Table E-5 gives the calculated values of f. for the example case for 10- and 15-percent
obtained from Table E-3; trucks, respectively. For reference purposes, the reader is referred to Table E-3, which
expresses f. as a function of lane width and shoulder width.
fW adjustment factor for the presence of heavy vehicles in the traffic
stream, computed as: Note that, for very low traffic volumes (less than 750 vpd), f. is very low—lower, in
fact, than the corresponding value for 9-ft lanes and 0-ft shoulders. This confirms that, under
I/[I + P,(E,- 1) + P,(E,- 1) + P,(E,- 1)]. low-traffic-volume conditions, the capacity effects of lane and shoulder width are not a factor
in decisions regarding design widths.
The relationship in Eq. E-I expresses the service flow rate for any given level of service
as a function of the geometric and traffic viariables that influence it. Among these are the As design traffic increases, however, particularly in more difficult terrain, and where
combined effect of lane and shoulder width, f., as summarized in Table E-3. Normal use of trucks are a greater proportion of the traffic stream, values for f. begin to reach limiting
the equation is to solve for a service flow rate, thereby determining level of service. values. Consider, for example, Table E-5, 15 percent trucks with a design traffic of 1,500 -
vpd. For an arterial in rolling terrain, for lane and shouldd width to produce a minimally
One can use Eq. E-I in a more unconventional manner to test the extent to which design consistent operation in terms of capacity, a value of 0.78 or greater for f. is necessary.
standards for lane width and shoulder width are consistent with other design values and Referfing to Table E-3, this translates to either a 10-ft lane with 5-ft shoulders (interpolating to
guidelines. Consider the following example: a value of 0.80 for f.) or an I I-ft lane width with a 3-ft shoulder (again interpolating to a
value of 0.80 for f.). For mountainous terrain, f. is 0.93; indicating a need for 1 I-ft lanes
and 6-ft shoulders, or 12-ft lanes and 4-ft shoulders.
sf~
f. - [2800 x (v1c), x fd x fHvl
Tables E-6 and E-7 summarize the minimum lane and shoulder width combinations to
produce operationally consistent conditions for design. These are used, in combination with
One can solve for f., given assumed values for the other variables, relating to the full the other research results, to produce recommended minimum design values for lanes and
range of design conditions. Foi SFi, it is necessary to establish a level of service. The 1990 shoulders on low-volume, two-lane rural highways.
 Table E-1
Level of Service for General Two-Lane Highway Segments

VIC RATIO'

LEVEL TERRAIN ROLLING TERRAIN MOU14TAINOUS TERRAIN

PERCENT PERCENT NO PASSING ZONES b PERCENT NO PASSING ZONES

b PERCENT NO PASSING ZONES AVG
b AVG
TIME AVG
DELAY SPEED 0 20 40 60 80 100 SPEED 0 20 40 60 80 100 SPEED 0 .20 40 60 80 100
LOS

> 58 0.15 0.12 0.09 0.07 0.05 0.04 > 57 0.15 0.10 0.07 0.05 0.04 0.03 > 56 0.14 0.09 0.07 0.04 0.02 0.01
A .5 30
15 45 > 55 0.27 0.24 0.21 0.19 0.17 0.16 > 54 0.26 0.23 0.19 0.17 0.15 0.13 > 54 0.25 0.20 0.16 0.13 0.12 0.10
B
5 60 > 52 0.43 0.39 0.36 0.34 0.33 0.32 > 51 0.42 0.39 0.35 0.32 0.30 0.28 >: 49 0.39 0.33 0.28 0.23 0.20 0.16
C
D < 75 > 50 0.64 0.62 0.60 0.59 0.58 0.57 > 49 0.62 0.57 0.52 0.48 0.46 0.43 > 45 0.58 0.50 0.45 0.40 0.37 0.33
~! 45 .1 1.00 1.00 1.00 1.00 1.00 1.00 > 40 0.97 0.94 0.92 0.91 0.90 0.90 > 35 0.91 0.87 0.84 0.82 0.80 0.78
E > 75
100 < 45 - - - - - - < 40 - - - - - - < 35 - - - - - -
F

' Ratio of flow rate to an ideal capacity of 2,800 pcph in both directions.
b These speeds are provided for information only and apply to roads with design speeds of 60 mph or higher

Source: Transportation Research Board. "Highway Capacity Manual," Special Report 209, 1985.

