Potential MUTCD Criteria for Selecting the Type of Control for Unsignalized Intersections (2015)

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.

57 CHAPTER 6: ECONOMIC ANALYSIS INTRODUCTION This chapter describes the activities related to the economic analysis procedure. It contains discussion of the sources used as a basis for the analysis, as well as steps taken to collect the necessary information and perform the analysis. SELECTION OF TYPE OF COSTS TO CONSIDER IN ANALYSIS Creating an economic analysis procedure requires consideration of what benefits and costs to include in the analysis. Several documents provide guidance on how to determine total costs for traffic control, including:  AASHTO’s User and Non-User Benefit Analysis for Highways (commonly known as the Red Book) (73).  NCHRP Project 3-110, Estimating the Life-Cycle Cost of Intersection Designs: Interim Report (74).  FHWA’s Highway Economic Requirements System—State Version: Technical Report (HERS-ST) (75).  NCHRP Web-Only Document 193, Development of Left-Turn Lane Warrants for Unsignalized Intersections (76).  Upchurch’s “Guidelines for Use of Sign Control at Intersections to Reduce Energy Consumption” (50) and Development of an Improved Warrant for Use of Stop and Yield Control at Four-Legged Intersections (77). Table 23 lists the costs that these references suggest should be considered when evaluating a change in an intersection’s design or operations. The key costs considered in most of these documents are user delay, crash, and vehicle operating costs. Depending upon the source, other costs are considered such as pollution or travel time reliability. How each cost is calculated also varies depending upon the source. The AASHTO Red Book notes that, in general, control devices yield higher travel time costs and operating and ownership costs, which are offset by safety-related benefits. Operating costs include fuel, oil, maintenance, and tires. Ownership costs include insurance, license and registration fees and taxes, economic depreciation, and finance changes. To calculate the effect of the change in traffic control, the costs need to be calculated both before and after the change. Ownership costs are typically considered on a per-mile basis; however, because intersection traffic control will not change the total distance, these costs should not vary between the alternatives being considered in this study and, therefore, were not included in the analysis. The AASHTO Red Book also provides fuel costs as a function of time rather than a function of travel speed for those analyses where an improvement—such as a change in intersection traffic control —results in traffic delay. The Red Book notes that although these factors are a function of delay, the fuel consumption is due primarily to acceleration of vehicles after being delayed, rather than fuel consumed idling during delay periods.

58 Table 23. Costs Suggested for Evaluating Changes to Intersections from Several Sources. Source Costs AASHTO Red Book (73)  Travel time (delay) costs  Crash costs  Vehicle operating and ownership costs NCHRP 3-110, Estimating the Life-Cycle Cost of Intersection Designs Interim Report (74) User costs:  Construction  User delay at the intersection  Travel time reliability  Safety  Operating (e.g., fuel, oil, maintenance, and tires) Other costs:  Delay to travelers on other parts of the network  Emissions  Effects on businesses  Right-of-way acquisition  Public safety Upchurch (50, 77)  Vehicle operating costs  Delay  Crashes  Air pollution  Sign material, installation, and maintenance costs  Noise pollution FHWA Highway Economic Requirements System—State Version: Technical Report (75) Constant speed and excess speed cost components:  Fuel consumption  Oil consumption  Tire wear  Maintenance and repair  Depreciable value NCHRP Web-Only Document 193 (76)  User delay at the intersection  Safety HERS-ST was developed to estimate highway system performance for various investment levels. It contains detailed equations for estimating constant and variable speed operating costs for seven vehicle types by determining the estimated costs associated with fuel, oil, tire wear, maintenance and repair rate, and depreciation for each vehicle type. In addition, the equations consider grade, pavement condition adjustments, and other adjustment factors such as fuel efficiency. The equations for estimating the effect of speed-change cycles calculate the excess operating costs due to STOP signs; however, these equations only consider maximum speed during the speed-change cycle. Because the research team has the estimated change in delay at the intersection associated with the change in the traffic control, the AASHTO Red Book methodology was used for estimating vehicle operating costs. The NCHRP 3-110 Interim Report on Estimating the Life-Cycle Cost of Intersection Designs (74) recommends the consideration of construction and travel time reliability in addition to the costs already discussed. For the scenarios being considered, construction should be nominal with the exception of a conversion to a roundabout, and consideration of those construction costs was

59 added to the analysis. For this analysis, the research team assumed that the change in intersection traffic control at the volumes being considered would have no impact on travel time reliability. Upchurch (50, 77) recommended that pollution (both air and noise) and sign material, installation, and maintenance costs should be considered along with the costs discussed above. Researchers investigated the applicability of tools such as the Environmental Protection Agency’s Motor Vehicle Emissions Simulator (MOVES) to quantify the effects of air pollution from emissions, but some of the underlying conditions and assumptions used in these tools were not directly applicable to the unsignalized intersection scenario in this project, and identifying ways to adapt to this project proved very difficult at best. In addition, preliminary results obtained from MOVES indicated that costs associated with pollution from emissions would be very low compared to other costs in the analysis. The NCHRP 3-110 methodology includes pollution (emissions) along with several other costs as non-user costs. These non-user costs are costs endured by users elsewhere on the network or societal costs associated with the use of the network. Because of the low annual costs for signs, pollution, and societal costs as compared to other costs, they were not included in the analysis. Based upon the discussions from the sources referenced above, the research team selected the following costs for consideration in this project:  User delay (travel time).  Crash.  Vehicle operating.  Construction (for roundabouts). To obtain the information needed to calculate delay, which is needed for both user (time) and vehicle operating costs, simulation models were run for several scenarios. The HSM (24) was used to determine the crash prediction estimates. The following section present information on these efforts. SIMULATION Base Models To conduct the operational analysis, a microsimulation model (VISSIM) was used to measure the impact of intersection traffic control on intersection delay for cars, trucks, and pedestrians. The base models for three-leg intersections included the following:  All-way stop control (AW3).  Two-way stop control (TW3). The four-leg base intersections were:  All-way stop control (AW4).  Two-way stop control (TW4).  Roundabout (RO4).

60 Table 24 lists the values of the variables that were not modified between simulation runs. Table 25 lists the values of the variables that were modified, except for volume, which is provided in Table 26. Table 24. Non-changing Simulation Variable Values. Variable Value Major or Minor No. of Legs Approach segment length 2640 ft Both 3 and 4 Bicycle free-flow speed 15 mph Both 3 and 4 Critical gap for pedestrians 6 sec Both 3 and 4 Critical gap for vehicles 3 sec Both 3 and 4 Dedicated left-turn lane None Both 3 and 4 Dedicated right-turn lane None Both 3 and 4 Lane width 12 ft Both 3 and 4 Median type None Both 3 and 4 Heavy-vehicle percent 5% Major 3 and 4 Through percent 80% Major 3 Turn (either left or right) percent 20% Major 3 Left-turn percent 15% Major 4 Right-turn percent 15% Major 4 Through percent 70% Major 4 Left-turn percent 50% Minor 3 Right-turn percent 50% Minor 3 Through percent 20% Minor 4 Heavy-vehicle percent 1% Minor 3 and 4 Number of lanes on approach 1 lane Minor 3 and 4 Table 25. Changing Simulation Variable Values. Variable Value Major or Minor Geometry Three legs, four legs, or roundabout Intersection Traffic control Two-way stop, all-way stop, or roundabout Intersection Number of lanes 2- or 4-lane roads (1- or 2-lane approach) Major Posted speed limit 25 or 40 mph Minor Posted speed limit 25, 40, or 55 mph Major Directional bicycle flow rate 0, 10 bikes/hr Both Directional pedestrian flow rate 5, 10, or 20 ped/hr Both Table 26. Major and Minor Approach Volume Pairs. Major (veh/hr/approach) 210 300 450 500 600 700 750 1000 Minor (veh/hr/approach) 140 200 300 300 400 400 350 500 Assumptions Assumptions for the simulation runs included:  Arrival is random.  The standard deviation for speeds is 5 mph.  Driveways or unsignalized intersections do not exist along any of the approaches except for the one intersection of interest.

