Skip to contentUnited States Department of Transportation - Federal Highway Administration FHWA Home
Research Home
Report
This report is an archived publication and may contain dated technical, contact, and link information
Publication Number: FHWA-HRT-11-024
Date: April 2011

Safety Evaluation of the Safety Edge Treatment

Chapter 4. Analysis Results for Safety Evaluation

This chapter presents the safety evaluation approach, the development of SPFs, and the safety evaluation results. The safety evaluation results include the findings of a before-period compatibility study, a before-after evaluation using the EB technique, a cross-sectional analysis, and an analysis of shifts in crash severity.

4.1 Evaluation Approach

Two statistical approaches were used to evaluate the safety effectiveness of the safety edge treatment: (1) a before-after comparison of the effect of pavement resurfacing with and without the safety edge treatment using the EB technique and (2) a cross-sectional comparison of the effect of pavement resurfacing with and without the safety edge treatment based on after-period data only. These two evaluation approaches were applied concurrently to provide alternative statistical approaches to the key issues being addressed. The following discussion describes these evaluations, including issues related to the specific nature of the safety edge treatment.

A key objective of the evaluation was to determine the safety effectiveness of the safety edge treatment while avoiding the potential confounding effects of regression to the mean and the safety effect of pavement resurfacing. Regression to the mean is a characteristic of repeated measures data in which observations move toward the mean value over time. That is, if an observation in a year is unusually high, then the observation in the following year will nearly always be lower (and vice versa), returning to the mean. This phenomenon often leads to an overestimation or underestimation of safety for some sites. Thus, the effect of the treatment is likely to be partially confounded with the expected decrease or increase in crash experience from regression to the mean. Regression to the mean can only be accounted for with knowledge of the "normal" or expected value of before-period crash experience at the treated sites. The EB technique has the advantage of compensating for regression to the mean. The cross-sectional approach does not explicitly compensate for regression to the mean. This concern is lessened by the availability of 3 years of crash data for the period after resurfacing.

The second potential confounding effect is the safety effect of pavement resurfacing since it is always used in conjunction with the safety edge treatment. Previous research has indicated that pavement resurfacing by itself may have an effect on safety, increasing crashes because of increased speeds. This effect was found in one study to be statistically significant but was found to persist for only 12-30 months after resurfacing.(4) However, a more recent, larger study in National Cooperative Highway Research Program Project 17-9(2) found inconsistent results; increases in crash frequency with resurfacing were found in some States, but decreases in crash frequency with resurfacing were found in others.(5) Therefore, the safety effects of the pavement resurfacing and installation of the safety edge treatment will be confounded, at least for some time, following resurfacing.

The study design was developed to address the safety effect of resurfacing and the safety edge treatment as well as the confounding effect of resurfacing. First, the study period after resurfacing was selected to be 3 years. This is sufficiently long as to extend beyond the duration of any short-term resurfacing effect. Annual interim evaluations to monitor time trends were conducted to address this issue. Thus, the results for safety effectiveness of the safety edge treatment in the first- and second-year interim reports may be confounded by the safety effect of pavement resurfacing, but it is expected that this confounding effect is lessened in the final results. Second, resurfaced sites both with and without the safety edge treatment were considered. The ratio of safety between resurfaced sites with and without the safety edge treatment (i.e., the treatment and comparison sites) may represent an effect of the safety edge treatment as long as the sites can be assumed comparable in other respects.

The first evaluation approach is an observational before-after comparison using the EB technique, as formulated by Hauer.(6,7) The specific version of the EB technique used in this evaluation was developed for the FHWA SafetyAnalyst software tools.(8) The primary objective of the before-after evaluation is to compare the observed number of crashes after the treatment is implemented to the expectednumber of crashes in the after period had the countermeasure not been implemented. This provides an estimate of the overall safety effectiveness of the countermeasure expressed as a percent change in the crash frequency.

When performing before-after evaluations using the EB technique, it is typical to collect data at sites where countermeasures were implemented (i.e., treatment sites) and at sites similar to the treatment sites with respect to area type (rural/urban), geometric design, and traffic volumes, but where no countermeasures were installed. Data from this comparison group of sites are used to create SPFs, which are then used with the observed crash counts at the treated sites in the before period to estimate the number of crashes that would have occurred at the treated sites in the after period if no improvement had been made. These SPFs are discussed in section 4.2.

The comparability before resurfacing of the two types of sites (treatment and comparison sites) is critical to interpreting the difference of the two estimated before-after effects as an effect of the safety edge treatment. For example, if one of the site types had a higher mean in the before period and both site types had the same mean in the after period, then the effectiveness of one treatment may be presumed greater than the other treatment. The comparability of sites was established through analysis of the before-period crash data. These analyses are discussed in section 4.3.1.

The EB before-after evaluation produced separate estimates of the effectiveness of resurfacing with the safety edge (treatment sites) and resurfacing only (comparison sites) for each target crash type in each State. From each pair of estimated percent changes in safety (treatment and comparison), the effect of the safety edge alone was estimated as the ratio between the two measures of effectiveness. For every combination of site characteristics under consideration, the mean and standard error of the percent change in target crash frequency and its statistical significance are presented in section 4.3.2.

It was anticipated that the effectiveness measure for the safety edge treatment would be relatively small since it was expected that the safety edge treatment would affect only certain crash types and would have the greatest impact on two-lane highways with no paved shoulders. Most such sites have relatively low traffic volume and therefore are not expected to have a high frequency of run-off-the-road and drop-off-related crashes.

The EB-based before-after comparison technique is theoretically the strongest approach to evaluations of this type. However, because of the confounding of the pavement resurfacing effect and the safety edge treatment effect, it cannot be assured that this approach correctly identifies the treatment effectiveness. Therefore, an alternative cross-sectional comparison was also conducted.

A cross-sectional evaluation of the after data at the treated sites was conducted to directly compare the crash data between the two types of treatment-resurfacing with the safety edge treatment and resurfacing without the safety edge treatment. Assuming that all roadway factors except resurfacing are held constant, one could hypothesize that the differences in either after-period crash frequencies or crash severity distributions between treatment and comparison sites are due to the provision of the safety edge treatment. This comparison was made with a cross-sectional approach using data for the period after resurfacing while accounting for the effects of AADT.

The cross-sectional comparison of crash data for the period after resurfacing was conducted using negative binomial regression models to compare the crash frequencies for the period after resurfacing for the sites with the safety edge treatment to those of the sites resurfaced without the safety edge treatment. Site type (i.e., treatment versus comparison) was the main factor of interest in the analysis. The effect of AADT was accounted for in this approach by quantifying the relationship between AADT and specific target crash types. When significant, the effect of lane width was also accounted for in the model. The safety edge treatment effect and its standard error were then calculated for each target crash type. The treatment effect was converted to a percent change in crash frequency for ease in interpreting the results. The results of the cross-sectional analysis are presented in section 4.4.3.

In addition to evaluating mean crash frequencies, a comparison of the before-after data by crash severity level was performed to determine shifts in the crash severity distribution. These comparisons were accomplished by calculating a confidence interval for the average difference in proportions across all sites at a preselected significance level of 10 percent. However, a non-parametric statistical test, the Wilcoxon signed-rank test, was also applied as the differences in proportions may not follow a normal distribution. Results from this analysis are presented in section 4.4.4.(9)

4.2 Safety Performance Functions

This section documents the SPFs and calibration factors developed for use in the before-after EB evaluation of the safety effectiveness of the safety edge treatment. SPFs are regression relationships between target crash frequencies and traffic volumes that can be used to predict the long-term crash frequency for a site. SPFs are used in the before-after EB evaluation to estimate what the safety performance of a treated site would be in the after period if the treatment had not been implemented.

Negative binomial regression models were developed using data from the reference group of untreated sites for use in three categories of target crashes (all crash types combined, run-off-road crashes, and drop-off-related crashes) and two severity levels (total crashes and fatal and injury crashes). Thus, a total of six dependent variables were considered. Traffic volume and lane width were the only independent variables considered in the SPFs. Separate models were developed for Georgia and Indiana for each of the three classifications, as follows:

  • Rural multilane highways with paved shoulders with widths of 4 ft or less.
  • Rural two-lane highways with paved shoulders with widths of 4 ft or less.
  • Rural two-lane highways with no paved shoulders (i.e., unpaved shoulders only).

Regression models were not developed for New York due to the limited number of treated sites.

All regression models were developed to predict target crash frequencies per mile per year as a function of traffic volume and, in some cases, lane width in the functional forms shown in equation 1 and equation 2.

N equals exponential open parenthesis a plus b natural log AADT close parenthesis. (1)

N equals exponential open parenthesis a plus b natural log AADT plus c times L times W close parenthesis. (2)

Where:

N = predicted number of target crashes per mile per year

AADT = average daily traffic volume (vehicles per day) for the roadway segment

LW = lane width for the roadway segment (ft)

a, b, c = regression coefficients

The AADT in the regression models was statistically significant in all cases. The lane width term was included in the regression model only when it was statistically significant.

Two generalized linear modeling techniques were used to fit the data. The first method used a repeated measures correlation structure to model yearly crash counts for a site. In this method, the covariance structure, assuming compound symmetry, is estimated before final regression parameter estimates are determined by general estimating equations. Consequently, model convergence for this method is dependent on the covariance estimates as well as parameter estimates. When the model failed to converge for the covariance estimates, an alternative method was considered. In this method, yearly crash counts for a site were totaled and annual daily traffic (ADT) values were averaged to create one summary record for a site. Regression parameter estimates were then directly estimated by maximum likelihood without an additional covariance structure being estimated.

Both methods produced an estimate of the overdispersion parameter, the estimate for which the variance exceeds the mean. Overdispersion occurs in traffic data when a number of sites being modeled have zero accident counts, which creates variation in the data. When the estimate for dispersion was very small or even slightly negative, the model was refit assuming a constant value. Both methods were accomplished with the GENMOD procedure of SAS®.(3)

Statistically significant models were not found for all dependent variables for some road type/ shoulder type combinations. In these three cases, the intercept coefficient of the total crashes or fatal and injury crashes model was adjusted by the proportion of the applicable dependent variable to produce the final model. The model coefficients with their standard errors are presented in table 10 for Georgia and in table 11 for Indiana. All AADT coefficients shown are significant at the 10 percent significance level or better. Lane width coefficients shown are significant at the 20 percent significance level or better. Total crash and fatal and injury crash SPFs are illustrated in figure 3 for Georgia and in figure 4 for Indiana.

Table 10. SPFs for Georgia sites.

