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


Skip to content
Facebook iconYouTube iconTwitter iconFlickr iconLinkedInInstagram

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-14-065    Date:  February 2015
Publication Number: FHWA-HRT-14-065
Date: February 2015

 

Evaluation of Pavement Safety Performance

CHAPTER 6. BEFORE-AFTER ANALYSIS

AGGREGATE RESULTS

Table 24 through table 31 provide the estimated CMFs and standard errors for the various treatments, broken down by crash type, State, and road class. A general discussion follows the presentation of all of the aggregate results.

Chip Seal Results

The results are shown in table 24. For multilane roads, there are significant benefits overall for wet-road crashes, due largely to reductions in California. There was an estimated increase in dry-road crashes on these roads, which contributed to a significant (5-percent level) increase in total crashes.

Table 24 . Estimates of CMFs for chip seal treatment.


Group

Mi

Crashes After

Estimated CMF (standard error)

Total Crashes

Injury Crashes

ROR Crashes

Wet-Road Crashes

Dry-Road Crashes

Wet-Road ROR Crashes

California

948

3,272

0.908 (0.020)

0.892
(0.028)

0.870
(0.032)

0.830
(0.053)

0.918
(0.022)

0.709
(0.074)

Minnesota

274

179

1.255 (0.103)

1.005
(0.134)

1.271
(0.173)

1.604
(0.312)

1.201
(0.108)

0.862
(0.355)

North Carolina

765

2,149

1.011 (0.029)

1.011
(0.039)

0.655
(0.066)

0.937
(0.055)

1.027
(0.033)

0.682
(0.141)

Pennsylvania

570

1,271

0.949
(0.031)

0.959
(0.041)

1.053
(0.069)

0.999
(0.062)

1.256
(0.044)

1.004
(0.125)

All Freeway

15

94

0.832
(0.102)

0.570
(0.119)

0.638
(0.202)

Too few crashes

0.948
(0.122)

Too few crashes

All Multilane

95

619

1.147
(0.059)

1.105
(0.085)

0.959
(0.094)

0.775
(0.116)

1.206
(0.066)

0.373
(0.157)

Multilane California

70

425

1.046
(0.065)

1.039
(0.093)

0.935
(0.098)

0.423
(0.096)

1.141
(0.075)

0.222
(0.130)

Multilane Minnesota

23

94

1.519
(0.178)

1.067
(0.221)

1.214
(0.342)

Too few crashes

1.412
(0.186)

0.997
(0.705)

Multilane North Carolina

1

100

1.385
(0.172)

1.656
(0.327)

1.004
(0.708)

Too few crashes

1.390
(0.188)

Too few crashes

All Two-Lane

2448

6,158

0.939
(0.015)

0.934
(0.020)

0.883
(0.028)

0.950
(0.035)

0.937
(0.017)

0.829
(0.062)

Two-Lane California

863

2,753

0.892
(0.022)

0.884
(0.030)

0.865
(0.034)

0.927
(0.063)

0.888
(0.023)

0.775
(0.083)

Two-Lane Minnesota

251

85

1.050
(0.121)

0.960
(0.166)

1.285
(0.199)

1.092
(0.349)

1.045
(0.129)

Too few crashes

Two-Lane North Carolina

764

2,049

0.997
(0.029)

0.995
(0.040)

0.650
(0.066)

0.650
(0.066)

1.014
(0.034)

0.666
(0.141)

Two-Lane Pennsylvania

570

1,271

0.949
(0.031)

0.959
(0.041)

1.053
(0.069)

0.999
(0.062)

0.933
(0.036)

1.004
(0.124)

CMF = Crash modification factor
ROR = Run-off road

For chip seal on two-lane roads, there was a small benefit overall (significant at the 10-percent level) for wet-road crashes due mainly to reductions in California and North Carolina. For dry-road crashes, there was a small benefit overall (significant at the 5-percent level) due mainly to reductions in California and Pennsylvania. These benefits contribute to an overall benefit for all crashes and States combined for chip seal on two-lane roads.