61  The pedestrian will wait until a sufficient gap is present, either created because a vehicle stopped or due to available headway within the traffic stream. If a marked crosswalk is present, drivers should yield or stop to a pedestrian in the crosswalk, even if a STOP sign is not present. Previous research (22), however, has demonstrated that few drivers will yield to pedestrians in an uncontrolled yet marked crosswalk. Therefore, the assumption for this simulation is that pedestrians on uncontrolled approaches will wait and only cross when there is a sufficient gap. Pedestrians will have no delay when crossing a stop-controlled approach. Modeling Runs A series of simulation modeling runs were conducted. Initially, the range of speed was a variable of emphasis to be able to determine warrants for a range of posted speed limits or for rural (high- speed) and urban (low-speed) conditions. Examining the results from these earlier runs revealed, however, that delay did not vary greatly due to posted speed limit. Table 27 shows the result for a subset of the trials where the major, minor, pedestrian, and bicycle volumes were constant and the major and minor speeds were varied. For cars within the trials shown in Table 27, the maximum average intersection delay was 7.8 sec, and the minimum delay was 6.7 sec, representing a range of only 1.1 sec. When compared to the variation in delay due to a change in volume, the variation in delay due to changing speed is nominal. Because delay was not as affected by speed, later simulation efforts focused on varying vehicle volume and the number of pedestrians. Table 27. Simulation Results Illustrating Variation Due to Speed Limit. Trials A B C D E F Average Major Speed (mph) 25 40 40 55 55 55 Varies Minor Speed (mph) 25 25 40 25 40 55 Varies Car, Average Intersection Delay (sec/car) 6.7 7.5 7.8 7.5 7.6 7.7 7.5 Truck, Average Intersection Delay (sec/truck) 5.9 5.5 6.7 7.0 6.9 7.2 6.5 Pedestrian, Average Intersection Delay (sec/pedestrian) 4.3 6.0 5.8 6.9 7.3 7.0 6.2 Car, Average Minor Road Delay (sec/car) 18.7 21.1 22.0 20.9 21.0 21.4 20.8 Other input values: 500 veh/hr/approach on major, 250 veh/hr/ln on minor, 20 ped/hr all approaches, 0 bike/hr all approaches, TWSC, four lanes on major, two lanes on minor, four legs. Findings from Simulation Figure 14 shows plots of the delay findings for cars, while Figure 15 shows delay results for pedestrians. The entering volume is the sum of the volume of vehicles, pedestrians, and bicycles on each approach for the hour of simulation.

(a) Two (c) Tw -Lane Maj o-Lane Maj or Road, Th or Road, F ree Legs our Legs Figure 14. 62 (b) (d) Car Delay R Four-Lane Four-Lane esults. Major Roa Major Roa d, Three Le d, Four Le gs gs

(a) Two (c) Tw -Lane Maj o-Lane Maj or Road, Th or Road, F Fig ree Legs our Legs ure 15. Ped 63 (b) (d) estrian Del Four-Lane Four-Lane ay Results. Major Roa Major Roa d, Three Le d, Four Le gs gs

64 For intersections with four lanes on the major road, average car delay begins to increase above 1,500 units/hr. The increase in average car delays begins at a slightly lower entering volume when there are only two lanes on the major road (i.e., one-lane approaches), as expected. As illustrated in Figure 14(a) and (c), the average per-car delay at AWSC intersections with only one-lane approaches on the major road exceeds the average per-car delay at the other intersection types (TWSC and roundabout intersections). Delay at TWSC and roundabouts also increases with higher entering volumes, but not as much as it does for the AWSC condition. The pedestrian delay results illustrate the benefits of AWSC to pedestrians because the delay is minimal for pedestrians across all volume levels. For the higher volumes levels, the delay incurred by the pedestrians waiting for an adequate gap at the higher volume—whether at a TWSC or a roundabout intersection—can be seen in Figure 15. The average pedestrian delay is higher at roundabouts because pedestrians must search for a gap on all approaches at a roundabout, while the pedestrians at a TWSC intersection only search for a gap on two of the four approaches. The results from the VISSIM runs were reviewed, and the volume combinations were identified where the average minor-road delay was greater than 60 sec (see Table 28). These results reflect the average vehicle delay for the minor road rather than the average vehicle delay for the entire intersection. The combinations listed in Table 28 are higher than the current peak-hour signal warrant; for example, the signal warrant is 240 units/hr on the minor approach when the major volume is assumed to be 600 veh/hr/approach or 1,200 veh/hr total for both approaches. Table 28. Volume Combinations Used in VISSIM Resulting in More Than 35 sec or 60 sec of Minor-Road Average Vehicle Delay during the Simulated Hour. Delay (sec) Number of Legs Number of Lanes on Major Approach Major Volume (veh/hr/approach) Minor Volume (veh/hr/approach) >35, less than 60 3 1 500 300 >35 3 1 and 2 600 400 >35 3 1 and 2 700 400 >35 3 1 and 2 750 350 >35 3 2a 1,000 500 >35, less than 60 4 1 500 300 >35 4 1 and 2 600 400 >35 4 1 and 2 700 400 >35 4 1 and 2 750 350 >35 4 2a 1,000 500 a Volume combination not used with a one-lane major-road approach. Table 28 also provides the volume combinations included in the VISSIM simulation where greater than 35 sec of delay per vehicle was observed on the minor-road approach. The value of 35 sec of delay corresponds to LOS E in the HCM (23). The lowest volume combination with more than 35 sec delay per veh was 500 veh/hr/approach on the major road (or 1,000 veh/hr for both approaches) and 300 veh/hr/approach on the minor road for a two-lane major road. For a four-lane major road, the lowest volume combination is 600 on the major road and 400 on the minor road. However, these volumes (500/300 or 600/400) would both plot above the relevant

65 peak-hour signal warrant curve. For a two-lane major road, the signal warrant for 1,000 veh/hr on the major road is 200 veh/hr. For a four-lane major road, the signal warrant of 1,200 veh/hr on the major road is 225 veh/hr on the minor road. COSTS Based upon the review of several contributing sources, as discussed previously in this chapter, the research team selected the following costs for consideration in evaluating changes in intersection traffic control:  User delay.  Crash.  Vehicle operating.  Roundabout construction. User Delay Costs To evaluate the change in user delay, the results from the simulation runs were used. The delay results from the AWSC and roundabout scenarios were compared to the delay determined with only TWSC present. The intersection-wide measure of performance used was average delay per car, per truck, or per pedestrian for the network, measured in seconds. To determine the consequences of changing the intersection traffic control, the difference between the average total delay before (i.e., TWSC) and after (i.e., either the AWSC or roundabout scenario) the change was calculated. The difference could be positive or negative with the following meaning:  Negative difference in delay means that more user delay is occurring due to the change.  Positive difference in delay means that there is a delay savings due to the change. For example, assume that the intersection traffic control at a four-leg intersection with four lanes on the major road was changed from TWSC to AWSC. The peak-hour volume is 300 veh/hr/approach on the major road, 200 veh/hr/approach on the minor road, 10 ped/hr/approach, and 0 bikes/hr/approach. The estimated delays per hour for the scenarios being used in this example are shown in Table 29. When TWSC is replaced with AWSC, the delay for cars and trucks becomes worse (as illustrated by the negative values in Table 29), while delay for pedestrians improves (as illustrated by the positive value in the TW4-AW4 row of Table 29). Per-hour delays available from the simulation are converted into hours of delay per year and then multiplied by the assumed vehicle occupancy (for cars and trucks) and the assumed value of time (for cars, trucks, and pedestrians).