Roadway type

Shoulder type

Number of site-years

Intercept (standard error)

AADT coefficient (standard error)

Lane width coefficient (standard error)

Overdispersion parameter

R2LR (%)

Total crashes

Multilane

Paved

192

-4.801 (1.608)

0.642 (0.172)

 

0.487

9.2

Two-lane

Paved

582

-8.921 (1.189)

1.108 (0.141)

 

0.724

36.4

Two-lane

Unpaved

792

-7.730 (0.783)

0.978 (0.095)

 

0.425

25.1

Fatal and injury crashes

Multilane

Paved

192

-2.204 (1.752)

0.252 (0.184)

 

0.588

0.2

Two-lane

Paved

582

-7.818 (1.116)

0.853 (0.132)

 

0.401

21.3

Two-lane

Unpaved

792

-8.556 (0.796)

0.958 (0.098)

 

0.346

16.0

PDO crashes

Multilane

Paved

192

-6.611 (1.747)

0.787 (0.189)

 

0.540

14.0

Two-lane

Paved

582

-11.414 (1.397)

1.349 (0.164)

 

0.982

34.6

Two-lane

Unpaved

792

-8.470 (0.981)

1.011 (0.119)

 

0.623

19.3

Total run-off-road crashes

Multilane

Paved

192

-3.475 (2.145)

0.360 (0.228)

 

0.213

1.9

Two-lane

Paved

582

-2.625 (1.710)

0.783 (0.134)

-0.376 (0.109)

0.464

19.9

Two-lane

Unpaved

132

-4.405 (1.443)

0.757 (0.141)

-0.199 (0.106)

0.472

14.8

Fatal and injury run-off-road crashes

Multilane

Paved

192

-3.425(1.752)

0.252 (0.184)

 

0.588

0.2

Two-lane

Paved

582

-1.848(1.618)

0.544 (0.128)

-0.339 (0.110)

0.374

8.1

Two-lane

Unpaved

132

-5.556(1.543)

0.743 (0.139)

-0.151 (0.115)

0.341

15.8

PDO run-off-road crashes

Multilane

Paved

192

-7.742(3.004)

0.750 (0.320)

 

0.117

5.6

Two-lane

Paved

582

-5.029(2.236)

1.033 (0.154)

-0.406 (0.144)

0.598

19.2

Two-lane

Unpaved

132

-4.544(1.709)

0.752 (0.173)

-0.238 (0.126)

0.636

9.7

Total drop-off-related crashes

Multilane

Paved

192

-3.583(2.126)

0.318 (0.226)

 

0.131

1.6

Two-lane

Paved

582

-4.586(2.069)

0.884 (0.169)

-0.327 (0.125)

0.585

16.3

Two-lane

Unpaved

132

-4.140(1.495)

0.770 (0.141)

-0.270 (0.114)

0.427

14.0

Fatal and injury drop-off-related crashes

Multilane

Paved

192

-2.344(1.974)

0.113 (0.141)

 

0.294

0.1

Two-lane

Paved

582

-3.297(1.894)

0.604 (0.154)

-0.290 (0.121)

0.558

6.2

Two-lane

Unpaved

132

-4.869(1.654)

0.699 (0.148)

-0.209 (0.127)

0.357

11.9

PDO drop-off-related crashes

Multilane

Paved

192

-6.690(3.194)

0.574 (0.340)

 

0.101

2.7

Two-lane

Paved

582

-8.291(3.272)

1.269 (0.217)

-0.359 (0.195)

0.754

16.3

Two-lane

Unpaved

792

-4.345(3.899)

0.872 (0.157)

-0.388 (0.290)

0.565

6.6

Note: Blank cells indicate lane width coefficient was not significant.

Table 11. SPFs for Indiana sites.

Road type

Shoulder Type

Number of site-years

Intercept (standard error)

AADT coefficient (standard error)

Lane width coefficient (standard error)

Overdispersion parameter

R2LR (%)

Total crashes

Two-lane

Paved

100

-5.500(1.317)

0.737(0.154)

 

0.444

15.3

Two-lane

Unpaved

98

-3.865(1.118)

0.701(0.146)

-0.156(0.086)

0.654

15.5

Fatal and injury crashes

Two-lane

Paved

100

-6.279(1.977)

0.642(0.233)

 

0.563

5.1

Two-lane

Unpaved

196

-2.707(1.305)

0.427(0.139)

-0.198(0.098)

0.211

7.2

PDO crashes

Two-lane

Paved

100

-5.572(1.373)

0.718(0.161)

 

0.398

14.8

Two-lane

Unpaved

98

-4.348(1.153)

0.694(0.148)

-0.128(0.089)

0.661

15.9

Total run-off-road crashes

Two-lane

Paved

100

-3.250(1.962)

0.303(0.231)

 

0.413

1.5

Two-lane

Unpaved

196

-1.700(1.221)

0.490(0.119)

-0.278(0.103)

0.438

10.9

Fatal and injury run-off-road crashes

Two-lane

Paved

296

-3.127(1.034)

0.346(0.105)

-0.132(0.078)

0.154

2.5

Two-lane

Unpaved

196

-1.467(1.432)

0.331(0.129)

-0.284(0.102)

0.027

6.4

PDO run-off-road crashes

Two-lane

Paved

100

-4.764(2.398)

0.426(0.286)

 

0.212

2.5

Two-lane

Unpaved

196

-2.752(1.260)

0.573(0.133)

-0.279(0.112)

0.540

8.6

Total drop-off-related crashes

Two-lane

Paved

100

-4.477(3.598)

0.313(0.421)

 

0.738

0.6

Two-lane

Unpaved

98

-2.352(1.489)

0.356(0.192)

-0.232(0.111)

0.310

1.5

Fatal and injury drop-off-related crashes

Two-lane

Paved

100

-7.772(1.977)

0.642(0.233)

 

0.563

5.1

Two-lane

Unpaved

98

-2.943(1.989)

0.227(0.258)

-0.167(0.147)

0.276

0.3

PDO drop-off-related crashes

Two-lane

Paved

100

-7.464(5.554)

0.597(0.653)

 

0.623

1.4

Two-lane

Unpaved

98

-3.006(1.593)

0.419(0.209)

-0.266(0.122)

0.069

1.7

Note: Blank cells indicate lane width coefficient was not significant.

This figure is a multiple line graph that shows the six SPFs developed for Georgia data. The x-axis has a range in AADT of 0 to 25,000 veh/day while the y-axis has a range of 0 to 8 crashes per mile per year. All of the lines are slightly curvilinear (i.e., depicting logarithmic or exponential growth) and increasing. The order of magnitude in lines is given by: (1) total crash severity for two-lane roadways with paved shoulders, (2) total crash severity for two-lane roadways with unpaved shoulders, (3) total crash severity for multilane roadways with paved shoulders, (4) F&I crash severity for two-lane roadways with unpaved shoulders, (5) F&I crash severity for two-lane roadways with paved shoulders, and (6) F&I crash severity for multilane roadways with paved shoulders.

Figure 3. Graph. Comparison of Georgia SPFs by crash severity and roadway and shoulder type.

This figure is a multiple line graph that shows the four SPFs developed for Indiana data. The x-axis has a range in AADT of 0 to 15,000 veh/day while the y-axis has a range of 0 to 5 crashes per mile per year. All of the lines are slightly curvilinear (i.e., depicting logarithmic or exponential growth) and increasing. The order of magnitude in lines is given by: (1) total crash severity for two-lane roadways with paved shoulders, (2) total crash severity for two-lane roadways with unpaved shoulders, (3) F&I crash severity for two-lane roadways with paved shoulders, and (4) F&I crash severity for two-lane roadways with unpaved shoulders.

Figure 4. Graph. Comparison of Indiana SPFs by crash severity and roadway and shoulder type.

As noted earlier, the proportion of run-off-road and drop-off-related crashes (developed from reference sites) was sometimes needed to adjust total or fatal and injury SPFs for prediction of those crash types. Table 12 presents these proportions estimated from the reference site data.

Table 12. Run-off-road and drop-off-related crash frequencies as a proportion of total crashes.

State

Roadway type

Shoulder type

Crash severity Level

Proportion of run-off-road crashes

Proportion of drop-off-related crashes

GA

Multilane

Paved

Total

0.215

0.127

FI

0.295

0.194

PDO

0.162

0.084

Two-lane

Paved

Total

0.371

0.230

FI

0.511

0.368

PDO

0.298

0.158

Unpaved

Total

0.473

0.300

FI

0.574

0.410

PDO

0.398

0.219

IN

Two-lane

Paved

Total

0.272

0.091

FI

0.566

0.225

PDO

0.199

0.058

Unpaved

Total

0.321

0.119

FI

0.645

0.248

PDO

0.253

0.091

FI = Fatal and injury crashes.

PDO = Property-damage-only crashes.

Additionally, yearly calibration factors were developed from the SPFs to provide a better yearly prediction in the methodology. These factors are needed because the SPFs are developed as an average of all years. The yearly calibration factor is determined as the ratio of the sum of observed crashes for all sites for a specific roadway type/shoulder type combination to the sum of the predicted crashes for the same sites using the AADT and crash count values for that year. These factors are presented in table 13 for Georgia and in table 14 for Indiana.

Table 13. Georgia SPF calibration factors.

Roadway type

Shoulder type

Crash severity level

Yearly calibration factors

2001

2002

2003

2004

2005

2006

2007

2008

Total crashes

Multilane

Paved

Total

0.956

1.023

1.071

0.943

1.078

1.178

0.993

0.983

FI

0.908

1.091

0.950

1.153

1.168

1.170

0.959

0.942

PDO

0.998

1.005

1.155

0.849

1.049

1.203

1.031

1.021

Two-lane

Paved

Total

0.856

0.949

0.919

1.044

0.990

1.045

1.025

1.023

FI

0.926

0.979

0.996

1.114

1.139

1.167

1.115

1.075

PDO

0.823

0.933

0.873

0.998

0.905

0.977

0.969

0.990

Unpaved

Total

0.996

0.876

0.884

1.061

1.068

1.112

0.895

1.024

FI

1.056

0.999

0.840

1.106

1.318

1.202

1.031

1.167

PDO

0.964

0.804

0.910

1.036

0.922

1.062

0.817

0.943

Run-off-road crashes

Multilane

Paved

Total

0.958

1.135

1.174

0.891

1.094

0.974

0.962

0.819

FI

1.167

1.268

1.267

1.267

1.425

1.216

1.048

1.064

PDO

0.928

1.168

1.241

0.731

0.987

0.917

0.999

0.747

Two-lane

Paved

Total

1.192

1.389

1.131

1.397

1.307

1.542

1.458

1.378

FI

1.302

1.188

1.226

1.502

1.481

1.688

1.474

1.416

PDO

1.110

1.581

1.058

1.318

1.168

1.430

1.451

1.355

Unpaved

Total

1.107

1.064

1.089

1.201

1.335

1.280

1.046

1.183

FI

1.150

1.167

0.828

1.241

1.405

1.232

1.114

1.265

PDO

1.003

0.905

1.282

1.095

1.194

1.256

0.923

1.036

Drop-off-related crashes

Multilane

Paved

Total

1.040

1.102

1.134

1.101

1.034

1.003

1.156

0.774

FI

0.925

1.320

0.989

1.121

0.989

1.254

1.117

0.860

PDO

1.156

0.874

1.275

1.075

1.075

0.741

1.188

0.683

Two-lane

Paved

Total

1.203

1.410

1.144

1.385

1.364

1.609

1.491

1.511

FI

1.290

1.135

1.270

1.449

1.549

1.652

1.543

1.636

PDO

1.111

1.746

1.001

1.312

1.154

1.564

1.429

1.368

Unpaved

Total

1.129

1.035

1.133

1.240

1.397

1.409

1.194

1.303

FI

1.217

1.212

0.818

1.186

1.426

1.409

1.345

1.393

PDO

0.997

0.794

1.506

1.285

1.335

1.384

0.982

1.165

FI = Fatal and injury crashes.

PDO = Property-damage-only crashes.

Table 14. Indiana SPF calibration factors.

Roadway type

Shoulder type

Crash severity level

Yearly calibration factors

2003

2004

2005

2006

2007

2008

Total crashes

Two-lane

Paved

Total

0.932

0.944

0.579

0.605

0.320

0.384

FI

0.918

1.006

0.456

0.586

0.343

0.326

PDO

0.943

0.936

0.616

0.615

0.317

0.402

Unpaved

Total

1.268

1.011

0.629

0.556

0.365

0.265

FI

0.914

1.002

0.471

0.472

0.287

0.182

PDO

1.322

0.968

0.650

0.557

0.373

0.279

Run-off-road crashes

Two-lane

Paved

Total

1.092

0.936

0.607

0.551

0.304

0.497

FI

1.266

1.097

0.489

0.651

0.407

0.448

PDO

1.074

0.911

0.713

0.535

0.268

0.558

Unpaved

Total

1.002

0.863

0.479

0.363

0.279

0.177

FI

0.850

1.041

0.503

0.446

0.232

0.154

PDO

1.068

0.754

0.457

0.313

0.300

0.186

Drop-off-related crashes

Two-lane

Paved

Total

0.994

0.946

0.646

0.431

0.431

0.690

FI

0.729

0.722

0.362

0.434

0.290

0.290

PDO

1.016

0.929

0.777

0.310

0.467

0.934

Unpaved

Total

1.289

1.038

0.544

0.545

0.520

0.249

FI

1.265

0.989

0.661

0.441

0.385

0.220

PDO

1.298

1.066

0.459

0.613

0.610

0.267

FI = Fatal and injury crashes.

