U.S. Department of Transportation
Federal Highway Administration
1200 New Jersey Avenue, SE
Washington, DC 20590
202-366-4000


Skip to content
FacebookYouTubeTwitterFlickrLinkedIn

Federal Highway Administration Research and Technology
Coordinating, Developing, and Delivering Highway Transportation Innovations

 
REPORT
This report is an archived publication and may contain dated technical, contact, and link information
Back to Publication List        
Publication Number:  FHWA-HRT-16-054    Date:  October 2016
Publication Number: FHWA-HRT-16-054
Date: October 2016

 

Investigating the Impact of Lack of Motorcycle Annual Average Daily Traffic Data in Crash Modeling and the Estimation of Crash Modification Factors

 

Chapter 4. Data Collection and Summary

This chapter describes the data collection procedures undertaken to support the analyses outlined in chapter 3. In general terms, these data are primarily roadway inventory, crash data, total traffic volumes, and motorcycle traffic volumes. Other data collection includes further information on motorcycle licenses, registrations, and other sociodemographic data at the county level.

To identify which States could provide a substantial sample of motorcycle volume counts, the project team asked various contacts for this information, primarily through the following channels:

These inquiries quickly confirmed that few States have substantial motorcycle volume estimates. For many States, the only available motorcycle counts come from permanent counting stations, where full class counts are performed and the interest is focused on truck traffic, not motorcycles. No attempt is made to estimate motorcycle counts for nearby segments in these instances.

There were, however, a number of States reported by these contacts as having a large number of classification count locations from permanent and short-term count programs. These States focused on for determining which States could provide the best datasets for the analyses.

The prioritization of States for data collection considered several factors, including the following:

For each method applied, this section describes the source of data, variables acquired, and steps taken to assemble the data into an appropriate format for analysis.

Avenue A Databases

For developing the avenue A models, data were collected from Florida and Pennsylvania. The project team selected these States because they had a large number of locations with an estimated motorcycle AADT; were able to provide linkable roadway inventory, traffic, and crash data; and expressed an interest in providing data in a timely manner. They also provided a degree of geographical diversity. The project team acquired data from Virginia to validate the models developed.

The following sections provide further details on the data acquired and data manipulation.

Florida

Florida provided statewide roadway inventory, traffic volume, and crash data. The Florida Department of Transportation (FDOT) provided the roadway inventory and traffic volume data on the 2012 Florida Transportation Information DVD. Roadway and traffic volume data variables included the following:

Crash data were provided for 2008 to 2012, including the following:

The query included total, motorcycle, multi-vehicle motorcycle, and single-vehicle motorcycle crashes.

Roadway identification numbers and mileposts linked all roadway, traffic, and crash data. Segments were defined so that all variables were homogeneous for the entire length.

Further data collected included the number of motorcycle licenses and registrations by county from 2008 to 2012. These data were available from the Florida Department of Highway Safety and Motor Vehicles and were linked to the other data by county.

Sociodemographic data were obtained from the U.S. Census Bureau at the county level. These variables included the following:

These data were also linked to the other data by county. Segments were initially grouped into the following six categories for analysis:

Table 3 provides the variable definitions of the database created. Table 4 shows the total length of roadway and total number of crashes by crash type for the six site types in Florida. Table 5 through table 8 provide summary statistics for other variables included in the Florida dataset by site type.

Table 3. Florida data variable definitions.
Variable Definition
AVGMOTO Motorcycle AADT
AVGAADT Total vehicle AADT
NOLANES Number of lanes
SPDLIMT Posted speed limit (mi/h)
SURFWIDTH Surface width in ft
DIVUND Divided versus undivided
OUTSHLDWID Outside shoulder width in ft
MEDWIDTH Median width in ft
INSHLDWID Inside shoulder width in ft
CURVLENGTH Length of horizontal curves in segment in mi
LENGTH Total segment length in mi
TOT Total crash frequency
MOTO Motorcycle crash frequency
MOTOSINGLE Single-vehicle motorcycle crash frequency
MOTOMULTI Multi-vehicle motorcycle crash frequency
VEHREG Number of registered motorcycles in county
LIC Number of motorcycle-licensed individuals in county
PST045213 Population, 2013 estimate
AGE135213 Persons under 5 years, percent, 2013
AGE295213 Persons under 18 years, percent, 2013
AGE775213 Persons 65 years and over, percent, 2013
SEX255213 Female persons, percent, 2013
EDU635213 High school graduate or higher, percent of persons age 25+, 2009–2013
EDU685213 Bachelor’s degree or higher, percent of persons age 25+, 2009–2013
LFE305213 Mean travel time to work (minutes), workers age 16+, 2009–2013
HSG010213 Housing units, 2013
HSG445213 Homeownership rate, 2009–2013
HSD310213 Persons per household, 2009–2013
INC110213 Median household income, 2009–2013
PVY020213 Persons below poverty level, percent, 2009–2013
NES010212 Non-employer establishments, 2012
SBO001207 Total number of firms, 2007
LND110210 Land area in mi2, 2010
POP060210 Population per mi2, 2010
Table 4. Total length and crash frequency for Florida data.
Type LENGTH (mi) TOT MOTO MOTOSINGLE MOTOMULTI
1 494.2 10,611 174 111 63
2 552.6 49,453 921 385 536
3 3,068.2 21,400 1,031 449 582
4 3,425.8 269,689 9,539 2,341 7,198
5 640.8 698 33 20 13
6 1,041.2 2,800 152 57 95

1 mi = 1.6 km.