66 Table 29. Example of Delay by User. Scenario Car Intersection Delaya (sec/car) or Delay Costs/Savings ($/yr)b Truck Intersection Delaya (sec/truck) or Delay Costs/Savings ($/yr)b Pedestrian Intersection Delaya (sec/ped) or Delay Costs/Savings ($/yr)b TW4 3.6 2.1 4.0 AW4 9.3 12.5 0.3 TW4-AW4 (Change from TW4 to AW4) −5.7 −10.4 3.7 TW4-AW4 Costs per Year $(155,566) $(63,347) $3,950 a Average intersection delay for four-leg intersection with four lanes on major road and when volume is 300 veh/hr/approach on the major road, 200 veh/hr/approach on the minor road, 10 ped/hr/approach, and 0 bikes/hr/approach. b Average annual delay costs/savings determined using the methodology discussed in this document. The parentheses with dollars represent a negative amount and indicate that more delay is occurring due to the intersection control change for cars and trucks. Positive delay costs/savings indicate less delay is occurring for pedestrians due to the intersection control change. Table 30. Factors Used to Convert Seconds/Vehicle Delay to Hour/Intersection Delay for a Year. Traffic Period Number of Hours in Weekday Number of Hours in Weekend Hours per Yeara Hourly Percent of ADT during Periodb Typical Hourly Volume If AADT = 1,000 veh/day Weekday Peak Period 3 0 751 7.8 78 Weekday, Near-Peak Hour, and Weekend Typical Period 7 11 3,014 6.1 61 Weekday and Weekend Off-Peak 8 8 2,920 3.7 37 Night 6 5 2,075 0.7 7 Total 24 24 8,760 1,000 a Assume 52.16 weeks/year with 4.8 days having weekday traffic distribution and 2.2 days having weekend traffic distribution (the typical 5 weekdays and 2 weekend days were adjusted to reflect 10 holidays). b Assumed hourly percent of traffic for given traffic period. Delay for Entire Year The simulation provides predictions of delay measured in seconds per user. This value needs to be converted to delay at the intersection for the entire year. To perform the conversion, the assumed number of hours along with the percent of the ADT represented by each traffic period is needed. Table 30 provides the assumptions used in this project to convert seconds-per-user delay into hours of delay for the year at the intersection. The hourly percent of ADT values was determined using hourly traffic distributions available in the 2012 Urban Mobility Report (78). The distributions for non-freeway, AM and PM peak periods for both no/low congestion and moderate congestion were considered to obtain the weekday values. The non-freeway, weekend

67 traffic distribution was used to obtain the weekend data. While hourly factors were available for each hour of the day, hours were grouped as shown in Table 30 to facilitate calculations. Travel Time Delay Cost The national congestion constants used in the 2012 Urban Mobility Report (78) are shown in Table 31. The values represent 2011 dollars. The value of person-time used in the Urban Mobility Report is based on the value of time, rather than the average or prevailing wage rate. The average cost of time was assumed to be $16.79 per person-hour for 2011. The 2011 value of time was adjusted using the Consumer Price Index (CPI) for 2011 and 2013 available from the U.S. Bureau of Labor Statistics (79). The ratio of the 2013 to 2011 CPI value is 232.957 divided by 224.939, which is 1.04. The ratio 1.04 multiplied by $16.79 gives a 2013 value of time of $17.39, which represents the average cost of time per person. To convert to an average cost of time per vehicle, the cost of time per person is multiplied by the vehicle occupancy factor of 1.25 persons per vehicle. The CPI was also applied to the commercial vehicle operating cost to obtain a 2013 hourly value of $89.90. The vehicle occupancy for trucks was assumed to be 1.0. Table 31. National Congestion Constants Used in the 2012 Urban Mobility Report (78).a Constant Value Vehicle Occupancy (Passenger Vehicles) Average Cost of Time (2011) Commercial Vehicle Operating Cost (2011) 1.25 persons per vehicle $16.79 per person-hourb $86.81 per vehicle-hourb, c a Source: 2012 Urban Mobility Report methodology, http://tti.tamu.edu/documents/mobility-report-2012- wappx.pdf. b Adjusted annually using the Consumer Price Index. c Adjusted periodically using industry cost and logistics data. Crash Costs Crash Prediction The predicted average crash frequency for an intersection can be determined from equations in the HSM (24). These equations, called safety performance functions, are regression models for estimating the predicted average crash frequency of individual roadway segments or intersections for a set of specific base conditions. As discussed in the HSM, each SPF in the predictive method was developed with observed crash data for a set of similar sites. The SPFs, like all regression models, estimate the value of a dependent variable as a function of a set of independent variables. In the SPFs developed for the HSM, the dependent variable estimated is the predicted average crash frequency for a roadway segment or intersection under base conditions, and the independent variables are the AADTs of the roadway segment or intersection legs (and, for roadway segments, the length of the roadway segment). The SPFs applicable to the rural conditions in this study are listed in Table 32, while the SPFs applicable to the urban conditions are listed in Table 33. Table 34 lists the definitions for the variables listed in Table 33. Table 35 lists the acceptable ranges for AADT for each equation. These ADT ranges were not exceeded in the evaluations.

68 Table 32. Safety Performance Functions for Rural Highways for Total Crashes. Number of Lanes Number of Legs Equation Two Three Nspf 2 ln, 3st = exp[−9.86 + 0.79 × ln(AADTmaj) + 0.49 × ln(AADTmin)] (1) Two Four Nspf 2 ln, 4st = exp[−8.56 + 0.60 × ln(AADTmaj) + 0.61 × ln(AADTmin)] (2) Four Three Nspf 4 ln, 3st = exp[−12.526 + 1.204 × ln(AADTmaj) + 0.236 × ln(AADTmin)] (3) Four Four Nspf 4 ln, 4st = exp[−10.008 + 0.848 × ln(AADTmaj) + 0.448 × ln(AADTmin)] (4) Where: Nspf 2 ln, 3st = estimate of intersection-related predicted average crash frequency for base conditions for a rural two-lane highway with three-leg stop-controlled intersections. Nspf 2 ln, 4st = estimate of intersection-related predicted average crash frequency for base conditions for a rural two-lane highway with four-leg stop-controlled intersections. Nspf 4 ln, 3st = estimate of intersection-related predicted average total crash frequency for base conditions for a rural four-lane highway with three-leg stop-controlled intersections. Nspf 4 ln, 4st = estimate of intersection-related predicted average total crash frequency for base conditions for a rural four-lane highway with four-leg stop-controlled intersections. AADTmaj = AADT (vehicles per day) on the major road. AADTmin = AADT (vehicles per day) on the minor road. Table 33. Safety Performance Functions for Urban and Suburban Arterial Intersections for Total Crashes. Number of Legs Crash Type Equation Intersections with Stop Control on the Minor Approach Three Multiple Nspf U/S-MV, 3st = exp[−13.36 + 1.11 × ln(AADTmaj) + 0.41 × ln(AADTmin)] (5) Four Multiple Nspf U/S-MV, 4st = exp[−8.90 + 0.82 × ln(AADTmaj) + 0.25 × ln(AADTmin)] (6) Three Single Nspf U/S-SV, 3st = exp[−6.81 + 0.16 × ln(AADTmaj) + 0.51 × ln(AADTmin)] (7) Four Single Nspf U/S-SV, 4st = exp[−5.33 + 0.33 × ln(AADTmaj) + 0.12 × ln(AADTmin)] (8) Three Multiple and Single Nspf U/S, 3st, M&S = (Nspf U/S-MV, 3st + Nspf U/S-SV, 3st) = (5) + (7) (9) Four Multiple and Single Nspf U/S, 4st, M&S = (Nspf U/S-MV, 4st + Nspf U/S-SV, 4st) (10) Three Pedestrian Nspf U/S-Ped, 3st = 0.021 × (Nspf U/S, 3st, M&S) (11) Four Pedestrian Nspf U/S-Ped, 4st = 0.022 × (Nspf U/S, 4st, M&S) (12) Three Bike Nspf U/S-Bike, 3st = 0.016 × (Nspf U/S, 3st, M&S) (13) Four Bike Nspf U/S-Bike, 4st = 0.018 × (Nspf U/S, 4st, M&S) (14) Three All Nspf U/S, 3st = Nspf U/S, 3st, M&S + Nspf U/S-Ped, 3st + Nspf U/S-Bike, 3st (15) Four All Nspf U/S, 4st = Nspf U/S, 4st, M&S + Nspf U/S-Ped, 4st + Nspf U/S-Bike, 4st (16) Note: Equations and coefficients obtained from Section 12.6.2 of the HSM (24). Variable descriptions are in Table 34.