PDO = Property-damage-only crashes.

4.3 Safety Evaluations

As previously discussed, four types of safety evaluations were performed as part of this study: (1) a safety comparison of treatment and comparison sites in the period before resurfacing; (2) an EB before-after evaluation; (3) a cross-sectional analysis; and (4) an analysis of shifts in the severity distribution from before to after resurfacing. The findings of these evaluations are presented in this section.

4.3.1 Safety Comparison of Treatment and Comparison Sites in the Period Before Resurfacing

An evaluation was conducted to compare the safety performance of treatment and comparison sites before resurfacing for specific States and roadway type/shoulder type combinations. This evaluation is critical to the interpretation of the safety differences between the treatment and comparison sites as an effect of the safety edge treatment. If the safety performance of the two types of sites differed in the period before resurfacing, the comparison of treatment and comparison sites in the period after resurfacing may be influenced.

Initial comparisons were made by examination of scatter plots of crashes and traffic volumes (crashes per mile per year versus lnAADT). Ideal plots would contain no discernable differences between treatment and comparison sites nor any extreme points. Separation of the data points between the two groups may indicate a potential concern in the subsequent analyses. Furthermore, if one group had systematically higher crash frequencies in the period before resurfacing, then the analysis for the period after resurfacing might need to account for this difference. Finally, large variation in crash frequencies for the same AADT values could inhibit crash analysis of the treatment and comparison groups. Inspection of these plots with data from year 3 (see appendix C) showed an improvement in the plots from year 1 and year 2.

Yearly total crash and target crash distributions were also presented in box plots to review data consistency from year to year. Ideal plots would have approximately the same distribution for crashes each year within a given site type and between site types. Additionally, potential concerns for the crash analysis-specifically, a regression to the mean or resurfacing effect- may be identified if the period after resurfacing is included.

Since crash frequencies are known to experience random variation around the mean or regression to the mean, the average over several years for the period before resurfacing should be compared to the average of several years for the period after resurfacing. Therefore, if the after period data are within the range of yearly crash means but numerically higher than the before period average, then safety analyses might show an increase in crash frequency due to the treatment (provided AADT growth was minimal). Conversely, if the after implementation year data are lower than the before period average, then the treatment effect would be a decrease in crash frequency. Examination of these graphs indicated that the after period data were almost always higher than the average of the before years but within the range of variation in yearly crash totals for both types of treated sites.

The apparent increase in crashes was examined to determine if it could be attributed to resurfacing. A resurfacing effect occurs when the reference sites remain the same or decrease in crashes while the treatment and comparison sites both increase. This effect was observed in nearly all of the plots.

One additional potential problem was found in this analysis. One treatment site on a two-lane highway with paved shoulders in Georgia site doubled in crash frequency from the before to the after period. Subsequent investigation found that this site was reconstructed during the second year after resurfacing, and therefore, it was excluded from the safety analysis presented in this report.

Formal crash frequency comparisons of means between the treatment and comparison sites for the period before resurfacing were conducted for each State/roadway type/shoulder type combination and target crash type. Two types of comparisons were made, a comparison of EB-adjusted expected crash frequencies and a comparison of observed crash frequencies. Both comparisons were performed using PROC GENMOD, a generalized linear model procedure available in SAS®, assuming a negative binomial crash distribution.(3) This procedure uses predictive modeling to test the means between the two treatment groups for statistical significance.

The results of these analyses are presented in table 15 and table 16. For the EB-adjusted crash analysis, results are provided only for those roadway type/shoulder type combinations for which SPFs could be developed. However, all target crash types were considered as they can be estimated by the EB technique. Regression coefficients with their standard errors are shown in the tables for each independent variable, including AADT and the treatment versus comparison site effect. The significance and p-value for each effect are also presented. Blank rows in the tables represent models that did not converge.

Table 15. Evaluation of treatment versus comparison site effect for the period before resurfacing using EB-adjusted crash frequencies.

State

Roadway type

Shoulder type

Crash type and severity level

Number of site-years

Intercept

AADT effect

Lane width effect

Treatment versus comparison site effect

Dispersion parameter

R2LR%

Coefficient

Standard error

p-value

Statistically significant?1

Coefficient

Standard error

p-value

Statistically significant?1

Coefficient

Standard error

p-value

Statistically significant?1

GA

Multilane

Paved

TOT

102

-3.965

0.572

0.143

0.001

Yes

       

-0.497

0.452

0.272

No

0.237

31.4

FI

102

-2.115

0.239

0.165

0.148

No

       

-0.391

0.394

0.322

No

0.031

9.4

PDO

102

-5.926

0.740

0.172

0.001

Yes

       

-0.531

0.525

0.312

No

0.282

35.3

rorTOT

102

-5.253

0.559

0.179

0.002

Yes

       

-0.149

0.191

0.434

No

0.010

5.9

rorFI

102

-3.843

0.328

0.158

0.038

Yes

       

-0.293

0.169

0.083

Yes

0.010

22.2

rorPDO

102

-7.761

0.757

0.189

0.001

Yes

       

-0.043

0.239

0.856

No

0.010

8.1

doTOT

102

-4.265

0.430

0.127

0.001

Yes

       

-0.095

0.132

0.469

No

0.010

21.2

doFI

102

-3.620

0.285

0.104

0.006

Yes

       

-0.134

0.106

0.204

No

0.010

24.1

doPDO

102

-6.240

0.569

0.172

0.001

Yes

       

-0.060

0.177

0.734

No

0.010

19.1

Two-lane

Paved

TOT

264

-12.086

1.475

0.089

0.000

Yes

       

0.154

0.177

0.384

No

0.010

56.9

FI

264

-11.367

1.306

0.099

0.000

Yes

       

-0.104

0.129

0.420

No

0.010

33.5

PDO

264

-13.244

1.534

0.095

0.000

Yes

       

0.302

0.222

0.175

No

0.010

48.0

rorTOT

264

-5.358

1.133

0.107

0.000

Yes

-0.361

0.093

0.001

Yes

-0.259

0.141

0.067

Yes

0.010

25.3

rorFI

264

-4.377

0.973

0.132

0.001

Yes

-0.381

0.070

0.001

Yes

-0.338

0.126

0.007

Yes

0.010

5.2

rorPDO

264

-7.053

1.173

0.100

0.000

Yes

-0.314

0.128

0.014

Yes

-0.190

0.167

0.255

No

0.010

15.9

doTOT

264

-7.238

1.221

0.116

0.000

Yes

-0.303

0.067

0.001

Yes

-0.312

0.120

0.009

Yes

0.010

19.0

doFI

264

-6.870

1.207

0.124

0.000

Yes

-0.366

0.102

0.000

Yes

-0.369

0.158

0.020

Yes

0.010

1.7

doPDO

264

-10.155

1.290

0.105

0.000

Yes

-0.189

0.107

0.078

Yes

-0.169

0.123

0.171

No

0.010

12.1

Unpaved

TOT

318

-8.117

1.001

0.098

0.000

Yes

       

0.611

0.149

0.001

Yes

0.010

61.2

FI

318

-8.026

0.899

0.102

0.000

Yes

       

0.237

0.137

0.084

Yes

0.010

39.1

PDO

318

-9.051

1.031

0.079

0.000

Yes

       

0.902

0.180

0.001

Yes

0.010

57.8

rorTOT

318

-7.230

0.819

0.088

0.000

Yes

       

0.358

0.159

0.024

Yes

0.010

37.3

rorFI

318

-6.816

0.703

0.077

0.000

Yes

       

0.179

0.133

0.179

No

0.010

20.6

rorPDO

318

-8.895

0.909

0.071

0.000

Yes

       

0.574

0.208

0.006

Yes

0.010

30.0

doTOT

318

-7.444

0.801

0.084

0.000

Yes

       

0.271

0.142

0.056

Yes

0.010

29.7

doFI

318

-6.545

0.636

0.085

0.001

Yes

       

0.180

0.127

0.157

No

0.010

12.8

doPDO

318

-10.351

1.026

0.107

0.000

Yes

       

0.414

0.196

0.035

Yes

0.010

21.4

See notes at end of table.

Table 15. Evaluation of treatment versus comparison site effect for the period before resurfacing using EB-adjusted crash frequencies-Continued.

State

Roadway type

Shoulder type

Crash type and severity level

Number of site-years

Intercept

AADT effect

Lane width effect

Treatment versus comparison site effect

Dispersion parameter

R2LR%

Coefficient

Standard error

p-value

Statistically significant?1

Coefficient

Standard error

p-value

Statistically significant?1

Coefficient

Standard error

p-valuev

Statistically significant?1

IN

Two-lane

Paved

TOT

42

-10.904

1.409

0.056

0.000

Yes

       

-0.517

0.307

0.092

Yes

0.150

7.3

FI

42

-21.302

2.546

0.222

0.000

Yes

       

-1.076

0.338

0.001

Yes

0.010

9.3

PDO

42

-2.772

0.422

0.114

0.000

Yes

       

-0.184

0.405

0.650

No

0.152

4.7

rorTOT

42

-2.431

0.208

0.283

0.463

No

       

-0.054

0.208

0.793

No

0.010

23.8

rorFI

42

-4.735

0.361

0.061

0.001

Yes

       

-0.073

0.078

0.352

No

0.010

34.0

rorPDO

42

                             

doTOT

42

                             

doFI

42

-5.918

0.391

0.059

0.001

Yes

       

-0.269

0.159

0.090

Yes

0.010

35.2

doPDO

42

                             

Unpaved

TOT

68

-0.578

0.787

0.231

0.001

Yes

-0.506

0.117

0.001

Yes

0.097

0.273

0.723

No

0.137

43.8

FI

68

-1.688

0.470

0.141

0.001

Yes

-0.312

0.061

0.001

Yes

-0.063

0.165

0.701

No

0.010

15.9

PDO

68

-1.128

0.932

0.264

0.000

Yes

-0.584

0.141

0.001

Yes

0.172

0.302

0.570

No

0.212

40.4

rorTOT

68

0.889

0.588

0.208

0.005

Yes

-0.585

0.103

0.001

Yes

-0.045

0.219

0.837

No

0.010

34.6

rorFI

68

-1.126

0.283

0.035

0.001

Yes

-0.278

0.010

0.000

Yes

-0.039

0.040

0.328

No

0.010

17.1

rorPDO

68

0.902

0.879

0.321

0.006

Yes

-0.838

0.174

0.001

Yes

-0.015

0.323

0.964

No

0.010

35.7

doTOT

68

-0.837

0.211

0.106

0.047

Yes

-0.242

0.069

0.000

Yes

-0.433

0.169

0.011

Yes

0.010

5.1

doFI

68

-1.842

0.139

0.056

0.013

Yes

-0.190

0.036

0.001

Yes

-0.246

0.092

0.008

Yes

0.010

21.4

doPDO

68

-1.212

0.259

0.161

0.108

Yes

-0.285

0.107

0.008

Yes

-0.565

0.258

0.028

Yes

0.010

3.5

1 At the 0.20 level.

TOT = total crashes (all severity levels combined).

Fl = fatal and injury crashes.

PDO = property-damage-only crashes.

ror = run-off-road crashes.

do = drop-off-related crashes.

Note: Blank cells represent models that did not converge.

Table 16. Evaluation of treatment versus comparison site effect for the period before resurfacing using observed crash frequencies.

