Guidelines for Integrating Safety and Cost-Effectiveness into Resurfacing, Restoration, and Rehabilitation (3R) Projects (2021)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

19   Managing a 3R Program to Reduce Crash Frequency and Severity Resurfacing, restoration, and rehabilitation programs are managed by highway agencies as part of their effort to preserve and extend the service life of pavements. However, 3R projects also provide an opportunity to consider the need for design improvements to reduce crash frequency and severity. When Federal-Aid Highway funding is involved, safety improvement needs must be considered in 3R projects, and most highway agencies also consider safety improvement needs in 3R projects that do not involve federal funds. This chapter shows how 3R programs can be managed not only to preserve pavements and extend their service life, but also to be a key part of each highway agency’s efforts to reduce the frequency and severity of traffic crashes on the roadway system. This goal can be achieved through a structured design process that considers the need for design improvements in each 3R project, with the objective of implementing such improvements in locations where a crash history review identifies an existing crash pattern or where an economic analysis indicates that a design improvement would be cost-effective. Prioritizing improvement needs is critical to using available funds efficiently and maximizing the crash reduction benefits from 3R improvements, so it is important that design improvements be made as part of 3R projects where needed and also that design improvements not be made where there is no demonstrated need and where such improvements would not be cost-effective. 4.1 Role of 3R Projects in Overall Safety Management Programs of Highway Agencies Highway agencies and their safety partners undertake a broad range of activities to reduce crash frequency and severity. These may include • Monitoring the crash history of the roadway system and making design or traffic control improvements where potentially correctable crash patterns are identified; • Conducting systemic reviews of the roadway system and making design or traffic control improvements where such improvements appear to be cost-effective on the basis of their potential reduction of expected or predicted future crashes, whether or not there is a history of observed crashes; • Reviewing plans for highway infrastructure projects, as they develop, to identify opportunities for incorporating design or traffic control changes in specific projects to reduce crash frequency or severity; • Coordinating with law enforcement agencies to prioritize enforcement activities and focus enforcement at locations where it is most needed; • Conducting public education and driver training programs to promote highway safety awareness and safe driving practices among the traveling public; and C H A P T E R   4

20 Guidelines for Integrating Safety and Cost-Effectiveness into Resurfacing, Restoration, and Rehabilitation (3R) Projects • Encouraging improvement of emergency medical services to assist crash victims in receiving on-site and in-hospital treatment as soon as practical after a crash occurs. All types of highway infrastructure improvement projects, including 3R projects, help to address the first three aspects of highway safety management listed above. The list of highway safety activities presented above makes clear that, while highway safety management includes monitoring crash history and developing projects to address identified crash patterns, it is in no way limited to that. Similarly, highway safety management should not be seen as limited to projects funded from safety-specific funding sources such as the Highway Safety Improvement Program. Any project, regardless of why it is initiated and how it is funded, can include features intended specifically to reduce crash frequency and severity. Thus, all types of projects, including 3R projects, can play an important role in reducing crash frequency and severity. Rational management of safety considerations in 3R projects requires an understanding of the crash reduction effectiveness of specific design features that may be considered as safety improvements in 3R projects. The crash reduction effectiveness of design features is typically presented in the form of CMFs, as explained in Section 4.2. 4.2 Quantifying the Crash Reduction Effectiveness of 3R Improvements: Crash Modification Factors Management of a 3R program to effectively reduce crashes requires an understanding of the crash reduction effectiveness of design improvements that can potentially be made as part of 3R projects. The following discussion introduces the concept of CMFs to represent the crash reduction effectiveness of design improvements, defines CMFs, and describes their use. The values of CMFs for specific 3R improvement types are presented in Section 4.3. 4.2.1 Introduction to the Concept of Crash Modification Factors CMFs represent the relative change in crash frequency due to a change in a roadway design feature or any other specific condition. A CMF represents the ratio of the crash frequency of a site under two different conditions, for example, after a project in comparison with before a project. Therefore, a CMF serves as an estimate of the effect of a particular geometric design or traffic control feature or the effectiveness of a particular treatment or condition. Some CMFs apply to specific crash severities, crash types, and roadway types or may only be applicable to specific ranges of traffic volumes. As an example, if a change in a specific design feature is expected to reduce the crash frequency at a site from 10 to 8 crashes per year, the corresponding CMF value would be 8/10 = 0.80. A CMF value less than 1.0 indicates that the design feature or treatment is expected to reduce crashes. A CMF value greater than 1.0 indicates that the design feature or treatment would be expected to increase crashes. A CMF value equal to 1.0 indicates that the design feature or treatment would not be expected to have any effect on crashes. CMFs are a useful form for expressing the expected crash reduction effectiveness of design features or treatments, because the expected crash frequency for a site after a project can be estimated as the expected crash frequency before the project multiplied by the CMF. The crash reduction effectiveness of design features or treatments can also be expressed as a crash reduction factor, which represents the expected percentage of reduction in crashes result- ing from a design feature or treatment. The CMF value of 0.80 presented above corresponds to a crash reduction factor of 20%, determined as follows:

Managing a 3R Program to Reduce Crash Frequency and Severity 21   ( )= − × =crash reduction factor 1.00 0.80 100 20% (1) Similarly, a CMF value of 1.20 corresponds to a crash reduction factor of −20% (that is, an increase in crash frequency of 20%), determined as follows: ( )= − × = −crash reduction factor 1.00 1.20 100 20% (2) 4.2.2 Quality and Uncertainty Issues Related to Crash Modification Factors CMFs represent the expected or average effects of geometric design features on crash frequency and severity. CMFs are developed from data for individual sites at which the crash counts and the change in crash counts from before to after a project may vary substantially. CMFs represent the average effect of future infrastructure improvements on crash counts, but as in the data used to develop CMFs, such effects are likely to vary substantially from site to site. Some CMFs are more precise and reliable than others. For many of the CMFs presented in HSM Part D, “Crash Modification Factors” (2), an accompanying standard error value that represents the degree of uncertainty associated with predictions made with the CMF is also presented. For example, HSM Part D gives the CMF applicable to all types and all severities of crashes for installing centerline rumble strips on rural two-lane highways as CMF = 0.86, standard error = 0.05. The CMF value of 0.86 indicates that installation of centerline rumble strips would be expected to reduce crashes by 14%. This can be demonstrated by replacing the value of 0.80 in Equation 1 with 0.86. The standard error represents the precision of the CMF value. The confidence range around the CMF value permits assessment of its accuracy. Confidence limits at the 95% confidence level typically range from the CMF value minus twice the standard error to the CMF value plus twice the standard error. This CMF value of 0.86 is considered reliable because a CMF value of 1.0 (no effect on crashes) is not included within the range 0.86 ± 2 p (0.05), or 0.76 to 0.96. This range indicates that the effectiveness of center- line rumble strips can range from a CMF of 0.76 to a CMF of 0.96, or a reduction of crashes in the range of 4% to 24%. Such uncertainties are typical of even the most reliable CMFs, given the variations inherent in crash data. While the crash reduction effectiveness of centerline rumble strips at individual sites may vary over the range shown (4% to 24%), the overall effectiveness of applying centerline rumble strips to a broad range of rural two-lane highway sites is likely to be about 14%. Highway agencies can proceed with planning 3R programs with the expectation that the available CMFs will represent the average results to be expected. The HSM has used a formal inclusion rule to determine whether CMFs are of sufficient quality to be applied reliably (2, 12). With limited exceptions, only CMFs that meet the inclusion rule are presented in the HSM, Part D. Thus, CMFs from the HSM can be relied upon in planning 3R programs. CMFs are also available from the FHWA Crash Modification Factors Clearinghouse (13). The FHWA Clearinghouse assigns star ratings to CMFs to represent their quality and identi- fies the facility type(s) to which each CMF applies. Star ratings for CMFs range from one star (lowest quality) to five stars (highest quality). The ratings are based on an assessment of the research approach and the data used in developing the CMFs. The philosophy of the FHWA Clearinghouse is to include all CMFs reported in the literature, regardless of quality, and to let the star ratings indicate the quality of individual CMFs. Thus, CMFs rated as one or two stars in the FHWA Clearinghouse are derived from research or data of low quality and should