69 Table 34. Definitions for Variables in Table 33. Variable Definition Nspf U/S-MV, 3st Estimate of multiple-vehicle predicted average crash frequency for base conditions for urban/suburban arterial three-leg intersections with stop control on the minor-road approach (3ST) Nspf U/S-MV, 4st Estimate of multiple-vehicle predicted average crash frequency for base conditions for urban/suburban arterial four-leg intersections with stop control on the minor-road approaches (4ST) Nspf U/S-SV, 3st Estimate of single-vehicle predicted average crash frequency for base conditions for urban/suburban arterial three-leg intersections with stop control on the minor-road approach Nspf U/S-SV, 4st Estimate of single-vehicle predicted average crash frequency for base conditions for urban/suburban arterial four-leg intersections with stop control on the minor-road approaches Nspf U/S, 3st, M&S Estimate of multiple- and single-vehicle predicted average crash frequency for urban/suburban arterial three-leg intersections with stop control on the minor-road approach Nspf U/S, 4st, M&S Estimate of multiple- and single-vehicle predicted average crash frequency for urban/suburban arterial four-leg intersections with stop control on the minor-road approaches Nspf U/S-Ped, 3st Estimate of pedestrian predicted average crash frequency for urban/suburban arterial three-leg intersections with stop control on the minor-road approach Nspf U/S-Ped, 4st Estimate of pedestrian predicted average crash frequency for urban/suburban arterial four-leg intersections with stop control on the minor-road approaches Nspf U/S-Bike, 3st Estimate of bicycle predicted average crash frequency for urban/suburban arterial three-leg intersections with stop control on the minor-road approach Nspf U/S-Bike, 4st Estimate of bicycle predicted average crash frequency for urban/suburban arterial four-leg intersections with stop control on the minor-road approaches Nspf U/S, 3st Estimate of predicted average crash frequency for urban/suburban arterial three-leg intersections with stop control on the minor-road approach Nspf U/S, 4st Estimate of predicted average crash frequency for urban/suburban arterial four-leg intersections with stop control on the minor-road approaches AADTmaj AADT (vehicles per day) on the major road AADTmin AADT (vehicles per day) on the minor road Table 35. Minimum and Maximum AADT for HSM Equations. Intersection Characteristics Major-Approach Minimum to Maximum AADT Minor-Approach Minimum to Maximum AADT Rural Two-Lane Highway with Three-Leg Stop-Controlled Intersections 0 to 19,500 veh/day 0 to 4,300 veh/day Rural Two-Lane Highway with Four-Leg Stop-Controlled Intersections 0 to 14,700 veh/day 0 to 3,500 veh/day Rural Four-Lane Highway with Three-Leg Stop-Controlled Intersections 0 to 78,300 veh/day 0 to 23,000 veh/day Rural Four-Lane Highway with Four-Leg Stop-Controlled Intersections 0 to 78,300 veh/day 0 to 7,400 veh/day Urban and Suburban Arterial Intersections with Three-Leg Stop-Controlled Intersections 0 to 45,700 veh/day 0 to 9,300 veh/day Urban and Suburban Arterial Intersections with Four-Leg Stop-Controlled Intersections 0 to 46,800 veh/day 0 to 5,900 veh/day Figure 16 shows an illustration of predicted crash frequency at stop-controlled rural and urban arterial intersections. The graph shows the predicted crashes for a range of major-road volumes when the minor-road ADT is 2,000 veh/day. The predicted number of crashes for intersections on rural four-lane highways and rural two-lane four-leg intersections is higher than the crash prediction for urban and suburban arterials. The crash prediction in this illustration for rural four-

lane three major-ro F Crash M To obtain is applied considere stop cont MUTCD all-way s follow M effective study als “no less e approach given the warrants An altern A study p several C beacons) character assumed -leg interse ad ADT. igure 16. Il odification an estimate to the pred d for use in rol to AWSC warrants. A top convers UTCD warr for total ent o showed th ffective wh es.” Since th findings fro restriction i ative is to u ublished in MFs. They to AWSC ( istics presen for this stud ctions is sim lustration o Factor of the num icted crash f this analysi available i study by Pe ion in urban ants. When ering volum at for total a en approach is NCHRP m Persaud s not critical se a CMF av 2010 by Sim determined without flash t within the y (e.g., urba ilar to urban f Predicted ber of crash requency (to s are listed i n the HSM rsaud (34) f areas is not analyzing to es less than nd right-ang volumes ar study is to i (34), the res . ailable on t pson and H a CMF of 0 ing beacon developed 0 n and rural, 70 and suburb Crash Fre es at an AW tal crashes) n Table 36. includes a re ound result limited to a tal and righ 6,000 per d le crashes, a e unbalance dentify volu earchers ass he Crash M ummer (32 .393 for the s) for interse .393 CMF volume ran an three-leg quency Usi SC intersec determined The CMF fo striction tha s that showe certain rang t-angle cras ay as it is fo ll-way stop d as when th mes for war umed that m odification F ) using Nort conversion o ctions with matches the ge, crash typ intersectio ng HSM Eq tion or a rou from the SP r convertin t the volum d that the ef e of enterin hes, it “can r higher vol conversion ey are equa ranting a sto eeting the M actors Clea h Carolina d f TWSC (w four legs. T characterist e, and seve ns for a give uations. ndabout, a C F. The CM g a minor-ro es must mee fectiveness g volumes t be just as umes.” The in urban are l on all p control an UTCD ringhouse (2 ata determi ithout flash he roadway ics being rity) except n MF Fs ad t the of hat as is d 9). ned ing for

