A BENEFIT/COST ANALYSIS OF PAVED SHOULDER INSTALLATION ON HIGH 
PRIORITY ROAD SEGMENTS ON TWO-LANE RURAL HIGHWAYS IN 
ALABAMA 
 
 
 
 
Except where reference is made to the work of others, the work described in this thesis is 
my own or was done in collaboration with my advisory committee.  This thesis does not 
include proprietary or classified information. 
 
 
 
 
 
_________________________ 
Renee? Danielle Ayala 
 
 
Certificate of Approval: 
 
 
 
 
_______________________     ________________________ 
Robert L. Vecellio      Rod E. Turochy, Chair 
Associate Professor      Associate Professor 
Civil Engineering      Civil Engineering 
 
 
 
 
 
_______________________     ________________________ 
David H. Timm      George T. Flowers 
Associate Professor      Interim Dean 
Civil Engineering      Graduate School 
 
A BENEFIT/COST ANALYSIS OF PAVED SHOULDER INSTALLATION ON HIGH 
PRIORITY ROAD SEGMENTS ON TWO-LANE RURAL HIGHWAYS IN 
ALABAMA 
Renee? Danielle Ayala 
 
A Thesis 
Submitted to 
the Graduate Faculty of 
Auburn University 
in Partial Fulfillment of the  
Requirements for the 
Degree of  
Master of Science 
 
 
Auburn, Alabama 
August 9, 2008 
 
 
iii 
A BENEFIT/COST ANALYSIS OF PAVED SHOULDER INSTALLATION ON HIGH 
PRIORITY ROAD SEGMENTS ON TWO-LANE RURAL HIGHWAYS IN 
ALABAMA 
 
 
Renee? Danielle Ayala 
 
Permission is granted to Auburn University to make copies of this thesis at its discretion, 
upon requests of individuals or institutions and at their expense.  The author reserves all 
publication rights. 
 
 
 
 
_________________________ 
     Renee? Danielle Ayala 
 
 
_________________________ 
       Date of Graduation 
 
 
 
iv 
VITA 
Renee? Danielle (Cunningham) Ayala, daughter of Kenneth Allen and Adele 
(Hurtado) Cunningham, was born September 17, 1979, in Escondido, California.  She 
graduated from Orange Glen High School in June, 1997.  She joined the U.S. Army in 
August, 1997, and served at the U.S. Army Marksmanship Unit until August, 2001.  She 
married Armando Ayala, son of Ernesto and Eva (Urias) Ayala, on June 7, 1999.  She 
attended Columbus State University, Georgia, for one year, then entered Auburn 
University in August, 2002, and graduated magna cum laude with a Bachelor of Science 
degree in Biosystems Engineering in May, 2006.  As an undergraduate student she 
participated in the Co-operative Education Program, Auburn University, where she 
worked as an Engineering Aide for the U.S.  Forest Service, Auburn, Alabama.  She 
entered Graduate School, Auburn University, in May, 2006.  Her daughter Mila Cecilia 
Ayala was born September 16, 2006 at Fort Benning, Georgia. 
 
v 
THESIS ABSTRACT 
A BENEFIT/COST ANALYSIS OF PAVED SHOULDER INSTALLATION ON HIGH 
PRIORITY ROAD SEGMENTS ON TWO-LANE RURAL HIGHWAYS IN 
ALABAMA 
 
Renee? Danielle Ayala 
 
Master of Science, August 9, 2008 
(B.B.S.E., Auburn University, 2006) 
 
89 Typed Pages 
 
Directed by Rod E. Turochy 
 
 There are currently more than 40,000 deaths a year that occur as the result of 
motor vehicle traffic accidents, which have been ranked among the leading causes of 
death for Americans.  In addition, a reported 55% of all roadway fatalities are the result 
of roadway departures.  In an effort to reduce the number of accidents that occur on two-
lane rural highways the Alabama Department of Transportation established a new policy 
in 2006 to begin installing 2-ft of paved shoulders on both sides of the highways during 
resurfacing and reconstruction projects.  This type of countermeasure provides an
 
vi 
additional recovery area for vehicles that have veered out of the traveled way as well as a 
reduction in accidents.  
Since the policy that ALDOT established in 2006 has only been in effect for two 
years, a before-and-after study to quantify the benefits of installing 2-ft paved shoulders 
was not possible.  However, a hypothetical benefit/cost analysis was needed to evaluate 
the potential benefits and lay the foundation for future research in this area.  As part of 
this analysis a detailed methodology was created to assist in the identification of high 
priority segments that could experience a significant reduction in accidents if this 
treatment were implemented.  The CARE 8 software program that has been developed 
and maintained by the University of Alabama CARE Research and Development 
Laboratory was used to conduct a high accident analysis.  Based on this analysis, two 
different lists of high priority segments were created which contained accident and 
location information for the top 5% of the road segments that were identified based on 
two criteria: crash rate and severity.   
The economic benefits associated with the installation of paved shoulders at these 
locations were calculated based on a range of crash reduction factors.  The cost of 
installing 2-ft paved shoulders was determined based on 41 projects that had already been 
let to bid by ALDOT.  The resulting benefit/cost ratios for the high priority segments in 
each of the criteria ranged from 0.08:1 to 2.20:1 for the segments identified based on 
crash rate and 1.22:1 to 8.18:1 for those segments identified based on severity.  The 
results of this preliminary analysis indicated that the installation of 2-ft paved shoulders 
on two-lane rural highways in Alabama is a cost effective countermeasure that has the 
potential to significantly reduce accidents and improve safety.  
vii 
ACKNOWLEDGEMENTS 
 I would like to thank my husband Armando for all of the years of patience, 
understanding, and love.  I would also like to express my appreciation for my family and 
friends who have been supportive and have encouraged me every step of the way.  I 
would also like to express my gratitude to Dr. Rod Turochy for the opportunity to study 
under him and for all of his wisdom, guidance, and support throughout this project.  
Finally, I would like to acknowledge the Alabama Department of Transportation for 
funding this project and providing me with an opportunity to conduct this benefit/cost 
analysis. 
 
 viii 
Style manual used: Publication Manual of the American Psychological Association, Fifth 
Edition 
Computer software used: Microsoft Word, Microsoft Excel, CARE 8 
 
 
 ix 
 
 
 
 
TABLE OF CONTENTS 
ACKNOWLEDGEMENTS .............................................................................................. vii 
LIST OF TABLES ............................................................................................................ xii 
LIST OF FIGURES ......................................................................................................... xiv 
CHAPTER ONE:  INTRODUCTION .................................................................................1 
1.1   Background ............................................................................................................. 1 
1.2   Objectives ................................................................................................................ 3 
1.3   Scope ....................................................................................................................... 3 
1.4   Outline ..................................................................................................................... 4 
CHAPTER TWO:  LITERATURE REVIEW .....................................................................7 
2.1   Effects of Paved Shoulders and Shoulder Widths on Accident Rates .................... 7 
2.1.1   Comparisons between Paved Shoulders and Unpaved Shoulders ................... 8 
2.1.2   Examination of the Effects of Shoulder Width .............................................. 18 
2.2   Crash Reduction/Modification Factors ................................................................. 22 
2.3   Economic Evaluation of Accidents ....................................................................... 27 
 
 x 
2.4   Summary of Findings ............................................................................................ 30 
CHAPTER THREE:  METHODOLOGY .........................................................................32 
3.1   Data Collection ...................................................................................................... 34 
3.1.1   Filter Creation and Implementation ............................................................... 34 
3.1.2   High Accident Analysis ................................................................................. 35 
3.1.3   Verification of Traffic Data ........................................................................... 35 
3.1.4   Combining Road Segments............................................................................ 37 
3.2   Data Analysis ........................................................................................................ 40 
3.2.1   Determination of High Priority Segments ..................................................... 40 
3.2.2   Determination of CRF Values ....................................................................... 44 
3.3   Economic Evaluation of Potential Benefits .......................................................... 46 
3.4   Costs of Shoulder Paving ...................................................................................... 49 
3.5   Benefit/Cost Analysis ............................................................................................ 52 
CHAPTER FOUR:  RESULTS .........................................................................................53 
4.1   Identification of High Priority Segments .............................................................. 54 
4.2   Calculated Benefits ............................................................................................... 54 
4.3   Calculated Costs .................................................................................................... 56 
 
 xi 
4.4   Benefit/Cost Analysis ............................................................................................ 56 
CHAPTER FIVE:  CONCLUSIONS AND RECOMMENDATIONS .............................58 
5.1   Conclusions ........................................................................................................... 58 
5.2   Recommendations for ALDOT ............................................................................. 59 
5.3   Recommendations for Future Research ................................................................ 59 
REFERENCES ..................................................................................................................61 
APPENDIX A:  ACCIDENT MODIFICATION FUNCTION (NCHRP, 2005) ..............67 
APPENDIX B:  ECONOMIC COSTS OF ACCIDENTS AND TOTAL POLICE-
REPORTED INCIDENCES (NHTSA, 2002) ................................................................68 
APPENDIX C:  FILTER CREATION CARE 8 (CRDL, 2007) .......................................69 
APPENDIX D:  HIGH ACCIDENT ANALYSIS CARE 8 (CRDL, 2007) ......................70 
APPENDIX E:  DECISION RULE FREQUENCY AND CUMULATIVE FREQUENCY 
DISTRIBUTIONS ..........................................................................................................71 
APPENDIX F:  GDP IMPLICIT PRICE DEFLATOR TRENDS AND VALUES 
(Bloomberg, 2008) ..........................................................................................................72 
APPENDIX G:  ALDOT SHOULDER WIDENING PROJECTS ...................................73 
 
 
 xii 
 
 
 
 
LIST OF TABLES 
Table 2 ? 1   Paved shoulder widening CRF values ..........................................................24 
Table 2 ? 2   CRF values used by State DOTs...................................................................24 
Table 2 ? 3   CRF values obtained from within the state of Kentucky ..............................25 
Table 2 ? 4   CRF values obtained from other State DOTs ...............................................25 
Table 2 ? 5   Comprehensive costs of police reported crashes ..........................................28 
Table 2 ? 6   Total unit costs of injuries on the AIS 90 scale ............................................30 
Table 3 ? 1   Example data from actual road segment .......................................................36 
Table 3 ? 2   Example ALDOT?s traffic data ....................................................................36 
Table 3 ? 3   Comparison of consecutive road segments ...................................................38 
Table 3 ? 4   Determination of the distance between consecutive segments .....................38 
Table 3 ? 5   List of high priority segments based on crash rate .......................................42 
Table 3 ? 6   List of high priority segments based on severity ..........................................43 
Table 3 ? 7   CRF values and expected number of crashes based on crash rate ................46 
Table 3 ? 8   CRF values and expected number of crashes based on severity...................46 
 
 xiii 
Table 3 ? 9   Total costs of accidents and injuries on crash rate high priority 
 segments ..........................................................................................................................49 
Table 3 ? 10   Total costs of accidents and injuries on severity high priority  
 segments ..........................................................................................................................49 
Table 3 ? 11   Statistical values used in cost determination ..............................................50 
Table 4 ? 1   Potential savings if shoulder treatment is implemented on high priority 
segments based on crash rate ..........................................................................................55 
Table 4 ? 2   Potential savings if shoulder treatment is implemented on high priority 
segments based on severity .............................................................................................55 
Table 4 ? 3   Benefit/Cost ratios for high priority segments based on crash rate ..............57 
Table 4 ? 4   Benefit/Cost ratios for high priority segments based on severity .................57 
 
 xiv 
 
 
 
 
LIST OF FIGURES 
Figure 3 ? 1   Frequency and cumulative frequency distributions of road segments less 
than 2.1-miles apart .........................................................................................................39 
Figure 3 ? 2   Frequency and cumulative frequency distributions used in high priority 
segment identification based on crash rate .....................................................................41 
Figure 3 ? 3   Frequency and cumulative frequency distributions used in high priority 
segment identification based on severity ........................................................................41 
Figure 3 ? 4   Frequency and cumulative frequency distributions of shoulder paving 
costs-per-mile ..................................................................................................................51 
 
1 
CHAPTER ONE 
INTRODUCTION  
1.1 Background 
Motor vehicle accidents have been reported by the Center for Disease Control?s 
(CDC) National Center for Health Statistics (NCHS) as one of the leading causes of death 
in the United States (Kochanek et al., 2004).  In this report by the NCHS, motor vehicle 
accidents were all categorized under unintentional injuries.  As a result, the National 
Highway Traffic Safety Administration (NHTSA) conducted its own study to further 
examine the proportion of deaths that occur from motor vehicle accidents.  NHTSA?s 
National Center for Statistics and Analysis (NCSA) ?determined that motor vehicle 
traffic crashes were the leading cause of death for every age 3 through 33? and were the 
cause of 44,065 deaths in 2002 (NHTSA, 2005a).? Furthermore, it was determined to be 
the cause of more than 50% of the deaths that resulted from unintentional injuries as well 
as the leading external cause of death for both males and females in the U.S. (NHTSA, 
2005b).   
It has also been shown that the highest proportions of fatal crashes occur in the 
southeastern portion of the U.S., defined here as eight states including Alabama (Dixon, 
2005).  These states collectively contribute over 26% of the nation?s fatal motor vehicle 
crashes annually and exceed the national average by ?30 fatalities per million vehicle
 
 2 
miles traveled? (MVMT) (Dixon, 2005).  More than 40% of fatal motor vehicle crashes 
occur in rural areas and approximately 60% of them are the result of a roadway departure 
(NHTSA, 2005b and FHWA, 2006).  In addition, there are fewer vehicle miles of travel 
in rural areas and as a result the crash rates are often higher than those of urban areas.  
Crashes in rural areas have also been shown to be more severe, cause greater injuries, and 
result in more fatalities (NHTSA, 2005b). 
Many studies of countermeasures that can be implemented to reduce the number 
and severity of motor vehicle accidents that occur on rural two-lane highways have been 
conducted.  The findings of several different studies have identified shoulder paving or 
widening as an effective measure in reducing accidents and accident severity on two-lane 
rural highways (Heimbach et al., 1974; Ogden, 1997; Rogness et al., 1982; Turner et al., 
1981, Zegeer et al., 1988, and Zegeer et al., 1995).  This countermeasure provides 
vehicles with an additional recovery area that has been shown to effectively reduce run-
off-the-road and head-on crashes.  It is thought to be most effective where shoulder edge 
drop offs are eliminated by the addition or widening of paved shoulders (NCHRP, 2003, 
AAA Foundation for Traffic Safety, 2005).   
The Alabama Department of Transportation (ALDOT) established a new policy in 
2006 to reduce the number of run-off-road crashes.  As a result ALDOT has been 
installing 2-ft of paved shoulders as part of their resurfacing and reconstruction projects 
for two-lane rural highways throughout the state.   
 