There were too few crashes on freeways with this treatment to obtain a definitive result, although there are indications of an overall benefit for total crashes.

Diamond Grinding Results

For diamond grinding, the results in table 25 indicate that there was an overall benefit (significant at the 5-percent level) for both wet- and dry-road crashes, which resulted in a significant overall benefit for total crashes.

Table 25 . Estimates of CMFs for diamond grinding treatment.

Group

Mi

Crashes After

Estimated CMF (standard error)

Total Crashes

Injury Crashes

ROR Crashes

Wet-Road Crashes

Dry-Road Crashes

Wet-Road ROR Crashes

California

85

12,267

0.950
(0.012)

0.973
(0.020)

0.606
(0.043)

0.866
(0.037)

0.957
(0.012)

0.703
(0.113)

Minnesota

8

119

0.899
(0.099)

1.127
(0.204)

1.221
(0.256)

Few crashes

0.792
(0.098)

Few crashes

North Carolina

24

139

0.641
(0.057)

0.525
(0.091)

Few crashes

0.576
(0.058)

Pennsylvania

33

105

0.720
(0.081)

0.769
(0.115)

0.106
(0.106)

0.480
(0.136)

0.898
(0.104)

All Freeway

141

12,518

0.943
(0.011)

0.967
(0.020)

0.642
(0.043)

0.869
(0.036)

0.950
(0.012)

0.869
(0.120)

Freeway California

76

12,155

0.951
(0.012)

0.975
(0.020)

0.595
(0.044)

0.862
(0.037)

0.959
(0.012)

0.700
(0.115)

All Multilane

8

108

Insufficient sites

All Two-Lane

1

4

Insufficient sites

CMF = Crash modification factor
ROR = Run-off road

Thin HMA Results

For thin HMA, the results in table 26 indicate that there were benefits (significant at the 5-percent level) for wet-road crashes for multilane roads and freeways, and no effect overall for dry-road crashes. (For the latter crash type, there was an increase in California and a decrease in North Carolina, both results significant at the 5-percent level.)

Table 26 . Estimates of CMFs for thin HMA treatment.

Group

Mi

Crashes After

Estimated CMF (standard error)

Total Crashes

Injury Crashes

ROR Crashes

Wet-Road Crashes

Dry-Road Crashes

Wet-Road ROR Crashes

California

584

20,275

1.091
(0.010)

1.087
(0.017)

0.972
(0.034)

0.938
(0.032)

1.104
(0.011)

0.772
(0.075)

Minnesota

204

43

0.907
(0.148)

0.963
(0.220)

1.103
(0.243)

0.531
(0.310)

0.957
(0.163)

0.750
(0.533)

North Carolina

3,154

39,579

1.073
(0.009)

1.125
(0.014)

1.278
(0.033)

1.069
(0.018)

1.074
(0.010)

0.999
(0.047)

Pennsylvania

7

29

1.102
(0.252)

0.906
(0.294)

1.471
(0.931)

0.674
(0.367)

1.401
(0.335)

No crashes

All Freeway

259

18,323

1.021
(0.011)

0.986
(0.018)

0.973
(0.042)

0.910
(0.028)

1.039
(0.012)

0.797
(0.065)

Freeway California

164

13,326

1.043
(0.012)

1.019
(0.021)

0.666
(0.040)

0.903
(0.038)

1.054
(0.013)

0.551
(0.091)

Freeway North Carolina

87

5,068

0.967
(0.023)

0.908
(0.037)

1.405
(0.097)

0.914
(0.039)

0.990
(0.029)

0.871
(0.083)

Freeway Pennsylvania

7

29

Insufficient crashes

All Multilane

279

15,776

0.988
(0.013)

1.021
(0.021)

1.420
(0.066)

0.865
(0.028)

1.010
(0.015)

1.149
(0.108)

Multilane California

72

4,241

1.188
(0.027)

1.191
(0.040)

1.051
(0.098)

0.955
(0.075)

1.209
(0.028)

0.680
(0.195)

Multilane Minnesota

6

7

Very few sites and crashes

Multilane North Carolina

201

11,528

0.930
(0.015)