22 Guidelines for Integrating Safety and Cost-Effectiveness into Resurfacing, Restoration, and Rehabilitation (3R) Projects not be relied upon unless no better information is available. CMFs rated one or two stars, and some rated three stars, in the FHWA Clearinghouse would likely not pass the inclusion rule for incorporation in HSM Part D. The FHWA Clearinghouse also indicates which CMFs presented in the Clearinghouse also meet the HSM inclusion rule. Application of CMFs from the FHWA Clearinghouse should focus on those CMFs rated three stars, or preferably four or five stars, or those identified in the FHWA Clearinghouse as meeting the HSM inclusion rule. 4.2.3 Development of Crash Modification Factors CMFs are developed primarily through analysis of crash data. The two most common types of crash analyses for developing CMFs are • Observational before–after studies and • Cross-sectional studies. Observational before–after studies are conducted by identifying projects in which a design improvement has been implemented at multiple sites. Crash history and traffic volume data are then obtained for time periods before and after the implementation of each project (typically at least 3 years of data for the period before project implementation and 3 years of data for the period after project implementation). Observational before–after evaluations are typically conducted with the EB method, which compensates for potential biases in crash data. An observational before–after evaluation compares the observed crash count in the period after implementation of each project to an estimate of the crash count that would have been expected if the project had not been implemented. The crash count that would have been expected if the project had not been implemented is estimated with a crash prediction model known as a safety performance function (SPF). SPFs are developed with a modeling technique known as negative binomial regression analysis, which is well suited to modeling crash data. HSM Chapter 9, “Safety Effectiveness Evaluation,” presents a computational procedure for observa- tional before–after evaluations that uses the EB method. The analysis results provide a CMF that represents the crash reduction effectiveness of the project. Figure 2 illustrates a typical SPF for a roadway segment developed with negative binomial regression that could be used in developing CMFs with the EB method. This particular SPF is from HSM Figure 11-3 and applies to roadway segments on rural multilane undivided high- ways (nonfreeways). SPFs can have either linear or curvilinear functional forms. The horizontal axis for the SPF represents the AADT for the roadway segment. The vertical axis represents the predicted number of crashes per mile per year on the roadway segment. Cross-sectional studies are conducted when observational before–after evaluations are not feasible. Cross-sectional modeling uses negative binomial regression modeling to quantify the effect of a specific geometric design feature on crash counts by considering sites with and without that feature or where that feature is present in varying dimensions. Cross-sectional analyses must be carefully designed and conducted because, where specific design features or other factors are correlated with one another, the effect of other factors can potentially be mistaken for an effect of the geometric design feature of interest. HSM Chapter 9 also presents a computational procedure for developing CMFs with cross-sectional modeling. Further guidance on CMF development is found in HSM Chapter 9 (2) and in the FHWA report A Guide to Developing Quality Crash Modification Factors (14). 4.2.4 Sources for Obtaining Crash Modification Factor Values As indicated above, the two most extensively used catalogs of CMF values for design and traffic control improvements are the HSM (2) and the FHWA Crash Modification Factors

Managing a 3R Program to Reduce Crash Frequency and Severity 23   Clearinghouse (13). All of the CMFs in the HSM and the FHWA Clearinghouse were developed in research projects and then reviewed and assessed for inclusion. CMFs can also be obtained from review of individual research reports, although this should generally be done only for new sources that have not yet been assessed for inclusion in the HSM or the FHWA Clearinghouse. 4.3 Crash Modification Factors for Specific Types of 3R Improvements This section presents the CMF values most commonly used to represent the crash reduction effectiveness of specific types of 3R improvements. The following discussion is organized by roadway type and, within roadway type, by design feature. The CMFs presented here are, whenever possible, those used in the HSM (2), which is the most widely utilized source for such information. Where no CMF is available in the HSM, high-quality CMFs from other sources are utilized. While these other CMFs were identified in the FHWA Crash Modification Factors Clearinghouse (13), the original sources have also been reviewed. 4.3.1 Rural Two-Lane Highways 4.3.1.1 Lane Width HSM Chapter 10, “Rural Two-Lane, Two-Way Roads,” presents CMFs for lane widths on rural two-lane highways (2). The CMF is calculated with the equations shown in Table 3 on the basis of the lane width and AADT for the roadway in question. A 12-ft lane is considered to be the base condition (CMF = 1.0). The lane width CMF is illustrated graphically in Figure 3. Source: HSM Figure 11-3 (2). Figure 2. Example of safety performance function for undivided roadway segments on rural multilane highways.

24 Guidelines for Integrating Safety and Cost-Effectiveness into Resurfacing, Restoration, and Rehabilitation (3R) Projects Lane Width (ft) CMF AADT < 400 veh/day AADT = 400 to 2,000 veh/day AADT > 2,000 veh/day ≤9 1.05 1.05 + 2.81 × 10−4(AADT − 400) 1.50 10 1.02 1.02 + 1.75 × 10−4(AADT − 400) 1.30 11 1.01 1.01 + 2.5 × 10−5(AADT − 400) 1.05 ≥12 1.00 1.00 1.00 Note: To determine the CMF for changing lane width or AADT, divide the CMF for the new condition by the CMF for the existing condition. The standard error of the CMF is unknown. The collision types related to lane width to which these CMFs apply are single-vehicle run-off-the-road and multiple-vehicle head-on, opposite-direction sideswipe, and same-direction sideswipe crashes. veh/day = vehicles per day. Source: Based on HSM Tables 10-8 and 13-2. Table 3. CMFs for lane width on rural two-lane roadway segments (2, 15–17). Note: CMFra = CMF for the effect of lane width on target or “related” crashes. Source: Based on HSM Figure 10-7. Figure 3. CMFra for lane width on undivided roadway segments on rural two-lane highways (2, 15). The lane-width CMF illustrated in Table 3 and Figure 3 applies only to single-vehicle run- off-the-road and multiple-vehicle head-on, opposite-direction sideswipe, and same-direction sideswipe crashes. Equation 3 is used to adjust the lane-width CMF for these target or “related” crash types to a lane-width CMF applicable to total crashes: ( )= − × +CMF CMF 1.0 1.0 (3)pra ra where CMFra is the crash modification factor for the effect of lane width on target or “related” crashes (i.e., single-vehicle run-off-the-road and multiple-vehicle head-on, opposite-direction