Table 5. Summary statistics for Florida data.
Type Statistic AVGMOTO AVGAADT (mi) NOLANES SPDLIMT (mi/h) SURFWIDTH (ft) DIVUND OUTSHLDWID (ft)
1 No. Segments 482 482 482 482 482 482 482
1 MIN 18.9 16,050.0 4.0 65.0 48.0 2.0 6.0
1 MAX 339.9 98,700.0 8.0 70.0 96.0 2.0 21.0
1 MEAN 105.6 34,845.0 4.5 69.9 53.4 2.0 10.0
1 STD 72.2 19,303.0 0.9 0.8 10.3 0.0 1.2
2 No. Segments 952 952 952 952 952 952 952
2 MIN 18.9 6,120.0 2.0 30.0 24.0 0.0 2.0
2 MAX 39,488.9 316,000.0 10.0 70.0 128.0 2.0 35.0
2 MEAN 862.1 84,174.3 5.7 64.8 68.5 2.0 10.0
2 STD 3,795.4 56,359.9 1.6 6.8 20.0 0.1 3.0
3 No. Segments 3,243 3,243 3,243 3,243 3,243 3,243 3,243
3 MIN 1.4 450.0 2.0 25.0 18.0 0.0 2.0
3 MAX 507.8 43,000.0 6.0 70.0 72.0 2.0 26.0
3 MEAN 51.0 8,044.0 2.6 54.4 30.7 0.9 5.3
3 STD 52.5 6,820.3 0.9 7.6 11.0 1.0 2.2
4 No. Segments 8,721 8,721 8,721 8,721 8,721 8,721 8,721
4 MIN 0.6 170.0 2.0 20.0 18.0 0.0 1.0
4 MAX 2,416.4 92,785.0 10.0 65.0 120.0 2.0 32.0
4 MEAN 171.5 26,198.8 4.0 44.4 47.4 1.7 4.6
4 STD 141.2 15,797.2 1.6 6.9 18.3 0.7 2.9
5 No. Segments 622 622 622 622 622 622 622
5 MIN 1.4 100.0 1.0 25.0 12.0 0.0 1.0
5 MAX 274.4 31,000.0 4.0 60.0 52.0 2.0 12.0
5 MEAN 32.3 4,416.6 2.1 47.5 23.7 0.3 6.1
5 STD 37.9 4,211.4 0.3 8.6 4.6 0.7 3.1
6 No. Segments 2,345 2,345 2,345 2,345 2,345 2,345 2,345
6 MIN 1.9 170.0 2.0 15.0 14.0 0.0 1.0
6 MAX 362.1 55,000.0 6.0 55.0 78.0 2.0 25.0
6 MEAN 63.8 8,498.3 2.4 37.9 27.3 0.8 6.3
6 STD 51.1 6,679.8 0.9 7.6 10.1 1.0 3.9

STD = Standard deviation.