8
102

Table E-2
Adjustment Factors for Directional Distribution on General Terrain Segments

Directional Distribution 100/0 90/10 80/20 70/30 60/40 50150

Adjustment Factor, fd 0.71 0.75 0.83 0.89 0.94 1.00
Source: Transportation Research Board, "Highway Capacity Manual," Special Report 209, 1985.

Table E-3
Adjustment Factors for the Combined Effect of Narrow Lanes
and Restricted Shoulder Width (fw)

12-foot Lanes 11-foot Lanes 10-foot Lanes 9-foot Lanes

Usable
Shoulder LOS LOS LOS LOS LOS LOS LOS* LOS
Width" (feet) 1 A-D EbI I A-D Eb A-D e I A-D e I
~~- 6 1.00 1.00 0.93 0.94 0.84 0.87 0.70 0.76
4 0.92 0.97 0 * 85 0.92 0.77 0.85 0.65 0.74
2 0.81 0.93 0.75 0.88 0.68 0.81 0.57 0.70
0 0.70 0.88 0.65 0.82 0.58 0.75 0.49 0.66

'Where shoulder width is different on each side of the roadway, use the average shoulder width.
bFactor applies for all speeds less than 45 mph.

Source: Transportation Research Board, "Highway Capacity Manual," Special Report 209, 1985.
 103

Table E4
Analysis Assumptions for Evaluation of
Capacity Effects on Lane and Shoulder Widths
FbVd
Rural Highway LOSa Percent Truck
Type/Terrain No-Passingb V/Cc Fdc pCEc 10% Trucks F~rucks PHF KDES b
Arterial
-Level B 20 0.24 0.94 2.2 0.893 0.847 0.92 0.15
-Rolling B 40 0.19 0.94 5.0 0.714 0.625 0.92 0.15
- Mountainous C 60 0.23 0.94 10.0 0.526 0.426 0.94 0.15
-
Collector
Level C 20 0.39 0.94 2.2 0.893 0.847 0.94 0.15
- Rolling C 40 0.35 0.94 5.0 0.714 0.625 0.94 0.15
- Mountainous D 60 0.40 0.94 12.0 0.476 0.377 0.95 0.15

Local
-Leve D 20 0.62 0.94 2.2 0.909 0.870 0.95 0.15
-Rolling D 40 0.52 0.94 5.0 0.714 0.625 0.95 0.15
- Mountainous D 60 0.40 0.94 12.0 0.476 0.377 0.95 0.15

'Source: AASHTO, A Policy on Geometric Design of Highways and Streets, 1990.

bnese values are assumed.
"Source: Transportation Research Board, "Highway Capacity Manual," Special Report 209, 1985.
d
These values are calculated.

Table E-5
Minimum Values for Capacity Factor F. to Accomodate
Typical Design Assumptions for Two-Lane, Rural, Low-Volume Roads

Capacity Factor (f.) for Average Daily Traffic Volumes

400 vpd 750 vpd 1,500 vpd 2,000 vpd

Rural Highway
Type/Terrain 10% 15% 10% TZZ 7 -
15 . Trucks 10% 15% 10% Trucks
TrucksTTrucks Trucks I Trucks I

Arterial
-Level 0.12 0.12 0.22 0.23 0.43 0.46 0.58 0.61
-Rolling 0.18 0.21 0.34 0.39 0.68 0.78 0.91 1.04
- Mountainous 0.20 0.25 0.38 0.46 0.75 0.93 1.00 1.24

Collector
- Level 0.07 0.07 0.13 0.14 0.26 0.28 0.35 0.37
-Rolling 0.10 0.11 0.18 0.21 0.36 0.42 0.49 0.55
- Mountainous 0.13 0.16 0.24 0.30 0.47 0.60 0.63 0.79
1

Local
Leve 0.04 0.04 0.08 0.08 0.16 0.17 0.21 0.22
Rolling 0.06 0.07 0.12 0.14 0.24 0.28 0.32 0.37
Mountainous 0.13 0.16 0.24 0.30 0.47 0.60 0.63 0.79
 Table E-6 Table F,7
Minimum Lane and Shoulder Width Values to Produce Consistency Minimum Lane and Shoulder Width Values to Produce Consistency
in Design Capacity ror Low-Volume Roads (Trucks :!r 1 0% of ADT)' in Design Capacity for Low-Volume Roads (Trucks > io% or ADT)'