Managing a 3R Program to Reduce Crash Frequency and Severity 25   sideswipe, and same-direction sideswipe crashes), such as the CMFs for lane width shown in Table 3, and pra is the proportion of total crashes constituted by crash types related to lane and shoulder width. The proportion of related crashes, pra, (i.e., single-vehicle run-off-the-road, and multiple- vehicle head-on, opposite-direction sideswipe, and same-direction sideswipes crashes) is estimated as 0.574 (i.e., 57.4%) on the basis of the default distribution of crash types presented in HSM Table 10-4. This default crash type distribution and, therefore, the value of pra may be updated from local data as part of the calibration process. 4.3.1.2 Shoulder Width and Shoulder Type HSM Chapter 10 presents CMFs for shoulder width and shoulder type on rural two-lane roadways (2). The shoulder width effect, represented by CMFwra, is calculated with the equations shown in Table 4. A 6-ft shoulder represents the base condition (CMF = 1.0). Shoulders wider than 6 ft (e.g., 8-ft shoulders) have CMFs less than 1.0, and shoulders narrower than 6 ft have CMFs greater than 1.0. The shoulder width CMF for rural two-lane highways (CMFwra) is illustrated in Figure 4. The base condition for shoulder type is a paved shoulder (CMF = 1.0). Table 5 presents values for CMFtra, which adjusts for the safety effects of shoulder types (paved, gravel, turf, and composite shoulders). A combined CMF for shoulder width and type is computed as: ( )= × − × +CMF CMF CMF 1.0 1.0 (4)pwra tra ra where CMFwra is the crash modification factor for shoulder width from the equations in Table 4 and CMFtra is the crash modification factor for shoulder type from Table 5. If the shoulder types or widths for the two directions of a roadway segment differ, the CMFs are determined separately for the shoulder type and width in each direction of travel and the resulting CMFs are then averaged. The CMFs for shoulder width and type shown below apply only to the collision types that are most likely to be affected by shoulder width and type: single-vehicle run-off-the-road and multiple-vehicle head-on, opposite-direction sideswipe, and same-direction sideswipe crashes. The CMFs expressed on this basis are, therefore, adjusted to total crashes with Equation 4. The HSM default value for pra for two-lane highways in Equation 4 is 0.574. Shoulder Width (ft) CMF AADT < 400 veh/day AADT = 400 to 2,000 veh/day AADT > 2,000 veh/day 0 1.10 1.10 + 2.5 × 10−4(AADT − 400) 1.50 2 1.07 1.07 + 1.43 × 10−4(AADT − 400) 1.30 4 1.02 1.02 + 8.125 × 10−5(AADT − 400) 1.15 6 1.00 1.00 1.00 ≥8 0.98 0.98 − 6.875 × 10−5(AADT − 400) 0.87 Note: To determine the CMF for changing paved shoulder width or AADT, divide the CMF for the new condition by the CMF for the existing condition. The standard error of the CMF is unknown. The collision types related to lane width to which these CMFs apply include single-vehicle run-off-the-road and multiple-vehicle head-on, opposite-direction sideswipe, and same-direction sideswipe crashes. The values from this table are used as CMFwra in Equation 4. Source: Based on HSM Tables 10-9 and 13-7. Table 4. CMFs for shoulder width on rural two-lane roadway segments (2, 15, 16, 18).

26 Guidelines for Integrating Safety and Cost-Effectiveness into Resurfacing, Restoration, and Rehabilitation (3R) Projects It is noted in Table 5 that CMFs for composite shoulders are based on the assumption that 50% of the surface width of a composite shoulder is paved and 50% is unpaved (turf). The CMFs for composite shoulders in Table 5 are simply an average of the CMFs for paved and turf shoulders (i.e., a weighted average with equal weight given to paved and unpaved shoulders). While the assumption that a composite shoulder consists of equal widths of paved and unpaved shoulder has been implemented in the HSM, this assumption of equal widths seems unneces- sarily limiting for 3R design guidelines. The logic used to determine the composite shoulder CMFs in the HSM can be generalized so that a CMF value can be determined for any combination of paved and unpaved shoulder Source: Based on HSM Figure 10-8. Figure 4. Crash modification factor for shoulder width on roadway segments for two-lane highway (2, 15, 16, 18). Shoulder Type CMF by Shoulder Width 0 ft 1 ft 2 ft 3 ft 4 ft 6 ft 8 ft Paved 1.00 1.00 1.00 1.00 1.00 1.00 1.00 Gravel 1.00 1.00 1.01 1.01 1.01 1.02 1.02 Composite 1.00 1.01 1.02 1.02 1.03 1.04 1.06 Turf 1.00 1.01 1.03 1.04 1.05 1.08 1.11 Note: The values for composite shoulders in this table represent a shoulder for which 50% of the shoulder width is paved and 50% is turf. The standard error of the CMF is unknown. Source: Based on HSM Table 10-10. Table 5. CMFs for shoulder types and shoulder width on roadway segments (CMFtra) (2, 15, 16, 18).

Managing a 3R Program to Reduce Crash Frequency and Severity 27   widths. This approach is an extension of the HSM logic, which does not appear in the HSM but is conceptually consistent with the approach used to determine the composite shoulder CMFs in Table  5. In this approach, the composite shoulder CMF would be a weighted average of the paved and unpaved (turf) shoulder CMFs in Table 5, with weights based on the relative proportions of the paved and unpaved shoulder widths. The CMF for shoulder type, CMFtra, can be determined with Equations 5 through 12, which serve as a replacement for Table 5. For shoulders with total width of 0 ft (i.e., no shoulder present), =CMF 1.00 (5)tra For shoulders with total width of 1 ft, ( )= + −CMF 1.00 1.01 1 (6)paved pavedp ptra For shoulders with total width of 2 ft, ( )= + −CMF 1.00 1.03 1 (7)paved pavedp ptra For shoulders with total width of 3 ft, ( )= + −CMF 1.00 1.04 1 (8)paved pavedp ptra For shoulders with total width of 4 ft, ( )= + −CMF 1.00 1.05 1 (9)paved pavedp ptra For shoulders with total width of 6 ft, ( )= + −CMF 1.00 1.08 1 (10)paved pavedp ptra For shoulders with total width of 8 ft, ( )= + −CMF 1.00 1.11 1 (11)paved pavedp ptra For shoulders with total width of 10 ft, ( )= + −CMF 1.00 1.14 1 (12)paved pavedp ptra where CMFtra is the crash modification factor for shoulder type and ppaved is the proportion of shoulder width that is paved (i.e., paved shoulder width divided by total shoulder width). Since ppaved is a proportion, its value must, by definition, be in the range from 0 to 1. The value of CMFtra for a paved shoulder can be determined by setting ppaved equal to 1. The value of CMFtra for an unpaved shoulder can be determined by setting ppaved equal to 0. The value of CMFtra for a composite shoulder can be determined by setting ppaved equal to an appropriate value between 0 and 1. Thus, the value of ppaved can be used both to specify whether a shoulder is paved or unpaved and to specify the allocation between paved and unpaved surfaces for composite shoulders.

28 Guidelines for Integrating Safety and Cost-Effectiveness into Resurfacing, Restoration, and Rehabilitation (3R) Projects 4.3.1.3 Realignment of Horizontal Curves HSM Chapter 10 presents the following CMF for radius and length of horizontal curves on rural two-lane roads (2): ( ) ( ) ( ) = × +     − × × CMF 1.55 80.2 0.012 1.55 (13) L R S L c c where Lc = length of horizontal curve (mi), including length of spiral transitions if present; R = radius of curvature (ft); and S = spiral transition curve = 1 if present and 0 if not present. The base condition (CMF = 1.0) is a tangent alignment with no curvature. This CMF applies to total crashes and is based on research by Zegeer et al. (19). The CMF computed with Equation 13 can be used to determine the crash reduction effectiveness of realigning a horizontal curve on a rural two-lane highway with an increased radius. 4.3.1.4 Superelevation Restoration/Improvement on Horizontal Curves HSM Chapter 10 presents a CMF for superelevation variance on rural two-lane highways (2), as follows: = <CMF 1.00 for SV 0.01 (14) ( )= + × − ≤ <CMF 1.00 6 SV 0.01 for 0.01 SV 0.02 (15) ( )= + × − ≥CMF 1.06 3 SV 0.02 for SV 0.02 (16) where SV is the superelevation variance (ft/ft), which represents the superelevation rate presented in the AASHTO Green Book (4) minus the actual superelevation rate of the curve. This CMF applies to total roadway segment crashes located on horizontal curves. 4.3.1.5 Centerline Rumble Strips HSM Chapter 10 includes a CMF for the effects of a centerline rumble strip on a rural two-lane highway (2). The value of the CMF is 0.94 for crashes of all types. 4.3.1.6 Shoulder Rumble Strips NCHRP Report 641 found that shoulder rumble strips on rural two-lane highways reduce the target crash type—single vehicle run-off-the-road crashes—by 15%, corresponding to a CMF of 0.85 (20). On the basis of the estimates of crash-type proportions from HSM Chapter 10 (2), the estimated CMF for the effect of shoulder rumble strips on total crashes is 0.92. 4.3.1.7 Striping and Delineation Resurfacing, restoration, and rehabilitation projects often include improvement of the strip- ing and delineation of the roadway. Replacement of pavement markings is an inherent part of most 3R projects, since pavement markings must be renewed when a roadway is resurfaced. No long-term safety benefit is likely if the existing pavement markings are simply replaced with equivalent pavement markings. Research for the Missouri Department of Transportation (DOT) in the report, Benefit/Cost Evaluation of MoDOT’s Total Striping and Delineation Program