71 the number of legs. The North Carolina data were based on four-leg intersections; therefore, an assumption was made that the 0.393 CMF would also be valid for three-leg intersections. The CMF for the Simpson and Hummer study was selected over the CMF in the HSM because the CMF in the HSM is only for injury crashes, and all severity crashes are used within the economic analysis. For a rural setting, the assumed CMF was 0.52, which is the value available in the HSM and in the clearinghouse (see Table 36). Table 36. CMFs Considered for This Analysis. Treatment Source Setting Crash Type (Severity) CMF Convert Minor-Road Stop Control to All-Way Stop Control HSM Table 14-5 based on work by Lovell and Hauer (33) Urban (MUTCD warrants are met) All types (injury) 0.30 Convert Minor-Road Stop Control to All-Way Stop Control HSM Table 14-5 based on work by Lovell and Hauer (33) Rural (MUTCD warrants are met) All types (all severities) 0.52 Convert Two-Way (without Flashing Beacons) to All-Way Stop Control (without Flashing Beacons) CMF Clearinghouse (29) based on work by Simpson and Hummer (32) All All (all) 0.393 Convert Minor-Road Stop Control to All-Way Stop Control CMF Clearinghouse (29) based on work by Harwood et al. (43) Rural All (all) 0.52 Convert Intersection with Minor-Road Stop Control to Modern Roundabout HSM Table 14-4 based on work by Rodegerdts et al. (80) Suburban (one or two lanes) All types (all severities) 0.68 Convert Intersection with Minor-Road Stop Control to Modern Roundabout HSM Table 14-4 based on work by Rodegerdts et al. (80) Rural (one lane) All types (all severities) 0.29 2013 Value of a Statistical Life by Crash Severity In 2013, a memorandum was released by the U.S. Department of Transportation regarding the treatment of the economic value of a statistical life (VSL) in developmental analyses (81). The memorandum “identifies $9.1 million as the VSL to be used for Department of Transportation analyses assessing the benefits of preventing fatalities and using a base year of 2012.” Researchers developed an estimate of the VSL in 2013 dollars by crash severity using the methodology documented in Council et al. (82) and subsequently implemented in the HSM. Table 37 shows the resulting comprehensive society costs by crash severity.

72 Table 37. 2013 Comprehensive Societal Cost Estimates (2013 Dollars). Crash Severity Comprehensive Societal Cost (Low) Comprehensive Societal Cost (Mid-range) Comprehensive Societal Cost (High) Fatality (K) $5,291,800 $9,260,700 $13,127,800 Disabling Injury (A) $284,900 $498,500 $706,700 Evident Injury (B) $104,200 $182,300 $258,500 Possible Injury (C) $59,200 $103,600 $146,800 Property Damage Only (PDO) $9,700 $17,100 $24,200 Note: Values are rounded after spreadsheet calculations. Typical Crash Cost for Unsignalized Intersections The cost per crash at an unsignalized intersection requires knowing the distribution of crash severity for the different intersection configurations. Table 10-5 in the HSM (24), which is reproduced as Table 38 in this report, provides the default proportions for crash severity levels for three-leg and four-leg stop-controlled rural intersections. Table 38. Default Distribution of Crash Severity Level at Rural Two-Lane Two-Way Intersections from the HSM (24). Crash Severity Level Percentage of Total Crashes Three-Leg Stop- Controlled Intersections Four-Leg Stop- Controlled Intersections Four-Leg Signalized Intersections Fatality Incapacitating Injury Nonincapacitating Injury Possible Injury Property Damage Only 1.7 4.0 16.6 19.2 58.5 1.8 4.3 16.2 20.8 56.9 0.9 2.1 10.5 20.5 66.0 Total 100.0 100.0 100.0 Also needed for the analysis is the conversion of the cost per person to a cost per crash; for that information, the number of individuals killed or injured in a crash must be known. A TxDOT study examined crashes at rural intersections. Data available for 595 rural intersections provided the distributions shown in Table 39. For the 1,198 crashes in the dataset, the number of injured persons per crash ranged from 1.22 to 2.30. The fatal crashes had 1.09 deaths per crash at the four-leg intersections and 1.46 deaths per crash at the three-leg intersections. Reflecting the multiple conflict points at an intersection, the average number of vehicles involved at a crash ranged from 1.48 to 2.36 veh/crash for rural intersections.

73 Table 39. Injuries or Deaths per Crash for Rural Two-Way or One-Way Stop Control Intersections. Severity Injuries or Deaths/Crasha Number of Persons/Crash Number of Vehicles/Crash Three Legs Four Legs Three Legs Four Legs Three Legs Four Legs K 1.46 deaths/crash 0.31 A injuries/crash 1.15 B injuries/crash 0.00 C injuries/crash 0.31 no injuries/crash 0.15 unk. injuries/crash 1.09 deaths/crash 0.55 A injuries/crash 0.55 B injuries/crash 0.36 C injuries/crash 0.36 no injuries/crash 0.36 unk. injuries/crash 3.38 3.27 1.54 2.36 A 1.17 A injuries/crash 0.29 B injuries/crash 0.12 C injuries/crash 0.38 no injuries/crash 0.06 unk. injuries/crash 1.40 A injuries/crash 0.47 B injuries/crash 0.43 C injuries/crash 1.83 no injuries/crash 0.10 unk. injuries/crash 2.01 4.23 1.48 2.00 B 1.30 B injuries/crash 0.18 C injuries/crash 0.59 no injuries/crash 0.09 unk. injuries/crash 1.42 B injuries/crash 0.48 C injuries/crash 1.08 no injuries/crash 0.08 unk. injuries/crash 2.15 3.06 1.55 1.87 C 1.22 C injuries/crash 0.89 no injuries/crash 0.13 unk. injuries/crash 1.34 C injuries/crash 1.20 no injuries/crash 0.09 unk. injuries/crash 2.24 2.64 1.53 1.82 PDO 0.00 injuries/crash 0.00 injuries/crash 2.15 2.48 1.61 1.88 a Findings based on 1,189 crashes at 595 rural Texas intersections for the time period of 2003 to 2008. Unk. = unknown. While information on crash distribution and number of persons per crash is available for rural intersections and portions are available for urban conditions (e.g., information is available for red-light running at signalized intersections [83]), the concern is that unsignalized intersections, especially intersections with lower speeds, would have a very different distribution. For this analysis, researchers contacted representatives from cities of various sizes in different parts of the country to request data from their crash databases. Researchers initially requested data for the last available 7 years at all unsignalized intersections on streets with posted speed limits of 40 mph or less. Three cities were able to provide usable data within the time frame of the analysis: Bryan, Texas; Lawrence, Kansas; and Phoenix, Arizona. Only Phoenix was able to provide 7 years’ worth of data, but the other cities were able to share at least 3 years. The resulting database contained information on 10,208 crashes from 6,374 unsignalized intersections; the number of crashes by city and year is shown in Table 40.

74 Table 40. Crashes in Database by City and Year. City 2006 2007 2008 2009 2010 2011 2012 2013 Total Bryan N/Aa N/A N/A N/A 280 251 276 243 1,050 Lawrence N/A N/A N/A N/A N/A 107 126 106 339 Phoenix 1,687 1,634 1,327 1,108 1,041 1,039 983 N/A 8,819 Total 1,687 1,634 1,327 1,108 1,321 1,397 1,385 349 10,208 a N/A = data not available for use in this study. The cities submitted crash data for intersections with a variety of control types. Researchers assigned codes to the control types as follows:  0W (“zero-way” stop control, an uncontrolled intersection).  1W (one-way stop control, a three-leg intersection with stop control only on the minor approach).  2W (two-way stop control, a four-leg intersection with stop control only on the minor approaches).  3W (three-way stop control, a three-leg intersection with all-way stop control).  4W (four-way stop control, a four-leg intersection with all-way stop control).  Y (yield control on the minor approaches). Based on these control types, researchers determined the number of crashes per intersection per year in each of the three cities; that information is summarized in Table 41. Researchers reviewed the crashes by severity and by type of intersection control to determine the distribution of injuries and deaths per crash in the database. In comparison to Table 38, the distribution of crash severity by intersection control is shown in Table 42. The proportion of injury and fatality crashes in the database is lower for all injury severities than the default distribution for rural intersections in the HSM. Correspondingly, the share of PDO (non-injury) crashes is higher; roughly three-fourths of the crashes in the database were non-injury crashes. Crash data from the 6,374 intersections provided the distributions of injuries per crash shown in Table 43. For the 10,208 crashes, the number of injured persons per crash ranged between 1.00 and 1.48, not counting unknown injuries. The fatal crashes resulted in 1.14 deaths per crash at uncontrolled intersections and 1.00 death per crash at intersections with minor-road stop control. There were no fatal crashes at intersections with AWSC or yield control. In Table 44, the number of injuries per crash is summed to show the number of persons involved in each crash. Most of the crash categories on the 0W, 1W, and 2W intersections had an average number of persons per crash less than 2.0, suggesting a sizeable portion of single-vehicle crashes at those intersections.