Design ADT Design ADT

1500 vpd 2000 vpd 1500 vpd 2000 vpd

Highway Type b
Min~i~
mumLane Width/Shoulder Mini.. Lane Width/Shoulder Width Highway Type Lane Width/Sh.ulder Lane Width/Shoulder Width
Width (feet) f.b (feet) fb
f~ Width (feet) (feet)
L
Arterial Arterial

-Level 0.43 9/0 0.58 9/3 -Level 0.46 9/0 0.61 9/3
10/0 10/1
9/6 1015
-Rolling 0.68 10/2 0.91 11/6 -Rolling 0.78 11/3 1.04~ 12/6
1 1/1 12/4 12/2

10/4 -Mountainous 0.93 11/6 1.241 12/6

-Mountainous 0.75 11/2 1.00 12/6 12/5
12/1
Collector
Collector
-Level 0.28 9/0 0.37 9/0
-Level 0.26 9/0 0.35 9/0
-Rolling 0.42 9/0 0.55 9/12
-Rolling 0.36 9/0 0.49 9/0 10/0

9/4 10/5
-Mountainous 0.47 9/0 0.63 10/1 -Mountainous 0.60 9/3 0.79 1113
11/0 10/1 12/2

Local Local

-Level 0.16 9/0 0.21 9/0 -Level 0.17 9/0 0.22 9/0

-Rolling 0.24 9/0 0.32 9/0 -Rolling 0.28 9/0 0.37 9/0

9/4 -Mountainous 0.60 9/3 0.79 1015

-Mountainous 0.47 9/0 0.63 10/1 10/1 11/3
11/0 12/2

'Table computed assuming trucks are 10 percent of ADT 'Table computed assuming trucks are 15 percent of ADT
'From Table E-5 'From Table E-5
Waximum lane and shoulder width values do not provide sufficient capacity for assumed conditions
APPENDIX F
GUIDELINES FOR DERIVATION OF CONSISTENCY
BETWEEN DESIGN SPEED AND LANE WIDTH

Figure F- I is derived from research by Lamm and Choueiri on the effects of alignment DESIGN SPEEDS OF 40 MPH
and cross section on speeds of vehicles on two-lane rural highways (12). It is used to assist in
establishing a consistency between lane width design standards and design speed. Ten-foot lane widths produce predicted speeds within the 5-mph criterion for up to
Superimposed on Figure F-I are data points describing maximum horizontal curves for design about 80 percent of controlling curvature. Eleven-foot lane widths produce 45-mph predicted
speeds of 30, 40, 50, and 60 mph for AASHTO e_, values of 0.08. speeds at controlling curvature only; and higher speeds (above the criterion) at all milder
curves.
Wherever the predicted 85th-percentile speed from Lamm is significantly above the
value for design speed, the corresponding lane width value is considered to potentially The following conclusions are noted: - (1) 12-ft lanes are inappropriate, in that they
contribute to "over driving" of the curve, and is therefore termed "inconsistent." encourage speeds much higher than 40 mph; (2) 1 I-ft lanes are appropriate only where terrain
and geometry are such that controlling curvature is frequent and continuous, thereby keeping
In general, it is desirable that predicted 85th-percentile speeds be as close to the design speeds within an acceptable range; (3) 10-ft lanes are appropriate; and (4) 9-11 lanes are
*
speed as possible. A suggested criterion is that predicted speeds be within 5 mph of the design appropriate -(extrapolating).
speed. The following discussion summarizes design speed and lane width considerations for
60-, 50-, 40-, and 30-mph design speeds.
DESIGN SPEEDS OF 30 MPH

DESIGN SPEEDS OF 60 MPH . Very low design speeds of 30 mph tend to be used in severely rolling or mountainous
terrain. Controlling curvature is frequent. Ten-foot lane widths produce predicted speeds
Ten- and I I-ft lane widths result in predicted speeds less than '55 mph for controlling within the 5-mph criterion for up to about 80 percent of controlling curvature. Eleven-foot
(i.e., maximum for desigri~speed) curves. For curves flatter than controlling curves, 11-ft lane lane widths produce speeds within the criterion only at controlling curves.
widths produce speeds'within the 5-mph criterion.
.
The following conclusions are noted: (1) 12-ft lanes are inappropriate; (2)' 1 1-ft lanes
The following conclusions are noted: (1) 12-ft lanes are appropriate; (2) Il-ft lanes *are appropriate only where terrain and geometry are such that controlling curvature is frequent
may be appropriate in cases where controlling curvature is infrequent (i.e., level terrain); and and continuous; (3) 10-ft lanes are appropriate; and (4) 9-ft lanes are appropriate
(3) lanes of 10-ft width and narrower are inappropriate- for a 60-mph design speed. (extrapolating).