State

Roadway type

Shoulder type

Crash type and severity level

Number of site-years

Intercept

AADT effect

Lane width effect

Treatment versus comparison site effect

Dispersion parameter

R2LR%

Coefficient

Standard error

p-value

Statistically significant?1

Coefficient

Standard error

p-value

Statistically significant?1

Coefficient

Standard error

p-value

Statistically significant?1

GA

Multilane

Paved

TOT

102

-9.014

1.128

0.282

0.000

Yes

       

-0.878

0.505

0.082

Yes

0.378

27.7

FI

102

-7.293

0.812

0.288

0.005

Yes

       

-0.826

0.516

0.109

No

0.338

10.1

PDO

102

-10.881

1.286

0.321

0.000

Yes

       

-0.885

0.527

0.093

Yes

0.505

27.3

rorTOT

102

-7.749

0.822

0.268

0.002

Yes

       

-0.295

0.225

0.188

No

0.015

15.0

rorFI

102

-7.005

2.654

0.660

0.288

Yes

       

-0.547

0.259

0.035

Yes

0.010

5.6

rorPDO

102

-9.207

3.372

0.911

0.353

Yes

       

-0.119

0.282

0.672

No

0.010

16.2

doTOT

102

-8.374

2.175

0.844

0.229

Yes

       

-0.355

0.180

0.048

Yes

0.010

11.9

doFI

102

-8.785

2.656

0.809

0.287

Yes

       

-0.442

0.221

0.046

Yes

0.010

5.0

doPDO

102

-9.387

3.466

0.884

0.364

Yes

       

-0.304

0.272

0.264

No

0.010

7.6

Two-lane

Paved

TOT

264

-8.045

0.982

0.130

0.000

Yes

       

0.222

0.176

0.207

No

0.247

35.7

FI

264

-7.646

0.852

0.143

0.000

Yes

       

-0.121

0.152

0.426

No

0.018

21.3

PDO

264

-10.106

1.147

0.139

0.000

Yes

       

0.447

0.200

0.025

Yes

0.436

30.8

rorTOT

264

-3.346

0.666

0.121

0.000

Yes

-0.220

0.161

0.172

Yes

-0.231

0.206

0.261

No

0.166

13.9

rorFI

264

-1.188

0.465

0.156

0.003

Yes

-0.303

0.161

0.059

Yes

-0.593

0.229

0.010

Yes

0.073

7.2

rorPDO

264

-8.299

0.834

0.118

0.000

Yes

       

0.110

0.203

0.587

No

0.360

10.8

doTOT

264

-6.541

0.673

0.173

0.000

Yes

       

-0.297

0.173

0.086

Yes

0.177

10.3

doFI

264

-1.063

0.414

0.202

0.040

Yes

-0.300

0.170

0.077

Yes

-0.752

0.249

0.003

Yes

0.173

5.7

doPDO

264

-10.600

1.038

0.204

0.000

Yes

       

-0.016

0.223

0.942

No

0.156

9.9

Unpaved

TOT

318

-8.615

1.059

0.104

0.000

Yes

       

0.610

0.174

0.000

Yes

0.389

35.9

FI

318

-8.473

0.940

0.097

0.000

Yes

       

0.258

0.177

0.143

No

0.318

21.3

PDO

318

-9.950

1.148

0.119

0.000

Yes

       

0.864

0.197

0.000

Yes

0.419

34.3

rorTOT

318

-7.022

0.774

0.106

0.000

Yes

       

0.441

0.194

0.023

Yes

0.309

19.4

rorFI

318

-7.358

0.740

0.109

0.000

Yes

       

0.226

0.220

0.304

No

0.487

9.7

rorPDO

318

-8.611

0.874

0.150

0.000

Yes

       

0.653

0.228

0.004

Yes

0.385

16.8

doTOT

318

-7.106

0.736

0.132

0.000

Yes

       

0.397

0.212

0.061

Yes

0.247

15.6

doFI

318

-6.937

0.645

0.139

0.000

Yes

       

0.270

0.245

0.272

No

0.548

6.9

doPDO

318

-9.469

0.922

0.188

0.000

Yes

       

0.554

0.252

0.028

Yes

0.361

12.5

See notes at end of table.

Table 16. Evaluation of treatment versus comparison site effect for the period before resurfacing using observed crash frequencies-Continued.

State

Roadway type

Shoulder type

Crash type and severity level

Number of site-years

Intercept

AADT effect

Lane width effect

Treatment versus comparison site effect

Dispersion parameter

R2LR%

Coefficient

Standard error

p-value

Statistically significant?1

Coefficient

Standard error

p-value

Statistically significant?1

Coefficient

Standard error

p-value

Statistically significant?1

IN

Two-lane

Paved

TOT

42

-3.824

0.588

0.250

0.019

Yes

       

-0.380

0.364

0.296

No

0.416

6.6

FI

42

-7.523

0.850

0.606

0.161

No

       

-0.842

0.533

0.114

No

0.629

7.9

PDO

42

-3.076

0.465

0.235

0.048

Yes

       

-0.205

0.356

0.565

No

0.369

4.1

rorTOT

42

-6.756

0.736

0.546

0.178

No

       

-0.468

0.396

0.237

No

0.446

5.1

rorFI

42

-2.953

4.414

0.188

0.515

No

       

-0.830

0.457

0.069

Yes

0.010

6.0

rorPDO

42

-8.070

0.815

0.589

0.167

No

       

-0.092

0.432

0.831

No

0.461

5.0

doTOT

42

-13.860

1.420

1.212

0.241

No

       

-0.996

0.503

0.048

Yes

0.478

6.0

doFI

42

                           

0.0

doPDO

42

-23.901

2.478

0.959

0.010

Yes

       

-0.477

0.529

0.368

No

0.010

11.9

Unpaved

TOT

68

-0.761

0.918

0.248

0.000

Yes

-0.587

0.140

0.000

Yes

0.188

0.297

0.525

No

0.435

35.8

FI

68

-0.041

0.732

0.347

0.035

Yes

-0.640

0.225

0.005

Yes

-0.050

0.347

0.884

No

0.093

23.4

PDO

68

-1.612

0.998

0.270

0.000

Yes

-0.594

0.155

0.000

Yes

0.269

0.304

0.377

No

0.527

31.5

rorTOT

68

1.418

0.806

0.316

0.011

Yes

-0.783

0.200

0.000

Yes

-0.068

0.331

0.838

No

0.221

34.7

rorFI

68

1.478

3.119

0.435

0.358

No

-0.608

0.212

0.004

Yes

-0.268

0.340

0.431

No

0.010

18.7

rorPDO

68

0.475

1.120

0.398

0.005

Yes

-0.974

0.313

0.002

Yes

0.091

0.365

0.804

No

0.377

29.2

doTOT

68

1.029

0.101

0.386

0.794

No

-0.312

0.246

0.204

No

-1.107

0.596

0.063

Yes

0.010

21.7

doFI

68

                           

0.0

doPDO

68

-4.194

0.270

0.560

0.630

No

       

-1.090

0.699

0.119

No

0.584

6.4

NY

Two-lane

Paved

TOT

36

-5.328

0.674

0.085

0.000

Yes

       

0.127

0.182

0.484

No

0.486

24.9

FI

36

-6.943

0.766

0.113

0.000

Yes

       

0.308

0.172

0.074

Yes

0.674

19.3

PDO

36

-5.467

0.625

0.083

0.000

Yes

       

-0.030

0.204

0.884

No

0.813

15.7

rorTOT

36

-4.846

0.480

0.085

0.000

Yes

       

0.577

0.140

0.000

Yes

0.243

19.6

rorFI

36

-5.333

0.486

0.122

0.000

Yes

       

0.643

0.175

0.000

Yes

0.410

14.4

rorPDO

36

-5.784

0.372

0.467

0.048

Yes

       

0.475

0.105

0.000

Yes

0.010

13.0

doTOT

36

                             

doFI

36

                             

doPDO

36

                             

1 At the 0.20 significance level.

TOT = total crashes (all severity levels combined).

Fl = fatal and injury crashes.

PDO = property-damage-only crashes.

ror = run-off-road crashes.

do = drop-off-related crashes.

Note: Blank cells indicate models that did not converge.

Results from the analysis of EB-adjusted crash frequencies in table 15 show that there tended to be significant differences between treatment and comparison site crash frequencies for Georgia sites with unpaved shoulders in the period before resurfacing. Comparison sites that had unpaved shoulders had lower crash rates than treatment sites. There is also evidence of differences in drop-off-related and run-off-road crashes for Georgia paved shoulder locations. Similarly, Indiana unpaved shoulder locations differed for drop-off-related crashes. These locations had treatment sites with lower crash rates.

Results from the analysis of observed crash frequencies somewhat confirmed the results of the EB-adjusted crashes. However, there tended to be fewer significant results and poorer fit of the models in general. This was to be expected because EB-adjusted crashes are smoothed by the SPF model predictions, causing smaller differences and less variation and leading to more significant results. Differences between treatment and comparison sites were confirmed for Georgia unpaved shoulder locations and drop-off-related crashes for paved shoulder locations. Additionally, New York locations, which were not tested by EB-adjusted crashes, showed differences for run-off-road crashes. All other significant differences were associated with poor models.

It was also desirable to confirm the existence of a cause-and-effect chain leading from the frequency and height of pavement-edge drop-offs to the likelihood of crashes. The drop-off height analysis reported in chapter 3 indicated that two-lane highway sites with unpaved shoulders and the multilane highway sites in Georgia did not have significant differences in the proportion of high drop-offs and therefore should have non-significant differences in crash frequency in the period before resurfacing. This expectation was not entirely supported by crash analysis results. However, for cases in which there were significant differences, these differences were in the same direction indicated in the drop-off analysis. That is, if drop-offs were more prevalent, then the sites had more crashes. Similarly, two-lane highway sites with paved shoulders in Georgia had comparison sites with a significantly higher probability of high drop-offs, and the crash analysis showed the comparison sites had more crashes, although the result was not significant.

Results for Indiana sites on two-lane highways with paved shoulders were consistent with the analysis of drop-off measurements, but the results for Indiana sites on two-lane highways with unpaved shoulders were not consistent with the analysis of drop-off measurements.

Overall, the treatment and comparison sites showed similar crash frequencies for paved shoulder sites in the period before resurfacing. By contrast, there were some statistically significant differences in crash frequencies between treatment and comparison sites for unpaved shoulders during the period before resurfacing. It should be noted that only 2 years of crash data were available for the period before resurfacing in Indiana, in comparison to 6 years for the period before resurfacing in Georgia. Thus, the variability of the Indiana crash frequencies was expected to be higher. In most cases (with the single exception previously noted), the differences in crash frequencies between treatment and comparison sites were similar to the differences in proportions of extreme drop-off heights for the period before resurfacing.

4.3.2 Before-After Evaluation Using the EB Technique

An observational before-after evaluation was conducted using the EB technique to estimate the safety effectiveness of the safety edge treatment. Separate before-after evaluations were conducted for resurfacing projects with the safety edge (treatment sites) and resurfacing projects without the safety edge (comparison sites). The ratio of these results was used to estimate the effect of the safety edge treatment.

All crash severity levels for total crashes, run-off-road crashes, and drop-off-related crashes were evaluated. The study period before resurfacing for these evaluations was the 4-year period from 2001 to 2004. The study period after resurfacing was the 3-year period from 2006 to 2008. The entire year in which resurfacing was performed, 2005, was excluded from the evaluation. The rationale for excluding crashes during the construction year is that it takes time for drivers to adjust to new driving conditions, and so the transition period is not necessarily representative of the long-term safety performance of the site. All of the crash data used in the evaluation were for complete calendar years so that there would be no opportunity for seasonal biases to affect the results.

The EB procedure was programmed and executed in SAS®.(3) Effectiveness estimates and their precision estimates, along with their statistical significance, are presented for specific crash types in table 17 through table 25.

Table 17. Before-after EB evaluation results for total crashes.

State

Roadway type

Shoulder type

Site type

Number of

sites

Odds ratio

Change in crash frequency from before to after resurfacing

Statistically significant?