0.956
(0.025)

1.566
(0.086)

0.853
(0.031)

0.946
(0.017)

1.222
(0.122)

All Two-Lane

3,411

25,827

1.194
(0.011)

1.247
(0.016)

1.180
(0.031)

1.256
(0.023)

1.181
(0.013)

1.007
(0.054)

Two-Lane California

348

2,808

1.203
(0.031)

1.167
(0.043)

1.262
(0.062)

1.018
(0.083)

1.223
(0.033)

0.993
(0.137)

Two-Lane Minnesota

198

36

0.930
(0.165)

0.881
(0.222)

1.042
(0.244)

Too few crashes

Too few crashes

Too few crashes

Two-Lane North Carolina

2,866

22,983

1.193
(0.012)

1.258
(0.017)

1.146
(0.036)

1.273
(0.024)

1.175
(0.014)

1.013
(0.058)

Two-Lane Pennsylvania

0

0

No sites

CMF = Crash modification factor
ROR = Run-off road

For two-lane roads, the thin HMA treatment was associated with highly significant increases overall in both wet- and dry-road crashes, a pattern that was consistent between California and North Carolina, the two States with large enough samples for a definitive result.

OGFC Results

For OGFC, the results in table 27 indicate a negligible effect on wet-road crashes for multilane and two-lane roads, but increases in dry-road crashes resulted in significant increases (5-percent level) in total crashes for these road types. By contrast, for freeways, there was a small but significant (5-percent level) decrease in total crashes, due in large part to highly significant and substantial reduction in wet-road crashes with no change in dry-road crashes for California and North Carolina combined.

Table 27 . Estimates of CMFs for open OGFC treatment.

Group

Mi

Crashes After

Estimated CMF (standard error)

Total Crashes

Injury Crashes

ROR Crashes

Wet-Road Crashes

Dry-Road Crashes

Wet-Road ROR Crashes

California

416

9,525

1.060
(0.014)

1.032
(0.014)

0.974
(0.036)

0.997
(0.039)

1.068
(0.015)

0.807
(0.080)

North Carolina

42

2,231

0.748
(0.028)

0.743
(0.049)

0.485
(0.083)

0.506
(0.036)

0.875
(0.038)

0.306
(0.077)

All Freeway

165

8,571

0.945
(0.015)

0.934
(0.025)

0.816
(0.041)

0.685
(0.031)

1.008
(0.017)

0.482
(0.066)

Freeway California

124

6,354

1.041
(0.017)

1.004
(0.027)

0.873
(0.046)

0.920
(0.046)

1.055
(0.018)

0.643
(0.099)

Freeway North Carolina

41

2,217

0.747
(0.028)

0.746)
(0.049)

0.481
(0.082)

0.508
(0.036)

0.873
(0.038)

0.307
(0.076)

All Multilane
(almost all California)

61

1,734

1.092
(0.036)

0.959
(0.051)

1.028
(0.100)

0.981
(0.086)

1.108
(0.039)

1.114
(0.246)

All Two-Lane
(California only)

232

1,451

1.109
(0.037)

1.128
(0.053)

1.107
(0.067)

1.038
(0.089)

1.120
(0.162)

0.878
(0.141)

CMF = Crash modification factor
ROR = Run-off road

Grooving Results

For grooving, there were two few sites to obtain a definitive result as indicated in table 28.

Table 28 . Estimates of CMFs for grooving treatment.

Group

Mi

Crashes After

Estimated CMF (standard error)

Total Crashes

Injury Crashes

ROR Crashes

Wet-Road Crashes

Dry-Road Crashes

Wet-Road ROR Crashes

California
(All Freeway)

5

119

0.776
(0.087)

0.746
(0.148)

0.674
(0.186)

2.034
(0.466)
(Few crashes)

0.615
(0.079)

1.311
(0.696)
(Few crashes)

CMF = Crash modification factor
ROR = Run-off road

Microsurfacing Results

The results are shown on table 29. For two-lane roads, there was a decrease in wet-road crashes and an increase in dry-road crashes overall (both results significant at the 5-percent level) resulting in a net increase in total crashes that was also significant at the 5-percent level. This trend was mainly due to results from Pennsylvania, which had the largest sample. For North Carolina, the sample was small but there are weak indications of decreases on both wet- and dry-road crashes. For California, by contrast, the indication is that there was an increase in both wet- and dry-road crashes for microsurfacing on two-lane roads.