75 Table 41. Crashes per Intersection for Cities in the Database. City  Control  Number of Legs Number of Intersections  Number of Crashes  Crashes per Intersection  Crashes per Intersection per Year Bryan  All  3 or 4 452 1,050 2.32  0.58   0W  3 41 50 1.22  0.30   1W  3 192 379 1.97  0.49   2W  4 174 536 3.08  0.77   3W  3 4 7 1.75  0.44   4W  4 15 45 3.00  0.75   Y  3 or 4 26 33 1.27  0.32 Lawrence  All  3 or 4 109 339 3.11  1.04   0W  -- 0 0 -- --   1W  -- 0 0 -- --   2W  4 89 269 3.02  1.01   3W  3 3 8 2.67  0.89   4W  4 17 62 3.65  1.22   Y  -- 0 0 -- -- Phoenix  All  3 or 4 5,813 8,819 1.52  0.22   0W  3 or 4 4,237 5,691 1.34  0.19   1W  3 552 926 1.68  0.24   2W  4 1,004 2,131 2.12  0.30   3W  3 8 13 1.63  0.23   4W  4 12 58 4.83  0.69   Y  -- 0 0 -- -- All  All  3 or 4 6,374 10,208 1.60  -- Note: Yearly crashes per intersection values are based on 4 years of data in Bryan, 3 years in Lawrence, and 7 years in Phoenix. -- = No data for the category Table 42. Distribution of Crash Severity Level at Urban and Suburban Unsignalized Intersections in Database (Posted Speed Limit ≤ 40 mph). Crash Severity Level Percentage of Total Crashes 0W 1W 2W 3W 4W Y All Count Fatality Incapacitating Injury Nonincapacitating Injury Possible Injury Property Damage Only 0.4 1.7 8.1 9.4 80.5 0.2 2.3 9.7 13.0 74.8 0.3 2.3 10.2 15.7 71.5 0.0 3.6 0.0 14.3 82.1 0.0 2.4 10.9 13.9 72.7 0.0 6.1 15.2 6.1 72.7 0.3 2.0 8.9 11.8 77.0 32 201 913 1200 7862 Total 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 Count 5,741 1,305 2,936 28 165 33 10,208 10,208

76 Table 43. Injuries or Fatalities per Crash for Crashes in Database. Severitya Injuries or Fatalities/Crash 0W 1W 2W 3W 4W Y K 1.14 K 0.00 A 0.00 B 0.00 C 0.05 N 0.29 U 1.00 K 0.00 A 0.00 B 0.00 C 0.50 N 0.00 U 1.00 K 0.00 A 0.11 B 0.00 C 0.00 N 0.67 U N/Ab N/A N/A A 1.00 A 0.00 B 0.00 C 0.01 N 0.26 U 1.00 A 0.10 B 0.20 C 0.27 N 0.17 U 1.07 A 0.01 B 0.03 C 0.25 N 0.40 U 1.00 A 0.00 B 0.00 C 0.00 N 0.00 U 1.00 A 0.00 B 0.00 C 0.75 N 0.25 U 1.00 A 0.00 B 0.00 C 1.00 N 0.00 U B 1.01 B 0.00 C 0.01 N 0.31 U 1.13 B 0.05 C 0.69 N 0.22 U 1.22 B 0.05 C 0.61 N 0.28 U N/A 1.17 B 0.06 C 0.72 N 0.22 U 1.20 B 0.40 C 1.00 N 0.00 U C 1.27 C 0.01 N 0.00 U 1.32 C 0.66 N 0.01 U 1.43 C 0.59 N 0.02 U 1.25 C 0.25 N 0.00 U 1.48 C 0.17 N 0.09 U 1.00 C 2.00 N 0.00 U PDO 0.00 0.00 0.00 0.00 0.00 0.00 a Crash Severity:  K = fatal  A = incapacitating injury  B = nonincapacitating injury  C = possible injury  N = no injury  U = unknown (not reported) b N/A = not applicable; no crashes in this category. Table 44. Number of Persons per Crash for Crashes in Database. Severity Number of Persons/Crash 0W 1W 2W 3W 4W Y K 1.48 1.50 1.78 N/Aa N/A N/A A 1.27 1.73 1.78 1.00 2.00 2.00 B 1.34 2.09 2.16 N/A 2.17 2.60 C 1.28 1.99 2.03 1.50 2.74 3.00 PDO 1.02 1.44 1.45 2.04 2.09 2.83 a N/A = not applicable; no crashes in this category. Researchers calculated typical crash cost using the ranges for comprehensive societal cost for all of the urban stop-control scenarios (i.e., 1W, 2W, 3W, and 4W) and rural three-leg and four-leg intersections. As an example of the calculation process, Table 45 shows the calculations to determine the typical crash cost using the ranges for comprehensive societal cost for four-leg urban intersections with minor-road stop control (i.e., 2W intersections). A summary of crash costs for all scenarios is shown in Table 46.

77 Table 45. Crash Cost Calculations for Urban 2W Intersections. R an ge (C os t) Crash Severity Injury Severity Cost a, b Convert Cost/Person to Cost/Crashc Cost per Crash Percent of Total Crashesd Extension M id -r an ge ($ 10 0, 00 0) Fatality K A B C $9,260,700 $498,500 $182,300 $103,600 1.00 0.00 0.11 0.00 $9,260,700 $0 $20,956 $0 0.31 $28,450 A A B C $498,500 $182,300 $103,600 1.07 0.01 0.03 $535,701 $2,721 $3,093 2.28 $12,357 B B C $182,300 $103,600 1.22 0.05 $223,150 $5,197 10.18 $23,255 C C $103,600 1.43 $147,872 15.70 $23,218 PDO PDO $17,100 1.00e $17,100 71.53 $12,231 Total (cost/crash) 100.00 $99,511 L ow ($ 57 ,0 00 ) Fatality K A B C $5,291,800 $284,900 $104,200 $59,200 1.00 0.00 0.11 0.00 $5,291,800 $0 $11,578 $0 0.31 $16,257 A A B C $284,900 $104,200 $59,200 1.07 0.01 0.03 $306,161 $1,555 $1,767 2.28 $7,062 B B C $104,200 $59,200 1.22 0.05 $127,549 $2,970 10.18 $13,292 C C $59,200 1.43 $84,498 15.70 $13,268 PDO PDO $9,700 1.00 e $9,700 71.53 $6,938 Total (cost/crash) 100.00 $56,817 H ig h ($ 14 1, 00 0) Fatality K A B C $13,127,800 $706,700 $258,500 $146,800 1.00 0.00 0.11 0.00 $13,127,800 $0 $28,722 $0 0.31 $40,330 A A B C $706,700 $258,500 $146,800 1.07 0.01 0.03 $759,439 $3,858 $4,382 2.28 $17,519 B B C $258,500 $146,800 1.22 0.05 $316,425 $7,365 10.18 $32,974 C C $146,800 1.43 $209,532 15.70 $32,900 PDO PDO $24,200 1.00 e $24,200 71.53 $17,309 Total (cost/crash) 100.00 $141,032 a Comprehensive societal cost for fatal crash is from “Guidance on Treatment of the Economic Value of a Statistical Life (VSL) in U.S. Department of Transportation Analyses,” Memorandum to Secretarial Officers, Modal Administrators, available at http://www.dot.gov/sites/dot.gov/files/docs/VSL Guidance_2013.pdf. b Comprehensive societal cost for crash severity A, B, C, or PDO is based on distribution determined using HSM data, with costs adjusted to 2013 dollars. c Factors from Table 43. d From Table 42. e No factor is needed. Assumption is that cost reflects cost per crash.