 3 
1.2 Objectives 
The objectives of this research include the development of a procedure that could 
be used by ALDOT to identify high priority road segments that have either high crash 
rates or severity.  These segments could potentially have the greatest reduction in run-off-
road crashes as a result of the installation of 2-ft paved shoulders.  In addition a 
benefit/cost analysis was conducted in an attempt to quantify the potential benefits of 
adding 2-ft of paved shoulders to high priority segments on two-lane rural highways in 
Alabama.  
1.3 Scope 
This project was conducted for the Alabama Department of Transportation as part 
of a preliminary analysis to quantify the potential benefits of installing 2-ft paved 
shoulders on two-lane rural highways within the state.  Pavement widening is often used 
as a countermeasure against lane departures which have been shown to result in high 
severity accidents.  The widening of paved shoulders on two-lane rural roads can provide 
errant vehicles with an additional recovery area that can reduce accident frequency and 
severity.   
This study established a methodology that ALDOT could use to identify high 
priority sections of roadway based on crash rate and severity.  These segments were then 
used in a hypothetical benefit/cost analysis to determine the potential benefits that could 
be obtained by installing 2-ft paved shoulders on these segments.  Since ALDOT?s 
shoulder paving policy has been in place for only 2 years, a hypothetical analysis was 
 
 4 
conducted in the absence of sufficient data after installation.  Common practice is to have 
a minimum of three years of after data when conducting a before-and-after analysis.   
The potential benefits were calculated based on the number of accidents that 
could be prevented if this shoulder treatment had been applied to the identified segments.  
The per-mile-cost of adding paved shoulders was calculated based on 41 projects that had 
been let to bid within the past two years.  The mean value of these project costs was used 
in this analysis which limits the results of this study to the State of Alabama. 
1.4 Outline 
 This thesis begins with background information regarding the need for this 
project, a list of the objectives, as well as how and where the research will be applied; 
which is presented in chapter one.  Chapter two contains an extensive literature review 
which consists of a summary of the studies that have been conducted to determine and 
analyze the effects that paved shoulders can have on accident rates and safety.  Included 
are publications examining the impacts of shoulder widening as well as comparisons 
between the accident rates experienced on similar roadways with different shoulder 
widths.  Following the examination of the relationship between paved shoulders on rural 
two-lane highways and accident occurrence, there is a synopsis of associated crash 
reduction factors (CRFs) currently being recommended for use by the U.S. Department of 
Transportation (DOT), National Cooperative Highway Research Program (NCHRP), and 
other state DOTs.  Finally, a section examining the economic costs incurred as a result of 
motor vehicle accidents was included.  In this section, a brief discussion regarding the 
 
 5 
rating scale that is used to characterize different levels of injury severity and a breakdown 
of their associated costs were outlined. 
 Chapter three contains the methodologies used to identify the high priority road 
segments as well as the details of the benefit/cost analysis.  The first section of this 
chapter provides a detailed outline of the methodology that was used during data 
collection.  Section two contains the methodology used during data analysis to identify 
the high priority segments and to determine an appropriate range of CRF values that were 
used to estimate the reductions in accidents that may be obtained from the installation of 
paved shoulders.  The final three sections of this chapter give the details of the 
determination of the potential benefits and costs associated with the installation of 2-ft 
paved shoulders on two-lane rural highways in Alabama.  The results of which were used 
in the hypothetical benefit/cost analysis to determine the potential benefits of 
implementing this countermeasure at the locations that were identified as high priority 
segments.   
The final two chapters of this thesis present the results and conclusions of this 
study.  In chapter four, the results of the high priority identification process based on the 
two criteria: crash rate and severity, were examined.  Consequently the use of these 
segments in the calculation of potential benefits and costs were also discussed.  The 
ranges of potential economic benefits that can result from the application of the shoulder 
treatment were identified.  These calculated benefits were based on the selected range of 
CRF values selected in the methodology section.  The construction costs associated with 
the installation of paved shoulders were also identified and used in the determination of 
 
 6 
the benefit/cost ratios for these two criteria.  Chapter five expands on these results and 
discusses the recommendations for the application of the outlined methodology for 
identifying high priority segments as well as recommendations for future research.   
 
 
 7 
CHAPTER TWO 
LITERATURE REVIEW 
 In an attempt to reduce accidents, the Alabama DOT has been installing 2-ft 
paved shoulders on each side of two-lane rural highways during resurfacing and 
reconstruction projects for the past two years.  Many related articles and publications 
were reviewed to quantify the potential accident reductions that may be obtained from 
this countermeasure.  Of particular interest were publications that examined the effects 
that the installation or presence of paved shoulders have been shown to have on accident 
rates.  Consequently, studies that have examined the relationship between different paved 
shoulder widths and accident rates or reductions were also reviewed.  In addition, a state-
of-the-practice approach was followed to determine the crash reduction or accident 
modification values that were currently being recommended for use regarding the 
installation or widening of paved shoulders.  Finally, publications that quantified the 
economic costs associated with accidents were used to determine the value of an injury, 
fatality, and property damage.   
2.1 Effects of Paved Shoulders and Shoulder Widths on Accident Rates 
Two different types of studies were reviewed to gain an understanding of the 
relationship between paved shoulders and accident rates.  The first type, were those that 
examined the presence of paved shoulders.  Seven studies were reviewed that compared 
the accident rates of road segments with paved shoulders with those that had unpaved 
 
 8 
shoulders.  Each of these studies established a methodology of characterizing road 
segments in an attempt to isolate the effects of paved shoulders on accident rates.  The 
second type of study followed a similar methodology in their evaluations of the accident 
rates on similar road segments that had different shoulder widths. 
2.1.1 Comparisons between Paved Shoulders and Unpaved Shoulders  
In 1974 a study was published which examined the cost-effectiveness of paving 
the shoulders of rural highways in North Carolina (Heimbach et al., 1974).  During this 
study the accident rates of road segments with paved shoulders were compared to those 
with unpaved shoulders.  The accident data included the number of accidents, fatalities, 
injuries, and property damage from 1966 to 1969.   
The accident data were initially grouped into two categories based on the presence 
or absence of paved shoulders and rated according to nine characteristics.  All sections of 
roads that contained an object that may change speed, flow, or in some way alter the 
traffic pattern were eliminated.  This included road segments that contained intersections, 
bridges, as well as varying roadway and shoulder widths.  This preliminary analysis 
resulted in the selection of 817 sections of roadway that had paved shoulders and roughly 
3,100 sections with unpaved shoulders. 
An extensive classification system was then employed to ensure that the road 
segments used in the analysis would be similar in all aspects except shoulder treatment.  
However, when the researchers attempted to group the road segments according to the 
rating system they found only a few roads to be similar.  As a result, an analysis of 
covariance was conducted to determine which of the variables used in the classification 
 
 9 
process were actually significant.  From this test it was determined that only five of the 
original nine variables were found to have a statistically significant relationship with the 
number of accidents: average daily traffic (ADT), access control, speed limit, number of 
lanes, and lane width.  Once the number of variables used in the classification system was 
reduced, 141 sections that had paved shoulders were paired against 985 sections that had 
unpaved shoulders.   
The accident data that were used in this study included only four years of 
information.  If the year of construction fell within this range, then the entire year of 
accident data was omitted resulting in only three years of before and after data combined.  
After the two sets of data were established, a statistical analysis was conducted to 
determine if there was a ?difference in accident experience between highway sections 
alike in all respects, except for the shoulder construction? (Heimbach et al, 1974).  The 
Chi-square test that was conducted showed there to be a statistically lower accident 
experience as well as accident severity on roadways with paved shoulders than roadways 
with unpaved shoulders.  The conclusion of this analysis was limited to highway road 
segments that had 3-ft to 4-ft of paved shoulders and ADT values between 2,000 and 
10,000 vehicles per day (vpd). 
One of the major drawbacks to this study is that only four years of accident data 
were available.  As a result, there was an inconsistent amount of data available for the 
time periods before and after shoulder construction.  The shoulder widths of the road 
segments studied also varied, although, the majority were reported to be between 3-ft and 
4-ft.  Another point that should be considered when examining the findings of this study 
 
 10 
is that many of the road segments that were classified as unpaved did have shoulder 
treatments such as rock or gravel.   
The Texas State Department of Highways and Public Transportation conducted 
two studies that both had similar results to the Heimbach et al. (1974) study (Turner et 
al., 1981 and Rogness et al., 1982).  The purpose of the first study was to determine the 
effect that paved shoulders had on accident rates.  During this study they compared rural 
two-lane highways that had full width paved shoulders with similar roads without paved 
shoulders.   
The accident data set used in this study was for the three years from 1975 to 1977.  
The accident data were reviewed and the accidents that were the result of a condition that 
was not normally characteristic of the road were eliminated.  Some examples include 
accidents that occurred in construction areas, or near railroad crossings.  The accident 
data were then divided into two categories: all accidents and non-intersection accidents.   
Road data for these locations were also collected.  Segments that contained 
divided roadways, major intersections, or other objects that may influence accident rates 
were eliminated.  Road segments with shoulders that were paved but less than 6-ft were 
also classified as not having a paved shoulder.  As a result, the analysis was conducted 
using 85 sections of similar roadways that were classified as either full-width paved 
shoulders or no shoulders.  The road characteristics that were initially created divided the 
road data for paved and unpaved shoulders into three categories based on three ADT 
ranges: 1,000 to 3,000, 3,000 to 5,000, and 5,000 to 7,000.  The researchers had 
originally intended to have ten sites in each ADT range.  However, the ADT range of 
 
 11 
5,000 to 7,000 consisted of only eight sites that met the criteria of being five or more 
miles in length with similar characteristics.   
From a regression analysis, it was determined that two-lane highways with full 
width paved shoulders had lower accident rates over all three ADT ranges than those 
without full width paved shoulders.  This was found to be true regardless of the presence 
of intersections.  They also concluded that ?two-lane highways without paved shoulders 
are very sensitive to intersection accidents, especially at high traffic volumes? (Turner et 
al., 1981).  The road sections with shoulders were also observed to have lower fatality 
rates, hit-other-car, run-off-road, and fixed-object accidents.   
The second study conducted by the Texas State Department of Highways and 
Public Transportation focused on the effect that the installation of full width shoulders 
had on accident frequency (Rogness et al., 1982).  During this before and after study the 
comparison was made between the number and type of accidents that occurred on a road 
segment before the installation of full width shoulders with those that occurred after 
construction was completed.   
The same classification system and a similar set of criteria were used to evaluate 
the road data and accident data as in the previous study by Turner et al.  (1981).  Road 
segments for this study were selected based on whether two years of before and after 
accident data were available.  As an additional requirement, there could not have been 
any other alterations or construction conducted on the road segment within this four year 
period.  Accidents that were not considered to be associated with roadway type were also 
excluded from the accident data even though the classification of all accidents and non-
 
 12 
intersection accidents were maintained.  In addition, some road segments were included 
even though they were less than five miles in length due to a small sample size in the 
3,000 to 5,000 and 5,000 to 7,000 ADT ranges.   
Rogness et al. (1982) concluded that for the frequency analysis the total number 
of accidents decreased after the installation of full width paved shoulders for both the all-
accident and non-intersection classifications.  They also conducted a paired t-test on the 
before and after data for each ADT range to determine if there was a statistically 
significant difference in either the type of accident or the severity.  From this analysis 
they concluded that there was no significant change in accident type.  However, there 
appeared to be a decrease in accident severity for the two lower ADT ranges and an 
increase in accident severity for the higher ADT range. 
In 1988, an article was published that again found there to be a reduction in 
accidents after shoulder widening of paved and unpaved shoulders (Zegeer et al., 1988).  
This study was conducted using road and accident data from seven different states.  The 
emphasis of the study was on the effects that certain road treatments had on safety.  One 
of the primary objectives of the study was to develop a model that would explain the 
relationship between accidents and different road characteristics.  In addition, the 
comparison was again made between the accident rates that occurred on road segments 
with paved and unpaved shoulders. 
Accident data were obtained from each of the seven states.  Several modifications 
were then made to the data sets to account for the different reporting measures of each 
state.  Five years of data were obtained for most of the sites.  There were a few sites in 
 