Table 29 . Estimates of CMFs for microsurfacing treatment.

Group

Mi

Crashes After

Estimated CMF (standard error)

Total Crashes

Injury Crashes

ROR Crashes

Wet-Road Crashes

Dry-Road Crashes

Wet-Road ROR Crashes

California

72

766

1.078
(0.049)

1.120
(0.067)

1.016
(0.094)

1.061
(0.153)

1.079
(0.052)

0.712
(0.217)

Minnesota

94

626

1.108
(0.065)

1.026
(0.092)

1.226
(0.136)

0.944
(0.127)

1.140
(0.074)

1.105
(0.247)

North Carolina

39

89

0.765
(0.090)

0.958
(0.158)

0.440
(0.186)

0.604
(0.160)

0.810
(0.106)

0.505
(0.366)

Pennsylvania

164

865

1.067
(0.045)

1.123
(0.062)

1.077
(0.117)

0.775
(0.070)

1.419
(0.065)

1.173
(0.219)

All Freeway

40

518

1.075
(0.071)

1.036
(0.103)

1.169
(0.152)

0.963
(0.128)

1.101
(0.084)

1.178
(0.278)

All Multilane

58

580

1.006
(0.052)

0.972
(0.071)

0.925
(0.125)

0.785
(0.116)

1.039
(0.058)

Few crashes(12)

All Two-Lane

273

1,263

1.090
(0.038)

1.180
(0.053)

1.114
(0.082)

0.867
(0.071)

1.142
(0.044)

1.018
(0.171)

Two-Lane California

46

443

1.300
(0.076)

1.419
(0.110)

1.140
(0.122)

1.810
(0.314)

1.255
(0.077)

Few crashes(8)

Two-Lane Minnesota

43

23

Insufficient crashes

Two-Lane North Carolina

32

60

0.718
(0.102)

0.838
(0.161)

Few crashes

0.516
(0.177)

0.769
(0.120)

Few crashes

Two-Lane Pennsylvania

152

737

1.040
(0.047)

1.099
(0.064)

1.088
(0.121)

0.761
(0.075)

1.129
(0.059)

1.122
(0.217)

CMF = Crash modification factor
ROR = Run-off road

For freeways, the results for microsurfacing were inclusive (i.e., there were no statistically significant effects), likely a result of the small sample size. For multilane roads, there was a decrease in wet-road crashes (significant at the 5-percent level) and a negligible effect on total and dry-road crashes.

Slurry Seal Results

For slurry seal, which was mostly on two-lane roads, almost all of which were in California, the results in table 30 indicate that there were benefits for wet-road crashes and weak (i.e., statistically insignificant) indications of a benefit for dry-road crashes.

Table 30 . Estimates of CMFs for slurry seal treatment.

Group

Mi

Crashes After

Estimated CMF (standard error)

Total Crashes

Injury Crashes

ROR Crashes

Wet-Road Crashes

Dry-Road Crashes

Wet-Road ROR Crashes

California

134

1,084

0.936
(0.037)

0.888
(0.052)

0.669
(0.059)

0.736
(0.091)

0.959
(0.039)

0.621
(0.143)

North Carolina

5

5

0.843
(0.403)

0.710
(0.520)

Insufficient crashes

All Freeway

19

200

Insufficient crashes

All Multilane

15

192

Insufficient crashes

All Two-Lane (almost all California)

105

697

0.931
(0.044)

0.972
(0.068)

0.578
(0.067)

0.802
(0.126)

0.943
(0.047)

Few crashes

CMF = Crash modification factor
ROR = Run-off road

UTBWC Results

The results are shown in table 31. For freeways, there was a small and marginally significant benefit overall for wet-weather crashes, due largely to the California treatments, which had a substantial and significant benefit. There was no effect for dry weather and for total crashes when this is considered.

