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 3. Preliminary Analysis Results for Field Measurements of Pavement-Edge Drop-Offs

This chapter presents preliminary analysis results for field measurements of pavement‑edge drop‑offs. Field measurements of drop-off heights were made to evaluate the comparability of existing pavement-edge drop-offs for the treatment and comparison sites in the period before resurfacing and to verify that the safety edge treatment does not encourage the development of pavement-edge drop-offs in the period after resurfacing.

Field data for pavement-edge drop-off heights were collected for each participating agency for both treatment and comparison sites in the period before resurfacing and during each year after resurfacing. The field data collection methodology is presented in appendix B. A few sites were resurfaced before field visits could be made. Consequently, these sites were excluded from the analysis of before-period drop-off height data presented in this chapter.

3.1 Comparison of Drop-Off Measurements for Treatment and Comparison Sites Before Resurfacing

A formal assessment of the comparability of the treatment and comparison sites with respect to the presence of pavement-edge drop-offs in the period before resurfacing was undertaken. The measure used for this comparison was the proportion of drop-off heights exceeding 2 inches. This criterion was used based on research indicating that pavement-edge drop-off heights exceeding 2 inches may affect safety.(1) It should be noted that this previous research was conducted on sites without the safety edge treatment.

It would be desirable if the proportion of sites with pavement-edge drop-off heights exceeding 2 inches were similar for the treatment and comparison sites in the period before resurfacing. An analysis to make this comparison was conducted by performing a logistic regression analysis using the LOGISTIC procedure in SAS®.(3) This procedure uses the Fisher scoring method to estimate the statistical significance of differences in proportions between the treatment and comparison sites.

Ideal results for this analysis would have been obtained if the difference between the proportions of drop-off heights exceeding 2 inches for the treatment and comparison sites were not statistically significant at some predetermined significance level. A statistically significant result would be indicated by an odds ratio point estimate that was significantly greater than or less than 1.0 (i.e., the confidence interval for the odds ratio does not contain 1.0). Conversely, for a difference that is not statistically significant, the odds ratio for the difference would contain 1.0. If the odds ratio could not be determined by maximum likelihood due to small sample size or poor variation of responses (i.e., identical responses for each site type or non-overlapping responses between site types), then an exact test was performed and a median unbiased estimate of the odds ratio was provided.

The results of this analysis for each State, roadway type, shoulder type, and treatment type combination, including the frequency and proportion of measurements exceeding 2 inches, the odds ratio point estimate, the odds ratio confidence interval, and the statistical significance of the odds ratio point estimate, are given in table 6. Odds ratio values above 1.0 in this table indicate that comparison sites had a greater probability of experiencing drop-offs exceeding 2 inches than treatment sites.

Table 6. Comparison of the proportions of drop-off heights exceeding 2 inches for the period before resurfacing.

State

Roadway type

Shoulder type

Site type

Drop-off heights that exceed 2 inches

Odds ratio point estimate

Lower confidence limit

Upper confidence limit

Statistically significant at 0.05 level?

Number

Proportion

GA

Multilane

Paved

T

2

0.07

0.909

0.184

6.596

No

C

5

0.06

Two-lane

Paved

T

10

0.03

4.591

2.211

10.259

Yes

C

25

0.14

Unpaved

T

23

0.09

1.557

0.876

2.799

No

C

29

0.13

IN

Two-lane

Paved

T

6

0.04

2.519

0.902

7.642

No

C

10

0.10

Unpaved

T

150

0.39

0.423

0.291

0.608

Yes

C

53

0.22

NY

Two-lane

Paved

T

36

0.38

0.028

0.000

1.620

No1

C

0

0.00

1 Indicates that median unbiased estimate was used.

T = Treatment sites resurfaced with safety edge.

C = Comparison sites resurfaced without safety edge.