Safety edge effect

Percent change

Direction

Standard error (%)

5% level

10% level

Effect (%)

Direction

Standard error (%)

5% level

10% level

GA

Two-lane

Paved

T

25

1.804

13.123

Increase

7.276

No

Yes

7.732

Decrease

9.596

No

No

C

19

2.262

22.602

Increase

9.993

Yes

Yes

Unpaved

T

22

2.246

-13.562

Decrease

6.038

Yes

Yes

11.361

Decrease

8.467

No

No

C

31

0.389

-2.483

Decrease

6.376

No

No

Combined

T

47

0.143

-0.670

Decrease

4.697

No

No

6.817

Decrease

6.459

No

No

C

50

1.217

6.597

Increase

5.421

No

No

IN

Two-lane

Paved

T

14

0.047

0.567

Increase

12.167

No

No

15.524

Decrease

14.422

No

No

C

7

1.333

19.048

Increase

14.293

No

No

Unpaved

T

16

1.524

29.925

Increase

19.639

No

No

-26.942

Increase

24.027

No

No

C

18

0.201

2.350

Increase

11.691

No

No

Combined

T

30

1.000

10.456

Increase

10.454

No

No

-0.235

Increase

12.622

No

No

C

25

1.120

10.197

Increase

9.104

No

No

GA & IN

Two-lane

Paved

T

39

1.601

10.027

Increase

6.262

No

No

9.485

Decrease

8.009

No

No

C

26

2.628

21.556

Increase

8.203

Yes

Yes

Unpaved

T

38

1.311

-7.657

Decrease

5.842

No

No

6.516

Decrease

7.910

No

No

C

49

0.218

-1.221

Decrease

5.604

No

No

Combined

T

77

0.360

1.546

Increase

4.293

No

No

5.674

Decrease

5.737

No

No

C

75

1.642

7.654

Increase

4.662

No

No

T = Treatment sites resurfaced with safety edge.

C = Comparison sites resurfaced without safety edge.

Table 18. Before-after EB evaluation results for fatal and injury crashes.

State

Roadway type

Shoulder type

Site type

Number of

sites

Odds ratio

Change in crash frequency from before to after resurfacing

Statistically significant?

Safety edge effect

Percent change

Direction

Standard error (%)

5% level

10% level

Effect (%)

Direction

Standard error (%)

5% level

10% level

GA

Two-lane

Paved

T

25

1.592

19.899

Increase

12.499

No

No

10.959

Decrease

13.779

No

No

C

19

2.244

34.656

Increase

15.446

Yes

Yes

Unpaved

T

22

0.232

2.761

Increase

11.887

No

No

-15.555

Increase

17.687

No

No

C

31

1.243

-11.072

Decrease

8.907

No

No

Combined

T

47

1.346

11.647

Increase

8.651

No

No

-5.982

Increase

11.462

No

No

C

50

0.676

5.345

Increase

7.905

No

No

IN

Two-lane

Paved

T

14

0.736

-18.579

Decrease

25.239

No

No

44.993

Decrease

21.492

Yes

Yes

C

7

1.359

48.020

Increase

35.335

No

No

Unpaved

T

16

1.340

-35.945

Decrease

26.829

No

No

43.548

Decrease

26.137

No

No

C

18

0.598

13.469

Increase

22.521

No

No

Combined

T

30

1.339

-24.947

Decrease

18.633

No

No

40.939

Decrease

17.167

Yes

Yes

C

25

1.403

27.078

Increase

19.294

No

No

GA & IN

Two-lane

Paved

T

39

1.295

14.685

Increase

11.337

No

No

16.528

Decrease

11.919

No

No

C

26

2.633

37.393

Increase

14.199

Yes

Yes

Unpaved

T

38

0.087

-0.961

Decrease

11.012

No

No

-6.361

Increase

15.147

No

No

C

49

0.827

-6.884

Decrease

8.328

No

No

Combined

T

77

0.924

7.328

Increase

7.93

No

No

1.667

Decrease

9.780

No

No

C

75

1.246

9.148

Increase

7.341

No

No

T = Treatment sites resurfaced with safety edge.

C = Comparison sites resurfaced without safety edge.

Table 19. Before-after EB evaluation results for PDO crashes.

State

Roadway type

Shoulder type

Site type

Number of

sites

Odds ratio

Change in crash frequency from before to after resurfacing

Statistically significant?

Safety edge effect

Percent change

Direction

Standard error (%)

5% level

10% level

Effect (%)

Direction

Standard error (%)

5% level

10% level

GA

Two-lane

Paved

T

25

0.975

9.276

Increase

9.511

No

No

2.554

Decrease

15.183

No

No

C

19

0.837

12.140

Increase

14.502

No

No

Unpaved

T

22

2.886

-20.980

Decrease

7.271

Yes

Yes

24.281

Decrease

10.078

Yes

Yes

C

31

0.437

4.359

Increase

9.963

No

No

Combined

T

47

1.150

-6.764

Decrease

5.881

No

No

13.201

Decrease

8.607

No

No

C

50

0.899

7.416

Increase

8.249

No

No

IN

Two-lane

Paved

T

14

0.287

4.045

Increase

14.108

No

No

6.579

Decrease

18.450

No

No

C

7

0.710

11.372

Increase

16.024

No

No

Unpaved

T

16

1.898

47.501

Increase

25.033

No

Yes

-50.902

Increase

33.486

No

No

C

18

0.160

-2.254

Decrease

14.046

No

No

Combined

T

30

1.449

18.175

Increase

12.540

No

No

-12.894

Increase

16.531

No

No

C

25

0.440

4.678

Increase

10.631

No

No

GA & IN

Two-lane

Paved

T

39

0.995

7.860

Increase

7.901

No

No

3.845

Decrease

11.632

No

No

C

26

1.127

12.173

Increase

10.803

No

No

Unpaved

T

38

1.469

-10.599

Decrease

7.215

No

No

12.795

Decrease

9.898

No

No

C

49

0.309

2.518

Increase

8.150

No

No

Combined

T

77

0.185

-0.995

Decrease

5.364

No

No

7.100

Decrease

7.601

No

No

C

75

1.007

6.572

Increase

6.529

No

No

T = Treatment sites resurfaced with safety edge.

C = Comparison sites resurfaced without safety edge.

Table 20. Before-after EB evaluation results for total run-off-road crashes.

State

Roadway type

Shoulder type

Site type

Number of

sites

Odds ratio

Change in crash frequency from before to after resurfacing

Statistically significant?

Safety edge effect

Percent change

Direction

Standard error (%)

5% level

10% level

Effect (%)

Direction

Standard error (%)

5% level

10% level

GA

Two-lane

Paved

T

25

3.375

27.777

Increase

8.230

Yes

Yes

13.721

Decrease

9.010

No

No

C

19

3.956

48.097

Increase

12.158

Yes

Yes

Unpaved

T

22

0.102

0.718

Increase

7.049

No

No

9.080

Decrease

8.652

No

No

C

31

1.487

10.777

Increase

7.248

No

No

Combined

T

47

2.594

14.006

Increase

5.400

Yes

Yes

7.872

Decrease

6.406

No

No

C

50

3.766

23.747

Increase

6.306

Yes

Yes

IN

Two-lane

Paved

T

14

3.152

63.675

Increase

20.201

Yes

Yes

8.183

Decrease

15.844

No

No

C

7

3.639

78.263

Increase

21.505

Yes

Yes

Unpaved

T

16

3.408

105.593

Increase

30.982

Yes

Yes

-46.895

Increase

27.796

No

Yes

C

18

2.502

39.959

Increase

15.968

Yes

Yes

Combined

T

30

4.573

78.116

Increase

17.083

Yes

Yes

-13.484

Increase

14.389

No

No

C

25

4.391

56.952

Increase

12.969

Yes

Yes

GA & IN

Two-lane

Paved

T

39

4.522

34.710

Increase

7.675

Yes

Yes

14.177

Decrease

7.593

No

Yes

C

26

5.351

56.962

Increase

10.645

Yes

Yes

Unpaved

T

38

1.631

11.534

Increase

7.071

No

No

4.786

Decrease

8.094

No

No

C

49

2.581

17.140

Increase

6.640

Yes

Yes

Combined

T

77

4.496

23.514

Increase

5.230

Yes

Yes

6.315

Decrease

5.654

No

No

C

75

5.576

31.840

Increase

5.710

Yes

Yes

T = Treatment sites resurfaced with safety edge.

C = Comparison sites resurfaced without safety edge.

Table 21. Before-after EB evaluation results for fatal and injury run-off-road crashes.

State

Roadway type

Shoulder type

Site type

Number of

sites

Odds ratio

Change in crash frequency from before to after resurfacing

Statistically significant?

Safety edge effect

Percent change

Direction

Standard error (%)

5% level

10% level

Effect (%)

Direction

Standard error (%)

5% level

10% level

GA

Two-lane

Paved

T

25

3.036

46.712

Increase

15.384

Yes

Yes

19.175

Decrease

12.659

No

No

C

19

3.856

81.517

Increase

21.142

Yes

Yes

Unpaved

T

22

1.891

28.134

Increase

14.877

No

Yes

-18.058

Increase

18.139

No

No

C

31

0.784

8.535

Increase

10.892

No

No

Combined

T

47

3.529

37.878

Increase

10.733

Yes

Yes

-3.609

Increase

11.192

No

No

C

50

3.295

33.075

Increase

10.037

Yes

Yes

IN

Two-lane

Paved

T

14

0.086

-2.553

Decrease

29.761

No

No

46.332

Decrease

20.632

Yes

Yes

C

7

1.931

81.574

Increase

42.252

No

Yes

Unpaved

T

16

0.140

6.109

Increase

43.637

No

No

-12.464

Increase

51.101

No

No

C

18

0.309

-5.651

Decrease

18.313

No

No

Combined

T

30

0.018

0.440

Increase

24.694

No

No

16.402

Decrease

23.965

No

No

C

25

1.137

20.146

Increase

17.714

No

No

GA & IN

Two-lane

Paved

T

39

2.877

39.851

Increase

13.853

Yes

Yes

23.123

Decrease

11.053

Yes

Yes

C

26

4.323

81.916

Increase

18.950

Yes

Yes

Unpaved

T

38

1.877

26.499

Increase

14.120

No

Yes

-20.037

Increase

17.137

No

No

C

49

0.573

5.383

Increase

9.388

No

No

Combined

T

77

3.407

33.764

Increase

9.911

Yes

Yes

-2.622

Increase

10.228

No

No

C

75

3.469

30.346

Increase

8.748

Yes

Yes

T = Treatment sites resurfaced with safety edge.

C = Comparison sites resurfaced without safety edge.

Table 22. Before-after EB evaluation results for PDO run-off-road crashes.

State

Roadway type

Shoulder type

Site type

Number of sites

Odds ratio

Change in crash frequency from before to after resurfacing

Statistically significant?

Safety edge effect

Percent change

Direction

Standard error (%)

5% level

10% level

Effect (%)

Direction

Standard error (%)

5% level

10% level

GA

Two-lane

Paved

T

25

1.821

18.575

Increase

10.200

No

Yes

5.160

Decrease

14.498

No

No

C

19

1.587

25.026

Increase

15.770

No

No

Unpaved

T

22

1.286

-10.527

Decrease

8.185

No

No

20.225

Decrease

10.451

No

Yes

C

31

1.158

12.156

Increase

10.494

No

No

Combined

T

47

0.535

3.462

Increase

6.472

No

No

11.524

Decrease

8.673

No

No

C

50

1.929

16.938

Increase

8.779

No

Yes

IN

Two-lane

Paved

T

14

3.154

82.392

Increase

26.120

Yes

Yes

-3.538

Increase

21.327

No

No

C

7

2.922

76.160

Increase

26.061

Yes

Yes

Unpaved

T

16

3.396

131.099

Increase

38.606

Yes

Yes

-31.229

Increase

30.365

No

No

C

18

2.699

76.104

Increase

28.193

Yes

Yes

Combined

T

30

4.549

99.719

Increase

21.920

Yes

Yes

-12.878

Increase

17.414

No

No

C

25

3.999

76.934

Increase

19.240

Yes

Yes

GA & IN

Two-lane

Paved

T

39

3.323

32.288

Increase

9.718

Yes

Yes

7.374

Decrease

11.235

No

No

C

26

3.108

42.820

Increase

13.778

Yes

Yes

Unpaved

T

38

0.657

5.574

Increase

8.485

No

No

16.029

Decrease

9.550

No

No

C

49

2.548

25.727

Increase

10.096

Yes

Yes

Combined

T

77

2.976

19.201

Increase

6.452

Yes

Yes

10.162

Decrease

7.404

No

No

C

75

3.995

32.684

Increase

8.181

Yes

Yes

T = Treatment sites resurfaced with safety edge.