Managing a 3R Program to Reduce Crash Frequency and Severity 29   presents CMFs for striping and delineation improvement packages that include replacement of conventional painted markings with more durable markings that often have higher retro- reflectivity, including wider edgelines (21). These striping and delineation packages often included placement of shoulder rumble strips as well, so the CMFs presented in this section should not be combined with the CMFs for shoulder rumble strips. The Missouri DOT evaluation found a 24.5% crash reduction in fatal-and-injury (FI) crashes on rural two-lane highways that were improved with striping and delineation packages, equivalent to a CMF of 0.76 (21). A CMF of 0.76 also represents the best available estimate of the effectiveness of striping and delineation packages for PDO crashes and for total crashes. 4.3.1.8 Roadside Slope Flattening Flattening the roadside slope refers to lessening the steepness of the slope, such as adjust- ing a 1V:3H slope to a 1V:4H or 1V:6H slope. Roadside slope flattening may be considered as a crash countermeasure in 3R projects. Flatter roadside slopes provide the driver with a higher probability of a safe recovery maneuver in a vehicle that has departed the roadway, thus lessening the chance of a rollover or a collision with a roadside object. Roadside slope flattening appears to be primarily applicable to roadside foreslopes in fill sections, but flattening of cut slopes may be considered as well. Research by Fitzpatrick et al. quantified the effect of roadside slope flattening on total crash reduction based on the CMFs shown in Table 6 (22). A roadside slope of 1V:3H represents the baseline condition in Table 6. 4.3.1.9 Guardrail Installation/Restoration Resurfacing, restoration, and rehabilitation projects frequently install new guardrail where it is found to be needed or restore existing guardrail by providing needed maintenance or replacing older guardrail with current designs. There are no formal CMFs for guardrail instal- lation or restoration. The cost-effectiveness of guardrail improvements is typically assessed with the Roadside Safety Analysis Program (RSAP) (23–25), which incorporates methods for quantifying the effectiveness of guardrail improvements. 4.3.1.10 Intersection Left- and Right-Turn Lanes HSM Chapter 10 presents CMFs for the installation of left- and right-turn lanes at an inter- section on rural two-lane highways (2). Table 7 presents CMFs for installing left-turn lanes and Table 8 presents CMFs for installing right-turn lanes. These CMFs apply to all crash severity levels. These CMFs apply to left- or right-turn lanes installed on signalized intersection approaches or uncontrolled major-road approaches to stop-controlled intersections, but do not apply to stop-controlled approaches to stop-controlled intersections. Roadside Slope CMF 1V:2H 1.01 1V:3H 1.00 1V:4H 0.95 1V:6H 0.89 Table 6. Roadside slope CMFs for rural two-lane highways (22). Intersection Type Intersection Traffic Control CMF by Number of Approaches with Left-Turn Lanesa One Approach Two Approaches Three Approaches Four Approaches Three-leg Minor road stop controlb 0.56 — — — Four-leg Minor road stop controlb 0.72 0.52 — — Traffic signal 0.82 0.67 0.55 0.45 aStop-controlled approaches are not considered in determining the number of approaches with left-turn lanes. bStop signs present on minor road approaches only. Source: Based on HSM Table 10-13. Table 7. CMFs for installation of left-turn lanes on intersection approaches (2).

30 Guidelines for Integrating Safety and Cost-Effectiveness into Resurfacing, Restoration, and Rehabilitation (3R) Projects 4.3.2 Rural Multilane Undivided Highways 4.3.2.1 Lane Width The CMF for lane widths on multilane undivided roadways is determined with the equations in Table 9 (2). The base condition for lane width on multilane undivided highways is 12 ft. Figure 5 illustrates graphically the CMFs for lane widths on rural multilane undivided highways shown in Table 9. The CMFs shown in Table 9 and Figure 5 are applicable to single-vehicle run-off-the-road, multiple-vehicle head-on, opposite-direction sideswipe, and same-direction sideswipe crashes. Equation 3 can be used to convert these crash-type-specific CMFs to a CMF for total crashes. The default value of pra in Equation 3 is 0.27 for rural multilane undivided highways. 4.3.2.2 Shoulder Width and Shoulder Type HSM Chapter 11, “Rural Multilane Highways,” presents CMFs for shoulder type and width on rural multilane undivided highways (2). CMFs for undivided sections of multilane highways are calculated by using the same equations as two-lane highways and the shoulder width effect shown in Table 4 (see also Figure 4) and the shoulder type effect presented in Equations 5 through 12. As in the case of rural two-lane highways, the base condition for the shoulder width CMF on a rural multilane highway is 6 ft and the base condition for shoulder type is a paved shoulder. Also, as is the case for rural two-lane highways, the shoulder width and type CMF for rural multilane highways is adjusted from related crashes to total crashes with Equation 4. The HSM default value for pra for rural multilane undivided highways used in Equation 4 is 0.27. Intersection Type Intersection Traffic Control CMF by Number of Approaches with Right-Turn Lanesa One Approach Two Approaches Three Approaches Four Approaches Three-leg Minor road stop controlb 0.86 — — — Four-leg Minor road stop controlb 0.86 0.74 — — Traffic signal 0.96 0.92 0.88 0.85 aStop-controlled approaches are not considered in determining the number of approaches with right-turn lanes. bStop signs present on minor road approaches only. Source: Based on HSM Table 10-14. Table 8. CMFs for installation of right-turn lanes on intersection approaches (2). Lane Width (ft) CMF AADT < 400 veh/day AADT = 400 to 2,000 veh/day AADT > 2,000 veh/day ≤9 1.04 1.04 + 2.13 × 10−4(AADT − 400) 1.38 10 1.02 1.02 + 1.31 × 10−4(AADT − 400) 1.23 11 1.01 1.01 + 1.88 × 10−5(AADT − 400) 1.04 ≥12 1.00 1.00 1.00 Note: To determine the CMF for changing lane width or AADT, divide the CMF for the new condition by the CMF for the existing condition. The standard error of the CMF is unknown. The collision types related to lane width to which these CMFs apply are single-vehicle run-off-the-road and multiple-vehicle head-on, opposite-direction sideswipe, and same-direction sideswipe crashes. Source: Based on HSM Tables 11-11 and 13-3. Table 9. CMFs for lane width on undivided rural multilane roadway segments (2, 26).

Managing a 3R Program to Reduce Crash Frequency and Severity 31   4.3.2.3 Realignment of Horizontal Curves No CMFs for horizontal curve length, radius, or superelevation on rural multilane undivided highways are presented in HSM Chapter 11, and there is no definitive research on this topic. 4.3.2.4 Superelevation Restoration/Improvement on Horizontal Curves No CMFs are available for superelevation variance or superelevation restoration for hori- zontal curves that apply specifically to rural multilane undivided highways. The CMFs shown in Equations 14 through 16 represent the best available estimate of the effect of super elevation variance on crash frequency for curves on rural multilane undivided highways. 4.3.2.5 Centerline Rumble Strips No CMFs are available for the effect of centerline rumble strips on rural multilane highways. For purposes of these guidelines, the safety effect of centerline rumble strips on rural multilane undivided highways has been estimated as a CMF of 0.94, the same value used in the HSM for rural two-lane highway segments. 4.3.2.6 Shoulder Rumble Strips No CMF for the effect of shoulder rumble strips on rural multilane undivided highways has been found, but for purposes of these guidelines, the CMF has been estimated as 0.92, the same as the CMF for rural two-lane undivided highways. 4.3.2.7 Striping and Delineation The Missouri DOT evaluation presents the results of striping and delineation evaluation for rural multilane undivided highways (21). The Missouri DOT evaluation found that FI crashes Source: Based on HSM Figure 11-8. Cr as h M od ifi ca tio n Fa ct or Figure 5. CMFra for lane width on undivided roadway segments on rural multilane highways (2, 26).