The results in table 6 indicate that in the period before resurfacing, there were relatively equal proportions of extreme drop-off heights between treatment and comparison sites for Georgia sites on multilane highways with paved shoulders and two-lane highways with unpaved shoulders. This finding indicates that these two types of sites were relatively well matched in terms of shoulder conditions in the period before resurfacing. By contrast, the findings for Georgia sites on two-lane highways with paved shoulders suggest that there was a statistically significant chance that comparison sites had greater proportions of drop-offs exceeding 2 inches.

For Indiana sites on two-lane highways with paved shoulders, there was a greater proportion of extreme drop-off heights for the comparison sites than for the treatment sites in the period before resurfacing, but the difference was not statistically significant. The opposite was the case for Indiana sites on two-lane highways with unpaved shoulders and for New York sites on two-lane highways with paved shoulders. In these cases, the treatment and comparison sites were not perfectly matched in terms of shoulder conditions in the period before resurfacing. For Indiana, this difference was statistically significant. Some differences of this sort may have been inevitable because resurfacing projects that received the safety edge treatment were not selected based on consideration of the existing shoulder condition. This is a potential confounding factor that should be considered in interpreting the research results.

3.2 Comparison of Drop-Off Measurements between the Before and After Resurfacing periods

The field measurement data for pavement-edge drop-offs were initially reviewed by State, roadway type, shoulder type, and treatment type. Table 7 presents summary descriptive statistics for these measures for each study period. Figure 2 presents histograms for a sample of the distributions and shows the impact of resurfacing for both treatment and comparison sites.

Table 7. Summary of pavement-edge drop-off height measurements.

State

Road type

Shoulder type

Site type

Before resurfacing

After resurfacing (Year 1)

Number of measurements

Drop-off height (inches)

Coefficient of variation %

Number of measurements

Drop-off height (inches)

Coefficient of variation %

Minimum

Mean

Median

Maximum

Standard deviation

Minimum

Mean

Median

Maximum

Standard deviation

GA

Multilane

Paved

T

30

0

0.783

0.750

2.000

0.618

79

59

0.375

1.047

0.875

2.875

0.504

48

C

82

0

0.811

0.750

3.000

0.710

88

86

0.250

1.038

1.000

2.375

0.467

45

Two-lane

Paved

T

291

0

0.546

0.500

3.750

0.611

112

289

0.000

0.960

1.000

2.375

0.495

52

C

178

0

0.912

0.750

4.000

0.912

100

150

0.000

0.887

0.875

1.875

0.471

53

Unpaved

T

270

0

0.881

0.750

3.750

0.695

79

273

0.000

0.941

0.875

2.500

0.495

53

C

229

0

1.076

1.000

4.750

0.804

75

466

0.000

0.945

0.875

2.875

0.556

59

IN

Two-lane

Paved

T

136

0

0.630

0.500

3.500

0.598

95

158

0.000

0.703

0.625

1.875

0.356

51

C

96

0

0.960

0.750

3.250

0.708

74

137

0.250

1.340

1.125

4.250

0.707

53

Unpaved

T

380

0

1.758

1.625

5.125

0.778

44

367

0.250

1.653

1.500

4.500

0.737

45

C

245

0

1.353

1.250

6.875

0.930

69

279

0.125

1.168

1.000

5.250

0.673

58

NY

Two-lane

Paved

T

94

0

1.681

1.500

5.125

1.270

76

77

0.000

1.110

0.875

4.000

0.886

80

C

42

0

0.777

0.750

1.750

0.487

63

83

0.000

1.065

1.000

2.750

0.480

45

State

Road type

Shoulder type

Site type

After resurfacing (Year 2)

After resurfacing (Year 3)

Number of measure-ments

Drop-off height (inches)

Coefficient of variation %

Number of measure-ments

Drop-off height (inches)