On two-lane roads, there was a substantial and highly significant benefit for wet-road crashes and a smaller, but significant (10-percent level), benefit for dry-road crashes.

Table 31 . Estimates of CMFs for UTBWC treatment.


Group

Mi

Crashes After

Estimated CMF (standard error)

Total Crashes

Injury Crashes

ROR Crashes

Wet-Road Crashes

Dry-Road Crashes

Wet-Road ROR Crashes

California

57

1,937

0.961
(0.027)

0.982
(0.046)

1.075
(0.098)

0.925
(0.083)

0.964
(0.029)

0.802
(0.208)

North Carolina

94

3,940

0.954
(0.019)

0.860
(0.032)

1.260
(0.093)

0.978
(0.043)

0.948
(0.021)

0.926
(0.109)

Pennsylvania

21

104

0.641
(0.073)

0.632
(0.100)

0.502
(0.198)

0.330
(0.082)

0.962
(0.118)

0.634
(0.381)

All Freeway

109

4,365

0.994
(0.019)

0.875
(0.031)

1.139
(0.070)

0.947
(0.041)

1.005
(0.021)

0.917
(0.102)

Freeway California

30

850

1.017
(0.044)

1.061
(0.078)

1.170
(0.129)

0.761
(0.102)

1.049
(0.048)

0.896
(0.274)

Freeway North Carolina

69

3,484

0.994
(0.021)

0.871
(0.036)

1.317
(0.100)

0.985
(0.046)

0.996
(0.024)

0.945
(0.113)

Freeway Pennsylvania

10

31

Insufficient crashes

All Multilane

21

103

Insufficient crashes

All Two-Lane

43

440

0.872
(0.051)

0.956
(0.081)

0.908
(0.169)

0.694
(0.103)

0.905
(0.058)

0.550
(0.254)

CMF = Crash modification factor
ROR = Run-off road

Summary of Aggregate Results

In summary, the combined results for all treatment types (except grooving, for which there were very few sites) suggest that the treatments resulted in benefits for wet-road crashes, with a few exceptions. The exceptions were for thin HMA on two-lane roads for both California and North Carolina, the two States with large enough samples for a definitive result, and for OGFC for two-lane and multilane roads, for which the effect was negligible.

For dry-road crashes, crashes increased for microsurfacing (except for North Carolina), thin HMA and OGFC on two-lane roads, and OGFC and chip seal on multilane roads. There were indications of a benefit for UTBWC, chip seal, and slurry seal on two-lane roads, and diamond grinding on freeways.

The estimated CMFs for treatments by road and crash type may be considered for use in the Highway Safety Manual and the CMF clearinghouse

Disaggregate Results

Effect of Age of Treatment

For some of the pavement treatments, it was of interest to investigate the possible change in safety effects as the pavement ages. Traffic and weather play a significant role in wearing pavements down over time, generally leading to a reduction in pavement texture and reduction in friction.(28) The cause can be a complex interaction of factors but intuitively, we understand that aggregates abrade, polish, and are broken off of the pavement surface, bituminous binders can bleed to the surface of a pavement over time, ruts can form, and porous surfaces can become clogged. Although there have not been many studies to confirm this link between treatment age and safety, the project team wanted to evaluate whether there is any correlation in the data analyzed.

For the following identified treatments, the effect of age was investigated where the sample size allows for wet-road crashes:

Table 32 presents the CMF estimates for all years of data and for years 1 to 3 for chip seal on two-lane roads. The results indicate that the positive safety effect of chip seal treatment on wet-weather crashes is greatest in the first year following treatment, with a declining benefit thereafter. This result is not entirely surprising for chip seal treatments. The two common "failure" mechanisms of chip seals are chip loss (raveling) and bleeding, both of which result in reduced surface texture and reduced friction, particularly in the wheelpaths where traffic has the most impact on the performance of the treatment. Figure 26 shows an example of an approximately 5-year-old chip seal on a heavily traveled roadway, with the loss of texture and friction apparent in the wheelpaths. Although it is not possible to say with certainty that this is the explanation of the results observed from this study (because each treatment site was not specifically investigated), the trend is consistent with observed performance of chip seals over time.