32 Guidelines for Integrating Safety and Cost-Effectiveness into Resurfacing, Restoration, and Rehabilitation (3R) Projects were reduced by 29.8% on rural multilane undivided highways that were improved with striping and delineation packages. This is equivalent to a CMF of 0.70. For purposes of these guidelines, the CMFs for PDO crashes have been assumed to be the same as the CMFs for FI crashes. Thus, the total crash CMF for striping and delineation packages on rural multilane undivided highways is 0.70. 4.3.2.8 Roadside Slope Flattening HSM Chapter 11 presents CMFs for roadside slope flattening on rural multilane undivided highways (2), as shown in Table 10. These CMFs apply to total crashes. The base condition for these crashes is a roadside slope of 1V:7H. 4.3.2.9 Guardrail Installation/Restoration Resurfacing, restoration, and rehabilitation projects frequently install new guardrail where it is found to be needed or restore existing guardrail by providing needed maintenance or replacing older guardrail with current designs. There are no formal CMFs for guardrail installa- tion or restoration. The cost-effectiveness of guardrail improvements is typically assessed with the RSAP (23–25), which incorporates methods for quantifying the effectiveness of guardrail improvements. 4.3.2.10 Intersection Left- and Right-Turn Lanes HSM Chapter 11 presents CMFs for the installation of left- and right-turn lanes at inter- sections on rural multilane highways (2). CMF values for installation of left- and right-turn lanes are presented in Tables 11 and 12, respectively. Separate CMFs apply to total and FI crashes. Roadside Slope CMF 1V:2H 1.18 1V:3H 1.15 1V:4H 1.12 1V:5H 1.09 1V:6H 1.05 1V:7H 1.00 Source: Based on HSM Table 11-14. Table 10. Roadside slope CMFs for rural multilane highways (2). Intersection Type Crash Severity CMF by Number of Approaches with Left-Turn Lanesa One Approach Two Approaches Three-leg minor road stop controlb Total 0.56 — Fatal and injury 0.45 — Four-leg minor road stop controlb Total 0.72 0.52 Fatal and injury 0.65 0.42 aStop-controlled approaches are not considered in determining the number of approaches with left-turn lanes. bStop signs present on minor road approaches only. Source: Based on HSM Table 11-22. Table 11. CMFs for installation of left-turn lanes on intersection approaches (2). Intersection Type Crash Severity CMF by Number of Approaches with Right-Turn Lanesa One Approach Two Approaches Three-leg minor road stop controlb Total 0.86 — Fatal and injury 0.77 — Four-leg minor road stop controlb Total 0.86 0.74 Fatal and injury 0.77 0.59 aStop-controlled approaches are not considered in determining the number of approaches with right-turn lanes. bStop signs present on minor road approaches only. Source: Based on HSM Table 11-23. Table 12. CMFs for installation of right-turn lanes on intersection approaches (2).

Managing a 3R Program to Reduce Crash Frequency and Severity 33   4.3.3 Rural Multilane Divided Highways (Nonfreeways) 4.3.3.1 Lane Width The CMF for multilane divided roadways is determined with the equations in Table 13 (2). The base condition for lane width on multilane divided highways is 12 ft. Figure 6 illustrates graphically the CMFs for lane widths on rural multilane divided highways. The CMFs shown in Table 13 and Figure 6 are applicable to single-vehicle run-off-the-road, multiple-vehicle head-on, opposite-direction sideswipe, and same-direction sideswipe crashes. Lane Width (ft) CMF AADT < 400 veh/day AADT = 400 to 2,000 veh/day AADT > 2,000 veh/day ≤9 1.03 1.03 + 1.38 × 10−4(AADT − 400) 1.25 10 1.01 1.01 + 8.75 × 10−5(AADT − 400) 1.15 11 1.01 1.01 + 1.25 × 10−5(AADT − 400) 1.03 ≥12 1.00 1.00 1.00 Note: To determine the CMF for changing lane width or AADT, divide the CMF for the new condition by the CMF for the existing condition. The standard error of the CMF is unknown. The collision types related to lane width to which these CMFs apply are single-vehicle run-off-the-road and multiple-vehicle head-on, opposite-direction sideswipe, and same-direction sideswipe crashes. Source: Based on HSM Tables 11-16 and 13-4. Table 13. CMFs for lane width on divided rural multilane roadway segment (2, 26). Source: Based on HSM Figure 11-10. Cr as h M od ifi ca tio n Fa ct or Figure 6. CMFra for lane width on divided roadway segments on rural multilane highways (2, 26).

34 Guidelines for Integrating Safety and Cost-Effectiveness into Resurfacing, Restoration, and Rehabilitation (3R) Projects Equation 3 can be used to convert these crash-type-specific CMFs to a CMF for total crashes. The default value of pra in Equation 3 is 0.50 for rural multilane divided highways. 4.3.3.2 Shoulder Width HSM Chapter 11 presents CMFs for outside shoulder width on rural multilane divided high- ways (2). CMFs for the width of the right (outside) shoulder of a rural multilane divided highway are given by the values presented in Table 14. The base condition (CMF = 1.0) for the divided highway shoulder width is represented by an 8-ft shoulder. This CMF applies to total crashes and does not need to be adjusted with a pra value. There is no CMF for shoulder type. Under HSM procedures, only the paved portion of the shoulder on a rural multilane divided highway is considered in crash prediction. There is no CMF available for the width or type of the left (inside) shoulder of a rural multilane divided highway (nonfreeway). 4.3.3.3 Realignment of Horizontal Curves HSM Chapter 11 does not present CMFs for horizontal curve length, radius, or superelevation on rural multilane divided highways (nonfreeways), and there is no definitive research on this topic. 4.3.3.4 Superelevation Restoration/Improvement on Horizontal Curves No CMFs were found for superelevation variance or superelevation restoration for horizontal curves on rural multilane divided highways. The CMFs shown in Equations 14 through 16 represent the best available estimate of the effect of superelevation variance on crash frequency for curves on rural multilane divided highways. 4.3.3.5 Shoulder Rumble Strips HSM Chapter 13, “Roadway Segments,” presents a CMF of 0.84 for the effect of shoulder rumble strips on total crashes on multilane divided highways (2). 4.3.3.6 Striping and Delineation The Missouri DOT evaluation also presents results for rural multilane divided highways (21). On rural multilane divided highways that were improved with striping and delineation pack- ages, FI crashes were reduced by 13.8%, equivalent to a CMF of 0.86. For the purpose of this research, the CMFs for PDO crashes were assumed to be the same as the CMF for FI crashes. Thus, the CMF for total crashes for striping and delineation packages on rural multilane divided highways is 0.86. 4.3.3.7 Roadside Slope Flattening No CMFs are available for rural multilane divided highways, but the CMFs presented in Table 10 likely represent the best available estimate. CMF by Average Paved Outside Shoulder Width 0 ft 2 ft 4 ft 6 ft ≥8 ft 1.18 1.13 1.09 1.04 1.00 Source: Based on HSM Table 11-17. Table 14. CMFs for paved right (outside) shoulder width on rural multilane divided highway segments (2, 27).