Coefficient of variation %

Minimum

Mean

Median

Maximum

Standard deviation

Minimum

Mean

Median

Maximum

Standard deviation

GA

Multilane

Paved

T

65

0.500

1.175

1.000

3.000

0.448

38

65

0.500

1.175

1.000

3.000

0.448

38

C

86

0.250

0.906

0.813

2.500

0.455

50

86

0.250

0.907

0.875

2.500

0.442

49

Two-lane

Paved

T

212

0.000

0.956

0.875

2.250

0.455

48

254

0.000

1.087

1.000

3.375

0.432

40

C

152

0.375

1.166

1.125

2.250

0.356

31

164

0.250

1.104

1.125

2.250

0.372

34

Unpaved

T

238

0.125

1.179

1.000

3.563

0.571

48

259

0.250

1.107

1.000

3.563

0.566

51

C

426

0.000

1.163

1.125

3.250

0.548

47

448

0.000

1.119

1.063

3.250

0.526

47

IN

Two-lane

Paved

T

187

0.000

0.788

0.750

2.250

0.379

48

189

0.125

0.780

0.750

2.250

0.398

51

C

102

0.250

1.456

1.250

4.375

0.857

59

147

0.000

1.344

1.250

3.875

0.609

45

Unpaved

T

370

0.250

1.916

1.750

6.875

0.993

52

373

0.250

1.584

1.375

4.500

0.774

49

C

280

0.000

1.353

1.250

5.500

0.764

56

290

0.125

1.236

1.125

4.500

0.676

55

NY

Two-lane

Paved

T

78

0.375

1.786

1.344

5.125

1.191

67

78

0.375

1.786

1.344

5.125

1.191

67

C

81

0.625

1.446

1.375

3.250

0.497

34

81

0.625

1.446

1.375

3.250

0.497

34

T = Treatment sites resurfaced with safety edge.

C = Comparison sites resurfaced without safety edge.

This figure shows a series of eight bar graphs representing the distribution of drop-off height measurements for each of the four years in the analysis (i.e., the year before resurfacing, the first year after resurfacing, the second year after resurfacing, and the third year after resurfacing) for both treatment and comparison sites of two-lane highways with paved shoulders in Georgia. The range of drop-off heights shown in all graphs is zero to four inches. Each graph resembles a normal curve or distribution with less spread in the data for the periods after resurfacing than for the period before resurfacing. The mean of each of the graphs is around 1 inch, with a slightly increasing mean in the periods after resurfacing.

Figure 2. Graph. Drop-off height measurement distributions for two-lane highways with paved shoulders in Georgia.

In each graph shown in figure 2, the vertical blue line marks a 2-inch drop-off height. The mean drop-off height did not vary between the before and after periods. For almost all roadway type/ shoulder type/treatment type combinations, the coefficient of variation (i.e., relative standard deviation) of drop-off height decreased substantially between before resurfacing and each of the first 2 years after resurfacing but increased again following the second year after resurfacing.

To formally assess whether the safety edge treatment has any effect on pavement-edge drop-offs, a trend analysis evaluating the change in drop-offs from before to after resurfacing was conducted. Specifically, the proportion of drop-off height measurements exceeding 2 inches was evaluated to determine if there were differences between the before and after study periods. This analysis was carried out using the same logistic regression approach presented in section 3.1. However, in this case, the proportions of drop-off heights exceeding 2 inches were compared between the periods before and after resurfacing for each type of site rather than between treatment and comparison sites.

The ideal trend for this analysis would be indicated by a substantial decrease in drop-off height for the first year after resurfacing, possibly followed by a slow increasing trend in the later years back to the drop-off height that existed before resurfacing. To evaluate this trend, all pairwise comparisons between years were evaluated for statistical significance. Four of the comparisons: before versus after year 1, after year 1 versus after year 2, after year 2 versus after year 3, and before versus after year 3 have been summarized.

For before versus after year 1, an odds ratio point estimate less than 1.0 indicates that after year 1 had more drop-off heights exceeding 2 inches than the period before resurfacing. A confidence interval for the odds ratio that does not contain the value 1.0 indicates statistical significance. Since the odds ratios were less than 1.0 in 3 of the 12 cases shown in table 8, the sites in after year 1 generally had fewer drop-off heights exceeding 2 inches than the sites in the period before resurfacing. Also, the three cases when after year 1 had more drop-off heights exceeding 2 inches than the period before resurfacing were not significant. Thus, it appears that resurfacing tends to reduce the proportion of extreme drop-off heights.