Table 32 . Estimates of CMFs for chip seal treatment for wet-road crashes on two-lane roads by period after treatment.

Group

Estimated CMF (standard error) by period after treatment

All Years

Year 1

Year 2

Year 3

All Two-Lane

0.950
(0.035)

0.830
(0.055)

0.872
(0.060)

0.952
(0.067)

CMF = Crash modification function

The photo was taken looking down the length  of a road. The left and right wheelpaths in both lanes are dark, lined by black.
Source: The Transtec Group, Inc.
Figure 26. Photo. Example of wear in wheelpaths over time for chip seal treatments, reducing surface texture and friction.

Table 33 provides the results for chip seal on all road types disaggregated by single versus double/triple seal applications. Data on single/double/triple seal were only available for North Carolina and Pennsylvania. For single applications, there is some indication that the safety benefit is greater in the first year after treatment than in later years; however, there is no such trend for double/triple seals.

Table 33 . Estimates of CMFs for single and multi-layer chip seal treatment for wet-road crashes (NC and PA only) by period after treatment.


Chip Seal Type

Estimated CMF (standard error) by period after treatment

All Years

Year 1

Year 2

Year 3

Single

1.015
(0.063)

0.845
(0.098)

1.115
(0.119)

1.029
(0.113)

Double/Triple

0.924
(0.055)

0.882
(0.098)

0.890
(0.102)

0.680
(0.097)

CMF = Crash modification function

Table 34 provides the results for diamond grinding on freeways. There is no clear time trend to be seen for the first 4 years.

Table 34 . Estimates of CMFs for diamond grinding treatment for wet-road crashes on freeways by period after treatment.

Group

Estimated CMF (standard error) by period after treatment

All Years

Year 1

Year 2

Year 3

Year 4

Freeways

0.869
(0.036)

0.916
(0.054)

0.779
(0.058)

0.923
(0.074)

0.940
(0.077)

CMF = Crash modification function

Table 35 provides the results for OGFC on freeways and two-lane roads. For freeways, there appears to be a trend of a decreasing CMF (increasing benefit) as the pavement age increases for the first 4 years. For two-lane roads, however, the trend is the opposite, and the benefits are seen to decline as the pavement ages.

Table 35 . Estimates of CMFs for OGFC treatment for wet-road crashes on freeways and two-lane roads by period after treatment.

Group

Estimated CMF (standard error) by period after treatment

All Years

Year 1

Year 2

Year 3

Year 4

Freeway

0.685
(0.031)

0.846
(0.050)

0.810
(0.051)

0.618
(0.051)

0.573
(0.060)

Two-Lane

1.038
(0.089)

0.975
(0.130)

1.148
(0.150)

1.237
(0.188)

few crashes

CMF = Crash modification function

Effect of Other Factors

A thorough disaggregate analysis was undertaken in which multiple variable regression modeling was used to investigate the effects on the CMF of a number of factors, including AADT, precipitation, expected crash frequency before treatment, and environment (urban/rural). The primary objective was to investigate whether CMFunctions could be developed to capture the effects of these factors and more precisely estimate CMFs for prospective treatments.

In the end, the CMFunctions developed were not robust enough to recommend them. The direction of effect for attempted variables was not always consistent, and the statistical significance of estimated parameters tended to be poor. Nevertheless, there were useful insights that suggest that it would be worthwhile to pursue the development of robust CMFunctions in future research. These insights suggest that there appears to be a relationship between CMFs and AADT and sometimes precipitation, urban versus rural setting, and crash frequency. However, the direction of the effect varies by crash type and treatment, so the future research will need to reconcile (i.e., explain), these apparent inconsistencies.

Appendix A summarizes the approach to CMFunction development and presents some of the more promising results.

 

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