Managing a 3R Program to Reduce Crash Frequency and Severity 35   4.3.3.8 Guardrail Installation/Restoration Resurfacing, restoration, and rehabilitation projects frequently install new guardrail where it is found to be needed or restore existing guardrail by providing needed maintenance or replacing older guardrail with current designs. There are no formal CMFs for guard- rail installation or restoration. The cost-effectiveness of guardrail improvements is typically assessed with RSAP (23–25) which incorporates methods for quantifying the effectiveness of guardrail improvements. 4.3.3.9 Intersection Left- and Right-Turn Lanes HSM Chapter 11 presents CMFs for the installation of left- and right-turn lanes at inter- sections on rural multilane highways. CMF values for installation of left- and right-turn lanes are presented in Tables 15 and 16, respectively. Separate CMFs apply to total and FI crashes. 4.3.4 Urban and Suburban Arterials 4.3.4.1 Lane Width There are no CMFs for lane width on urban and suburban arterials in HSM Chapter 12, “Urban and Suburban Arterials.” The AASHTO Green Book (4) provides broad flexibility for the use of 10-, 11-, and 12-ft lanes on urban and suburban arterials. Recent research found no substantial differences in safety performance between 10-, 11-, and 12-ft lanes, except in limited cases (28). Intersection Type Crash Severity CMF by Number of Approaches with Left-Turn Lanesa One Approach Two Approaches Three-leg minor road stop controlb Total 0.56 — Fatal and injury 0.45 — Four-leg minor road stop controlb Total 0.72 0.52 Fatal and injury 0.65 0.42 aStop-controlled approaches are not considered in determining the number of approaches with left-turn lanes. bStop signs present on minor road approaches only. Source: Based on HSM Table 11-22. Table 15. CMFs for installation of left-turn lanes on intersection approaches (2). Intersection Type Crash Severity CMF by Number of Approaches with Right-Turn Lanesa One Approach Two Approaches Three-leg minor road stop controlb Total 0.86 — Fatal and injury 0.77 — Four-leg minor road stop controlb Total 0.86 0.74 Fatal and injury 0.77 0.59 aStop-controlled approaches are not considered in determining the number of approaches with right-turn lanes. bStop signs present on minor road approaches only. Source: Based on HSM Table 11-23. Table 16. CMFs for installation of right-turn lanes on intersection approaches (2).

36 Guidelines for Integrating Safety and Cost-Effectiveness into Resurfacing, Restoration, and Rehabilitation (3R) Projects 4.3.4.2 Shoulder Type and Width There are no CMFs for shoulder width on urban and suburban arterials in HSM Chapter 12. The AASHTO Green Book provides broad flexibility in choosing shoulder widths on urban and suburban arterials, including use of curb-and-gutter sections with no shoulder on low- and intermediate-speed arterials (4). Research is underway as part of the FHWA Highway Safety Information System (HSIS) program to assess the safety performance of shoulder widths and curb-and-gutter sections on urban and suburban arterials. 4.3.4.3 Realignment of Horizontal Curves There are no CMFs for horizontal curve length or radius on urban and suburban arterials in HSM Chapter 12, and there is no definitive research on this topic. Research is underway as part of the FHWA HSIS program to assess the effects of horizontal curve design elements on urban and suburban arterials. 4.3.4.4 Superelevation There are no CMFs for superelevation of horizontal curves on urban and suburban arterials in HSM Chapter 12, and there is no definitive research on this topic. 4.3.4.5 Centerline Rumble Strips NCHRP Report 641: Guidance for the Design and Application of Shoulder and Centerline Rumble Strips, found that centerline rumble strips on urban two-lane arterials reduce total target crashes by 40%, corresponding to a CMF of 0.60 (20). Target crashes for centerline rumble strips are those involving a lane departure to the left. On the basis of the estimates of crash-type proportions from HSM Chapter 10 and the assumption that 50% of lane departures on rural two-lane highways are to the left, the estimated CMF for the effect of centerline rumble strips on total crashes on urban and suburban two-lane roads is 0.96. No CMFs are available for the safety effects of centerline rumble strips on multilane urban and suburban arterials. 4.3.4.6 Shoulder Rumble Strips No research was found on CMFs for shoulder rumble strips on urban and suburban arterials. For the purposes of these guidelines, the CMF for shoulder rumble strips is estimated as 0.92 for undivided arterials and 0.84 for divided arterials, consistent with the values found in research for rural highways. It is assumed that shoulder rumble strips would only be installed on urban and suburban arterials with open cross sections (i.e., with shoulder present) and not on urban and suburban arterials with curb-and-gutter sections. 4.3.4.7 Striping and Delineation There are no accepted CMFs for striping and delineation packages on urban and suburban arterials. 4.3.4.8 Intersection Left- and Right-Turn Lanes HSM Chapter 12 presents CMFs for the installation of left- and right-turn lanes at intersections on urban and suburban arterials (2). CMF values for installation of left- and right-turn lanes are shown in Tables 17 and 18, respectively. These CMFs apply to crashes of all severity levels. 4.3.5 Rural and Urban Freeways 4.3.5.1 Lane Width HSM Chapter 18, “Predictive Method for Freeways,” presents CMFs for lane width on through traffic lanes on freeways (3). The CMFs are the same for FI multiple-vehicle crashes and

Managing a 3R Program to Reduce Crash Frequency and Severity 37   FI single-vehicle crashes and are determined by using the following equations. The CMF is applicable to lane widths in the range of 10.5 to 14 ft. = <( )−CMF if 13 ft (17)12e Wa W ll = ≥CMF if 13 ft (18)b Wl where a = −0.0376, b = 0.963, and Wl = lane width (ft). The HSM shows no quantifiable effect of lane width on PDO crashes. 4.3.5.2 Inside Shoulder Width HSM Chapter 18 (3) presents a CMF for paved shoulder width on the inside or median side of freeways. The CMF is calculated with the following equation. = ( )−CMF (19)6ea Wis where a is the CMF coefficient (see Table 19) and Wis is the paved inside shoulder width (ft). This CMF is applicable to paved shoulder widths in the range of 2 to 12 ft. The CMF is computed separately for FI crashes and PDO crashes. Table 19 shows the coefficient values that are used in Equation 19. Note that for a given crash severity level, the coefficient values are the same for multiple- and single-vehicle crashes. Intersection Type Intersection Traffic Control CMF by Number of Approaches with Left-Turn Lanesa One Approach Two Approaches Three Approaches Four Approaches Three-leg Minor road stop controlb 0.67 — — — Traffic signal 0.93 0.86 — — Four-leg Minor road stop controlb 0.73 0.53 — — Traffic signal 0.90 0.81 0.73 0.66 aStop-controlled approaches are not considered in determining the number of approaches with left-turn lanes. bStop signs present on minor road approaches only. Source: Based on HSM Table 12-24. Table 17. CMFs for installation of left-turn lanes on intersection approaches (2). Intersection Type Intersection Traffic Control CMF by Number of Approaches with Right-Turn Lanesa One Approach Two Approaches Three Approaches Four Approaches Three-leg Minor road stop controlb 0.86 — — — Traffic signal 0.96 0.92 — — Four-leg Minor road stop controlb 0.86 0.74 — — Traffic signal 0.96 0.92 0.88 0.85 aStop-controlled approaches are not considered in determining the number of approaches with right-turn lanes. bStop signs present on minor road approaches only. Source: Based on HSM Table 12-26. Table 18. CMFs for installation of right-turn lanes on intersection approaches (2).