Table 8. Comparison of the proportions of drop-off heights exceeding 2 inches between the before and after resurfacing periods.

State

Roadway type

Shoulder type

Site Type

Test

Proportion Period 1

Proportion Period 2

Odds ratio point estimate

Lower 95% confidence limit

Upper 95% confidence limit

Statistically significant at the 0.05 level?

GA

Multilane

Paved

C

Period Before vs AfterY1

0.06

0.06

1.05

0.28

3.92

No

C

Period Before vs AfterY2

0.06

0.03

1.80

0.43

8.99

No

C

Period Before vs AfterY3

0.06

0.08

0.73

0.21

2.39

No

C

Period AfterY1 vs AfterY2

0.06

0.03

0.59

0.12

2.46

No

C

Period AfterY1 vs AfterY3

0.06

0.08

1.44

0.44

5.03

No

C

Period AfterY2 vs AfterY3

0.03

0.08

2.45

0.66

11.68

No

T

Period Before vs AfterY1

0.07

0.07

0.98

0.13

5.35

No

T

Period Before vs AfterY2

0.07

0.08

0.86

0.12

4.25

No

T

Period Before vs AfterY3

0.07

0.09

0.70

0.10

3.27

No

T

Period AfterY1 vs AfterY2

0.07

0.08

1.15

0.29

4.83

No

T

Period AfterY1 vs AfterY3

0.07

0.09

1.40

0.38

5.72

No

T

Period AfterY2 vs AfterY3

0.08

0.09

1.22

0.35

4.44

No

Two-lane

Paved

C

Period Before vs AfterY1

0.14

0

infinity

12.13

infinity

Yes

C

Period Before vs AfterY2

0.14

0.05

3.38

1.49

8.70

Yes

C

Period Before vs AfterY3

0.14

0.03

6.17

2.33

21.32

Yes

C

Period AfterY1 vs AfterY2

0

0.05

infinity

3.24

infinity

Yes

C

Period AfterY1 vs AfterY3

0

0.03

infinity

1.60

infinity

Yes

C

Period AfterY2 vs AfterY3

0.05

0.03

0.55

0.14

1.86

No

T

Period Before vs AfterY1

0.03

0.03

1.11

0.44

2.83

No

T

Period Before vs AfterY2

0.03

0.02

1.85

0.61

6.82

No

T

Period Before vs AfterY3

0.03

0

10.64

2.02

195.83

Yes

T

Period AfterY1 vs AfterY2

0.03

0.02

0.60

0.16

1.86

No

T

Period AfterY1 vs AfterY3

0.03

0

0.10

0.01

0.56

Yes

T

Period AfterY2 vs AfterY3

0.02

0

0.17

0.01

1.19

No

Unpaved

C

Period Before vs AfterY1

0.13

0.06

2.36

1.36

4.10

Yes

C

Period Before vs AfterY2

0.13

0.1

1.29

0.78

2.12

No

C

Period Before vs AfterY3

0.13

0.08

1.68

0.99

2.84

No

C

Period AfterY1 vs AfterY2

0.06

0.1

1.83

1.11

3.04

Yes

C

Period AfterY1 vs AfterY3

0.06

0.08

1.40

0.83

2.38

No

C

Period AfterY2 vs AfterY3

0.1

0.08

0.77

0.48

1.23

No

T

Period Before vs AfterY1

0.09

0.03

2.73

1.28

6.34

Yes

T

Period Before vs AfterY2

0.09

0.13

0.62

0.35

1.10

No

T

Period Before vs AfterY3

0.09

0.09

0.99

0.53

1.85

No

T

Period AfterY1 vs AfterY2

0.03

0.13

4.39

2.13

9.99

Yes

T

Period AfterY1 vs AfterY3

0.03

0.09

2.76

1.28

6.46

Yes

T

Period AfterY2 vs AfterY3

0.13

0.09

0.63

0.35

1.12

No

See notes at end of table.