DESIGN SPEEDS OF 50 MPH

All lane widths are associated with predicted speeds well within the 5-mph criterion at
the controlling curvature for 50 mph. Ten-foot lane widths produce predicted speeds within
5 mph for all curvature. Eleven- and 12-ft lane widths produce predicted speeds greater than
55 mph at about 50 to 65 percent of controlling curvature.

The following conclusions are noted: (1) 12-ft lanes are appropriate except in cases
where horizontal alignment is mild (i.e., level terrain)—in such terrain, speeds too high may be
encouraged; (2) 11-ft lanes are appropriate in essentially 0 cases; and (3) 10-ft lanes are
appropriate in all cases.
 70 8
CN

Kc

CL

50

w
CL
cn 40

z 30
w
0

CL
20

Lo
co
10

DEGREE OF CURVE

F'igLli- e F-1. Relationship between 85th percentile vehicle speed and horizontal CUrvatUre.
THE TRANSPORTATION RESEARCH BOARD is a unit of the National Research Coun-
cil, which serves the National Academy of'Sciences and the National Academy of Engineering. It
evolved in 1974 from the Highway Research Board which was established in 1920. The TRB
incorporates all former HRB activities and also performs additional functions under a broader scope
involving all modes of transportation and the interactions of transportation with society. The Board's
purpose is to stimulate research concerning the nature and performance of transportation systems,
to disseminate information that the research produces,
' and to encourage the application of appropriate
research findings. The Board's program is carried out by more than 270 committees, task forces,
and panels composed of more than 3,300 administrators, engineers, social scientists, attorneys,
educators, and others concerned with transpqrtation; they serve without compensation. The program
is supported by state transportation and highway departments, the modal administrations of the U.S.
Department of Transportation, the Association of American Railroads, the National Highway Traffic
Safety Administration, and other organizations and individuals interested in the development of
transportation.
The National Academy of Sciences is a private, nonprofit, self-perpetuating society of distin-
guished scholars engaged in scientific and engineering research, dedicated to the furtherance of
science and technology and to their use for the general welfare. Upon the authority of the charter
granted to it by the Congress in 1863, the Academy
' has a mandate that requires it to advise the
federal government on scientific and technical matters. Dr. Bruce M. Alberts is president of the
National Academy of Sciences.
The National Academy of Engineering was established in 1964, under the charter of the National
Academy of Sciences, as a parallel organization of outstanding engineers. It is autonomous in its
administration and in the selection of its members, sharing with the National Academy of Sciences
the responsibility for advising the federal government. The National Academy of Engineering also
sponsors engineering programs aimed at meeting national needs, encourages education and research
and recognizes the superior achievements of engineers. Dr. Robert M. White is president of the
National Academy of Engineering.
The Institute of Medicine was established in 1970 by the National Academy of.Sciences to s~cure
the services of eminent members of appropriate professions in the examination of policy matters
pertaining to the heaith of the public. The Institute acts under the responsibility given to the National
Academy of Sciences by its congressional charter to be an adviser to the federal government and,
upon its own initiative, to identify issues of medical care, research, and education. Dr. Kenneth I.
,Shine is president of the Institute of Medicine.

The National Research Council was organized by the National Academy of Sciences in 1916,to
associate the broad community of science and technology with the Academy's purpose of furthering
-knowledge and advising the federal government. Functioning in accordance with general policies
determined by the Academy, the Council has become the principal operating agency of both the
National Academy of Sciences and the National Academy of Engineering in providing services to the
government, the public, and the scientific and engineering communities. The Council is administered
jointly by both Academies and the Institute of Medicine. Dr. Bruce M. Alberts and Dr. Robert M.
White are chairman and vice chairman, respectively, of the National Research Council.
ink

IT sky
Eton -7-