 13 
which accident data were available for only two to three years.  The researchers then 
reduced the accident data set by including only those that they felt were the result of the 
specific road conditions of interest.  This was accomplished by conducting a Chi-square 
analysis on the different types of accidents and the selected roadway features.  Three 
types of single-vehicle accidents and four types of multi-vehicle accidents were found to 
be significant. 
Along with the accident data, road data were also obtained from the appropriate 
state agencies.  The road segments that were selected from rural areas varied in length 
from one to ten miles.  The length of each section was determined by significant changes 
in ADT, as well as lane and shoulder width, and several other factors that could affect 
traffic flow.  The important road characteristics were recorded and then analyzed using 
several statistical methods to determine whether they had a significant impact on the 
designated types of accidents.  Thirteen variables were identified and found to be 
significant based on a Chi-square, regression, and analysis of variance and covariance.   
A model was then developed based on the known variables and used to predict the 
number of accidents that would occur at specific shoulder widths.  The model was also 
used to develop accident reduction factors.  The percent of accident reduction that could 
be expected from adding a specified number of feet of paved shoulder was then 
calculated based on the predicted values of before and after accident data generated by 
the model.   
Zegeer et al. (1988) concluded from this study that unpaved shoulders were not as 
effective in reducing accidents in the specified categories as paved shoulders.  The 
 
 14 
percent reduction in accidents that could be expected from the addition of two, four, and 
six-ft of shoulder widening were 16%, 29%, and 40% reduction in accidents, 
respectively.  Although adding unpaved shoulders to both sides of the road did not reduce 
accident rates as much, they still reported reductions of 13%, 25%, and 35%, 
respectively.   
Although the results of this study are similar to those previously reviewed, it is 
difficult to make a direct comparison.  First, this study used road and accident data from 
several different states.  Secondly, the percent reduction in accident rates that were 
reported in this study were based on simulated before and after data.  Therefore, the 
results of this study were based on the model?s ability to accurately predict accident 
occurrence.  Any direct comparisons between the accident data that were collected and 
the simulated data were not discussed.  Different models were developed for each 
accident type with reported R2 values that ranged between 0.39 and 0.46.  Although many 
road characteristics were analyzed to determine their statistical significance, it is not 
possible to account for all of them.  Finally, the compounding effects that some of these 
characteristics may have when they occur together were not assessed.   
A study that was later conducted in Victoria, Australia, examined the benefits and 
costs of paving the shoulders of two-lane rural roads (Ogden, 1997).  The road segments 
selected for the study were all part of resealing projects that included the addition of a 
paved shoulder.  The paved shoulder that was added had to be at least 600-mm or 2-ft 
wide to be included in the study.  A corresponding control segment was also selected for 
each road segment and was required to be the same length, on the same highway, with the 
same speed limit, having similar geometric design, traffic volume, and preferably the 
 
 15 
same vehicular traffic.  In total, thirty-six road segments with corresponding control 
segments were selected.   
As part of this study a statistical analysis of the effects of paved shoulders was 
conducted by comparing the number of accidents that occurred on the treated segments, 
which had paved shoulders, with the accidents that occurred on the control segments 
without paved shoulders.  Accident data before and after the construction period were 
collected for both the treatment and control sites.  The financial year in which the road 
construction was conducted was eliminated from the analysis.  From this study Ogden 
concluded ?that paved shoulders have a positive effect on accidents on two-way rural 
highways? (1997).  Based on the results of a Chi squared test, he also concluded that 
there was a 41% reduction in casualty accidents per vehicle kilometer.   
However, some care should be taken when examining the results of these data 
because the rural roads in Australia that do not have paved shoulders also lack edge lines.  
In some instances raised reflective pavement markers were also added along with the 
edge lines and paved shoulders.  The lack of edge lines as well as the addition of the 
reflective pavement markers may have independently had an effect on the number of 
accidents that occurred before and after shoulder widening.   
There were two studies that found there to be no difference in the accident rates of 
road segments with paved shoulders when compared to unpaved shoulders.  The first 
study to come to this conclusion was published by Zegeer et al. in 1994.  Unlike the other 
studies, this study only examined data for low-volume roads having less than 2,000 vpd.  
During this study, an analysis of covariance was used to determine if there was a 
 
 16 
statistically significant relationship between accident rates of a certain type and shoulder 
width.  The only two that were found to have a significant relationship on low-volume 
two-lane roads were the single-vehicle accidents and opposite-direction accidents.  When 
a covariance model was used to compare unpaved shoulders with paved shoulders they 
concluded that there was no significant difference in the accident rates (Zegeer et al., 
1994). 
In 1999, a study was published for the Alabama Department of Transportation in 
which a safety effectiveness evaluation was conducted on two-lane rural roads (Bowman 
and Abboud, 1999).  There were a total of 57 road segments that were analyzed that had 
either a 2-ft or 4-ft paved shoulder installed where there had been none previously.  The 
accident rate and accident frequency for each segment were determined using three years 
of accident data before and after construction of the shoulders.  Control factors were 
developed based on county crash statistics for each location to help identify trends in the 
accident data.  Only three types of accidents were considered during this analysis: single 
vehicle, multiple vehicle same direction, and multiple vehicle opposite direction.   
A statistical analysis was conducted to determine whether there was a statistically 
significant difference in the accident rates per 100 million vehicle miles of travel before 
and after shoulder construction.  A paired t-test analysis showed that there were no 
statistically significant differences between the actual and expected crash rates as well as 
severity rates for either of the shoulder conditions.  However, ?there was a reduction in 
single vehicle and total crashes for both the two and four-foot shoulder installations? 
(Bowman and Abboud, 1999).  In addition, a two sample t-test was conducted to 
determine if there was a statistically significant difference in crash rate or severity rate 
 
 17 
between the two and four foot shoulder installations.  When compared at a significance 
level of five percent it was again concluded that there was no statistically significant 
difference between the crash rates associated with the two different shoulder widths.  
This was found to be true for single vehicle, multiple vehicle, and total accidents.   
Although five of the seven studies reviewed had similar findings, i.e. that roads 
with paved shoulders have lower accident rates than roads with unpaved shoulders 
(Heimbach et al., 1974; Ogden, 1997; Rogness et al., 1982; Turner et al., 1981; and 
Zegeer et al., 1988) the lack of continuity between these studies makes comparing their 
findings difficult.  Several key elements of the studies were different: width of shoulders 
evaluated, ADT values, classification system of roadway elements, and the statistical 
analyses used to determine the results.  The length of road segments evaluated varied not 
only between studies but often within a study as well.  Furthermore, each of the studies 
utilized a different set of criteria to evaluate the accident and road data.   
Most of the studies compared accident rates of roads with paved and unpaved 
shoulders by selecting roads based on an established set of criteria and using accident 
data for specific years (Heimbach et al., 1974; Turner et al., 1981; Zegeer et al., 1988; 
and Zegeer et al., 1994).  However, the number of years varied between studies, and 
sometimes even between sites within a study (Zegeer et al., 1988 and Zegeer et al., 1994).  
In contrast, three of the studies chose to conduct before and after analyses based on 
accident data for a specified number of years before and after the installation of paved 
shoulders (Rogness et al., 1982, Ogden, 1997, and Bowman and Abboud, 1999). 
 
 18 
Many of the unpaved shoulders evaluated in the studies had received some type of 
treatment such as gravel or rock but were obviously not paved.  Alternatively, in the 
studies conducted by Turner et al. (1981) and Rogness et al. (1982), road segments with a 
paved shoulder of less than 6-ft were classified as unpaved.  Although the results of these 
two studies were similar to the findings of the other studies, the criteria selected for 
shoulder classification makes them difficult to compare.  The 6-ft of paved shoulder may 
have influenced the number of accidents as well as their severity when compared to 
studies that examined entirely unpaved shoulders.   
Despite the differences in the studies, they all concluded that in general, rural 
roads with paved shoulders appear to be safer than rural roads with unpaved shoulders.  
The studies also suggest that paved shoulders reduce the number of accidents and the 
severity of the accidents that occurred on the roads they examined.  The two exceptions 
were the studies conducted by Zegeer et al. (1994) and Bowman and Abboud (1999), in 
which there was found to be no statistically significant reduction or difference in accident 
rates.  However, it should be noted that although not statistically significant, there were 
reported reductions in single vehicle and total crashes as the result of the installation of 
paved shoulders during the Bowman and Abboud study (1999). 
2.1.2 Examination of the Effects of Shoulder Width 
Several other studies have also been conducted to determine whether a 
relationship exists between different shoulder widths and accident rates.  One of the 
earlier studies was conducted by the Connecticut State Highway Department during 
which the relationship between accident rates and shoulder widths was examined 
 
 19 
(Perkins, 1956).  Four years of accident data from 1951-1954 were divided into 
categories based only on the paved shoulder width.  An analysis of the data was 
conducted which led to the conclusion that there was ?no significant relationship between 
accident rate and shoulder width? (Perkins, 1956).  However, there was no mention of the 
type of statistical analysis that was conducted to reach this conclusion.   
A study that was published just prior to the Perkins (1956) study was conducted in 
California by Belmont (1954).  His analyses included an examination of both paved and 
treated shoulders that were divided into three categories based on shoulder width.  The 
road segments were separated into shoulder width classifications of less than 6-ft, 6-ft, 
and greater than 6-ft.  Belmont proceeded to conduct a regression analysis which 
generated three different regression equations to explain the relationship between the 
three categories of shoulder widths and accident rates.  He concluded that 6-ft shoulders 
were safer than shoulders that were narrower than 6-ft, and shoulders wider than 6-ft 
when traffic volumes exceeded 5,000 vpd.   
Another study that found a similar relationship between paved shoulder width and 
accident frequency was conducted by the Oregon State Highway Department (Blensly 
and Head, 1960).  The majority of the road segments had ADT values that were less than 
5000 vpd, and they were categorized in increments of 1,000 vpd.  Two different 
statistical analyses were conducted to try and determine if there was a relationship 
between paved shoulder widths and personal injury, property damage, and total accidents.   
First, the relationship between accident frequency and the width of paved 
shoulders was examined using correlation procedures.  Blensly and Head concluded that 
 
 20 
there was a ?statistically reliable tendency toward an increase in accident frequency as 
paved shoulder width increased? (Blensly and Head, 1960).  However, the results of the 
partial correlation method that was conducted found that the 2,000 to 2,999 ADT range 
was the only one that showed there to be a significant relationship between accident 
frequency and paved shoulder width.   
In order to conduct the second set of analyses the data were sub-divided into two 
different categories: narrow paved shoulders and wide paved shoulders.  The narrow 
shoulder classification was given to all road segments that had paved shoulders of 4-ft 
and less, and the wide shoulder classification was assigned to segments that had paved 
shoulders of 8-ft or wider.  Then analyses of variance and co-variance procedures were 
used to assess whether there was a tendency for there to be an increase in any of the three 
types of accidents when shoulders were either narrow or wide.  The results of the analysis 
of variance ?found that without control over the effect of ADT there was a statistically 
reliable tendency for sections with wide shoulders to have a higher mean number of 
property damage and total accidents than sections with narrow paved shoulders? (Blensly 
and Head, 1960).  There were similar findings in their analysis of variance, where they 
concluded that when the road segments had wide shoulders there were higher incidents 
involving property damage and total accidents.   
The findings of the Zegeer et al. (1981) study were contradictory to the results of 
the correlation procedures conducted by Blensly and Head (1960) and the regression 
analysis of Belmont (1954).  Zegeer et al.  (1981) conducted a comparative analysis that 
found there to be a decrease in run-off-road and opposite-direction accidents as shoulder 
widths were increased from zero to 9-ft.  However, this study also determined that there 
 
 21 
was an increase in these types of accidents for shoulder widths from 10-ft to 12-ft.  It was 
also concluded that there were little changes or slight increases in accident rates for other 
types of accidents as the shoulder widths increased on these same road segments.   
The studies that examined the potential relationship between shoulder widths and 
accident rates all had substantially different conclusions.  Belmont (1954) concluded that 
two-lane rural roads with shoulder widths of 6-ft were safer than road segments that had 
either narrower or wider shoulders.  A study with similar findings that accident frequency 
increases as shoulder width increases was conducted by Blensly and Head (1960).  
However, they also concluded that only one ADT range in their study showed a 
statistically significant relationship between accident occurrence and shoulder width.  
Finally, the Zegeer et al. studies conducted in 1981 and 1988 contradicted the findings of 
both of the previous studies when they concluded that accident rates decreased as 
shoulder widths increased from zero to 9-ft.  However, the 1981 study also showed that 
when shoulder widths increased to 10-ft and12-ft there was a slight increase in accident 
rates (Zegeer et al., 1981).  In addition, at least two studies found there to be no 
statistically significant relationship between shoulder width and accident rates (Perkins, 
1956 and Bowman and Abboud, 1999).   
Although the findings of the studies were contradictory, several considerations 
should be made.  Each of the studies was conducted in a different state where the 
methods of reporting and recording accident information may differ.  Some of the studies 
included treated shoulders as well as paved shoulders which may have affected the 
number and types of accidents that occurred.  The studies also had different types of 
accidents that were examined rather than overall accidents.  Many of these studies 
 
 22 
separated the road segments into different classifications based on traffic volume and 
shoulder width.  Finally, each study used a different methodology for analyzing the data 
which may have contributed to the differences in their findings.   
2.2 Crash Reduction/Modification Factors 
A crash reduction factor (CRF) is a value that represents the expected percent 
reduction in crashes as a result of the implementation of a treatment or program.  These 
values are generally based on the quantitative results of research studies.  They are often 
used to determine the safety effectiveness of implementing geometric improvements or 
new countermeasures.  Formally, a CRF ?is the expected number of crashes with an 
action, divided by the expected number of crashes had the action not been taken (Gross 
and Jovanis, 2007).  This information can also be expressed as an accident modification 
factor (AMF) or crash modification factor (CMF), which is simply the expected reduction 
of crashes if the improvement or treatment is installed (Equation 2-1). 
AMF = CMF = 1-CRF        (2-1) 
The first attempt at assigning an AMF value to changes in shoulder width and 
shoulder type was determined based on an extensive literature review by a panel of 
experts (NCHRP, 2005).  In total, this panel determined AMF values and assigned a level 
of predictive certainty value (LOPC) to 100 different treatments.  The LOPC value was 
used to indicate to the reader the panel?s level of confidence in the AMF value assigned 
to each treatment.  This value was influenced by the type, amount, and credibility of the 
research that was reviewed.   
 