38 Guidelines for Integrating Safety and Cost-Effectiveness into Resurfacing, Restoration, and Rehabilitation (3R) Projects There is no CMF for inside shoulder type on freeways. Under HSM procedures, only the paved portion of the inside shoulder on a freeway is considered in crash prediction. 4.3.5.3 Outside Shoulder Width HSM Chapter 18 presents a CMF for paved shoulder width on the outside or right side of freeway roadways (3). The CMF is calculated with the following equation: ∑ ∑= −   +     ( ) ( ) = − = −CMF 1.0 2 2 (20), 1 10 , 1 10L L e L L ec i i m a W c i i m b Ws s where a and b = CMF coefficients (see Table 20), Ws = paved inside shoulder width (ft), Lc,i = length of curve i on freeway segment (mi), L = length of freeway segment (mi), and m = total number of curves on freeway section for both directions of travel combined. The curve lengths need to be summed for both directions of the freeway. The coefficient values used in Equation 20 are shown in Table 20. This CMF is only applicable to paved outside shoulder widths in the range of 4 to 14 ft. Also, this CMF is only applicable to single-vehicle crashes of all severity levels. Equation 20 is presented here in a slightly different form than it appears in the HSM, so that the effect of all curves for both directions of travel combined can be estimated in a single calculation. There is no CMF for outside shoulder type on freeways. Under HSM procedures, only the paved portion of the outside shoulder on a freeway is considered in crash prediction. 4.3.5.4 Shoulder Rumble Strips HSM Chapter 18 presents a CMF for the presence of shoulder rumble strips on the inside and outside shoulders of freeway roadways (3). The CMF is presented in Equations 21 and 22. ∑ ∑= −   +    = = CMF 1.0 2 2 (21), 1 tan , 1 L L f L L c i i m c i i m Crash Type Crash Severity CMF Coefficient a Multiple vehicle Fatal and injury −0.0172 Property damage only −0.0153 Single vehicle Fatal and injury −0.0172 Property damage only −0.0153 Table 19. CMF coefficients for inside shoulder width on freeways (3). Crash Severity CMF Coefficient a b Fatal and injury −0.0647 −0.0897 Property damage only 0.00 −0.0840 Note: Applicable to shoulder widths of 4 to 14 ft. Table 20. CMF coefficients for outside paved shoulder width in single-vehicle crashes on freeways (3).

Managing a 3R Program to Reduce Crash Frequency and Severity 39   [ ] [ ]( ) ( )= − + + − +0.5 1.0 0.811 0.5 1.0 0.811 (22)tanf P P P Pir ir or or where Pir is the proportion of the segment with rumble strips present on the inside shoulders and Por is the proportion of the segment with rumble strips present on the outside shoulders. Equation 21 is presented here in a slightly different form than it appears in the HSM so that the effect of all curves for both directions of travel combined can be estimated in a single calculation. 4.3.5.5 Median Barrier HSM Chapter 18 presents a CMF for the presence of a median barrier on a freeway (3). The CMF is computed with Equation 23. This CMF is applicable to both single- and multiple-vehicle collisions of all crash severity levels. ( )= − +CMF 1.0 (23)P P eib ib a Wicb where a = CMF coefficient (see Table 21), Pib = proportion of segment length with a barrier present in the median (see discussion below for calculation of Pib), and Wicb = distance from edge of inside shoulder to barrier face (see discussion below for calcula- tion of Wicb). If a continuous median barrier is centered in the median, Equations 24 and 25 are used to calculate Pib and Wicb. ∑∑ ( ) = − + − − − 2 2 0.5 2 (24) , , , , W L L W W L L W W W icb ib i off in i is ib i m is ib =1.0 (25)Pib where L = length of segment (mi), Lib,i = length of lane paralleled by inside barrier i (mi), Woff,in,i = horizontal clearance from the edge of the traveled way to the face of inside barrier i (ft), Wis = width of paved inside shoulder (ft), Wm = width of median (measured from near edge of traveled way in both directions) (ft), and Wib = width of inside barrier (measured from barrier face to barrier face) (ft). Crash Type Crash Severity CMF Coefficient a Multiple vehicle Fatal and injury 0.131 Property damage only 0.169 Single vehicle Fatal and injury 0.131 Property damage only 0.169 Table 21. CMF coefficients for presence of median barrier on freeways (3).

40 Guidelines for Integrating Safety and Cost-Effectiveness into Resurfacing, Restoration, and Rehabilitation (3R) Projects If a continuous median barrier is adjacent to one of the directions of travel but not centered in the median, Equations 26 and 27 are used to calculate Pib and Wicb. ∑∑ = − + − + − − − − 2 2 (26) near , , , , near W L L W W L W W L L W W W W icb is ib i off in i is ib i m is ib =1.0 (27)Pib where Wnear is the near horizontal clearance from the edge of the traveled way to the continuous median barrier (measured for both travel directions and the smaller distance used) (ft). For segments with some short sections of barrier in the median, Equations 28 and 29 are used to determine Pib and Wicb. ∑ ∑ = − (28), , , W L W W icb ib i off in i is ∑= 2 (29),P L L ib ib i Values for the median barrier CMF coefficient, a, are presented in Table 21. This coefficient is used in Equation 23. 4.3.5.6 Traffic Barrier on the Outside or Right Side of a Freeway HSM Chapter 18 presents a CMF for the presence of a traffic barrier on the outside or right roadside of a freeway (3). The CMF is calculated with Equation 30. This CMF is only applicable to single-vehicle collisions. The HSM shows no quantifiable effect of the presence of a median barrier on multiple-vehicle collisions. ( )= − +CMF 1.0 (30)P P eob ob a Wocb where a = CMF coefficient (see Table 22), Pob = proportion of segment length with a barrier present on the roadside, and Wocb = distance from edge of outside shoulder to barrier face (ft). Use Equations 31 and 32 to calculate Pob and Wocb. ∑ ∑ = − (31), , , , W L L W W ocb ob i ob i off o i s Crash Severity CMF Coefficient a Fatal and injury 0.131 Property damage only 0.169 Table 22. CMF coefficients for presence of outside barrier in single-vehicle crashes on freeways (3).

Managing a 3R Program to Reduce Crash Frequency and Severity 41   ∑= 2 (32),P L L ob ob i where L = length of segment (mi), Lob,i = length of lane paralleled by outside barrier i (mi), Woff,o,i = horizontal clearance from the edge of the traveled way to the face of outside barrier i (ft), and Ws = width of paved outside shoulder (ft). Values for the outside barrier CMF coefficient, a, are presented in Table 22. This coefficient is used in Equation 30. 4.3.5.7 Median Width HSM Chapter 11 presents a CMF for median width on rural multilane divided highways (nonfreeways) (2). This CMF is presented in HSM Table 11-18. However, median width on a rural multilane divided highway is unlikely to be changed in a 3R project, as such a project would be considered reconstruction, so the median width CMF for rural multilane divided highways is not needed in these guidelines. HSM Chapter 18 presents a CMF for median width on a freeway (3). Median width on a freeway is not likely to be changed in a 3R project. However, this CMF is presented here because the median width CMF is influenced by the width of the paved inside shoulder (Wis), which may be changed in a 3R project. The median width CMF is calculated with Equation 33. This CMF is applicable to both single- and multiple-vehicle collisions of all crash severity levels. ( )= − +( ) ( )− − −CMF 1.0 (33)2 48 2 48P e P eib a W W ib a Wm is icb where a = CMF coefficient (see Table 23), Wis = width of paved inside shoulder (ft), Wm = width of median (measured from near edge of traveled way in both directions) (ft), Pib = proportion of segment length with a barrier present in the median (see Equations 24 through 29 for calculation of this variable), and Wicb = distance from edge of inside shoulder to barrier face (ft) (see Equations 24 through 29 for calculation of this variable). Values for the median width CMF coefficient, a, are presented in Table 23. This coefficient is used in Equation 33. Crash Type Crash Severity CMF Coefficient a Multiple vehicle Fatal and injury −0.00302 Property damage only −0.00291 Single vehicle Fatal and injury 0.00102 Property damage only −0.00289 Table 23. Coefficients for median width on freeways (3).