Table 8. Comparison of the proportions of drop-off heights exceeding 2 inches between the before and after resurfacing periods-Continued.

State

Roadway type

Shoulder type

Site Type

Test

Proportion Period 1

Proportion Period 2

Odds ratio point estimate

Lower 95% confidence limit

Upper 95% confidence limit

Statistically significant at the 0.05 level?

IN

Two-lane

Paved

C

Period Before vs AfterY1

0.10

0.17

0.58

0.25

1.24

No

C

Period Before vs AfterY2

0.10

0.27

0.31

0.13

0.66

Yes

C

Period Before vs AfterY3

0.10

0.1

0.70

0.30

1.52

No

C

Period AfterY1 vs AfterY2

0.17

0.27

1.88

1.01

3.53

Yes

C

Period AfterY1 vs AfterY3

0.17

0.14

0.83

0.43

1.57

No

C

Period AfterY2 vs AfterY3

0.27

0.14

0.44

0.23

0.83

Yes

T

Period Before vs AfterY1

0.04

0

infinity

3.18

infinity

Yes

T

Period Before vs AfterY2

0.04

0.01

8.58

1.44

163.10

Yes

T

Period Before vs AfterY3

0.04

0.02

2.86

0.74

13.75

No

T

Period AfterY1 vs AfterY2

0.00

0.01

infinity

0.15

infinity

No

T

Period AfterY1 vs AfterY3

0.00

0.02

infinity

0.94

infinity

No

T

Period AfterY2 vs AfterY3

0.01

0.02

3.00

0.38

60.92

No

Unpaved

C

Period Before vs AfterY1

0.22

0.11

2.21

1.37

3.61

Yes

C

Period Before vs AfterY2

0.22

0.16

1.48

0.95

2.31

No

C

Period Before vs AfterY3

0.22

0.14

1.68

1.07

2.64

Yes

C

Period AfterY1 vs AfterY2

0.11

0.16

1.49

0.91

2.46

No

C

Period AfterY1 vs AfterY3

0.11

0.14

1.32

0.80

2.18

No

C

Period AfterY2 vs AfterY3

0.16

0.14

0.88

0.56

1.40

No

T

Period Before vs AfterY1

0.39

0.28

1.65

1.22

2.24

Yes

T

Period Before vs AfterY2

0.39

0.42

0.88

0.66

1.18

No

T

Period Before vs AfterY3

0.39

0.30

1.52

1.12

2.06

Yes

T

Period AfterY1 vs AfterY2

0.28

0.42

1.86

1.37

2.54

Yes

T

Period AfterY1 vs AfterY3

0.28

0.30

1.09

0.79

1.49

No

T

Period AfterY2 vs AfterY3

0.42

0.30

0.58

0.43

0.79

Yes

NY

Two-lane

Paved

C

Period Before vs AfterY1

0

0.02

-infinity

-infinity

3.18

No

C

Period Before vs AfterY2

0

0.12

-infinity

-infinity

0.37

Yes

C

Period Before vs AfterY3

0

0.18

-infinity

-infinity

0.23

Yes

C

Period AfterY1 vs AfterY2

0.02

0.12

5.70

1.44

37.92

Yes

C

Period AfterY1 vs AfterY3

0.02

0.18

9.07

2.44

58.83

Yes

C

Period AfterY2 vs AfterY3

0.12

0.18

1.59

0.67

3.89

No

T

Period Before vs AfterY1

0.38

0.31

2.79

1.39

5.84

Yes

T

Period Before vs AfterY2

0.38

0.27

1.68

0.88

3.26

No

T

Period Before vs AfterY3

0.38

0.27

1.72

0.91

3.30

No

T

Period AfterY1 vs AfterY2

0.18

0.27

1.66

0.78

3.63

No

T

Period AfterY1 vs AfterY3

0.18

0.27

1.62

0.77

3.52

No

T

Period AfterY2 vs AfterY3

0.27

0.27

0.98

0.49

1.98

No

T = Treatment sites resurfaced with safety edge.

C = Comparison sites resurfaced without safety edge.