 23 
The accident modification function that was given for shoulder width and type is 
shown in Equation 2-2 (NCHRP, 2005).   
AMF = (AMFWRAAMFTRA-1.0)PRA+1.0      (2-2) 
The accident modification factor for related accidents based on shoulder width, AMFWRA, 
and shoulder type, AMFTRA, can be obtained from the tables provided with the equation 
(Appendix A).  Both of these values were ?calculated by dividing the AMF for the after 
improvement condition by the AMF for the before condition? (NCHRP, 2005).  The PRA 
value is the proportion of total accidents that will be reduced.  This equation is capable of 
producing an accident modification factor that accounts for shoulder width as a function 
of ADT and shoulder type.  The function was also assigned a ?medium-high? LOPC 
rating indicating that the panel had a high level of confidence in the function?s predictive 
ability.  However, it should be noted that the reasons for the rating were not given. 
The ?Desktop Reference for Crash Reduction Factors? that was published in 
September 2007 by FHWA, provided several different crash reduction factors associated 
with shoulder widening (Bahar et al., 2007).  This technical report was again based on an 
extensive literature review that obtained most of its values from publications previously 
discussed.  The CRF values and functions that can be obtained from this document 
include countermeasures such as: changes in shoulder type and/or width, installation of 
shoulders, stabilization of shoulders, varying outside shoulder width, widening of 
shoulders, and widening of paved shoulders.  The values given for the widening of paved 
shoulders show an increase in CRF with increasing shoulder width (Table 2-1). 
 
 24 
Table 2 - 1 Paved shoulder widening CRF values (Bahar et al., 2007) 
Widen Paved Shoulder Crash Type Crash Severity Crash Reduction Factor (%) 
0 to 2 ft Fixed object & ROR All 16 
0 to 4 ft Fixed object & ROR All 29 
0 to 6 ft Fixed object & ROR All 40 
0 to 8 ft Fixed object & ROR All 49 
 
Several different CRF values were obtained from publications produced for state 
DOTs.  The Ohio and Oregon DOTs both recommended the use of the AMF function 
given in Equation 2-2 (Monsere et al., 2006 and Zwahlen and Suravaram, 2007).  Both of 
these state DOTs reported that the AMF function was obtained from an extensive 
literature review.  The decision to use this function was determined by each state?s DOT 
in conjunction with an expert panel.  The remaining values reportedly used by the state 
DOTs ranged from 12% to 20% (Table 2-2).   
Table 2 - 2 CRF values used by State DOTs 
State DOT Recommended CRF Value 
Alabama 20% 
Arizona 12% 
Kentucky 15% 
New York 17% 
North Carolina 18% 
South Dakota 20% 
 
The Arizona DOT reported the lowest value for shoulder widening which was 
12% for fatal and nonfatal injuries combined (Arizona DOT, 2004).  This recommended 
value was obtained from a 1996 Annual Report published by FHWA in which a value of 
21% reduction in accident rates was also reported for fatal crashes (FHWA, 1996).  The 
Arizona DOT also provided accident reduction factors based on their own accident data 
that were significantly higher than the FHWA recommended values.  The percent 
 
 25 
reduction in accident rates that they reported to be potentially attainable from shoulder 
widening was 57% (Arizona DOT, 2004).  This value was reported to be a statistically 
significant reduction for all types of accidents and injuries. 
A study was conducted for the Kentucky DOT that consisted of a survey of CRF 
values that were currently being used within the state as well as by other state DOTs for 
different treatments and improvements (Agent et al., 1996).  Shoulder widening values 
were examined as part of the construction/reconstruction category for several different 
shoulder treatments.  A range of CRF values were obtained for each of the treatments 
(Table 2-3 and 2-4).   
Table 2 - 3 CRF values obtained from within the state of Kentucky (Agent et al., 1996) 
Treatment No. of Values Range of Values (%) Average Value (%) 
Shoulder Widening 18 5 ? 50 19 
2 to 4-ft 2 15 ? 32 24 
Paved Shoulder 3 10 ? 25 18 
 
Table 2 - 4 CRF values obtained from other State DOTs (Agent et al., 1996) 
Treatment No. of Values Range of Values (%) Average Value (%) 
Shoulder Widening 17 8 ? 57 20 
2 to 4-ft 2 15 ? 16 15 
Paved Shoulder 1 20 20 
 
Based on this examination of reported CRF values, the Kentucky DOT 
recommended the following values be used within their state: shoulder widened to a 
width of less than 4-ft, 20% reduction, shoulder was stabilized or a shoulder drop off 
eliminated, 25% reduction, and if a paved shoulder was installed, 15% reduction in 
accidents (Agent et al., 1996).  However, there is no recommendation given for 
determining a CRF value for the combined effects of these shoulder treatments. 
 
 26 
The New York DOT reported only one CRF value for shoulder widening on two-
lane roads, 17% (NYSDOT, 2000).  This expected reduction in accidents was for the 
widening of existing shoulders only.  No additional information as to the determination of 
this value was given.  Similarly, the North Carolina DOT reported a CRF for total crashes 
on rural 2-lane roads to be 18% (North Carolina DOT, 2007).  This value was followed 
by recommended values for fatal crashes, non-fatal injury crashes, and PDO crashes of 
48%, 8%, and 23%, respectively.  These values were referenced from a meeting handout 
and no additional information was given. 
The value that was reportedly used by South Dakota?s DOT was based on an 
extensive literature review as well as a review panel that evaluated CRF values that were 
obtained from accident data.  The average CRF value that was reported for shoulder 
widening by this state DOT was a 20% reduction (Tople, 1998).  This was the same value 
that the Alabama DOT was currently using and the highest reported value that was found.   
Extreme caution should be exercised when using CRF values to determine when 
and where shoulder widening could be beneficial.  The user should be aware of how and 
where the CRF was developed.  The validity of a CRF when it is transferred from one 
location to another is not known.  The most accurate CRF values are obtained from 
before and after studies (Bahar et al., 2007, Gross and Jovanis, 2007, and NCHRP, 2005).  
However, the accuracy of the study or analysis process could still result in biased data 
that will in turn produce an inaccurate CRF value (NCHRP, 2005).  Although there are 
several sources for obtaining a CRF for shoulder widening or paved shoulder installation, 
the use and applicability to the desired project may be limited.  Consideration should also 
 
 27 
be given to any additional comments, information, or ratings that were assigned to the 
published CRF.   
2.3 Economic Evaluation of Accidents 
When motor vehicle accidents occur many people are affected and the costs 
associated with these traffic accidents are far reaching.  There are costs associated 
directly with the medical care that is received such as insurance premiums, medical 
expenses, and the cost of emergency services rendered.  There are many indirect costs to 
not only the individuals involved in the traffic accident but to society as well.  These 
costs include legal fees, lost wages, travel delay, property damage, and many others.  To 
determine the costs associated with an automobile accident several factors need to be 
taken into consideration.  Several examples include how many people were injured or 
killed, the extent of the injuries, and how much property damage resulted from the 
incident.   
A rating system has been produced by the Association for the Advancement of 
Automotive Medicine (AAAM) to pair the severity of an injury with an identification 
number.  This system is known as the Abbreviated Injury Scale (AIS) which assigns a 
value from 0 to 6 to indicate the threat-to-life that results from a particular injury 
(Zaloshnja et al., 2004).  The value of 0 represents the lowest level, or uninjured rating.  
The highest rating that can be received is a 6, indicating a fatal injury.  The AIS rating 
system is often used to determine the costs associated with automobile accidents (FHWA, 
2005, NHTSA, 2002, and Zaloshnja et al., 2004).   
 
 28 
In 1982, research was conducted by the FHWA to provide a cost scheme that 
could be used during economic evaluations of highway improvements.  The cost values 
were calculated to ?reflect the amount individuals are willing to pay to reduce the number 
and severity of accidents? (FHWA, 1994).  Later in 1985, the FHWA sponsored another 
research initiative that determined the comprehensive costs associated with injuries, 
fatalities, and property damage that resulted from motor vehicle crashes (FHWA, 1994).  
These analyses resulted in dollar values being assigned to the AIS rating system (Table 2-
5).   
These comprehensive costs consist of eleven different components that 
encompass the individual and societal costs associated with the severity of an injury.  The 
values in Table 2-5 were reportedly obtained in accordance with FHWA departmental 
guidance which recommended that the Gross Domestic Product (GDP) implicit price 
deflator be used to update the dollar amounts (FHWA, 1994). Following this procedure 
the 1988 dollar values were inflated by 18% and reported in terms of a 1994 dollar 
amount. 
Table 2 - 5 Comprehensive costs of police 
reported crashes (FHWA, 1994) 
Severity Cost Per Injury 
(1994 Dollars) 
AIS 1 $5,000 
AIS 2 $40,000 
AIS 3 $150,000 
AIS 4 $490,000 
AIS 5 $198,000 
AIS 6 $2,600,000 
 
In 2004 a study was published by Zaloshnja et al. which updated the crash cost 
estimates based on the 1990 version of the AIS codes (AIS90).  During this study an 
 
 29 
evaluation of the effects of using cost data that were based on the 1985 edition of the AIS 
codes to injuries that had been classified using the 1990 edition as applied during a 
highway crash cost analysis were examined.  From this study, it was concluded that 
adjusting the AIS85 costs based solely on inflation for injuries classified using the AIS90 
scale generally resulted in the overestimation of benefits.   
The AIS90 scale was assigned different severity codes than the AIS85 scale, and 
as a result, the error in the benefit calculations can be significant (Zaloshnja, 2004).  In 
addition, the costs associated with the AIS85 scale predated managed care in the U.S. 
which has greatly impacted health care services.  Managed care has changed the way care 
is administered.  In general managed care has reduced the length of hospitalization after 
an accident which has subsequently affected the cost of care.   
The most current documentation of the economic costs associated with motor 
vehicle crashes was published by NHTSA in 2002.  It was reported that in 2000 the total 
cost of motor vehicle crashes in the U.S. was $230.6 billion (NHTSA, 2002).  This is the 
present value of the total economic costs associated with all police-reported and 
unreported accidents that occurred in 2000.  In this publication a dollar amount was 
assigned to each of the six AIS90 codes as well as a value for incidents that resulted in 
property damage only (PDO) (Appendix B).  These unit costs were calculated in terms of 
the U.S. dollar in the year 2000 and included values for both injury and non-injury 
components (Table 2-6).  A breakdown of the number of police-reported and unreported 
accidents that occurred in 2000 were also given (Appendix B). 
 
 30 
Table 2 - 6 Total unit costs of injuries on the 
AIS 90 scale (NHTSA, 2002) 
Severity Cost 
(2000 Dollars) 
PDO $2,532 
MAIS 0 $1,962 
MAIS 1 $10,562 
MAIS 2 $66,820 
MAIS 3 $186,097 
MAIS 4 $348,133 
MAIS 5 $1,096,161 
Fatal $977,208 
 
2.4 Summary of Findings 
The literature review that was conducted yielded several key findings.  When the 
publications regarding the impacts that paved shoulders had on accidents and accident 
rates were examined in detail, there was found to be a considerable amount of variability 
between studies.  Inconsistencies in the evaluation criteria for accident and road data 
including shoulder classification, segment length, and shoulder width were found.  
Furthermore, the methodologies and statistical analyses used to determine the effects that 
the presence or absence of paved shoulders had on accidents and accident rates varied 
considerably.  Despite all of these differences, it was concluded, that in general, road 
segments with paved shoulders appeared to have fewer accidents and lower accident rates 
than segments without paved shoulders.   
There was an overall lack of consensus among the findings of the studies that 
examined the effects that shoulder widths have on accident rates.  One study concluded 
that shoulders that were 6-ft wide were safer than those that were wider or narrower 
(Belmont, 1954).  Another study concluded that accident frequency increases as shoulder 
width increases (Blensly and Head, 1960).  While two others concluded that the opposite 
 
 31 
was true, accident rates decreased as shoulder widths increased (Zegeer et al., 1981 and 
Zegeer et al., 1988).  The final two studies concluded that there were no statistically 
significant differences in the accident rates of different shoulder widths (Perkins 1956 
and Bowman and Abboud, 1999).  Again there were a considerable number of 
differences in the way these studies were conducted, the shoulder widths examined, the 
methodologies followed, and ultimately their results and conclusions. 
In the third section of this literature review a synopsis of CRF values that were 
currently being recommended for use was presented.  The CRF values from seven 
different state DOTs were identified and ranged from 12% to 20% (Table 2-2).  In 
addition, an equation that can be used to calculate an AMF value based on the shoulder 
width and shoulder materials was also examined (Equation 2-2) (NCHRP, 2005).   
Finally, an evaluation of the economic costs associated with automobile accidents 
was conducted.  These costs can be broken down into two categories: direct and indirect 
costs.  Direct costs can include tangible expenses such as medical costs, emergency 
services rendered, and legal fees.  Indirect costs however, may include expenses 
associated with travel delay and lost productivity.  Each of these categories are equally 
important when evaluating the economic implications that an accident may cause to both 
the individuals involved in the accident as well as society as a whole.  A unit cost was 
obtained for each level of injury severity on the AIS90 rating scale as well as for property 
damages (NHTSA, 2002).  These values were reported in terms of 2000 dollar amounts 
(Table 2-6).   
 