Table 6. Summary statistics for Florida data continued—roadway geometry.
Type Statistic MEDWIDTH (ft) INSHLDWID (ft) CURVLENGTH (mi) LENGTH (mi) TOT MOTO MOTOSINGLE MOTOMULTI
1 No. Segments 482 482 482 482 482 482 482 482
1 MIN 3.0 3.0 0.0 0.1 0.0 0.0 0.0 0.0
1 MAX 961.0 24.0 1.1 23.4 232.0 5.0 4.0 3.0
1 MEAN 147.2 6.5 0.1 1.0 22.0 0.4 0.2 0.13
1 STD 159.2 3.6 0.2 1.8 31.4 0.8 0.6 0.39
2 No. Segments 952 932 952 952 952 952 952 952
2 MIN 0.0 2.0 0.0 0.1 0.0 0.0 0.0 0.0
2 MAX 989.0 51.0 2.0 7.8 707.0 13.0 5.0 9.0
2 MEAN 84.8 9.3 0.1 0.6 51.9 1.0 0.4 0.56
2 STD 98.3 4.3 0.1 0.8 76.2 1.6 0.8 1.08
3 No. Segments 3,243 496 3,243 3,243 3,243 3,243 3,243 3,243
3 MIN 0.0 1.0 0.0 0.1 0.0 0.0 0.0 0.0
3 MAX 480.0 15.0 1.2 16.4 155.0 11.0 7.0 7.0
3 MEAN 13.8 2.9 0.0 0.9 6.6 0.3 0.1 0.18
3 STD 21.9 1.6 0.1 1.6 11.4 0.8 0.5 0.54
4 No. Segments 8,721 3,459 8,721 8,721 8,721 8,721 8,721 8,721
4 MIN 0.0 1.0 0.0 0.1 0.0 0.0 0.0 0.0
4 MAX 377.0 27.0 2.0 6.1 1,578.0 77.0 16.0 69.0
4 MEAN 20.8 2.3 0.0 0.4 30.9 1.1 0.3 0.83
4 STD 18.5 1.4 0.1 0.4 60.2 2.3 0.7 1.91
5 No. Segments 622 16 622 622 622 622 622 622
5 MIN 0.0 2.0 0.0 0.1 0.0 0.0 0.0 0.0
5 MAX 170.0 7.0 5.3 14.6 25.0 4.0 1.0 3.0
5 MEAN 2.5 2.4 0.0 1.0 1.1 0.1 0.0 0.02
5 STD 10.1 1.3 0.2 1.8 2.9 0.3 0.2 0.17
6 No. Segments 2,345 222 2,345 2,345 2,345 2,345 2,345 2,345
6 MIN 0.0 1.0 0.0 0.1 0.0 0.0 0.0 0.0
6 MAX 154.0 25.0 0.5 5.4 95.0 5.0 3.0 3.0
6 MEAN 7.0 2.3 0.0 0.4 1.2 0.1 0.0 0.0
6 STD 12.5 2.1 0.0 0.5 5.2 0.3 0.2 0.2
Table 7. Summary statistics for Florida data continued—model estimates for variables VEHREG, LIC, and a through i.
Type Statistic VEHREG LIC a b c d e f g h i
1 No. Segments 482 482 482 482 482 482 482 482 482 482 482
1 MIN 0.03 0.06 14,194.00 2.10 7.70 10.90 42.10 75.10 9.40 19.90 5,663.00
1 MAX 4.00 6.94 1,838,844.00 6.80 25.10 51.60 52.50 91.90 44.20 30.00 812,565.00
1 MEAN 0.77 1.45 232,353.61 5.22 19.57 21.67 49.71 84.83 21.65 25.57 115,870.27
1 STD 0.84 1.51 272,384.41 1.06 3.39 8.60 2.03 4.86 7.02 2.56 125,062.91
2 No. Segments 952 952 952 952 952 952 952 952 952 952 952
2 MIN 0.11 0.22 48,922.00 3.20 13.30 10.50 45.00 78.00 13.30 19.90 20,715.00
2 MAX 4.21 6.94 2,617,176.00 6.80 25.40 37.00 52.50 91.90 44.20 30.60 993,993.00
2 MEAN 2.43 4.26 1,034,493.61 5.62 20.89 17.54 51.27 86.72 27.49 25.63 450,529.30
2 STD 1.23 1.97 736047.64 0.76 2.23.00 5.84 0.70 3.27 5.66 2.40 286,103.34
3 No. Segments 3,243 3,243 3,243 3,243 3,243 3,243 3,243 3,243 3,243 3,243 3,243
3 MIN 0.01 0.03 8,349.00 2.10 7.70 10.50 35.30 64.20 7.80 18.80 3,282.00
3 MAX 4.21 6.94 2,617,176.00 7.90 28.30 51.60 52.50 93.20 44.20 32.10 993,993.00
3 MEAN 0.81 1.47 272,039.92 5.28 19.96 20.86 48.93 82.95 19.81 25.58 126,694.96
3 STD 0.94 1.61 426,048.86 0.94 3.11 7.30 3.30 6.67 8.10 2.81 176,082.45
4 No. Segments 8,721 8,721 8,721 8,721 8,721 8,721 8,721 8,721 8,721 8,721 8,721
4 MIN 0.04 0.09 22,857.00 2.10 7.70 10.50 43.60 64.20 8.80 18.80 9,516.00
4 MAX 4.21 6.94 2,617,176.00 7.90 28.30 51.60 52.50 93.20 44.20 32.10 993,993.00
4 MEAN 2.13 3.77 847,443.97 5.42 20.40 19.87 51.14 86.43 25.92 25.47 382,889.66
4 STD 1.25 2.06 694,603.48 0.81 2.55 6.35 1.07 4.09 6.24 2.53 281,445.36
5 No. Segments 622 622 622 622 622 622 622 622 622 622 622
5 MIN 0.01 0.03 8,349.00 2.10 7.70 10.90 35.30 71.30 7.80 18.80 3,298.00
5 MAX 4.21 6.17 2,617,176.00 6.70 25.40 51.60 52.50 91.80 44.20 32.10 993,993.00
5 MEAN 0.92 1.74 329,206.54 5.14 19.63 20.78 48.63 83.03 19.41 26.18 156,257.94
5 STD 0.99 1.80 426,773.89 0.83 2.97 7.78 3.64 5.12 7.38 2.79 191,564.69
6 No. Segments 2,345 2,345 2,345 2,345 2,345 2,345 2,345 2,345 2,345 2,345 2,345
6 MIN 0.07 0.14 26,850.00 2.10 7.70 10.50 43.70 64.40 9.20 18.80 10,837.00
6 MAX 4.21 6.94 2,617,176.00 7.90 28.30 51.60 52.50 91.90 44.20 30.60 993,993.00
6 MEAN 1.76 3.23 620,362.09 5.14 19.73 23.25 51.23 86.51 24.22 25.32 295,750.35
6 STD 1.03 1.73 504,284.65 0.83 2.71 6.64 0.89 3.15 5.76 2.40 216,888.03

a = PST045213.
b = AGE135213.
c = AGE295213.
d = AGE775213.
e = SEX255213.
f = EDU635213.
g = EDU685213.
h = LFE305213.
i = HSG010213.