C = Comparison sites resurfaced without safety edge.

Table 23. Before-after EB evaluation results for total drop-off-related crashes.

State

Roadway type

Shoulder type

Site type

Number of sites

Odds ratio

Change in crash frequency from before to after resurfacing

Statistically significant?

Safety edge effect

Percent change

Direction

Standard error (%)

5% level

10% level

Effect (%)

Direction

Standard error (%)

5% level

10% level

GA

Two-lane

Paved

T

25

3.817

32.758

Increase

8.582

Yes

Yes

10.293

Decrease

9.405

No

No

C

19

3.930

47.991

Increase

12.210

Yes

Yes

Unpaved

T

22

0.591

4.334

Increase

7.328

No

No

9.130

Decrease

8.698

No

No

C

31

1.970

14.817

Increase

7.522

No

Yes

Combined

T

47

3.250

18.272

Increase

5.623

Yes

Yes

6.603

Decrease

6.522

No

No

C

50

4.115

26.633

Increase

6.472

Yes

Yes

IN

Two-lane

Paved

T

14

3.681

86.217

Increase

23.425

Yes

Yes

8.759

Decrease

15.949

No

No

C

7

4.202

104.093

Increase

24.772

Yes

Yes

Unpaved

T

16

4.374

193.003

Increase

44.122

Yes

Yes

-79.564

Increase

34.036

Yes

Yes

C

18

3.357

63.175

Increase

18.820

Yes

Yes

Combined

T

30

5.509

117.38

Increase

21.305

Yes

Yes

-19.733

Increase

15.389

No

No

C

25

5.386

81.554

Increase

15.141

Yes

Yes

GA & IN

Two-lane

Paved

T

39

5.191

42.320

Increase

8.153

Yes

Yes

12.586

Decrease

7.803

No

No

C

26

5.657

62.812

Increase

11.104

Yes

Yes

Unpaved

T

38

2.489

18.854

Increase

7.574

Yes

Yes

4.455

Decrease

8.189

No

No

C

49

3.450

24.396

Increase

7.070

Yes

Yes

Combined

T

77

5.564

31.039

Increase

5.578

Yes

Yes

5.587

Decrease

5.739

No

No

C

75

6.428

38.794

Increase

6.035

Yes

Yes

T = Treatment sites resurfaced with safety edge.

C = Comparison sites resurfaced without safety edge.

Table 24. Before-after EB evaluation results for fatal and injury drop-off-related crashes.

State

Roadway type

Shoulder type

Site type

Number of

sites

Odds ratio

Change in crash frequency from before to after resurfacing

Statistically significant?

Safety edge effect

Percent change

Direction

Standard error (%)

5% level

10% level

Effect (%)

Direction

Standard error (%)

5% level

10% level

GA

Two-lane

Paved

T

25

3.082

48.356

Increase

15.691

Yes

Yes

15.135

Decrease

13.441

No

No

C

19

3.639

74.815

Increase

20.560

Yes

Yes

Unpaved

T

22

2.068

31.810

Increase

15.383

Yes

Yes

-18.357

Increase

18.246

No

No

C

31

1.014

11.366

Increase

11.214

No

No

Combined

T

47

3.683

40.587

Increase

11.019

Yes

Yes

-5.202

Increase

11.444

No

No

C

50

3.317

33.635

Increase

10.139

Yes

Yes

IN

Two-lane

Paved

T

14

1.727

120.288

Increase

69.662

No

Yes

42.488

Decrease

22.855

No

Yes

C

7

3.073

283.030

Increase

92.115

Yes

Yes

Unpaved

T

16

1.038

77.468

Increase

74.600

No

No

38.503

Decrease

28.727

No

No

C

18

3.208

188.581

Increase

58.789

Yes

Yes

Combined

T

30

2.037

105.168

Increase

51.641

Yes

Yes

36.972

Decrease

18.615

No

Yes

C

25

4.473

225.519

Increase

50.414

Yes

Yes

GA & IN

Two-lane

Paved

T

39

3.473

53.546

Increase

15.419

Yes

Yes

21.596

Decrease

11.453

No

Yes

C

26

4.604

95.839

Increase

20.818

Yes

Yes

Unpaved

T

38

2.255

34.086

Increase

15.115

Yes

Yes

-5.230

Increase

15.186

No

No

C

49

2.381

27.422

Increase

11.516

Yes

Yes

Combined

T

77

4.106

44.481

Increase

10.833

Yes

Yes

4.676

Decrease

9.672

No

No

C

75

4.980

51.568

Increase

10.356

Yes

Yes

T = Treatment sites resurfaced with safety edge.

C = Comparison sites resurfaced without safety edge.

Table 25. Before-after EB evaluation results for PDO drop-off-related crashes.

State

Roadway type

Shoulder type

Site type

Number of sites

Odds ratio

Change in crash frequency from before to after resurfacing

Statistically significant?

Safety edge effect

Percent change

Direction

Standard error (%)

5% level

10% level

Effect (%)

Direction

Standard error (%)

5% level

10% level

GA

Two-lane

Paved

T

25

2.281

24.794

Increase

10.870

Yes

Yes

2.550

Decrease

15.169

No

No

C

19

1.698

28.059

Increase

16.524

No

No

Unpaved

T

22

0.826

-7.063

Decrease

8.556

No

No

20.631

Decrease

10.466

No

Yes

C

31

1.550

17.095

Increase

11.032

No

No

Combined

T

47

1.194

8.149

Increase

6.827

No

No

10.829

Decrease

8.786

No

No

C

50

2.307

21.283

Increase

9.224

Yes

Yes

IN

Two-lane

Paved

T

14

3.242

81.224

Increase

25.055

Yes

Yes

-2.553

Increase

20.262

No

No

C

7

3.085

76.713

Increase

24.863

Yes

Yes

Unpaved

T

16

4.122

216.618

Increase

52.557

Yes

Yes

-131.652

Increase

50.289

Yes

Yes

C

18

1.922

36.678

Increase

19.087

No

Yes

Combined

T

30

5.032

118.845

Increase

23.618

Yes

Yes

-40.967

Increase

20.594

No

Yes

C

25

3.600

55.246

Increase

15.347

Yes

Yes

GA & IN

Two-lane

Paved

T

39

3.695

37.481

Increase

10.143

Yes

Yes

5.319

Decrease

11.529

No

No

C

26

3.217

45.205

Increase

14.052

Yes

Yes

Unpaved

T

38

1.390

12.718

Increase

9.148

No

No

7.890

Decrease

10.369

No

No

C

49

2.330

22.373

Increase

9.601

Yes

Yes

Combined

T

77

3.730

25.534

Increase

6.845

Yes

Yes

4.435

Decrease

7.789

No

No

C

75

3.926

31.360

Increase

7.988

Yes

Yes

T = Treatment sites resurfaced with safety edge.

C = Comparison sites resurfaced without safety edge.

The safety edge effect shown in the results tables is the ratio between the before-to-after change in crash frequency for the treatment sites and the before-to-after change in crash frequency for the comparison sites. This formulation of the safety effect was derived from the multiplicative nature of crash modification factors (CMFs), as shown in equation 3 and equation 4:

This figure is a multiple line graph that shows the six safety performance functions (SPFs) developed for the Georgia data. The x-axis has a range in annual average daily traffic (AADT) of 0 to 25,000 vehicles per day, and the y-axis has a range of 0 to 8 crashes per mile per year. All of the lines are slightly curvilinear (depicting logarithmic or exponential growth) and increasing. The order of magnitude in lines is given by: (1) total crash severity for two-lane roadways with paved shoulders, (2) total crash severity for two-lane roadways with unpaved shoulders, (3) total crash severity for multilane roadways with paved shoulders, (4) fatal and injury crash severity for two-lane roadways with unpaved shoulders, (5) fatal and injury crash severity for two-lane roadways with paved shoulders, and (6) fatal and injury crash severity for multilane roadways with paved shoulders. (3)

or

This figure is a multiple line graph that shows the four safety performance functions (SPFs) developed for the Indiana data. The x-axis has a range in annual average daily traffic (AADT) of 0 to 15,000 vehicles per day, and the y-axis has a range of 0 to 5 crashes per mile per year. All of the lines are slightly curvilinear (depicting logarithmic or exponential growth) and increasing. The order of magnitude in lines is given by: (1) total crash severity for two-lane roadways with paved shoulders, (2) total crash severity for two-lane roadways with unpaved shoulders, (3) fatal and injury crash severity for two-lane roadways with paved shoulders, and (4) fatal and injury crash severity for two-lane roadways with unpaved shoulders. (4)

The before-to-after percent change in crash frequency can be converted to a CMF for this calculation by dividing by 100 and adding a value of 1. Similarly, the final CMF for the safety edge can be converted back to a percent change by subtracting the ratio from 1 and multiplying by 100. When the increase in crashes with resurfacing was greater at the comparison sites than at the treatment sites, an indication that the safety edge treatment was effective, the safety edge effect is shown as a positive value. A precision estimate of the ratio was calculated and used to generate a confidence interval of the ratio. Confidence intervals excluding the value 1 indicate statistical significance. For instance, the safety edge effect for Georgia two-lane roadways with paved shoulders shown in table 17 is calculated by first converting the before-to-after changes to CMFs and then taking the ratio: (1+13.12/100)/(1+22.60/100) = 1.1312/1.2260 = 0.927. This CMF represents a (1-0.927)*100 = 7.73 percent decrease in crashes. Since both confidence intervals (i.e., 7.73±(9.60)*1.645 and 7.73±(9.60)*1.96, where 1.645 and 1.96 are critical confidence level values) contain the value 1, the estimate of the safety edge effect is not significant.

The EB results indicate that for all two-lane sites in Georgia and Indiana, the safety edge effect was 5.7 percent for total crashes, 6.3 percent for run-off-road crashes, and 5.6 percent for drop-off-related crashes. While none of these results is statistically significant, they do show a small, consistent benefit of the provision of the safety edge on rural two-lane highways.

When the results are examined separately for the two shoulder types (sites with paved shoulders having widths of 4 ft or less and sites with unpaved shoulders) the use of the safety edge shows more benefit for paved shoulder sites than for unpaved shoulder sites. The safety edge effect for sites with paved shoulders is 9.5 percent for total crashes, 14.2 percent for run-off-road crashes, and 12.6 percent for drop-off-related crashes. The results for run-off road crashes are statistically significant at the 10 percent significance level, but the result for total crashes and drop-off related crashes are not statistically significant. For sites with unpaved shoulders, the safety edge effect was 6.5 percent for total crashes, 4.8 percent for run-off-road crashes, and 4.5 percent for drop-off-related crashes. None of these results is significant.

The expectation was that the use of the safety edge treatment would produce larger benefits on highways with unpaved shoulders, since potential drop-offs at such sites are closer to the travel lanes than on highways with paved shoulders and therefore are expected to be driven over more frequently. However, the sites with unpaved shoulders in both States had much lower ADT than the sites with paved shoulders, and the lower numbers of crashes in the before and after resurfacing periods undoubtedly affected the effectiveness estimates.