78 Table 46. Calculated Crash Costs for Intersection Scenarios. Cost Range Crash Severity Urban 1W Urban 2W Urban 3W Urban 4W Rural Three-Leg Rural Four-Leg M id -r an ge Fatality $14,193 $28,450 $0 $0 $236,042 $189,106 A $12,355 $12,357 $17,804 $12,085 $25,942 $35,610 B $20,452 $23,255 $0 $23,830 $42,436 $49,992 C $17,783 $23,218 $18,500 $21,348 $24,267 $28,875 PDO $12,789 $12,231 $14,046 $12,436 $10,004 $9,730 Total $77,572 $99,511 $50,350 $69,699 $338,690 $313,313 L ow Fatality $8,110 $16,257 $0 $0 $134,881 $108,061 A $7,061 $7,062 $10,175 $6,907 $14,826 $20,351 B $11,690 $13,292 $0 $13,621 $24,255 $28,574 C $10,162 $13,268 $10,571 $12,199 $13,867 $16,500 PDO $7,255 $6,938 $7,968 $7,055 $5,675 $5,519 Total $44,278 $56,817 $28,714 $39,781 $193,504 $179,005 H ig h Fatality $20,119 $40,330 $0 $0 $334,610 $268,074 A $17,515 $17,519 $25,239 $17,132 $36,777 $50,482 B $29,001 $32,974 $0 $33,790 $60,171 $70,881 C $25,198 $32,900 $26,214 $30,250 $34,386 $40,916 PDO $18,099 $17,309 $19,879 $17,600 $14,157 $13,770 Total $109,932 $141,032 $71,332 $98,772 $480,101 $444,123 Note: Crash costs for urban 1W and 2W were used as the base condition in the economic analysis for urban scenarios; rural three-leg and four-leg values were applied to rural scenarios. Vehicle Operating Costs Vehicle operating costs reflect the expenses for users of the network for the operation of their vehicles. Operating costs are affected by changes in vehicle miles traveled (VMT) and user delay. The costs for VMT are calculated for the following components per HERS-ST (75): fuel, oil, maintenance, tires, and depreciation. Changes in operating speed and delay also affect operating costs by changing fuel consumption efficiency. Changes to an intersection’s operations impact fuel consumption based on changes in the average speed along with the number of times users must start and stop their vehicle. For example, two intersections with the same average speed but with differing numbers of starts and stops will result in different fuel consumption and different operating costs. As noted in the NCHRP 3-110 interim report (74), different intersection designs are not likely to cause differences in operating costs that measure the marginal cost of driving additional distance. Users are likely to travel the same distance regardless of intersection design. What is likely to vary among changes in intersection operations is the fuel consumed as a result of user delay. The simulation results can also be used to estimate fuel consumption. Delay is converted to minutes of delay per year and then multiplied by the number of gallons per minute rate available

79 from the AASHTO Red Book and the assumed cost for fuel. Table 47 shows the amount of fuel consumption per minute as a result of delays. For cars, the gallon-per-minute rate was assumed to be an average of the small and big vehicles. The three-axle single-unit vehicle rate was assumed for trucks. Low-speed condition was assumed to be represented by averaging the 30 to 40 mph data, while high-speed was represented by averaging the 45 to 55 mph data. The U.S. Energy Information Administration (EIA) projection, as of May 6, 2014, for the average retail price of regular-grade gasoline for May to December 2014 is $3.49 per gallon (84). EIA’s projection, as of May 6, 2014, for the average retail price of on-highway diesel fuel for May 2014 to December 2014 is $3.83 per gallon (85). These values were assumed to represent 2013 values of fuel. Table 47. Fuel Consumption (Gallons) per Minute of Delay by Vehicle Type (Table 5-6 in 73). Free-Flow Speed Small Car Big Car SUV 2-Axle Single-Unit 3-Axle Single-Unit Combo 20 0.011 0.022 0.023 0.074 0.102 0.198 25 0.013 0.026 0.027 0.097 0.133 0.242 30 0.015 0.030 0.035 0.122 0.167 0.284 35 0.018 0.034 0.037 0.149 0.203 0.327 40 0.021 0.038 0.043 0.177 0.241 0.369 45 0.025 0.043 0.049 0.206 0.280 0.411 50 0.028 0.048 0.057 0.235 0.321 0.453 55 0.035 0.054 0.065 0.266 0.362 0.495 60 0.037 0.060 0.073 0.297 0.404 0.537 65 0.042 0.066 0.083 0.328 0.447 0.578 70 0.047 0.073 0.094 0.360 0.490 0.620 75 0.053 0.080 0.105 0.392 0.534 0.661 Note: Values determined by ECONorthwest calculations based on HERS-ST model equations. From User and Non-User Benefit Analysis for Highways, 2010, by the American Association of State Highway and Transportation Officials, Washington, D.C. Used by Permission. Roundabout Construction Costs The average construction cost of roundabouts is estimated at approximately $250,000 (86) as reported on a 2010 FHWA website. Roundabouts discussed in an FHWA report (86) ranged in cost from $194,000 to just under $500,000, depending on their size and needed right-of-way acquisitions. Using a 20-year service life and a 4 percent return, the $250,000 construction cost can be converted to an annual cost of $18,395, while the $500,000 construction cost would be represented by a $36,790 annual cost. The $18,395 was assumed to represent rural conditions, while the $36,790 annual cost was assumed to represent urban conditions in consideration of the potentially higher right-of-way costs. Total Costs The total cost for a change in intersection traffic control would represent the summation of the crash, user delay, and vehicle operating costs. Table 48 shows the calculated cost when TWSC is

80 converted for a given set of volumes when three or four legs and 40 mph (urban) or 45 mph (rural) speeds are present. When the total cost for these combinations is positive, it indicates that the after treatment would be cost-effective for these conditions. For the volumes represented in Table 48, TWSC is more cost-effective as compared to AWSC except for rural (or higher-speed) scenarios. The higher crash cost for that situation justifies the higher level of control. The roundabout geometric form is more cost-effective compared to TWSC. Table 48. Summary of Costs for an Example Where the Peak Hour Is 300 veh/hr/Approach on the Major Road and 200 veh/hr/Approach on the Minor Road. Lanes on Major Approach Leg Change Rural or Urban Medium Crash Cost ($/yr) Car, Truck, and Pedestrian Delay Costs ($/yr) Car and Truck Operating Costs ($/yr) Initial Year of Construction Cost for Roundabouts ($/yr) Total Cost ($/Year) 2 3 TW3–AW3 Urban $33,898 $(185,877) $(57,513) $ - $(209,493) 2 4 TW4–AW4 Urban $74,918 $(186,378) $(58,284) $ - $(169,744) 2 4 TW4– RO4 Urban $34,248 $156,702 $41,018 $(36,790) $195,178 2 3 TW3–AW3 Rural $113,713 $(189,284) $(87,507) $ - $(163,078) 2 4 TW4–AW4 Rural $364,723 $(189,841) $(88,759) $ - $86,123 2 4 TW4–RO4 Rural $369,933 $163,754 $61,264 $(18,395) $576,556 1 3 TW3–AW3 Urban $33,898 $(270,394) $(83,552) $ - $(320,048) 1 4 TW4–AW4 Urban $74,918 $(209,609) $(65,695) $ - $(200,386) 1 4 TW4– RO4 Urban $34,248 $75,222 $20,716 $(36,790) $93,396 1 3 TW3–AW3 Rural $393,828 $(273,076) $(127,202) $ - $(6,451) 1 4 TW4–AW4 Rural $714,467 $(212,574) $(100,101) $ - $401,792 1 4 TW4– RO4 Rural $724,674 $82,274 $30,958 $(18,395) $819,511 Input variables: major volume = 300 veh/hr/approach, minor volume = 200 veh/hr/approach, pedestrian volume = 10 ped/hr/approach (urban) or 0 ped/hr/approach (rural), two lanes on minor (one-lane approach). TREATMENT RECOMMENDATIONS BASED ON COST Four-Lane Major Road Based on the results of the simulations and calculations described in the previous section, researchers identified the recommended traffic control for each combination of variables based on total cost. When the total cost of the change was positive, the after condition (e.g., all-way stop) was recommended. If total cost was negative, the before condition (i.e., two-way stop) would be recommended. Graphs were generated to illustrate when AWSC or TWSC would be justified for a given major and minor volume. The graphs are shown for the peak hour, which