Table 8. Summary statistics for Florida data continued—model estimates for variables J through r.
Type Statistic J k l m N o p q r
1 No. Segments 482 482 482 482 482 482 482 482 482
1 MIN 5,663.00 53.80 4,657.00 33,833.00 12.00 527.00 816.00 472.54 24.70
1 MAX 812,565.00 90.20 663,458.00 57,703.00 26.50 215,377.00 237,524.00 1,998.32 1,444.90
1 MEAN 115,870.27 73.38 88,345.35 44,336.88 17.38 179,14.67 21,492.63 838.31 260.78
1 STD 125,062.91 5.89 100,444.19 6,209.94 4.41 25,988.27 29,450.56 377.32 258.27
2 No. Segments 952 952 952 952 952 952 952 952 952
2 MIN 20,715.00 53.80 16,244.00 36,809.00 11.30 2,642.00 3,106.00 273.80 54.20
2 MAX 993,993.00 80.20 828,031.00 58,175.00 23.20 385,593.00 40,3672.00 1,998.32 3,347.50
2 MEAN 450,529.30 65.13 369,088.93 47,390.83 16.60 107,506.72 119,041.09 1,017.32 1,049.15
2 STD 286,103.34 6.99 240,329.25 3,870.59 2.60 109,657.00 114,337.80 486.79 764.17
3 No. Segments 3,243 3,243 3,243 3,243 3,243 3,243 3,243 3,243 3,243
3 MIN 3,282.00 53.80 2,305.00 32,497.00 9.60 318.00 247.00 243.56 10.00
3 MAX 993,993.00 90.20 828,031.00 64,876.00 29.60 385,593.00 403,672.00 1,998.32 1,444.90
3 MEAN 126,694.96 72.58 98,822.28 43,205.24 18.48 24,274.26 27,818.92 905.12 255.82
3 STD 176,082.45 6.72 143,902.59 6,734.04 4.79 55,716.44 58,978.97 419.83 289.70
4 No. Segments 8,721 8,721 8,721 8,721 8,721 8,721 8,721 8,721 8,721
4 MIN 9,516.00 53.80 7,463.00 32,497.00 9.60 976.00 1,363.00 273.80 21.60
4 MAX 993,993.00 90.20 828,031.00 64,876.00 29.60 385,593.00 403,672.00 1,998.32 3,347.50
4 MEAN 382,889.66 68.15 308,052.76 47,074.35 16.14 85,823.48 96,295.40 1,050.18 870.48
4 STD 281,445.36 6.85 234,944.41 4,691.20 3.01 97,161.36 102,438.99 515.71 789.18
5 No. Segments 622 622 622 622 622 622 622 622 622
5 MIN 3,298.00 53.80 2,305.00 32,780.00 9.80 411.00 354.00 243.56 10.00
5 MAX 993,993.00 90.20 828,031.00 59,482.00 29.60 385,593.00 403,672.00 1,998.32 1,315.50
5 MEAN 156,257.94 73.17 120,845.27 42,810.66 18.18 27,168.37 31,470.86 894.82 311.57
5 STD 191,564.69 6.68 152,005.26 6,072.52 3.98 48,519.98 52,539.47 441.17 316.27
6 No. Segments 2,345 2,345 2,345 2,345 2,345 2,345 2,345 2,345 2,345
6 MIN 10,837.00 53.80 8,857.00 34,963.00 11.30 1,229.00 1,899.00 273.80 34.00
6 MAX 993,993.00 90.20 828,031.00 58,175.00 29.60 385,593.00 403,672.00 1,998.32 3,347.50
6 MEAN 295,750.35 71.55 231,854.99 46,281.02 15.71 55,193.08 63,522.63 1,089.80 609.38
6 STD 216,888.03 5.71 178,603.72 4,641.58 2.49 63,611.04 68,193.67 529.46 578.72

j = HSG010213.
k = HSG445213.
l = HSD410213.
m = INC110213.
n = PVY020213.
o = NES010212.
p = SBO001207.
q = LND110210.
r = POP060210.

Pennsylvania

Pennsylvania provided statewide roadway inventory, traffic volume, and crash data. Roadway and traffic volume data variables included the following:

Crash data were provided for 2009–2013, including the following:

The query included total, motorcycle, multi-vehicle motorcycle, and single-vehicle motorcycle crashes. County, route number, segment number, and offset linked all roadway, traffic, and crash data. Segments were defined so that all variables were homogeneous for the entire length. The data for freeways are for one direction of travel only.

Further data collected included the number of motorcycle licenses and registrations by county from 2009–2013. These data were available from the Department of Motor Vehicles and linked to the other data by county.

The project team obtained sociodemographic data from the U.S. Census Bureau at the county level. These variables included the following:

These data were also linked to the other data by county. Segments were initially grouped into the following four categories for analysis:

Table 9 provides the variables’ definitions. Table 10 shows the total length of roadway and total number of crashes by crash type for the six site types in Pennsylvania. Table 11 to table 14 provide summary statistics for other variables included in the Pennsylvania dataset by site type.