 
 32 
CHAPTER THREE 
METHODOLOGY 
The project funds that can be used by a state DOT to install paved shoulders are 
often limited.  The development of a detailed process that could assist a state DOT in 
identifying road segments that could benefit from the installation of 2-ft paved shoulders 
could be useful.  The widening of paved shoulders can reduce the number of run-off-road 
crashes and provide a recovery area for errant vehicles.  If this treatment were installed at 
the appropriate locations there could possibly be a significant reduction in accident 
occurrence and severity.  These potential reductions can be converted to economic 
savings in terms of the amount of money that is saved when an accident or injury is 
prevented.  A benefit/cost analysis can then be conducted to assess the locations that 
would benefit most from this treatment.  The greatest benefits are obtained at those 
locations that are able to maximize public safety in terms of reducing accidents and 
saving lives at a lower cost to the state DOT.   
The main objectives of this study were to identify high priority segments based on 
crash rate and severity and conduct a hypothetical benefit/cost analysis to determine the 
potential benefits of adding 2-ft of paved shoulders to these segments.  This required a 
detailed methodology to assist ALDOT in the identification of high priority segments that 
would benefit most from the shoulder treatment.   
 
 33 
The initial accident data set was obtained from the CARE 8 program and 
contained accident and location information for the years 2003 through 2006 (CRDL, 
2007).  CARE is a computer software program that can be used to conduct statistical 
analyses and inquiries of accident data.  During this study the CARE 8 program was also 
used to create a filter to identify two-lane rural roads with unpaved shoulders and conduct 
a ?high accident analysis.?  Many of the locations that were identified by this analysis 
were within close proximity of another ?hotspot.?  In response, a decision rule was 
developed and used to combine segments that were within 0.5-mile of one another.   
This new data set was then used to identify high priority segments based on two 
criteria: crash rates and accident severity.  Once these segments were identified, a 
hypothetical benefit/cost analysis was conducted to evaluate the potential benefits of 
installing 2-ft paved shoulders at these locations.  The actual benefits were not 
quantifiable since this would require several sections of roadway to have been fully 
operational for a minimum of three years after paved shoulder installation.  However, a 
preliminary evaluation of the potential benefits of applying this treatment to high priority 
segments can be beneficial to the decision making process.   
To calculate these potential benefits, a range of CRF values for each of the criteria 
was selected.  The benefits were calculated in terms of the number of accidents that could 
potentially be prevented as well as their associated costs.  The costs of installing 2-ft 
paved shoulders as part of resurfacing and reconstruction projects on two-lane rural 
highways were obtained from ALDOT and evaluated to determine an appropriate cost 
value to be used in the benefit/cost analysis.  After careful consideration the mean per-
mile cost of the projects that had been bid over a two year period from 2006 to 2008 was 
 
 34 
selected.  Finally, a hypothetical analysis of the benefits and costs of adding paved 
shoulders to these locations was conducted. 
3.1 Data Collection 
The CARE program was developed and is maintained by the University of 
Alabama CARE Research and Development Laboratory (CRDL, 2007).  This software 
package was used to access the accident database for the state of Alabama, which 
contains information regarding the type, severity, location, as well as road characteristic 
data associated with each accident.  Four years of accident data were obtained from the 
CARE 8 database and used to identify high priority segments of two-lane rural highways. 
3.1.1 Filter Creation and Implementation 
The original data set contained over 571,000 accident records and needed to be 
reduced to include only the accidents that occurred specifically on two-lane rural 
highways without paved shoulders.  To accomplish this task a filter was created and 
checked in accordance with the CARE 7 User Manual and the CARE 8 Individualized 
Training Manual (CRDL, 2007) (Appendix C).  The filter consisted of several lines of 
code that facilitated the identification of accidents that occurred on road segments that 
met specific criteria.  The filter was created based on three primary criteria: two-lane 
rural roads, ALDOT maintained highways, unpaved shoulders of any condition.  Using 
the created filter, the original data set was reduced to 17,814 accident records. 
 
 35 
3.1.2 High Accident Analysis 
Once the filter had been applied to the original data set, a narrower data set was 
created.  This new data set was then used during a ?high accident? analysis that was 
conducted using the CARE 8 software package (CRDL, 2007).  This analysis 
subsequently produced a list of ?hotspots? that were defined as 0.5-mile segments of 
roadway that had a minimum of 5 crashes of any type occur within the designated four 
year period.  This minimum value of 5 crashes was the lowest value that could be 
selected within the CARE program.  The ?high accident? analysis resulted in the 
identification of 1,305 hotspots that occurred on 116 different routes within the state of 
Alabama (Appendix D).  The data set that was produced included location information as 
well as accident, annual average daily traffic (AADT), million vehicle miles of travel 
(MVMT), crash rate (C/MVMT), and severity values for each location.   
3.1.3 Verification of Traffic Data 
Thirty hotspot locations were randomly selected from the data set of 1,305 
highway segments using the sampling function in the program Microsoft Excel to verify 
the reported AADT, MVMT, and crash rate values.  This was done using traffic data 
obtained from ALDOT?s traffic statistics website to ensure that the reported values from 
CARE were reliable and sound (ALDOT, 2008a).  An example of the traffic data 
verification process can be seen in Tables 3-1 and 3-2.  First, location data from one of 
the randomly selected segments, ID #40, were taken and used to identify the three closest 
traffic counting stations on ALDOT?s website.   
 
 
 36 
Table 3 - 1 Example data from actual road segment  
ID # 40 Beg MP End MP AADT Total Crashes MVMT C/MVMT 
 145.3 145.8 5130 5 3.74 1.34 
 
The traffic counting station located at mile post 142.975 is located approximately 2.3-mi 
before the beginning of the ID #40 road segment.  The traffic counting station located at 
mile post145.665 is located within the segment, and the traffic counting station located at 
148.26 is approximately 2.5-mi after the end of the ID #40 road segment (Table 3-2).   
Table 3 - 2 Example of ALDOT?s traffic data 
ALDOT Before 
 
ALDOT In 
 
ALDOT After 
Mile Post 142.975 
 
Mile Post 145.665 
 
Mile Post 148.26 
AADT 2006 5200 
 
AADT 2006 5220 
 
AADT 2006 6680 
AADT 2005 5110 
 
AADT 2005 5130 
 
AADT 2005 6590 
AADT 2004 4900 
 
AADT 2004 4900 
 
AADT 2004 6320 
AADT 2003 5310 
 
AADT 2003 5270 
 
AADT 2003 6890 
Average 5130 
 
Average 5130 
 
Average 6620 
 
Initially, the average AADT value was calculated for the years 2003 through 2006 
using the traffic data at each of ALDOT?s counting station locations (Table 3-2).  These 
average values were then compared to the CARE 8 values which were found to be 
exactly the same as at least one of the data collection points for 28 of the 30 locations that 
were examined.  The AADT values that were calculated for the two remaining locations 
were within five percent of the average AADT values found on ALDOT?s website 
(ALDOT, 2008a).  The five percent difference was determined to be sufficient for this 
analysis.   
Once the AADT values of the randomly selected hotspots were determined to be 
acceptable, the MVMT and crash rate values were checked using equations 3-1 and 3-2, 
 
 37 
respectively.  All of the calculated MVMT and crashes per MVMT values were found to 
be exactly the same as the values obtained from the CARE 8 program (Table 3-1). 
MVMT = (AADT * 365 days/yr* 4 yr * 0.5 mi)     (3-1) 
    (1,000,000) 
C/MVMT = Total Crashes        (3-2) 
MVMT 
3.1.4 Combining Road Segments 
Once the data had been checked for errors, it was visually examined as part of the 
preliminary analyses.  It was noted that many of the ?hotspot? locations were located on 
the same routes and were often within close proximity of another ?hotspot?.  Since the 
road segment length was originally defined to be 0.5-mile during the high accident 
analysis, it was not possible to get a segment of greater length even if it was a continuous 
segment of roadway.  The data set also contained consecutive segments in which one or 
two sections had a substantially lower number of crashes occur within them when 
compared to the others around them.   
One such example is shown in Table 3-3, in which the total number of crashes 
that occurred on the road segment ID #22 was much lower than the three consecutive 
road segments before and after it.  In practice, this entire section would likely receive the 
same treatment.  In addition, paving projects are generally greater than 0.5-mile in length 
so a decision rule was needed to determine the greatest distance that two 0.5-mile road 
segments could be apart and still be combined into the same segment. 
 
 
 38 
Table 3 - 3 Comparison of consecutive road segments 
ID # City Route Beginning MP Ending MP Distance To Next Segment (mi) Total Crashes 
19 Dallas S-14 117.6 118.1 0 17 
20 Dallas S-14 118.1 118.6 0 16 
21 Dallas S-14 118.6 119.1 0 16 
22 Dallas S-14 119.1 119.6 0.1 5 
23 Dallas S-14 119.7 120.2 0 12 
24 Dallas S-14 120.2 120.7 0 10 
25 Dallas S-14 120.7 121.2  15 
 
In order to develop a decision rule, a histogram was created to show the frequency 
and cumulative frequency distributions of the distances between the ending and 
beginning milepost markers of all of the consecutive 0.5-mile road segments (Appendix 
E).  The distances between two road segments on the same route were obtained by 
subtracting the beginning milepost marker of the latter road segment from the previous 
road segment (Table 3-4).   
Table 3 - 4 Determination of the distance between consecutive segments 
City Route Beginning MP Ending MP Length (mi) Distance To Next Segment (mi) 
Pickens S-14 11.7 12.2 0.5 25.9 
Greene S-14 38.1 38.6 0.5 1.3 
Greene S-14 39.9 40.4 0.5 2.8 
Greene S-14 43.2 43.7 0.5 4.1 
Greene S-14 47.8 48.3 0.5 0.5 
Greene S-14 48.8 49.3 0.5 4.4 
 
Figure 3-1 shows the same frequency distributions as Appendix E, however, only 
the road segments that were within 2.1-miles of one another were examined to determine 
a decision rule.  From Figure 3-1, it can be seen that a large number of segments had 0.1-
mile or less between them.  It was determined through visual inspection, that any two 
road segments that were equal to or less than 0.5-mile apart would be considered the 
same segment of road for the rest of the analysis.  This breakpoint was selected based on 
 
 39 
the frequency distribution which showed a reduction in the number of segments that had 
distances of greater than 0.5-mile between them (Figure 3-1).  This trend is also shown in 
the reduction of the slope of the line on the cumulative frequency distribution between 
the 0.5 and 0.6-mile segments.  From the cumulative frequency distribution it can also be 
shown that approximately 60% of the segments are within 0.5-mi of the next closest 
hotspot location.   
 
Figure 3 - 1 Frequency and cumulative frequency distributions of road segments less than 2.1-miles 
apart 
 
Based on this decision rule, the data set was then reconfigured to reflect this 
change.  When road segments were combined, the values for length, total crashes, crash 
 
 40 
rate, and severity were recalculated.  Following this procedure the segments were reduced 
from 1,305 segments that were all 0.5-mile in length to 607 segments of varying lengths.   
3.2 Data Analysis 
During the data analysis portion of this project a methodology for identifying high 
priority segments was established based on two criteria: crash rate and severity.  In 
addition, a hypothetical analysis of the potential benefits associated with the installation 
of 2-ft paved shoulders on the identified segments was conducted.  This analysis required 
a range of CRF values to be used to estimate the potential reductions in accidents if the 
shoulder treatment were to be implemented.  Finally, a methodology to evaluate the 
potential economic benefits associated with these reductions was also developed.   
3.2.1 Determination of High Priority Segments 
A chart containing a frequency and cumulative frequency distribution was created 
to identify high priority road segments for each of the criteria (Figures 3-2 and 3-3).  
These segments were defined as those that were greater than the 95th percentile for either 
crash rate or severity.  This analysis resulted in the identification of 31 high priority 
segments based on crash rate and 41 based on severity (Tables 3-5 and 3-6).  The 
locations that were classified as high priority segments were located on a State Highway 
Map to further examine the location and to check the proximity to any urban areas.   
 