In considering the States individually, Georgia sites showed a safety edge effect of 6.8 percent for total crashes, 7.9 percent for run-off-road crashes, and 6.6 percent for drop-off-related crashes. None of the results were statistically significant. Indiana sites had safety edge effects of ‑0.2 percent for total crashes, -13.5 percent for run-off-road crashes, and -19.7 percent for drop-off-related crashes. The negative results are not statistically significant at the 10 percent significance level. The results for Indiana sites were affected by very low numbers of crashes in the before period.

Overall results for the EB evaluation are summarized in table 26 and compared to interim results obtained from analyses conducted 1 and 2 years after resurfacing. The analysis for data including 3 years after resurfacing, which are presented in this section of the report, includes additional comparisons because shoulder types and the two States were combined and compared. Fifty-six of the 81 results for year 3 showed positive safety edge effects; however, only 11 of these positive safety effects were statistically significant. While 25 of the observed effects were negative (e.g., comparison sites had fewer crashes than treatment sites), only 4 of these results were statistically significant.

Table 26. Summary of safety effects from year 3, year 2, and year 1 results for before-after EB safety evaluations.

Direction of safety effect

Statistically significant safety effect?

Number of cases

Year 3 analysis results

Year 2 analysis results

Year 1 analysis results

Positive

Yes

11

8

2

Positive

No

45

14

13

Negative

Yes

4

7

6

Negative

No

21

7

15

Total

81

36

36

The magnitude of the effects also changed with the addition of the year 3 data. The safety effects from the year 3 evaluation were smaller and less variable than the year 1 or year 2 results. The overall impact of the safety edge was expected to be small, since drop-off-related crashes are usually only a small percentage of the total non-intersection crashes on rural roads. The year 3 results presented above follow this trend and therefore are considered more reliable than the earlier results. However, in some cases the smaller magnitude of the safety edge effect makes it more difficult for the effect to be statistically significant.

Total crashes on all sites mainly increased; some of this increase may be due to a resurfacing effect that was very evident in the year 1 results but less so in later years.

The year 3 evaluation results presented above vary in magnitude and statistical significance. The overall evaluation results for total crashes in Georgia and Indiana combined show an average safety edge treatment effect of 5.7 percent. In other words, the sites treated with the safety edge appear to have lower crash frequencies after resurfacing than sites not treated with the safety edge. Although not statistically significant, this seems to be the most appropriate overall effectiveness measure for the safety edge treatment from the EB evaluation. The lack of statistical significance for this result is not surprising given the small magnitude of the effect.

Two trends were evident in the EB analysis of run-off-road and drop-off-related crashes. First, the safety edge treatment generally appears to have had a positive effect on safety for all site types except for sites with unpaved shoulders in Indiana. This variability in results has not been fully explained. Second, the negative safety edge effects for Indiana sites with unpaved shoulders may be explained by low frequencies of drop-off-related crashes on comparison sites in the period before resurfacing. The safety edge effect was statistically significant only for Indiana sites with unpaved shoulders (negative effect).

Georgia sites with paved shoulders showed safety edge treatment effects of approximately 14 percent for run-off-road crashes and 10 percent for drop-off-related crashes. Indiana sites with unpaved shoulders had safety edge effects of -31 to -47 percent for run-off-road crashes and -45 to -80 percent for drop-off-related crashes, and these effects were statistically significant. When data from both States were combined, the safety edge treatment effects for paved shoulders were 14 percent for run-off-road crashes and 12.6 percent for drop-off-related crashes. The effect for run-off-road crashes was statistically significant at the 10 percent significance level. The treatment effects for sites with unpaved shoulders were 5 percent for run-off-road crashes and 4.5 percent for drop-off-related crashes. These small, non-significant effects are probably influenced strongly by the Indiana sites with unpaved shoulders.

The effects for run-off-road and drop-off-related crashes are larger than the effects for total crashes in absolute magnitude but vary in sign and statistical significance. These evaluation results for run-off-road and drop-off-related crashes appear less stable, and thus less reliable, than the results for total crashes. Although not statistically significant, the single most reliable estimate of the effectiveness of the safety edge treatment is the 5.7 percent reduction in total crashes observed for two-lane highways in the combined data for sites with both paved and unpaved shoulders in both Georgia and Indiana (see the last row in table 17).

There are several potential biases and limitations that may influence these results, including the following:

  • There were some observed differences between treatment and comparison sites for the period before resurfacing which could confound the analysis results (see sections 3.1 and 4.3.1).
  • The sites with unpaved shoulders, where the safety edge treatment was expected to be most effective, also had the lowest crash frequencies. This increased the variability in the data and made the statistical test less powerful.

4.3.3 Cross-Sectional Analysis

A cross-sectional analysis of the crash data for the period after resurfacing at the treatment and comparison sites was conducted to directly compare the safety performance. This cross-sectional analysis is analogous to the analysis of safety differences for the period before resurfacing reported in section 4.3.1 but serves a different purpose. In this analysis, any observed difference in safety performance between the treatment and comparison sites is interpreted as an effect of the safety edge treatment. This interpretation should be made cautiously because, as noted in sections 3 and 4.3.1, there are other differences between the treatment and comparison sites that may affect the comparison.

The cross-sectional comparison of data for the period after resurfacing was conducted using analysis of covariance to assess the statistical significance of the treatment versus comparison site effects. This analysis was conducted for each State/roadway type/shoulder type combination with PROC GENMOD in SAS®.(3) Traffic volume and site type (treatment versus comparison) were the main factors of interest in the analysis. Lane width was also considered but was not found to be statistically significant. The analysis was conducted with the same negative binomial modeling techniques described in the discussion of SPFs in section 4.2.

The safety edge treatment effect and its standard error were calculated for each target crash type and adjusted for any covariates. The results are presented in table 27. The significance and p‑value for the treatment versus comparison site effects are also provided.

Where blank lines are shown in the table, the regression model did not converge, so no model could be developed. Table 27 shows that there were 44 models that converged for the final analysis. This is an improvement on the year 1 and year 2 analysis, for which only 20 and 35 models converged, respectively. Thus, as additional years of data have become available, more models have been obtained in the cross-sectional analysis. Table 27 shows that the crash frequencies for the treatment sites after resurfacing were generally lower than for the comparison sites, indicating that the safety edge treatment was effective. Statistically significant results for the safety edge effect (treatment versus comparison sites) were obtained for 19 of the 44 models shown in the table. In 15 of these cases, the safety performance of the treatment sites was better than the comparison sites, indicating that the safety edge was effective. However, in four cases (three of which were on two-lane highways with unpaved shoulders in Georgia), the safety performance of the comparison sites was better than the treatment sites.

In summary, the cross-sectional analysis results are similar to the results of the EB analysis, suggesting that the safety edge treatment is effective in reducing crashes for sites with paved shoulders and for sites in Indiana with unpaved shoulders. However, results for sites in Georgia with unpaved shoulders did not show that the safety edge was effective in reducing crashes.

The potential biases and limitations of this analysis are as follows:

  • There were some observed differences between treatment and comparison sites for the period before resurfacing which could confound the analysis results (see sections 3.1 and 4.3.1).
  • The sites with unpaved shoulders, where the safety edge treatment would be expected to be most effective, had the lowest crash frequencies, which increased the variability in the data and made the statistical test less powerful.
  • The cross-sectional approach does not explicitly compensate for regression to the mean.

Table 27. Cross-sectional analysis of safety edge treatment effect for the period after resurfacing.

State

Roadway type

Shoulder type

Crash type and severity level

Number of site-years

Intercept

AADT effect

Treatment effect

Dispersion parameter

R2LR%

Safety edge effect2

(%)

Coefficient

Standard error

p-Value

Statistically significant?1

Coefficient

Standard error

p-Value

Statistically significant?1

GA

Multilane

Paved

TOT

51

-13.212

1.542

0.309

0.000

Yes

-0.655

0.305

0.032

Yes

0.282

48.4

48.1

FI

51

-12.940

1.360

0.332

0.000

Yes

-0.293

0.257

0.254

No

0.027

30.4

25.4

PDO

51

-14.627

1.656

0.372

0.000

Yes

-0.703

0.345

0.042

Yes

0.404

44.0

50.5

rorTOT

51

-19.840

2.114

0.329

0.000

Yes

-0.946

0.201

0.000

Yes

0.010

66.5

61.2

rorFI

51

-15.748

1.562

0.391

0.000

Yes

-0.410

0.244

0.092

Yes

0.010

20.3

33.6

rorPDO

51

-23.547

2.462

0.308

0.000

Yes

-1.280

0.240

0.000

Yes

0.010

61.8

72.2

doTOT

51

-19.432

2.029

0.446

0.000

Yes

-0.882

0.219

0.000

Yes

0.010

48.8

58.6

doFI

51

-18.509

1.841

0.380

0.000

Yes

-0.610

0.121

0.000

Yes

0.010

23.6

45.7

doPDO

51

-20.271

2.061

0.552

0.000

Yes

-1.130

0.354

0.001

Yes

0.010

34.5

67.7

Two-lane

Paved

TOT

132

-8.695

1.104

0.121

0.000

Yes

0.111

0.183

0.545

No

0.178

51.0

-11.7

FI

132

-7.501

0.880

0.182

0.000

Yes

-0.197

0.264

0.457

No

0.273

27.4

17.8

PDO

132

-11.162

1.306

0.125

0.000

Yes

0.414

0.193

0.032

Yes

0.136

49.6

-51.2

rorTOT

132

-7.654

0.902

0.161

0.000

Yes

-0.120

0.260

0.645

No

0.279

30.3

11.3

rorFI

132

-5.201

0.546

0.197

0.006

Yes

-0.497

0.355

0.161

No

0.453

11.5

39.2

rorPDO

132

-12.208

1.322

0.216

0.000

Yes

0.342

0.328

0.298

No

0.283

29.3

-40.8

doTOT

132

-8.244

0.927

0.163

0.000

Yes

-0.153

0.301

0.611

No

0.287

25.9

14.2

doFI

132

-5.844

0.582

0.193

0.003

Yes

-0.368

0.383

0.337

No

0.589

8.7

30.8

doPDO

132

-14.467

1.518

0.276

0.000

Yes

0.328

0.415

0.428

No

0.209

24.9

-38.9

Unpaved

TOT

159

-10.116

1.253

0.173

0.000

Yes

0.581

0.225

0.010

Yes

0.555

37.0

-78.7

FI

159

-8.599

0.959

0.182

0.000

Yes

0.415

0.226

0.066

Yes

0.267

24.3

-51.5

PDO

159

-12.683

1.498

0.199

0.000

Yes

0.594

0.274

0.030

Yes

0.683

34.0

-81.1

rorTOT

159

-7.229

0.799

0.169

0.000

Yes

0.341

0.222

0.125

No

0.270

17.9

-40.6

rorFI

159

-8.063

0.834

0.193

0.000

Yes

0.275

0.265

0.301

No

0.217

14.4

-31.6

rorPDO

159

-8.374

0.840

0.177

0.000

Yes

0.365

0.255

0.152

No

0.262

12.0

-44.0

doTOT

159

-7.422

0.773

0.196

0.000

Yes

0.379

0.262

0.149

No

0.301

14.2

-46.0

doFI

159

-8.725

0.882

0.202

0.000

Yes

0.297

0.257

0.248

No

0.128

14.0

-34.6

doPDO

159

-8.048

0.728

0.243

0.003

Yes

0.411

0.344

0.231

No

0.402

6.8

-50.9

See notes at end of table.

Table 27. Cross-sectional analysis of safety edge treatment effect for the period after resurfacing-Continued.