The odds ratio for the treatment sites was less than 1.0 for one out of six cases, indicating that resurfacing with the safety edge treatment is effective in reducing the proportion of extreme drop-off heights. Resurfacing without the safety edge treatment was effective in reducing the proportion of extreme drop-off heights in four of six cases. Additionally, none of the observed odds ratios less than 1.0 and almost all of the observed odds ratios greater than 1.0 were statistically significant.

For after year 1 versus after year 2, an odds ratio point estimate greater than 1.0 indicates that the second year after resurfacing had more drop-off heights exceeding 2 inches than the first year after resurfacing. Since there were more drop-off heights exceeding 2 inches in after year 2 as compared to after year 1 (10 of the 12 cases shown in table 8), there appears to be deterioration of the shoulder condition in the second year after resurfacing. However, only about half of these observed differences in the proportion of drop-off heights exceeding 2 inches were statistically significant at the 5 percent significance level.

For after year 2 versus after year 3, an odds ratio point estimate greater than 1.0 indicates that the third year after resurfacing had more drop-off heights exceeding 2 inches than the second year after resurfacing. Since 7 of the 12 cases shown in table 8 have an odds ratio point estimate of 1.0 (or nearly 1.0), which indicates no change in the proportion of drop-off heights exceeding 2 inches, there appears to be minimal deterioration of the shoulder condition in the third year after resurfacing.

The before period drop-off height data were compared to the after year 3 drop-off height data to determine whether drop-off heights had increased to the levels that existed before resurfacing. For this comparison, an odds ratio point estimate less than 1.0 indicates that after year 3 had more drop-off heights exceeding 2 inches than the period before resurfacing. Since the odds ratios were greater than 1.0 in 7 of the 12 cases shown in table 8, there does not seem to be much evidence to suggest the proportion of high drop-offs after year 3 differs from the before period.

A final comparison of drop-off height data was made between sites resurfaced with and without the safety edge treatment in the third year after resurfacing to determine if the safety edge treatment has any role in development of drop-offs. The results of this analysis are given in table 9. Odds ratio values above 1.0 indicate that comparison sites had more drop-off heights exceeding 2 inches than treatment sites.

Table 9. Comparison of the proportions of drop-off heights exceeding 2 inches between treatment and comparison sites for the final period after resurfacing.

State

Road type

Shoulder type

Site type

Drop-off heights that exceed 2 inches

Odds ratio point estimate

Lower 95% confidence limit

Upper 95% confidence limit

Statistically significant

Number

Proportion

GA

Multilane

Paved

C

2

0.02

0.286

0.040

1.374

No

T

5

0.08

Two-lane

Paved

C

6

0.04

1.034

0.341

2.922

No

T

9

0.04

Unpaved

C

38

0.08

0.796

0.476

1.349

No

T

27

0.1

IN

Two-lane

Paved

C

21

0.14

10.332

3.470

44.394

Yes

T

3

0.02

Unpaved

C

41

0.14

0.384

0.256

0.567

Yes

T

112

0.3

NY

Two-lane

Paved

C

10

0.12

0.382

0.161

0.858

Yes

T = Treatment sites resurfaced with safety edge.

C = Comparison sites resurfaced without safety edge.

The results in table 9 indicate that there were no differences in extreme drop-offs between sites resurfaced with and without the safety edge in Georgia. In Indiana, sites with paved shoulders resurfaced with the safety edge had fewer drop-offs. However, sites with unpaved shoulders showed the reverse trend. In New York, sites resurfaced without the safety edge had fewer proportions of extreme drop-off heights. Taken together, these results are inconclusive.

The analysis of the field measurements of drop-off-heights suggests that resurfacing is effective in reducing the proportion of extreme drop-off heights. It also suggests that resurfacing with the safety edge treatment does not increase the number of extreme drop-off heights and is similar to resurfacing without the safety edge treatment in reducing the proportion of extreme drop-off heights over time.

ResearchFHWA
FHWA
United States Department of Transportation - Federal Highway Administration