 41 
 
Figure 3 ? 2 Frequency and cumulative frequency distributions used in high priority segment 
identification based on crash rate 
 
Figure 3 ? 3 Frequency and cumulative frequency distributions used in high priority segment 
identification based on severity 
 
 
 42 
Table 3 ? 5 List of high priority segments based on crash rate  
ID # Division City Route Length (mi) Beg MP End MP C/MVM  CRF (%) 
16 5 Perry S-14 0.5 93.7 94.2 19 2.8 
1285 5 Fayette S-107 0.5 7.4 7.9 18 2.0 
1282 1 Cherokee S-273 0.5 8.6 9.1 16 2.2 
1296 6 Bullock S-239 0.5 14.1 14.6 15 2.1 
525 5 Hale S-25 0.5 66 66.5 14 2.0 
912 4 Tallapoosa S-49 0.5 6.6 7.1 13 3.0 
1273 7 Coffee S-189 0.5 22.8 23.3 12 3.4 
950 2 Lawrence S-101 0.5 6.7 7.2 11 5.0 
1248 5 Chilton S-191 0.5 7.8 8.3 11 3.3 
1288 7 Barbour S-130 0.5 12.6 13.1 11 2.3 
1238 7 Geneva S-103 0.5 1.0 1.5 10 3.7 
1137 7 Dale S-105 1.0 8.9 9.9 10 3.5 
911 4 Tallapoosa S-49 0.5 5.4 5.9 10 2.5 
1274 7 Crenshaw S-189 0.5 31.2 31.7 10 2.5 
1224 5 Greene S-39 1.3 14.7 16.0 9 3.6 
522 5 Hale S-25 0.5 40.8 41.3 9 4.1 
1113 7 Covington S-12 0.5 119.6 120.1 9 5.3 
938 7 Coffee S-27 0.5 32.8 33.3 9 5.3 
949 7 Henry S-27 0.5 64.4 64.9 9 3.1 
1240 7 Geneva S-103 0.5 12.0 12.5 9 3.1 
526 5 Bibb S-25 0.5 80.8 81.3 8 2.8 
923 4 Tallapoosa S-49 1.3 36.7 38.0 8 3.3 
1135 7 Dale S-105 1.0 6.5 7.5 8 4.9 
1031 7 Coffee S-87 0.5 19.5 20.0 8 3.3 
924 4 Clay S-49 0.5 77.6 78.1 8 3.4 
1304 7 Dale S-92 0.5 3.3 3.8 8 3.4 
1225 5 Greene S-39 0.5 21.4 21.9 8 3.0 
523 5 Hale S-25 0.5 42.8 43.3 7 4.3 
1249 5 Chilton S-191 0.5 9.4 9.9 7 4.4 
1242 7 Pike S-125 1.3 19.7 21.0 7 3.6 
1190 5 Perry S-219 0.5 32.1 32.6 7 3.2 
 
 
 43 
Table 3 ? 6 List of high priority segments based on severity  
ID # Division City Route Length (mi) Beg MP End MP S/MVM CRF (%) 
926 7 Barbour S-51 0.5 28.5 29.0 34 4.5 
803 7 Geneva S-167 0.5 9.9 10.4 30 5.3 
349 8 Monroe S-21 0.5 51.9 52.4 30 5.3 
1185 2 Winston S-195 0.5 37.6 38.1 30 5.3 
1181 2 Colbert S-184 0.5 7.5 8.0 30 5.3 
444 5 Chilton S-3 0.5 225.0 225.5 30 5.3 
487 6 Bullock S-6 0.5 195.3 195.8 28 5.3 
1279 4 Coosa S-259 0.5 2.7 3.2 25 4.0 
335 4 Randolph S-22 0.5 151.1 151.6 24 5.3 
1280 4 Tallapoosa S-259 0.5 11.2 11.7 24 4.9 
14 5 Hale S-14 0.5 70.5 71.0 24 5.0 
90 6 Montgomery S-9 0.5 74.6 75.1 24 5.3 
6 5 Greene S-14 1.5 47.8 49.3 23 5.3 
645 6 Bullock S-15 0.5 138.6 139.1 23 5.3 
1094 5 Fayette S-18 0.5 36.0 36.5 23 5.3 
1155 1 Morgan S-157 1.2 23.6 24.8 23 5.3 
66 6 Elmore S-14 1.5 185.3 186.8 22 5.3 
1176 3 Shelby S-145 0.5 22.3 22.8 22 5.3 
555 5 Hale S-69 0.5 104.7 105.2 22 5.3 
1003 7 Crenshaw S-10 0.5 137.1 137.6 21 5.2 
953 2 Lawrence S-101 0.5 21.6 22.1 21 5.3 
81 7 Crenshaw S-9 0.5 58.6 59.1 21 5.3 
523 5 Hale S-25 0.5 42.8 43.3 21 4.3 
854 4 Tallapoosa S-50 1.6 2.5 4.1 21 4.1 
351 6 Lowndes S-21 0.5 103.8 104.3 20 4.1 
1211 7 Dale S-123 0.5 34.6 35.1 20 4.2 
639 7 Covington S-15 0.5 28.7 29.2 20 4.0 
1299 4 Russell S-26 0.5 8.5 9.0 20 5.3 
925 7 Barbour S-51 0.5 26.8 27.3 20 4.5 
133 4 Calhoun S-9 0.5 237.6 238.1 20 5.3 
1175 3 Shelby S-145 0.5 17.2 17.7 20 5.3 
1226 5 Hale S-60 0.5 1.2 1.7 20 5.3 
875 7 Coffee S-134 0.5 18.0 18.5 20 5.3 
979 4 Lee S-169 0.5 17.9 18.4 20 5.3 
870 7 Coffee S-134 0.5 7.4 7.9 20 5.3 
403 4 Calhoun S-21 0.5 276.1 276.6 20 5.3 
1152 1 Morgan S-157 0.5 20.0 20.5 20 5.3 
575 1 Cullman S-69 0.5 231.7 232.2 20 5.3 
773 5 Tuscaloosa S-171 2.3 8.7 11.0 19 5.3 
 
 44 
If a segment was considered to be in close proximity to a large urban area, or in 
an area that may have higher volumes of traffic it was located using aerial photography 
available from the Internet site, Google Earth (Google, 2008).  Each of the segments was 
examined to confirm that it was located on a rural two-lane highway.  This procedure was 
conducted to identify any segments that may have been widened during the four year 
analysis period.  One of the segments identified during the severity analysis was 
eliminated because it was located on a four-lane road.  This was considered to be the best 
alternative since the year that the widening was conducted was not known.   
3.2.2 Determination of CRF Values 
Once the two lists of high priority segments were located and visually examined 
using the Google Earth software, a range of CRF values were selected for each of the two 
criteria.  There were two primary reasons that a range of CRF values were selected rather 
than a single value.  First, the literature review had yielded a wide range of CRF values 
which were currently being recommended for use by different state and federal 
organizations.  Although no effective date was given, it was also known that ALDOT was 
planning to reduce their current CRF value from 20% down to 16% for shoulder paving 
projects conducted in the future (Sonya Baker, personal communication, May 28, 2008).   
The range of CRF values were determined by first calculating the CRF values for 
the individual segments on each of the corresponding lists, Tables 3-5 and 3-6, using 
Equations 2-1 and 2-2 (NCHRP, 2005).  These values were used to determine the lower 
limit CRF values which ranged from 2.0 to 5.3% for the segments that were selected 
based on crash rate and 4.0 to 5.3% for the high severity segments.  The lower limit CRF 
 
 45 
value for each of the criteria was set at the lowest calculated CRF value of the individual 
segments within it.  The upper limit was then set at 20% which was the highest value 
found in the literature, and the value that ALDOT was currently using.   
Once a range of CRF values was established, they were then applied to the total 
number of crashes that occurred on the high priority segments for each of the two criteria: 
crash rate and severity (Tables 3-7 and 3-8).  This was done to determine the potential 
reductions in accidents.  Tables 3-7 and 3-8 show the total number of crashes that 
occurred on all of the identified segments during the four-year analysis period based on 
each of the two criteria, which coincidentally happened to be 274 for both.  The lower 
and upper limit CRF values were then applied to the total number of crashes to determine 
the number of accidents that could be expected if the shoulder treatment were applied to 
all of the identified segments.  This process resulted in a lower and upper limit value for 
the number of expected crashes which ranged from 219 to 269 based on crash rate and 
219 to 263 based on severity.   
The upper limit actually has a lower number of crashes due to the upper limit 
CRF value.  This is because the higher CRF value corresponds to a greater reduction in 
accidents if 2-ft paved shoulders were to be installed at these locations.  The segments 
that were identified based on severity also show that there is potential to have 3 fewer 
accidents occur on these segments when compared to the segments identified based on 
crash rate.  This is due to the 2% higher lower limit value that was calculated from 
Equations 2-1 and 2-2 (NCHRP, 2005). 
 
 
 46 
Table 3 - 7 CRF values and expected   Table 3 ? 8 CRF values and expected 
number of crashes based on crash rate  number of crashes based on severity 
High Priority Segments by 
Crash Rate 
  High Priority Segments by 
Severity 
 
Total Crashes 274  Total Crashes 274 
CRF Lower Limit (LL) 2%  CRF Lower Limit (LL) 4% 
CRF Upper Limit (UL) 20%  CRF Upper Limit (UL) 20% 
Expected Crashes (LL) 269  Expected Crashes (LL) 263 
Expected Crashes (UL) 219  Expected Crashes (UL) 219 
 
3.3 Economic Evaluation of Potential Benefits 
To determine the potential benefits of adding 2-ft of paved shoulders to the high 
priority segments, the costs associated with each accident and injury were calculated.  
Initially, the reported values of the unit costs of an accident based on its severity were 
obtained from the 2002 NHTSA publication entitled The Economic Impact of Motor 
Vehicle Crashes (Table 2-6).  However, the injury data that was obtained from the CARE 
8 program only provided the number of injuries, fatalities, and reported property 
damages.  Since the unit costs in the NHTSA report were assigned for each level of the 
most current version of the abbreviated injury scale, AIS90, a weighted average of cost 
for the 1 through 6 ratings was calculated based on the total number of police-reported 
injury crashes that occurred in each category (Appendix B).  This resulted in the unit 
costs of each property damage, injury, and fatality in terms of a 2000 dollar amount, 
which were $2,532, $18,160, and $977,208 respectively.   
Once these unit costs were determined, they were adjusted using the gross 
domestic product (GDP) implicit price deflator to determine their average value over the 
four-year analysis period, 2003 through 2006.  The GDP is the dollar value of all services 
and goods that are produced within a country during an established period of time.  The 
 
 47 
implicit price deflator measures the year to year percent change in inflation or deflation 
with respect to the GDP.   
There were several factors that were taken into consideration when selecting an 
appropriate price index for this analysis.  It was important to select an index that would in 
some way account for the constant changes within the medical field to include 
pharmaceutical expenses, as well as goods and services rendered.  Since medical costs 
were only one contributor to the overall expense of an accident, it was equally important 
to select an index that would be representative of the other contributing expenses and 
their changes as well.  The GDP was selected based on the recommendations of the 
FHWA publication, Motor Vehicle Accident Costs (1994) and because it takes into 
account any new expenditure patterns within the country?s economy.  The GDP does this 
by accounting for not only the final goods and services produced, but their market value 
during a specific time period as well.   
Once a price index was selected, the inflation of the U.S. dollar from the year 
2000 to the average value over the analysis years, 2003 through 2006, was calculated.  
The GDP implicit price deflator values were obtained using Bloomberg Professional? 
service software for each quarter of every year from 2000 through 2006 (Bloomberg, 
2008) (Appendix F).  Several steps were followed to calculate the appropriate inflation 
value to apply to the unit cost data.  First the average value for each year was obtained.  
Next, the average inflation value for the years 2003 through 2006 was calculated and then 
applied to each of these years, respectively.  Then the sum of these average values was 
determined for the years 2001 through 2006 (Equation 3-3). 
 
 48 
Percent Inflation = ? average GDP implicit price deflator value for each year (3 ? 3) 
As a result, the inflation value to convert the unit cost data from 2000 dollars to the 
average value of the dollar over the four year analysis period was calculated to be 16%.  
When this inflation value was applied to the 2000 unit cost data it resulted in cost values 
of $2,937 per PDO, $21,065 per injury, and $1,133,561 per fatality.  These new unit costs 
were then used to determine a range of values for the total costs of the accidents and 
related injuries that occurred on the high priority segments (Tables 3-9 and 3-10). 
 The total number of PDOs, injuries, and fatalities that occurred over the four-year 
analysis period for each of the criteria are shown in Tables 3-9 and 3-10.  In addition, the 
expected number incidences at these locations after the installation of 2-ft paved 
shoulders are also shown.  These expected values were calculated using the CRF upper 
and lower limit values that were presented in section 3.2.2.  The costs of each PDO, 
injury, and fatality were then used to determine the total cost of these incidences on all of 
the segments combined based on the two criteria: crash rate and severity.  The results 
were a total combined cost that ranged from approximately $8,600,000 to $10,400,000 
for the crash rate high priority segments and $41,400,000 to $49,300,000 for the severity 
high priority segments (Tables 3-9 and 3-10).   
 
 49 
Table 3 - 9 Total costs of accidents and injuries on crash rate high priority segments 
 Total 
Expected based on 
CRF lower limit 
Expected based on 
CRF upper limit 
Total PDOs 172 169 138 
Total Cost: PDOs $      505,185  $      496,373  $      405,323  
Total Injuries 150 147 120 
Total Cost: Injuries $   3,159,785  $   3,096,589  $   2,527,828  
Total Fatalities 6 6 5 
Total Cost: Fatalities $  6,801,368  $   6,801,368  $   5,667,806  
Total Cost on the 
Segments $10,466,337  $10,394,330  $   8,600,957  
 
Table 3 - 10 Total costs of accidents and injuries on severity high priority segments 
 Total 
Expected based on 
CRF lower limit 
Expected based on 
CRF upper limit 
Total PDOs 73 71 59 
Total Cost: PDOs $    214,410 $     208,536 $     173,290 
Total Injuries 295 284 236 
Total Cost: Injuries $  6,214,244 $  5,982,527 $  4,971,395 
Total Fatalities 39 38 32 
Total Cost: Fatalities $44,208,890 $43,075,329 $36,273,961 
Total Cost on the 
Segments $50,637,544 $49,266,391 $41,418,646 
 
3.4 Costs of Shoulder Paving 
The costs of installing 2-ft paved shoulders vary from project to project.  This is 
primarily due to the variety in road geometry and site conditions that exist before 
construction begins.  In some locations there could be a severe pavement edge drop off 
that has to be reconstructed before the shoulder can be paved.  In other locations there 
may be a well-maintained gravel shoulder that requires a minimal amount of additional 
work other than paving.   
ALDOT has had a total of 41 resurfacing and reconstruction projects that have 
included the addition of 2-ft paved shoulders on two-lane rural highways from March 
 
 50 
2006 through February 2008.  The locations, project numbers, and percentage of the 
project funds that were designated for shoulder paving were provided by ALDOT (Clay 
McBrien, personal communication, June 9, 2008).  The bid amounts for 41 of the projects 
were obtained online from ALDOT?s Project Letting website (ALDOT, 2008b).  The 
road segments were located in 32 different counties and were of varying lengths, 2.04-
miles to 15.94-miles, and bid amounts, approximately $512,000 to $4,360,000 (Appendix 
G).  The cost of the shoulder paving portion of each project was calculated by 
multiplying the percentages provided by ALDOT by the total project amount.  These 
values were then divided by the length of the road segment to obtain the cost-per-mile for 
the shoulder treatment, which ranged from approximately $17,900 to $140,500 (Table 3-
11).   
Table 3 ? 11 Statistical values used in cost determination 
Per-mile Cost   Percentage of Project Cost  
Mean $44,901  Mean 19.4 
Median $44,500  Median 18.0 
Standard Deviation $22,666  Standard Deviation 7.4 
Maximum $140,516  Maximum 38 
Minimum $17,912  Minimum 7 
 
Since the cost-per-mile of installing 2-ft paved shoulders varied substantially, a 
frequency and cumulative frequency distribution was created to help determine an 
appropriate cost value to use in the analyses (Figure 3-4).  From the distributions it was 
shown that the majority of the projects fell within the $20,000 to $70,000 range with only 
three projects exceeding these cost-per-mile amounts.  The three projects that had a cost-
per-mile in excess of $70,000 were greater than 1 standard deviation away from the mean 
 
 51 
and were therefore eliminated from the dataset before any other cost analyses were 
conducted.   
 