42 Guidelines for Integrating Safety and Cost-Effectiveness into Resurfacing, Restoration, and Rehabilitation (3R) Projects 4.4 Investing Available 3R Funds for Maximum Reduction of Crash Frequency and Severity Since funds available for highway infrastructure improvements are limited, it is important that those funds, including funds for 3R projects, be invested to accomplish project objectives, including preserving the pavement and extending the pavement service life; improving traffic operations; and, to the extent practical, maximizing the potential reduction in crash frequency and severity. Maximizing the reduction in crash frequency and severity requires that the available funds be invested in a logical priority fashion that focuses on those projects where the maximum crash reduction benefit can be obtained for the least cost. Research has shown that investing available funds in safety improvement without careful planning can lead to suboptimal results (6). This is illustrated by the following example. 4.4.1 Example of Optimal and Suboptimal Strategies for Investing Available Funds in Design Improvements to Reduce Crash Frequency and Severity Highway agency users who apply the 3R design guidelines will apply them to one site at a time, but their appropriateness can be tested by applying them systemwide to rural highway systems (or urban freeways) and reviewing the results. As an example, four potential strategies for lane widening of the rural two-lane highway system of an actual state with data from the FHWA HSIS were tested in recent research (7). This rural two-lane highway system consists of 4,630.71 mi of road with AADTs up to 25,000 veh/day. The current lane width distribution for this road system is shown in Table 24. The following improvement strategies were considered and were applied as if the entire road system were a candidate for resurfacing in a single year: • Widen all lanes with widths of less than 12 ft to 12 ft. • Widen lanes where a need is indicated according to the TRB Special Report 214 (1) lane width criteria presented in Table 25. • Widen lanes on roadways where the AADT exceeds the minimum AADT criteria presented in Section 5.4. • Widen lanes where the net present benefits of the project exceed zero (i.e., where the benefits exceed the costs) according to the benefit–cost analysis procedure presented in Chapter 5. The benefits and cost of lane widening in this example are based on assumptions concerning crash costs, unit construction costs, project service life, and minimum attractive rate of return presented in Section 5.1. These values vary widely in current practice, but the assumptions used here are typical of the middle range of values currently used by highway agencies. Lane Width (ft) Total Length (mi) 9 18.29 10 251.74 11 2,087.38 12 2,273.30 Total 4,630.71 Table 24. Lane width distribution in the lane-widening example.

Managing a 3R Program to Reduce Crash Frequency and Severity 43   4.4.1.1 Lane-Widening Strategy: Widen All Lanes to 12 ft This strategy would select for widening the 2,357.41 mi of roadway with existing lane widths of less than 12 ft. The lane width distribution on the roadway system would be changed as shown in Table 26. This improvement program would provide benefits of $68,911,192 at a cost of $516,336,773— equivalent to a benefit–cost ratio of 0.13. Of the 2,357.41 mi of roadway improved, 97% consisted of projects with benefit–cost ratios less than 1.0. This high proportion of projects that were not cost-effective occurred because most of the extensive mileage of two-lane roadways consisted of 11-ft lanes, where lane widening provided only a limited benefit (see Section 4.3.1.1) 4.4.1.2 Lane-Widening Strategy: Base Widening of Lanes on Criteria in TRB Special Report 214 This strategy would select for widening the 832.28 mi of roadway that do not meet the TRB Special Report 214 lane width criteria shown in Table 25. This would change the lane-width distribution on the roadway system as shown in Table 27. This improvement program would provide benefits of $57,180,686 at a cost of $200,107,835— equivalent to a benefit–cost ratio of 0.29. Of the 832.28 mi of roadway improved, 96% consisted of projects with benefit–cost ratios less than 1.0. This strategy did a better job at focusing on the best projects and avoided many of the projects included in the 12-ft lane strategy that were not cost-effective, but not all of them. ≥10% Trucksc <10% Trucks Design year volume (ADT)a Running Speedb (mph) Lane Width (ft) Combined Lane and Shoulder Width (ft)d Lane Width (ft) Combined Lane and Shoulder Width (ft)d 1–750 <50 10 12 9 11 ≥50 10 12 10 12 751–2,000 <50 11 13 10 12 ≥50 12 15 11 14 >2,000 All 12 18 11 17 Note: This table is presented as an example of historical practice, but, as shown below, its use is no longer recommended. aDesign volume for a given highway feature should match average traffic anticipated over the expected performance of that feature. ADT = average daily traffic. bHighway segments should be classified as <50 mph only if most vehicles have an average speed of less than 50 mph over the length of the segment. cFor this comparison, trucks are defined as heavy vehicles with six or more tires. dOne foot less for highways in mountainous terrain. Table 25. Minimum lane and shoulder widths for rural two-lane highways from TRB Special Report 214 (1). Lane Width (ft) Total Length (mi) Before After 9 18.29 0.00 10 251.74 0.00 11 2,087.38 0.00 12 2,273.30 4,630.71 Total 4,630.71 4,630.71 Table 26. Change in lane width distribution for lane widening strategy: Widen all lanes to 12 ft.

44 Guidelines for Integrating Safety and Cost-Effectiveness into Resurfacing, Restoration, and Rehabilitation (3R) Projects 4.4.1.3 Lane-Widening Strategy: Minimum AADT Levels This strategy would select for widening the 54.88 mi of roadway that meet the minimum AADT criteria presented in Section 5.4. This would change the lane-width distribution on the roadway system as shown in Table 28. This improvement program would provide benefits of $12,391,936 at a cost of $14,824,774— equivalent to a benefit–cost ratio of 0.84. Of the 54.88 mi of roadway improved, 60% consisted of projects with benefit–cost ratios less than 1.0. This strategy did a better job at focusing on the best projects and avoided many of the projects that were not cost-effective. 4.4.1.4 Lane-Widening Strategy: Benefit–Cost Analysis for Individual Projects This strategy would select for widening 35.34 mi of roadway for which a need for lane widening was identified by a benefit–cost analysis using the benefit–cost analysis procedure presented in Section 5.2.2. This would change the lane-width distribution on the roadway system as shown in Table 29. Lane Width (ft) Total Length (mi) Before After 9 18.29 18.29 10 251.74 219.36 11 2,087.38 2,076.96 12 2,273.30 2,301.32 Total 4,630.71 4,630.71 Table 29. Change in lane width distribution for lane-widening strategy: Benefit–cost analysis for individual projects. Lane Width (ft) Total Length (mi) Before After 9 18.29 0.00 10 251.74 175.20 11 2,087.38 1,432.58 12 2,273.30 3,022.93 Total 4,630.71 4,630.71 Table 27. Change in lane width distribution for lane-widening strategy: Base widening of lanes on criteria in TRB Special Report 214. Lane Width (ft) Total Length (mi) Before After 9 18.29 18.29 10 251.74 231.25 11 2,087.38 2,052.99 12 2,273.30 2,328.18 Total 4,630.71 4,630.71 Table 28. Change in lane width distribution for lane-widening strategy: Minimum AADT levels.

Managing a 3R Program to Reduce Crash Frequency and Severity 45   This improvement program would provide benefits of $7,817,183 at a cost of $5,603,567— equivalent to a benefit–cost ratio of 1.40. Every portion of the 35.34 mi of roadway improved in this program was a cost-effective project, and the benefits for the program as a whole exceeded the costs. This strategy did the best job at focusing on the best projects and avoided many of the projects that were not cost-effective. 4.4.2 Summary of Findings from the Example of Lane-Widening Strategies The example of lane-widening strategies for a statewide system of rural two-lane highways shows that an improvement program based on benefit–cost analysis for individual projects would provide the greatest net present benefits and the greatest return per dollar spent and would avoid improvements that are not cost-effective. However, the results of benefit–cost analysis are not necessarily exact, and not every highway agency will have the data available and in a suitable form for a benefit–cost analysis. Minimum AADT criteria developed through a benefit–cost analysis come the closest of the other alternatives to providing optimal results. Both blanket lane widening to a minimum 12-ft lane width and lane widening based on the design criteria from TRB Special Report 214 (1) result in many lane-widening investments that would not be cost-effective. The primary drawback of strategies developed before publication of the HSM is likely to be a focus on widening 11-ft lanes to 12 ft, which is unlikely to be cost-effective except at very high volumes. Finally, it should be noted that, in actual practice, some roadways might be found to have experienced patterns of single-vehicle run-off-the-road, head-on, opposite-direction sideswipe, or same-direction sideswipe crashes that could indicate a need for lane widening regardless of the results of the benefit–cost analysis.