State

Roadway type

Shoulder type

Crash type and severity level

Number of site-years

Intercept

AADT effect

Treatment effect

Dispersion parameter

R2LR%

Safety edge effect2

(%)

Coefficient

Standard error

p-Value

Statistically significant?1

Coefficient

Standard error

p-Value

Statistically significant?1

IN

Two-lane

Paved

TOT

63

                       

FI

63

-1.982

0.117

0.647

0.856

No

-0.819

0.582

0.159

No

0.853

4.2

55.9

PDO

63

                       

rorTOT

63

                       

rorFI

63

                       

rorPDO

63

                       

doTOT

63

                       

doFI

63

-13.163

1.184

3.132

0.705

No

-1.599

0.729

0.028

Yes

0.010

3.0

79.8

doPDO

63

                       

Unpaved

TOT

102

-4.887

0.543

0.256

0.034

Yes

-0.069

0.211

0.742

No

0.653

7.4

6.7

FI

102

-5.650

0.493

0.313

0.115

No

-1.215

0.492

0.013

Yes

0.757

9.4

70.3

PDO

102

-5.657

0.594

0.269

0.027

Yes

0.206

0.231

0.373

No

0.697

6.1

-22.9

rorTOT

102

-3.429

0.273

0.396

0.491

No

-0.864

0.394

0.028

Yes

0.527

8.7

57.9

rorFI

102

-3.926

0.219

0.399

0.583

No

-1.689

0.610

0.006

Yes

0.247

10.2

81.5

rorPDO

102

-4.619

0.358

0.499

0.473

No

-0.486

0.417

0.244

No

0.972

3.4

38.5

doTOT

102

-5.486

0.488

0.363

0.178

No

-1.206

0.389

0.002

Yes

0.320

11.1

70.1

doFI

102

-9.490

0.869

0.355

0.014

Yes

-1.970

1.029

0.056

Yes

0.010

9.3

86.0

doPDO

102

-4.672

0.327

0.525

0.534

No

-0.990

0.407

0.015

Yes

0.718

5.4

62.8

NY

Two-lane

Paved

TOT

18

-3.595

0.510

0.186

0.006

Yes

-0.278

0.273

0.307

No

0.117

28.4

24.3

FI

18

-9.373

1.040

0.134

0.000

Yes

0.092

0.144

0.525

No

0.010

57.9

-9.6

PDO

18

-2.241

0.311

0.281

0.268

No

-0.440

0.405

0.277

No

0.214

16.9

35.6

rorTOT

18

-4.255

0.480

0.174

0.006

Yes

-0.128

0.271

0.638

No

0.010

28.5

12.0

rorFI

18

-9.553

1.035

0.101

0.000

Yes

0.004

0.135

0.976

No

0.010

48.7

-0.4

rorPDO

18

                       

doTOT

18

                       

doFI

18

                       

doPDO

18

                       

1 At the 0.10 level.

2 Percent difference between treatment and comparison sites.

TOT = total crashes (all severity levels combined).

Fl = fatal-and-injury crashes.

PDO = property-damage-only crashes.

ror = run-off-road crashes.

do = drop-off-related crashes.

Note: Blank cells represent models that did not converge.

4.3.4 Analysis of Shifts in the Crash Severity Distribution

An analysis was conducted to assess whether the safety edge treatment affected the proportion of severe crashes for specific crash types. This analysis compared fatal and injury crashes as a proportion of total crashes in the periods before and after resurfacing for each State/roadway type/ shoulder type combination. Results of this analysis are presented in table 28. The fatal and injury crash proportions were evaluated for run-off-road crashes, drop-off-related crashes, and all crash types combined. These comparisons were made by estimating the mean difference in proportions and its confidence interval across all sites at a 10 percent significance level.

Table 28. Comparison of proportions of fatal and injury crashes before and after resurfacing.

Crash type

State

Roadway type

Shoulder type

Site type

Average before proportion

Average after proportion

Estimated average difference

Number of sites

Estimated mean difference

Lower 90% confidence limit

Upper 90% confidence limit

Significant at the

0.10 level?

TOT

GA

Multilane

Paved

T

0.362

0.397

0.035

10

0.088

-0.115

0.208

No

C

0.353

0.370

0.017

6

-0.024

-0.272

0.334

No

Two-lane

Paved

T

0.276

0.246

-0.030

15

-0.030

-0.132

0.054

No

C

0.414

0.444

0.031

13

0.042

-0.167

0.296

No

Unpaved

T

0.209

0.476

0.267

20

0.238

0.088

0.480

Yes

C

0.384

0.317

-0.067

24

-0.025

-0.151

0.085

No

All

T

0.245

0.354

0.109

35

0.099

0.006

0.216

Yes

C

0.395

0.366

-0.030

37

-0.009

-0.122

0.095

No

IN

Two-lane

Paved

T

0.116

0.154

0.038

8

0.007

-0.172

0.286

No

C

0.222

0.165

-0.058

6

-0.034

-0.242

0.069

No

Unpaved

T

0.111

0.088

-0.023

7

-0.166

-0.276

0.218

No

C

0.233

0.271

0.038

14

0.044

-0.042

0.165

No

All

T

0.113

0.119

0.005

15

-0.047

-0.188

0.190

No

C

0.230

0.241

0.011

20

0.017

-0.072

0.106

No

NY

Two-lane

Paved

T

0.507

0.334

-0.172

3

-0.181

   

No Test

C

0.407

0.219

-0.188

3

-0.188

   

No Test

All

Two-lane

Paved

T

0.239

0.221

-0.018

26

-0.044

-0.116

0.045

No

C

0.367

0.354

-0.013

22

-0.032

-0.150

0.108

No

Unpaved

T

0.168

0.313

0.145

27

0.156

0.022

0.350

Yes

C

0.329

0.300

-0.028

38

0.008

-0.083

0.079

No

All

T

0.205

0.265

0.059

53

     

No

C

0.343

0.320

-0.023

60

     

No

ROR

GA

Multilane

Paved

T

0.340

0.459

0.119

9

0.168

-0.083

0.357

No

C

0.378

0.154

-0.224

6

-0.250

-0.400

-0.143

Yes

Two-lane

Paved

T

0.331

0.234

-0.097

17

-0.148

-0.297

0.035

No

C

0.386

0.361

-0.025

11

-0.035

-0.467

0.321

No

Unpaved

T

0.309

0.491

0.182

14

0.250

0.021

0.542

Yes

C

0.339

0.366

0.026

19

0.065

-0.126

0.250

No

All

T

0.321

0.355

0.034

31

0.035

-0.125

0.200

No

C

0.357

0.364

0.007

30

0.024

-0.142

0.214

No

See notes at end of table.

Table 28. Comparison of proportions of fatal and injury crashes before and after resurfacing-Continued.

Crash type

State

Roadway type

Shoulder type

Site type

Average before proportion

Average after proportion

Estimated average difference

Number of sites

Estimated mean difference

Lower 90% confidence limit

Upper 90% confidence limit

Significant at the

0.10 level?

ROR

IN

Two-lane

Paved

T

0.063

0.139

0.077

5

0.179

-0.171

0.750

No

C

0.486

0.193

-0.293

6

-0.317

-0.708

0.000

No

Unpaved

T

0.207

0.096

-0.111

8

-0.333

-0.667

0.313

No

C

0.367

0.413

0.046

11

0.042

-0.294

0.387

No

All

T

0.140

0.116

-0.024

13

-0.108

-0.333

0.333

No

C

0.400

0.351

-0.049

17

-0.113

-0.317

0.173

No

NY

Two-lane

Paved

T

0.685

0.519

-0.166

3

-0.156

   

No Test

C

0.628

0.635

0.007

3

-0.023

   

No Test

All

Two-lane

Paved

T

0.267

0.223

-0.044

25

-0.097

-0.229

0.065

No

C

0.435

0.349

-0.086

20

-0.133

-0.367

0.100

No

Unpaved

T

0.266

0.325

0.059

22

0.089

-0.110

0.333

No

C

0.350

0.383

0.034

30

0.061

-0.097

0.217

No

All

T

0.267

0.271

0.005

47

-0.011

-0.134

0.122

No

C

0.381

0.370

-0.011

50

-0.008

-0.142

0.117

No

DO

GA

Multilane

Paved

T

0.410

0.526

0.116

9

0.167

-0.083

0.333

No

C

0.401

0.186

-0.216

6

-0.250

-0.458

-0.057

Yes

Two-lane

Paved

T

0.416

0.313

-0.103

14

-0.200

-0.455

0.089

No

C

0.399

0.308

-0.091

12

-0.152

-0.500

0.250

No

Unpaved

T

0.305

0.562

0.257

17

0.375

0.104

0.563

Yes

C

0.285

0.355

0.070

18

0.151

-0.089

0.333

No

All

T

0.364

0.430

0.066

31

0.100

-0.075

0.292

No

C

0.328

0.337

0.009

30

0.000

-0.199

0.250

No

IN

Two-lane

Paved

T

0.000

0.071

0.071

1

1.000

   

No Test

C

0.238

0.097

-0.141

2

-0.492

   

No Test

Unpaved

T

0.141

0.063

-0.078

4

-0.438

-1.000

1.000

No

C

0.435

0.289

-0.146

11

-0.338

-0.583

0.125

No

All

T

0.075

0.067

-0.008

5

0.000

-1.000

1.000

No

C

0.380

0.236

-0.144

13

-0.375

-0.554

0.000

Yes

NY

Two-lane

Paved

T

0.667

0.000

-0.667

2

-1.000

   

No Test

C

0.667

0.167

-0.500

2

-0.750

   

No Test

All

Two-lane

Paved

T

0.295

0.210

-0.085

17

-0.211

-0.472

0.078

No

C

0.388

0.243

-0.145

16

-0.300

-0.550

0.000

Yes

Unpaved

T

0.236

0.352

0.116

21

0.250

0.000

0.508

No

C

0.340

0.331

-0.009

29

-0.003

-0.217

0.183

No

All

T

0.267

0.277

0.010

38

0.000

-0.181

0.250

No

C

0.358

0.298

-0.060

45

-0.113

-0.289

0.028

No

TOT = Total crashes.

ROR = Run-off-road crashes.

DO = Drop-off-related crashes.

T = Treatment sites resurfaced with safety edge.

C = Comparison sites resurfaced without safety edge.

Note: Blank cells indicate that the test of proportions could not be conducted or that the category had fewer than four sites.

These evaluations were performed with the Wilcoxon signed-rank test, a nonparametric test that does not require that the differences being considered follow a normal distribution. The Wilcoxon signed-rank test was programmed in SAS® using the algorithm developed for the FHWA SafetyAnalyst software.(8) The primary measures of interest presented in table 28 for differences in proportion of fatal and injury crashes are as follows:

  • Average proportion of fatal and injury crashes before resurfacing.
  • Average proportion of fatal and injury crashes after resurfacing.
  • Simple average difference in proportions (after-before).
  • Number of sites included in the analysis.
  • Estimated median before-after effect.
  • Lower confidence limit of median before-after effect.
  • Upper confidence limit of median before-after effect.
  • Summary of statistical significance.

The estimated average treatment effect is the difference between the proportions for the periods before and after resurfacing based only on those sites where the difference is not zero. Since the Wilcoxon signed-rank test uses only those sites with an observed non-zero change in the proportion of fatal and injury crashes, it estimates the median rather than the mean. Consequently, the test results are less influenced by extreme changes in proportions. Cases in which the test of proportions could not be conducted are left blank in the table.

A negative estimated median difference indicates that the proportion of fatal and injury crashes decreased. If the number of sites was less than four, no test was conducted.

The proportion of severe crashes after resurfacing was lower than the proportion of severe crashes before resurfacing in 31 out of 58 cases shown in table 28. Thirteen of the 31 positive results were for sites resurfaced with the safety edge treatment, and 18 were for sites resurfaced without the safety edge treatment. Only 4 of the 58 comparisons of severity proportions were statistically significant; all 4 of these cases were comparison sites. Overall, it appears that the proportion of severe crashes was reduced from before to after resurfacing, but only a few of the results were statistically significant, and there is no apparent difference in the shift in severity distributions between resurfacing with and without the safety edge treatment.

ResearchFHWA
FHWA
United States Department of Transportation - Federal Highway Administration