Figure 3 ? 4 Frequency and cumulative frequency distributions for shoulder paving costs-per-mile 
 
Next, a statistical analysis was conducted to determine a single value that could be 
used during the benefit/cost analysis.  The mean cost-per-mile value of $44,900 was 
considered to be representative of the projects based on several considerations.  The 
sample size was considered to be large enough to encompass a broad range of projects 
and conditions.  The projects were also dispersed throughout the state and were 
representative of most locations.  Finally, with the exception of 3 projects, the per-mile 
cost of shoulder paving on the remaining 38 projects fell within the range of $20,000 to 
 
 52 
$70,000.  Furthermore, the median value of this range was $45,000, which is close to the 
calculated mean of $44,900 (Table 3-11). 
3.5 Benefit/Cost Analysis 
 The benefit/cost ratio is calculated by dividing the total benefits by the total costs.  
In this hypothetical benefit/cost analysis the benefits were defined in terms of the 
economic costs that would potentially be saved through accident prevention.  Since the 
cost to install the treatment varied, and the exact number of accidents that would be 
prevented if the shoulder treatment were installed on each of the high priority segments 
was not known, the calculated range of values for the benefits and the mean value for the 
associated costs were used.  A range of benefits was selected rather than a single value 
due to the broad range of CRF values that were reviewed.   
 
 53 
CHAPTER FOUR 
RESULTS 
 It has been shown that lane departure accidents are among the most severe and 
often result in run-off-the-road and head-on crashes.  Since a large percentage of fatal 
motor vehicle accidents have been shown to occur in rural areas, ALDOT established a 
new policy to install 2-ft of paved shoulders on two-lane rural highways throughout the 
state.  This type of shoulder treatment provides an additional recovery area for vehicles 
that leave the travel lane, and reduces the chances of overcorrection that often result in 
head-on collisions and roll-over accidents.   
A limited number of projects that have included the installation of 2-ft paved 
shoulders on two-lane rural highways have been initiated over the last two years.  As a 
result, a before and after study to quantify the actual reductions in accidents and severity 
in response to this shoulder treatment was not possible.  Since funding for improvements 
such as shoulder paving can be limited and the cost of such projects can be significant, a 
preliminary benefit/cost analysis was conducted to determine the potential benefits of the 
treatment.  This hypothetical analysis was based on the identification of high priority 
segments that could possibly benefit the most from the installation of 2-ft paved 
shoulders.   
 
 54 
4.1 Identification of High Priority Segments 
During this study, high priority segments were defined as those that had either a 
high crash rate or severity.  The two lists of high priority segments contained 
approximately the top 5% of the segments that were originally obtained from the high 
accident analysis conducted using the CARE 8 software program.  There were a total of 
31 road segments identified based on crash rate and 40 based on severity (Tables 3-5 and 
3-6).  Only one of the identified segments appears on both lists, ID # 523, which is 
located on State Route 25 in Hale County.   
4.2 Calculated Benefits 
To quantify the potential benefits of widening the shoulders of the high priority 
segments, the economic savings in terms of the number of accidents prevented by the 
treatment were calculated.  A range of CRF values were used in this analysis to calculate 
the hypothetical number of accidents that would be prevented if this shoulder treatment 
were installed on all of the high priority segments.  The savings that would be obtained 
from the prevention of all of the associated injuries, fatalities, and PDOs were calculated 
on each segment.  These values along with the segment length were utilized to determine 
a per-mile savings.  The total lengths of the segments designated by each of the criteria 
were 18.90-mi for those identified based on crash rate and 25.10-mi for the severity 
segments.  The total savings for each criteria and CRF limit were divided by the total 
length of all of the segments within each of the criteria to get a savings per-mile value 
($1,865,380/18.90-miles = $98,697/mile) (Tables 4-1 and 4-2).   
 
 55 
Table 4 - 1 Potential savings if shoulder treatment is implemented on high 
priority segments based on crash rate 
 
Expected based on 
CRF lower limit 
Expected based on 
CRF upper limit 
Total PDOs 3 34 
Total Savings: PDOs  $                     8,811   $                  99,862  
Total Injuries 3 30 
Total Savings: Injuries  $                   63,196   $                631,957  
Total Fatalities 0 1 
Total Savings: Fatalities  $                          -     $             1,133,561  
Total Savings on the Segments  $                   72,007   $             1,865,380  
Total Savings per mile  $                     3,810   $                  98,697  
 
Table 4 - 2 Potential savings if shoulder treatment is implemented on high 
priority segments based on severity 
 
Expected based on 
CRF lower limit 
Expected based on 
CRF upper limit 
Total PDOs 2 14 
Total Savings: PDOs  $                  5,874   $                41,120  
Total Injuries 11 59 
Total Savings: Injuries  $              231,718   $           1,242,849  
Total Fatalities 1 7 
Total Savings: Fatalities  $           1,133,561   $           7,934,929  
Total Savings on the Segments  $           1,371,153   $           9,218,897  
Total Savings per mile  $                54,628   $              367,287  
 
It can be seen in Table 4-1 that, applying the CRF upper limit would result in a 
total expected reduction of 34 PDO crashes, 30 injuries, and 1 fatality when applied to 
the high priority segments that were identified based on crash rate.  The expected savings 
associated with the prevented PDOs, injuries, and fatalities were, $99,862, $631,957, and 
$1,133,561, respectively (Table 4-1).  It should also be noted that the injury value was 
calculated based on the weighted average of injury severity from police-reported 
accidents that occurred within the U.S. in the year 2000.  As a result, these dollar 
amounts may be conservative since it has been shown that the types of accidents that are 
 
 56 
generally prevented by installing paved shoulders tend to be more severe.  The calculated 
total savings-per-mile ranged from approximately $3,800 to $98,700 based on crash rate 
and $54,600 to $367,300 based on severity.   
4.3 Calculated Costs 
There were a total of 41 ALDOT projects that included the installation of 2-ft 
paved shoulders that had been let to bid since 2006 for which cost data were available.  
The cost-per-mile for the shoulder installation on these projects ranged from 
approximately $17,900 to $140,500 (Appendix G).  Since such a broad range of values 
existed, a statistical analysis was conducted and the mean cost-per-mile value of $44,900 
was selected for use in the benefit/cost analysis.   
4.4 Benefit/Cost Analysis 
 The benefit/cost analysis determined the associated benefits in terms of the 
economic savings associated with accidents, injuries, and fatalities that would be 
expected to be prevented by the installation of 2-ft paved shoulders.  The ratio of benefits 
to costs for the range of economic savings values ranged from 0.08 to 2.20 for the high 
priority segments that were identified based on crash rate, and 1.22 to 8.18 for the 
segments that were identified based on severity (Tables 4-3 and 4-4).   
The hypothetical results of the benefit/cost analyses indicate that the 
implementation of the shoulder treatment on either set of high priority segments would be 
beneficial to ALDOT as well as the traveling public.  However, it appears that the 
greatest benefit will result from the installation of 2-ft paved shoulders on the high 
 
 57 
priority segments that were selected based on severity.  The midpoint values for each of 
the criteria were 1.14:1 and 4.70:1, for the crash rate and severity respectively.   
Table 4 - 3 Benefit/Cost ratios for high priority segments based on crash rate 
 
Expected based on CRF lower limit Expected based on CRF upper limit 
Total Savings per mile $  3,810 $98,697 
Mean cost per mile $44,900 $44,900 
Benefit/Cost ratio 0.08 2.20 
Table 4 - 4 Benefit/Cost ratios for high priority segments based on severity 
 
Expected based on CRF lower limit Expected based on CRF upper limit 
Total Savings per mile $54,628 $367,287 
Mean cost per mile $44,900 $44,900 
Benefit/Cost ratio 1.22 8.18 
 
 
 58 
CHAPTER FIVE 
CONCLUSIONS AND RECOMMENDATIONS 
The overall results showed that the installation of 2-ft paved shoulders on two-
lane rural highways in the State of Alabama has the potential to greatly reduce accidents. 
However, the results and conclusions of this study were based on a hypothetical analysis 
of the potential benefits since the new policy to install this type of shoulder treatment was 
only implemented by ALDOT in 2006.  Although further research has been 
recommended to help quantify the results of this study, a minimum of three years of after 
data is needed.  In addition, the methodology given in this study could also be used in the 
identification of high priority segments for this future research.   
5.1 Conclusions 
The high priority segments that were identified based on crash rates and severity 
both yielded positive benefits that ranged from a total savings-per-mile of approximately 
$3,800 to $98,700 and $54,600 to $367,300, respectively.  These hypothetical savings 
reflect the total dollar amounts that could be saved through the prevention of property 
damage, injuries, and fatalities on these segments.  Utilizing the mean cost-per-mile of 
$44,900 for the installation of 2-ft paved shoulders resulted in benefit/cost ratios that 
ranged from 0.08:1 to 2.20:1 based on crash rate, and from 1.22:1 to 8.18:1 based on 
severity 
 
 59 
5.2 Recommendations for ALDOT 
The methodology that was used to identify high priority segments based on the 
two criteria, crash rate and severity, appears to be a useful tool in selecting road segments 
to receive this type of shoulder treatment.  The hypothetical analysis revealed that there 
are potential savings that can be obtained if 2-ft paved shoulders are applied to road 
segments that fall within the top 5% based on the established criteria.  ALDOT could use 
this analysis to further expand either list of locations beyond the top 5%.  The 
methodology provided in this study can also be used to examine a different analysis 
period that may be of particular interest to ALDOT.  It is also recommended that road 
segments with high severity crashes should be given extra consideration for this treatment 
since the results of the hypothetical evaluation suggest that they will experience the 
greatest benefits in terms of accident prevention.   
5.3 Recommendations for Future Research 
After sufficient time has passed, minimum three years from completion of 
construction, it is recommended that a before-and-after study be conducted to quantify 
the reductions or changes in accidents, injuries, and property damages that resulted from 
the installation of 2-ft paved shoulders.  A before-and-after study could also be used to 
determine a more accurate range of CRF values.  In the future it would be beneficial to 
apply the treatment to several of the high priority segments that were identified during 
this analysis to determine whether the benefit/cost ratios fall within the calculated ranges. 
A life-cycle cost analysis is also recommended since this study only examined the 
savings based on a four-year analysis period and the initial construction costs.  It is 
 
 60 
suspected that the savings over the lifetime of the pavement when compared to the 
installation and maintenance costs could potentially yield an even higher benefit/cost 
ratio.  This type of analysis could also be useful in determining long term savings.   
 