81 was determined as 7.8 percent of the daily volume used in the cost calculations. The following figures were generated for a four-lane major road:  Figure 17 shows the graph for three-leg urban intersections.  Figure 18 shows the graph for four-leg urban intersections.  Figure 19 shows the graph for three-leg rural intersections.  Figure 20 shows the graph for four-leg rural intersections. Within these graphs, symbols are used to indicate which type of stop control is more economical:  TW3, shown with a blue diamond, identifies conditions where TWSC is more economical for three-legged intersections.  AW3, shown with a red square, identifies conditions where AWSC is more economical for three-legged intersections.  TW4, shown with a blue diamond, identifies conditions where TWSC is more economical for four-legged intersections.  AW4, shown with a red square, identifies conditions where AWSC is more economical for four-legged intersections. To provide a comparison between the findings from the economic analysis and the MUTCD peak-hour signal warrant, the signal warrant criteria were added to each graph (shown with a green solid line). In all urban cases with a four-lane road when using total cost (crash, user, and vehicle operations), the intersection warrants a signal before an all-way stop. The economic analysis approach resulted in roundabouts being a more cost-effective geometry than TWSC for all volume combinations studied when the assumed right-of-way and construction costs are less than $500,000 for the roundabout.

Figure Three-L Hour S Figure Four-L Hour S 17. Recomm eg Urban I ignal War 18. Recomm eg Urban In ignal War ended Uns ntersection rant for Wh ended Uns tersection rant for Wh ignalized I on a Four- en There A One Lane o ignalized I on a Four-L en There A One Lane o 82 ntersection Lane Majo re Two or n the Mino ntersection ane Major re Two or n the Mino Traffic Co r Road; Als More Lane r Road. Traffic Co Road; Als More Lane r Road. ntrol Based o Shown Is s on the Ma ntrol Based o Shown Is s on the Ma on Costs fo the 2009 P jor Road a on Costs fo the 2009 Pe jor Road a r a eak- nd r a ak- nd

Figure Three-L Hour S Figure Four-L Hour S 19. Recomm eg Rural I ignal War 20. Recomm eg Rural In ignal War ended Uns ntersection rant for Wh ended Uns tersection rant for Wh ignalized I on a Four-L en There A One Lane o ignalized I on a Four-L en There A One Lane o 83 ntersection ane Major re Two or n the Mino ntersection ane Major re Two or n the Mino Traffic Co Road; Als More Lane r Road. Traffic Co Road; Also More Lane r Road. ntrol Based o Shown Is s on the Ma ntrol Based Shown Is s on the Ma on Costs fo the 2009 Pe jor Road a on Costs fo the 2009 Pe jor Road a r a ak- nd r a ak- nd

84 Two-Lane Major Road The results of the simulations and calculations described in the previous section were also used to generate recommendations for two-lane major roads. Graphs were generated to illustrate when AWSC or TWSC would be justified for a given major and minor volume. The graphs are shown for the peak hour, which was determined as 7.8 percent of the daily volume used in the cost calculations. The following figures were generated for a two-lane major road:  Figure 21 shows the graph for three-leg urban intersections.  Figure 22 shows the graph for four-leg urban intersections.  Figure 23 shows the graph for three-leg rural intersections.  Figure 24 shows the graph for four-leg rural intersections. Within these graphs, symbols are used to indicate which type of stop control is more economical:  TW3, shown with a blue diamond, identifies conditions where TWSC is more economical for three-legged intersections.  AW3, shown with a red square, identifies conditions where AWSC is more economical for three-legged intersections.  TW4, shown with a blue diamond, identifies conditions where TWSC is more economical for four-legged intersections.  AW4, shown with a red square, identifies conditions where AWSC is more economical for four-legged intersections. When the major road has two lanes, an all-way stop is not justified in the urban environment for both three-leg and four-leg intersections. When fewer major road lanes are present to accommodate the volume, much higher vehicle (car and truck) delay costs are present for the AWSC scenario. The larger number of crashes associated with rural two-lane highways along with higher crash costs for that environment due to higher speeds, resulting in more several crashes, presents a very different recommendation. All-way stops are recommended at four-leg intersections (see Figure 24), except for lower minor-road approach volumes. Similar to the finding from four-lane roads, the economic analysis approach resulted in roundabouts being a more cost-effective geometry than TWSC for all volume combinations studied when the assumed right-of-way and construction costs are less than $500,000 for the roundabout.

Figure Three-L Hour S Figure Four-L Hour S 21. Recomm eg Urban I ignal War 22. Recom eg Urban I ignal War ended Uns ntersection rant for Wh mended Un ntersection rant for Wh ignalized I on a Two-L en There A One Lane o signalized on a Two-L en There A One Lane o 85 ntersection ane Major re Two or n the Mino Intersection ane Major re Two or n the Mino Traffic Co Road; Als More Lane r Road. Traffic Co Road; Also More Lane r Road. ntrol Based o Shown Is s on the Ma ntrol Base Shown Is s on the Ma on Costs fo the 2009 Pe jor Road a d on Costs f the 2009 Pe jor Road a r a ak- nd or ak- nd

Figure Three-L Hour S Figure Four-L Hour S 23. Recomm eg Rural I ignal War 24. Recomm eg Rural In ignal War ended Uns ntersection rant for Wh ended Uns tersection rant for Wh ignalized I on a Two-L en There A One Lane o ignalized I on a Two-L en There A One Lane o 86 ntersection ane Major re Two or n the Mino ntersection ane Major re Two or n the Mino Traffic Co Road; Also More Lane r Road. Traffic Co Road; Also More Lane r Road. ntrol Based Shown Is s on the Ma ntrol Based Shown Is t s on the Ma on Costs fo the 2009 Pe jor Road a on Costs fo he 2009 Pe jor Road a r a ak- nd r a ak- nd

87 Discussion Regarding Use of Speed and Number of Legs As part of the research, the effects of speed were to be considered within the warrant development. Crash prediction equations are not sensitive to speed; however, there are different equations for the rural and urban conditions. Therefore, development environment (i.e., rural or urban) was used as a surrogate for speed. In addition, the cost of a crash is higher in rural areas due to increased severity associated with the higher speeds. Therefore, crash costs are influential in generating different results for urban (low-speed) and rural (high-speed) conditions. Initial simulation runs did include a range of major and minor speeds; however, as was illustrated in Table 27, the variability in delay for the range of speeds and volumes being considered was minimal. The influence of speed was included in the development of vehicle operating costs because vehicles consume more fuel at higher speeds. Previous MUTCD warrants were not sensitive to the number of legs at the intersection. Primarily due to different crash predictions for three- and four-leg intersections, the warrant criteria should also reflect the number of legs present. The economic analysis considered the number of legs at the intersection.