Table 9. Pennsylvania data variable definitions.
Variable Definition
AVGMOTO Motorcycle AADT
AVGAADT Total vehicle AADT
NOLANES Number of lanes
SPDLIMT Posted speed limit (mi/h)
SURFWIDTH Surface width in ft
DIVUND Divided versus undivided
LSHLDWID Left side shoulder width in ft
MEDWIDTH Median width in ft
RSHLDWID Right side shoulder width in ft
WIDTH Total lane widths in ft
LENGTH Total segment length in mi
TOT Total crash frequency
MOTO Motorcycle crash frequency
MOTOSINGLE Single-vehicle motorcycle crash frequency
MOTOMULTI Multi-vehicle motorcycle crash frequency
VEHREG Number of registered motorcycles in county
LIC Number of motorcycle licenced individuals in county
PST045213 Population, 2013 estimate
AGE135213 Persons under 5 years, percent, 2013
AGE295213 Persons under 18 years, percent, 2013
AGE775213 Persons 65 years and over, percent, 2013
SEX255213 Female persons, percent, 2013
EDU635213 High school graduate or higher, percent of persons age 25+, 2009–2013
EDU685213 Bachelor’s degree or higher, percent of persons age 25+, 2009–2013
LFE305213 Mean travel time to work (minutes), workers age 16+, 2009–2013
HSG010213 Housing units, 2013
HSG445213 Homeownership rate, 2009–2013
HSD310213 Persons per household, 2009–2013
INC110213 Median household income, 2009–2013
PVY020213 Persons below poverty level, percent, 2009–2013
NES010212 Non-employer establishments, 2012
SBO001207 Total number of firms, 2007
LND110210 Land area in mi2, 2010
POP060210 Population per mi2, 2010
Table 10. Summary of length and crash frequency for Pennsylvania data.
Type LENGTH (mi) TOT MOTO MOTOSINGLE MOTOMULTI
1 2,136 15,603 234 184 50
2 1,666 31,853 615 382 233
3 20,559 82,666 3,749 2,432 1,317
4 6,252 125,642 3,893 1,530 2,363

Note: Type 1 and type 2 data are for one direction of travel only.
1 mi = 1.6 km.

Table 11. Summary statistics for Pennsylvania data.
Type Statistic SURFWID (ft) WID (ft) SPDLIMT (mi/h) NOLANES LSHLDWID (ft) RSHLDWID (ft) AVGAADT (mi) AVGMOTO (mi) DIVUND LENGTH (mi)
1 No. Segments 4,350 4,350 4,350 4,350 4,350 4,350 4,350 4,350 4,350 4,350
1 MIN 52.0 12.0 35.0 1.0 0.0 0.0 390.7 0.1 1.0 0.1
1 MAX 98.0 52.0 70.0 4.0 15.0 14.0 27,673.3 735.0 1.0 1.1
1 MEAN 64.4 24.5 63.5 2.0 7.0 6.8 4,515.0 72.0 1.0 0.5
1 STD 5.2 2.7 3.8 0.1 3.1 3.1 4,010.6 81.6 0.0 0.1
2 No. Segments 3,459 3,459 3,459 3,459 3,459 3,459 3,459 3,459 3,459 3,459
2 MIN 52.0 12.0 35.0 1.0 0.0 0.0 347.3 1.1 1.0 0.1
2 MAX 98.0 72.0 65.0 5.0 24.0 20.0 118,193.2 817.0 1.0 1.2
2 MEAN 65.6 26.6 57.0 2.1 6.4 6.7 10,524.0 125.3 1.0 0.5
2 STD 6.8 6.8 6.0 0.4 3.5 3.5 9,871.5 127.4 0.0 0.1
3 No. Segments 43,914 43,914 43,914 43,914 43,914 43,914 4,3908 43,914 43,914 43,914
3 MIN 20.0 8.0 0.0 1.0 0.0 0.0 14.8 0.2 0.0 0.1
3 MAX 99.0 70.0 55.0 3.0 14.0 14.0 27,958.8 727.0 0.0 1.3
3 MEAN 57.2 20.9 45.2 2.0 1.5 1.5 1,753.0 22.1 0.0 0.5
3 STD 7.1 3.4 8.2 0.1 2.1 2.2 2,345.1 36.2 0.0 0.1
4 No. Segments 14,438 14,438 14,438 14,438 14,438 14,438 14,438 14,438 14,438 14,438
4 MIN 40.0 10.0 0.0 1.0 0.0 0.0 71.2 0.8 0.0 0.1
4 MAX 99.0 80.0 60.0 5.0 15.0 22.0 36,010.6 2,078.0 0.0 0.8
4 MEAN 59.0 26.0 37.8 2.0 2.1 2.2 6,599.0 64.8 0.0 0.4
4 STD 8.1 7.7 7.6 0.2 2.6 2.7 4,969.9 71.5 0.0 0.2
Table 12. Summary statistics for Pennsylvania data continued—roadway geometry.
Type Statistic TOT MOTO MOTOSINGLE MOTOMULTI
1 No. Segments 4,350 4,350 4,350 4,350
1 MIN 0.0 0.0 0.0 0.0
1 MAX 34.0 3.0 2.0 2.0
1 MEAN 3.6 0.1 0.0 0.0
1 STD 3.0 0.2 0.2 0.1
2 No. Segments 3,459 3,459 3,459 3,459
2 MIN 0.0 0.0 0.0 0.0
2 MAX 125.0 6.0 6.0 3.0
2 MEAN 9.2 0.2 0.1 0.1
2 STD 12.2 0.5 0.4 0.3
3 No. Segments 43,914 43,914 43,914 43,914
3 MIN 0.0 0.0 0.0 0.0
3 MAX 83.0 9.0 9.0 3.0
3 MEAN 1.9 0.1 0.1 0.0
3 STD 3.1 0.3 0.3 0.2
4 No. Segments 14,438 14,438 14,438 14,438
4 MIN 0.0 0.0 0.0 0.0
4 MAX 129.0 9.0 7.0 6.0
4 MEAN 8.7 0.3 0.1 0.2
4 STD 10.8 0.6 0.4 0.5
Table 13. Summary statistics for Pennsylvania data continued—model estimates for variables VEHREG, LIC, and A through H.
Type Statistic VEHREG LIC A B C D E F G H
1 No. Segments 5,817 5,817 5,817 5,817 5,817 5,817 5,817 5,817 5,817 5,817
1 MIN 574.4 1,388.6 14,670.0 3.9 15.5 12.3 44.7 81.5 11.5 18.6
1 MAX 27,476.4 61,992.6 1,231,527.0 6.6 24.3 20.2 52.1 93.3 40.4 41.5
1 MEAN 5,287.3 10,998.6 135,337.1 5.2 20.5 17.9 50.4 88.1 20.5 23.9
1 STD 3,602.3 7,474.1 105,230.9 0.6 1.6 1.8 1.1 2.4 6.3 4.4
2 No. Segments 4,340 4,340 4,340 4,340 4,340 4,340 4,340 4,340 4,340 4,340
2 MIN 16,55.6 3,090.8 37,838.0 4.1 15.5 12.3 44.7 81.2 11.5 18.6
2 MAX 27,476.4 61,992.6 1,553,165.0 7.0 24.3 20.2 52.7 93.5 48.5 38.9
2 MEAN 11,897.6 25,884.1 459,256.6 5.4 20.9 16.7 51.0 89.5 28.3 24.5
2 STD 7,327.3 16,611.3 415,431.1 0.6 1.8 2.0 0.9 3.4 9.2 3.6
3 No. Segments 58,095 58,095 58,095 58,095 58,095 58,095 58,095 58,095 58,095 58,095
3 MIN 306.6 655.6 4,886.0 1.7 8.8 12.3 33.1 81.5 7.8 15.2
3 MAX 27,476.4 61,992.6 1,231,527.0 6.6 24.3 26.5 52.1 93.5 48.5 41.5
3 MEAN 5,525.3 11,481.7 141,715.9 5.1 20.4 18.2 50.0 87.8 19.6 24.5
3 STD 5,092.4 10,551.0 152,302.6 0.6 2.0 2.0 1.9 2.4 6.6 4.0
4 No. Segments 19,074 19,074 19,074 19,074 19,074 19,074 19,074 19,074 19,074 19,074
4 MIN 731.0 1,633.0 18,541.0 4.0 16.0 12.0 45.0 81.0 12.0 17.0
4 MAX 27,476.0 61,993.0 1,553,165.0 7.0 24.0 20.0 53.0 94.0 49.0 39.0
4 MEAN 12,526.0 27,127.0 444,862.0 5.0 21.0 17.0 51.0 90.0 28.0 25.0
4 STD 7,315.0 16,543.0 359,173.0 1.0 2.0 2.0 1.0 3.0 10.0 3.0