 61 
REFERENCES 
AAA Foundation for Traffic Safety.  (2005).  Safety impacts of pavement edge drop-offs.  
Washington D.C.:  AAA Foundation for Traffic Safety. 
Agent, K.  R., Stamatiadis, N., & Jones, S.  (June 1996).  Development of accident 
reduction factors.  Retrieved May 21, 2008, from 
http://www.its.pdx.edu/CRF/reports/Kentucky_CRF.pdf 
Alabama Department of Transportation.  (2008a).  Alabama Traffic Statistics Version 2.0.  
Retrieved May 07, 2008, from 
http://aldotgis.dot.state.al.us/trafficvolume/viewer.htm 
Alabama Department of Transportation.  (2008b).  Project Letting Information.  
Retrieved June 10, 2008, from 
http://www.dot.state.al.us/docs/Bureaus/Office+Engineer/Project+Letting/Project
+Letting.htm 
Arizona Department of Transportation.  (June 2004).  ADOT Traffic engineering policies, 
guidelines, and procedures Section 200 ? Traffic studies 231 Benefit/Cost ratio 
economic analysis.  Retrieved May 22, 2008, from 
http://www.azdot.gov/Highways/Traffic/standards/PGP/TM231.pdf 
 
 62 
Bahar, G., Masliah, M., Wolff, R., & Park, P.  (2007).  Desktop reference for crash 
reduction factors.  (U.S. Department of Transportation Federal Highway 
Administration.  Report No.  FHWA-SA-07-015).  Springfield, VA: National 
Technical Service. 
Belmont, D.M.  (1954)  Effect of shoulder width on accidents on two-lane tangents.  
Highway Research Board, 91, 29-31. 
Blensly, R.C., & Head, J.A.  (1960).  Statistical determination of effect of paved shoulder 
width on traffic accident frequency.  Highway Research Board, 240, 1-23. 
Bloomberg Finance L.P.  (2008).  Bloomberg Professional? Service Software [computer 
software].  New York: New York City.  Retrieved June 11, 2008. 
Bowman, B. L., & Abboud, N.  (1999).  Evaluation of two and four foot shoulders on 
rural two lane Alabama roadways.  Auburn University Report No.  IR-99-02, 1-49. 
CARE Research & Development Laboratory.  (2007).  CARE 8 [computer software].  
Alabama: Birmingham.  Retrieved September 28, 2007, from 
http://care.cs.ua.edu/default.aspx. 
Dixon, K.  (2005).  Final Summary Report: Southeastern United States fatal crash study.  
Report No.  FHWA-GA-05-9815, 1-4. 
Federal Highway Administration.  (1994).  Motor Vehicle Accident Costs.  (FHWA 
Technical Advisory T 7570.2).  Retrieved May 21, 2008, from 
http://www.fhwa.dot.gov/legsregs/directives/techadvs/t75702.htm 
 
 63 
Federal Highway Administration.  (1996).  The 1996 annual report on highway safety 
improvement programs.  (U.S. DOT Publication No. FHWA-SA-96-040).  
Springfield, VA: National Technical Information Service. 
Federal Highway Administration.  (2005).  Crash cost estimates by maximum police-
reported injury severity within selected crash geometries.  (U.S. DOT Publication 
No.  FHWA-HRT-05-051).  Springfield, VA: National Technical Information 
Service. 
Federal Highway Administration.  (2006).  Priority, Market-Ready Technologies and 
Innovations.  (U.S. DOT Publication No.  FHWA-HRT-06-048).  Retrieved June 
18, 2008 from http://www.fhwa.dot.gov/crt/lifecycle/rumblestrips.pdf 
Google.  (2008).  Google Earth 4.3 [computer software].  Retrieved September 19, 2007 
from http://earth.google.com/download-earth.html 
Gross, F., & Jovanis, P.P.  (2007).  Estimation of the safety effectiveness of lane and 
shoulder width: Case-control approach.  Journal of Transportation Engineering, 
June 2007, 362-369. 
Heimbach, C.L., Hunter, W.W., & Chao, G.C.  (1974).  Paved highway shoulders and 
accident experience.  Transportation Engineering Journal, 100, 889-907. 
Kochanek, K. D., Murphy, S. L., Anderson, R., N., & Scott, C.  (2004).  Deaths: final 
data for 2002.  National Vital Statistics Report, 53(5), 2. 
 
 64 
Monsere, C. M., Bertini, R. L., Breakstone, A., Bonner, C., Bosa, P., Houssaye, D., et al.  
(2006).  Update and enhancement of ODOT?S crash reduction factors, final 
report (ODOT Publication No. SPR 162).  Retrieved May 22, 2008, from 
http://www.oregon.gov//ODOT/TD/TP_RES/ 
National Cooperative Highway Research Program.  (2003).  Guidance for 
implementation of the AASHTO strategic highway safety plan (NCHRP Report 
500 (6): A guide for addressing run-off-road collisions).  Washington, DC: 
Transportation Research Board. 
National Cooperative Highway Research Program.  (2005).  Crash reduction factors for 
traffic engineering and intelligent transportation system (ITS) improvements: 
state-of-knowledge report.  Research Results Digest, 299, 1-31. 
National Highway Traffic Safety Administration.  (2002).  The economic impact of motor 
vehicle crashes 2000.  (U.S.  DOT Publication No.  DOT-HS-809-446).  
Washington, DC: National Highway Traffic Safety Administration. 
National Highway Traffic Safety Administration.  (2005b).  Contrasting rural and urban 
fatal crashes 1994-2003.  (NHTSA Publication No. DOT HS 809 896).  
Springfield, VA: National Technical Information Service.   
National Highway Traffic Safety Administration.  (2005a).  Motor vehicle traffic crashes 
as a leading cause of death in the U.S., 2002 ? A demographic perspective 
(NHTSA Publication No.  DOT HS 809 843).  Springfield, VA: National 
Technical Information Service.   
 
 65 
North Carolina Department of Transportation.  (2007).  North Carolina Project 
Development Crash Reduction Factor Information.  Retrieved May 22, 2008, 
from 
http://www.ncdot.org/doh/PRECONSTRUCT/traffic/Safety/ses/project_guide/reg
ionalfactors.pdf 
New York State DOT.  (2000).  Accident Reduction Factors.  Retrieved May 22, 2008, 
from 
https://www.nysdot.gov/portal/page/portal/divisions/operating/osss/highway/accid
ent-reduction 
Ogden, K.W.  (1997).  The effects of paved shoulders on accidents on rural highways.  
Accident Analysis and Prevention, 29, 353-362. 
Perkins, E.T.  (1956).  Relationship of accident rate to highway shoulder width.  Highway 
Research Board, 151, 13-14. 
Rogness, R.O., Fambro, D.B., & Turner, D.S.  (1982).  Before-after accident analysis for 
two shoulder upgrading alternatives.  Transportation Research Record, 855,41-
47. 
Tople, A. R.  (1998).  Development of South Dakota accident reduction factors.  South 
Dakota Department of Transportation.  Retrieved May 22, 2008 from 
http://www.state.sd.us/Applications/HR19ResearchProjects/Projects/SD1998_13_
final_report.pdf 
 
 66 
Turner, D.S., Fambro, D.B., & Rogness, R.O.  (1981).  Effects of paved shoulders on 
accident rates for rural Texas highways.  Transportation Research Record, 819, 
30-37. 
Zaloshnja, E., Miller, T., Romano, E., & Spicer, R.  (2004).  Crash costs by body part 
injured, fracture involvement, and threat-to-life severity, United States, 2000.  
Accident Analysis and Prevention, 36, 415-427. 
Zegeer, C.V., & Council, F.M.  (1995).  Safety relationships associated with cross-
sectional roadway elements.  Transportation Research Record, 1512, 29-36. 
Zegeer, C.V., Deen, R.C., & Mayes, J.G.  (1981).  Effect of lane and shoulder widths on 
accident reduction on rural, two-lane roads.  Transportation Research Record, 
806, 33-43. 
Zegeer, C.V., Reinfurt, D.W., Hummer, J., Herf, L., & Hunter, W.  (1988).  Safety effects 
of cross-section design for two-lane roads.  Transportation Research Record, 
1195, 20-32. 
Zegeer, C.V., Stewart, R., Council, F., & Neuman, T.R.  (1994).  Accident relationships 
of roadway width on low-volume roads.  Transportation Research Record, 1445, 
160-168. 
Zwahlen, H. T., & Suravaram, K. R.  (May 2007).  Crash Reduction Factors for 
Education and Enforcement.  Retrieved May 22, 2008, from 
http://www.dot.state.oh.us/research/2007/Safety/134220-ES.pdf 
 
67 
APPENDIX A 
ACCIDENT MODIFICATION FUNCTION (NCHRP, 2005) 
 
 
68 
APPENDIX B 
ECONOMIC COSTS OF ACCIDENTS AND TOTAL POLICE-REPORTED 
INCIDENCES (NHTSA, 2002) 
 
 
69 
APPENDIX C 
FILTER CREATION CARE 8 (CRDL, 2007) 
 
 
70 
APPENDIX D 
HIGH ACCIDENT ANALYSIS CARE 8 (CRDL, 2007) 
 
 
71 
APPENDIX E 
DECISION RULE FREQUENCY AND CUMULATIVE FREQUENCY 
DISTRIBUTIONS 
 
 
72 
APPENDIX F 
GDP IMPLICIT PRICE DEFLATOR TRENDS AND VALUES (Bloomberg, 2008) 
 
 
 
73 
APPENDIX G 
ALDOT SHOULDER WIDENING PROJECTS 
Letting 
Date Projects in Contract 
Project 
Length 
(mi) Low Bid Percentage  
Widening 
Cost 
Per-mile 
Cost 
3/31/06 
STPSA-0006(518)        
99-307-036-006-601 8.143 $1,957,472 17 $332,770 $40,866 
3/31/06 
STPSA-0036(500)         
99-302-402-036-601 2.75 $511,788 22 $112,593 $40,943 
3/31/06 
STPSA-0176(501)        
99-301-253-176-601 13.67 $2,539,434 17 $431,704 $31,580 
4/28/06 
STPSA-0007(511)         
99-305-323-007-601 7.875 $1,514,114 20 $302,823 $38,454 
4/28/06 
STPSA-0014(512)            
99-306-015-014-601 14.719 $1,837,189 16 $293,950 $19,971 
4/28/06 
STPSA-0034(500)         
99-303-582-034-611 4.622 $1,057,505 11.22 $118,652 $25,671 
4/28/06 
STPSA-0036(501)          
99-301-521-036-602 5.25 $1,064,375 17 $180,944 $34,465 
4/28/06 
STPSA-0139(501)         
99-305-114-139-602 3.458 $704,092 10 $70,409 $20,361 
4/28/06 
STPSA-0147(501)                
99-304-093-147-601 4.18 $1,139,020 20 $227,804 $54,499 
4/28/06 
STPSA-0172(500)          
99-302-473-172-601 5.139 $1,315,008 7 $92,051 $17,912 
4/28/06 
STPSA-0139(500)         
99-305-044-139-601 9.126 $2,020,803 15 $303,120 $33,215 
 
74 
Letting 
Date Projects in Contract 
Project 
Length 
(mi) Low Bid Percentage 
Widening 
Cost 
Per-mile 
Cost 
5/26/06 
STPSA-0253(500)        
99-302-473-253-601 5.866 $1,086,158 10 $108,616 $18,516 
6/30/06 
STPSA-0048(500)                
99-304-144-048-601                    
STPSA-0048(501)                
99-304-564-048-601 15.489 $3,375,620 25 $843,905 $54,484 
6/30/06 
STPSA-0204(513)         
99-304-082-204-601 9.29 $2,196,101 30 $658,830 $70,918 
7/28/06 
STPSA-0001(528)         
99-301-285-001-603 7.11 $2,210,367 12.2 $269,665 $37,928 
7/28/06 
STPSA-0025(510)                  
99-308-462-025-606 5.484 $1,333,912 24.5 $326,808 $59,593 
7/28/06 
STPSA-0019(503)             
99-302-302-019-601 3.56 $985,596 8 $78,848 $22,148 
7/28/06 
STPSA-0025(511)       
99-303-595-025-617 8.08 $1,595,567 12.22 $194,978 $24,131 
7/28/06 
STPSA-0021(512)                
99-306-434-021-601 11.16 $1,636,319 18 $294,537 $26,392 
8/25/06 
STPSA-0015(512)                
99-306-062-015-601 6.396 $1,584,405 13 $205,973 $32,203 
11/3/06 
STPSA-0245(500)             
ST-007-245-001              
99-306-074-245-601 2.044 $992,342 13 $129,004 $63,114 
12/1/06 
STPSA-0091(502)            
99-301-224-091-703 12.97 $2,768,234 16.79 $464,786 $35,836 
1/12/07 
STPSA-0015(514)             
99-306-442-015-701 14.536 $2,633,906 18 $474,103 $32,616 
1/12/07 
STPSA-0079(505)             
99-301-484-079-708 10.44 $1,835,240 17.3 $317,497 $30,412 
1/12/07 
STPSA-0001(529)            
99-301-285-001-705 6.37 $1,597,510 9 $143,776 $22,571 
 
 
75 
Letting 
Date Projects in Contract 
Project 
Length 
(mi) Low Bid Percentage 
Widening 
Cost 
Per-mile 
Cost 
2/28/07 
STPSA-0021(515)               
99-306-434-021-701 10.929 $1,677,728 24 $402,655 $36,843 
2/28/07 
STPSA-0185(500)                
99-306-074-185-701 7.249 $1,297,203 32 $415,105 $57,264 
3/28/07 
STPSA-0140(500)             
99-306-245-022-701           
99-306-245-140-701 13.825 $2,761,286 25 $690,322 $49,933 
3/28/07 
HSIP-0069(512)           
99-308-134-069-708           
99-508-134-069-715 7.074 $2,088,546 20 $417,709 $59,049 
3/28/07 
STPSA-0025(513)        
99-308-462-025-703      
99-308-662-025-702 4.579 $1,296,533 22 $285,237 $62,292 
3/28/07 
STPSA-0089(500)     
99-308-663-089-706     3.135 $671,467 27 $181,296 $57,830 
4/27/07 
HSIP-0026(500)          
99-304-576-026-701 14.545 $3,871,216 34 $1,316,213 $90,493 
11/2/07 
HSIP-0021(518)          
99-308-503-021-802 5.95 $1,620,960 24 $389,030 $65,383 
11/2/07 
HSIP-0047(501)           
99-308-503-047-804 5.23 $1,174,571 23 $270,151 $51,654 
1/18/08 
HSIP-0005(518)          
99-305-044-005-801 6.434 $1,638,500 18 $294,930 $45,839 
1/18/08 
HSIP-0025(515)          
99-305-335-025-801 11.782 $4,356,727 38 $1,655,556 $140,516 
1/18/08 
HSIP-0107(500)          
99-305-291-107-801 15.944 $3,281,717 24 $787,612 $49,399 
1/18/08 
HSIP-0028(504)         
99-308-692-028-806 6.281 $1,522,572 24 $365,417 $58,178 
2/29/08 
HSIP-0134(502)           
99-307-234-134-801 6.833 $1,577,698 15 $236,655 $34,634 
2/29/08 
HSIP-0051(501)           
99-307-162-051-801 5.762 $1,047,532 32 $335,210 $58,176 
2/29/08 
STPSA-0141(501)        
99-307-162-141-802       
STPSA-0166(500)           
99-307-162-166-803 7.169 $1,552,237 16 $248,358 $34,643