A = PST045213.
B = AGE135213.
C = AGE295213.
D = AGE775213.
E = SEX255213.
F = EDU635213.
G = EDU685213.
H = LFE305213.

Table 14. Summary statistics for Pennsylvania data continued—model estimates for variables I through Q.
Type Statistic I J K L M N O P Q
1 No. Segments 5,817 5,817 5,817 5,817 5,817 5,817 5,817 5,817 5,817
1 MIN 70,71.0 59.6 5,965.0 39,115.0 5.4 941.0 1,238.0 130.2 33.9
1 MAX 587,831.0 84.2 526,004.0 76,555.0 20.5 73,833.0 95,698.0 1,228.6 1,675.6
1 MEAN 60,026.7 73.4 53,304.0 47,603.9 13.6 7,672.9 10,309.8 731.6 196.3
1 STD 43,747.3 4.5 40,928.5 6,013.2 2.8 6,441.7 8,289.7 233.1 147.2
2 No. Segments 4,340 4,340 4,340 4,340 4,340 4,340 4,340 4,340 4,340
2 MIN 16,057.0 53.3 14,397.0 37,192.0 5.4 1,541.0 2,203.0 134.1 44.2
2 MAX 667,571.0 80.0 580,017.0 86,050.0 26.5 77,675.0 95,698.0 1,228.6 11,379.5
2 MEAN 199,906.7 70.6 180,965.5 54,722.2 12.6 28,434.4 36,364.3 695.1 1,108.6
2 STD 189,067.9 5.9 167,709.4 12,955.0 4.7 25,000.5 31,022.7 245.8 2,312.7
3 No. Segments 58,095 58,095 58,095 58,095 58,095 58,095 58,095 58,095 58,095
3 MIN 4,382.0 59.6 2,001.0 36,556.0 5.4 250.0 0.0 130.2 12.8
3 MAX 587,831.0 84.2 526,004.0 86,050.0 20.5 73,833.0 95,698.0 1,228.6 1,675.6
3 MEAN 62,360.2 74.5 55,271.9 48,178.5 13.2 8,516.2 11,287.3 769.3 198.4
3 STD 62,676.0 4.1 58,856.9 7,852.3 2.8 10,328.0 13,173.8 250.1 220.8
4 No. Segments 19,074 19,074 19,074 19,074 19,074 19,074 19,074 19,074 19,074
4 MIN 8,003.0 53.0 7,233.0 37,192.0 5.0 1,042.0 1,353.0 130.0 39.0
4 MAX 667,571.0 80.0 580,017.0 86,050.0 27.0 77,675.0 95,698.0 1,229.0 11,380.0
4 MEAN 190,998.0 72.0 174,372.0 56,612.0 12.0 28,809.0 37,125.0 678.0 905.0
4 STD 163,509.0 5.0 146,785.0 13,132.0 4.0 23,597.0 29,774.0 243.0 1,517.0

I = HSG010213.
J = HSG445213.
K = HSD410213.
L = INC110213
M = PVY020213.
N = NES010212.
O = SBO001207.
P = LND110210.
Q = POP060210.

Virginia

Virginia provided statewide roadway inventory, traffic volume, and crash data. The roadway inventory and total AADT data were provided from the Virginia Department of Transportation (VDOT) EYROAD data files. Motorcycle AADTs, where available, were queried specifically for the project and provided by VDOT staff. Crash data were downloaded from the VDOT Web site.

Roadway and traffic volume data variables included the following:

Crash data were provided for 2010–2014, including the following:

Total, motorcycle, multi-vehicle motorcycle, and single-vehicle motorcycle crashes were queried. All roadway, traffic, and crash data were linked together by a route name variable unique to each segment and milepost. Segments were defined so that all variables were homogeneous for the entire length.

Segments were grouped into the following four categories for validation:

Table 15 provides variable definitions in the database. Table 16 shows the total length of roadway and total number of crashes by crash type for the four site types in Virginia. Due to the low numbers of motorcycle crashes for rural and urban freeways (18 and 90, respectively), the data for freeways were not used for validation of the avenue A models. Table 17 and table 18

provide summary statistics for other variables included in the Virginia dataset for rural arterials (type 3) and urban arterials (type 4).

Table 15. Virginia data variable definitions.
Variable Definition
AVGMOTO Motorcycle AADT
AVGAADT Total vehicle AADT
NOLANES Number of lanes
SURFWIDTH Surface width in ft
PAVEMENTWIDTH Pavement width in ft
OUTSHLDWID Outside shoulder width in ft
MINMEDWIDTH Minimum median width in ft
MAXMEDWIDTH Maximum median width in ft
INSHLDWID Inside shoulder width in ft
LENGTH Total segment length in mi
TOT Total crash frequency
MOTO Motorcycle crash frequency
MOTOSINGLE Single-vehicle motorcycle crash frequency
MOTOMULTI Multi-vehicle motorcycle crash frequency
Table 16. Total length and crash frequency for Virginia data.
Type LENGTH (mi) TOT MOTO MOTOSINGLE MOTOMULTI
1 22.30 246 18 16 2
2 278.25 4,700 90 50 40
3 4,441.53 33,325 799 410 388
4 3,323.39 125,313 2,705 836 1,865

1 mi = 1.6 km.

Table 17. Summary statistics for Virginia data.
Type Statistic AVGMOTO (mi) AVGAADT (mi) NOLANES SURFWIDTH (ft) PAVEMENTWIDTH (ft) OUTSHLDWID (ft)
3 No. Segments 12,574 12,574 12,574 12,574 12,574 12,574
3 MIN 1.00 305.00 1.00 12.00 0.00 0.00
3 MAX 245.00 58,715.40 7.00 86.00 90.00 33.00
3 MEAN 34.72 8,146.82 2.80 32.65 32.38 5.29
3 STD 30.69 6,816.04 0.99 12.86 15.40 2.94
4 No. Segments 25,390 25,390 25,390 25,390 25,390 25,390
4 MIN 1.00 131.00 1.00 0.00 0.00 0.00
4 MAX 509.00 130,076.50 9.00 108.00 138.00 33.50
4 MEAN 54.27 17,565.15 3.28 40.14 37.13 2.19
4 STD 55.20 14,331.02 1.28 15.54 24.13 3.12
Table 18. Summary statistics for Virginia data continued.
Type Statistic MINMEDWIDTH (ft) MAXMEDWIDTH (ft) INSHLDWID (ft) LENGTH (ft) TOT MOTO MOTOSINGLE MOTOMULTI
3 No. Segments 12,574 12,574 12,574 12,574 12,574 12,574 12,574 12,574
3 MIN 0.00 0.00 0.00 0.01 0.00 0.00 0.00 0.00
3 MAX 120.00 460.00 12.00 8.59 143.00 19.00 14.00 5.00
3 MEAN 11.66 20.66 1.16 0.35 2.65 0.06 0.03 0.03
3 STD 19.76 45.31 2.02 0.47 4.72 0.35 0.25 0.19
4 Segments 25,390 25,390 25,390 25,390 25,390 25,390 25,390 25,390
4 MIN 0.00 0.00 0.00 0.01 0.00 0.00 0.00 0.00
4 MAX 100.00 460.00 16.50 4.40 155.00 7.00 3.00 6.00
4 MEAN 4.79 9.34 0.41 0.13 4.94 0.11 0.03 0.07
4 STD 10.11 20.58 1.31 0.17 8.53 0.37 0.19 0.30

Avenue B Databases

The databases used for the avenue B analyses are a combination of actual and simulated data. The roadway inventory used total AADT and motorcycle AADT collected for the avenue A methods in Florida and Pennsylvania. This ensured that realistic combinations of roadway geometry and traffic volumes were represented. For motorcycle crashes, the crash counts were simulated using the SPFs developed in the A1 models as a starting point. Further details on this approach follow for both the B1 and B2 models.

Model B1 Approach

The model B1 approach was to develop CMFs using the EB before-after method. Using the simulated data, a countermeasure was assumed with a known value of its CMF. The CMF was applied to the simulated after period crashes for a selection of sites. Then, the CMF was estimated twice, once using the motorcycle AADTs and once for total AADT. A comparison was then made to see how the lack of motorcycle AADT affected the estimate of the CMF and its variance. The following steps discuss this process in greater detail:

  1. Estimate the existing roadway geometry and traffic volume file for a given class of road (e.g., rural freeway). The appropriate SPF from the A1 modeling, including the overdispersion parameter k, is used to estimate the mean motorcycle crash frequency for each site, mu. This represents the expected mean value of all sites with the same road geometry and traffic volumes. The SPFs used are documented in chapter 5.

  2. Estimate the site-specific mean crash frequency mi by simulating a multiplier, r, to be applied to mu. The multiplier r is generated from a gamma distribution with the shape and rate parameters equal to the overdispersion parameter, k. This is done using the equation: mi = r × mu.

  3. Simulate the observed crash count at a site. The count in year j is assumed to follow a Poisson distribution about its mean, mi. In this step, the number of motorcycle crashes for each site i in year j , Xij, are simulated using the Poisson distribution and the site mean, mi. Counts for 6 years are simulated.

  4. Select a subset of the sites as the treatment group, with the remainder to be used as a reference group. For the treatment sites, a CMF is assumed for a fictitious treatment with a known value. The first 3 years are used as a before period and the second 3 years as an after period. The crash counts for the after period are simulated by first applying the CMF value to the site-specific mean, CMF x mi. Then, simulate the after period crashes at the treated sites using this new expected mean as was done for all sites in step 4. For the treated sites, these new crash counts in the after period replace those simulated in step 4.

  5. Use the reference group sites to estimate two SPFs to predict motorcycle crashes. The first uses motorcycle AADT as an exposure variable, and the second uses total AADT. The geometric and volume data adopted for the B1 analysis came from Florida. The A1 models developed for Florida did not always support both motorcycle and non-motorcycle volumes in the same model, and where they were both included, the model performance was roughly the same as the model including only motorcycle AADT. For this reason, the B1 analysis did not apply models including both motorcycle and non-motorcycle AADT.

  6. Use the SPFs from step 6 with the treatment site data to estimate the expected number of crashes in the after period and estimate the CMF and its variance using the EB before-after study methodology.

  7. Compare the estimated CMFs and standard errors (SEs) of these estimates from step 7 to each other as well as to the assumed CMF value to determine whether using total AADT SPFs is less accurate than using SPFs with motorcycle AADTs and, if so, the magnitude of this bias.

  8. Repeat steps 2 through 8 multiple times so that conclusions can be made with confidence and have broad applicability.

Model B2 Approach

The model B2 approach was to develop CMFs using cross-sectional regression modeling. In this approach, an assumed relationship was defined between one or more geometric variables and added to the SPFs developed in model A1. The relationship was defined in terms of a CMF. This modified SPF was then used to simulate the motorcycle crash counts. Then, a GLM was used to re-estimate the SPF using motorcycle AADT and then using total AADT only. A comparison was then made to see how the lack of motorcycle AADT affected the estimate of the CMF and its variance. The following steps discuss this process in greater detail:

  1. Adopt the existing roadway geometry and traffic volume file for a given class of road (e.g., rural freeway). Re-estimate the appropriate SPF from the A1 modeling, including the overdispersion parameter k, after first assuming a CMF and including this CMF in the SPF as an offset in the model formulation. This ensures that the predicted motorcycle frequencies are realistic, given the assumed CMF value.

  2. Use the new SPF from step 1 to estimate the mean motorcycle crash frequency for each site, mu. This represents the expected mean value of all sites with the same road geometry and traffic volumes.

  3. Estimate the site-specific mean crash frequency mi by simulating a multiplier, r, and applied to mu. The multiplier r is generated from a gamma distribution with the shape and rate parameters equal to the overdispersion parameter, k. This is done using the equation: mi = r × mu.

  4. Simulate the observed crash count at a site. The count is assumed to follow a Poisson distribution about its mean, mi. In this step, use the Poisson distribution and site mean mi to simulate the number of motorcycle crashes for each site for a 5-year period.

  5. Use the data with simulated motorcycle crash counts to develop SPFs with the same model form as from step 1, first using motorcycle AADT as an exposure variable and then using total AADT.

  6. Compare the CMFs and SEs of these estimates derived from the parameter estimates of the SPFs developed in step 6 to each other as well as to the assumed CMF value to determine whether using total AADT SPFs is less accurate than using SPFs with motorcycle AADTs and, if so, the magnitude of this bias. Repeat steps 3 through 7 so that conclusions can be made with confidence and have broad applicability.

 

 

Federal Highway Administration | 1200 New Jersey Avenue, SE | Washington, DC 20590 | 202-366-4000
Turner-Fairbank Highway Research Center | 6300 Georgetown Pike | McLean, VA | 22101