• Federal Highway Administration >
• Publications >
• Research Publications >
• 02089 >
• 06.Cfm >
• Safety Effectiveness of Intersection Left- and Right-Turn Lanes

Safety Effectiveness of Intersection Left- and Right-Turn Lanes

PDF files can be viewed with the Acrobat® Reader®

6. EVALUATION RESULTS

This section of the report presents and interprets the results of the evaluations conducted using the YC, CG, and EB approaches. Detailed results of all evaluations conducted as part of the study are presented in appendix C. This section focuses on those evaluation results that were found to be statistically significant. All tests of statistical significance in this report were performed at the 5 percent significance level (95 percent confidence level) unless otherwise stated.

The evaluation results are tabulated in several different ways in this section. First, results tables for each dependent variable and target area are presented. Then, the same results are tabulated and reviewed by project type.

Evaluation Results for Specific Safety Measures

Tables 28 through 40 present the evaluation results for specific safety measures. The results for four-leg intersections are presented first. There are more statistically significant analysis results for four-leg intersections than for three-leg intersections because there were more treated sites and more accidents per site for four-leg intersections. The tables and the safety measures presented for four-leg intersections include:

• Table 28—total intersection accidents.
• Table 29—fatal and injury intersection accidents.
• Table 30—project-related intersection accidents.
• Table 31—project-related fatal and injury intersection accidents.
• Table 32—total accidents for individual intersection approaches.
• Table 33—fatal and injury accidents for individual intersection approaches.
• Table 34—project-related accidents for individual intersection approaches.

Each table shows all treatment effectiveness measures that were obtained for the YC, CG, and EB approaches. Only those results that were statistically significant are included. There is no table for project-related fatal and injury accidents on individual intersection approaches because none of the evaluation results for that safety measure were statistically significant for the YC approach and, because of low accident frequencies, appropriate regression models to conduct the CG and EB approaches could not be developed.

The tables and the safety measures presented for three-leg intersections include:

• Table 35—total intersection accidents.
• Table 36—fatal and injury intersection accidents.
• Table 37—project-related intersection accidents.

Note: Only statistically significant evaluation results are shown.

Note: Only statistically significant evaluation results are shown.

Note: Only statistically significant evaluation results are shown.

Note: Only statistically significant evaluation results are shown.

^a Because of small accident frequencies, regression models could not be developed for this safety measure. Therefore, the CG and EB evaluations could not be performed.

Note: Only statistically significant evaluation results are shown.

Note: Only statistically significant evaluation results are shown.

Note: Only statistically significant evaluation results are shown.

Note: Only statistically significant evaluation results are shown.

Note: Only statistically significant evaluation results are shown.

Note: Only statistically significant evaluation results are shown.

• Table 38—total accidents for individual intersection approaches.
• Table 39—fatal and injury accidents for individual intersection accidents.
• Table 40—project-related accidents for individual intersection approaches.

These tables are comparable to those presented above for four-leg intersections. There is no table for project-related fatal and injury accidents at three-leg intersections because no evaluation results for that accident type were statistically significant.

Evaluation Results for Specific Project Types

The evaluation results for specific project types are presented in tables 41 through 46. Specifically, the evaluation results by project type for four-leg intersections are presented in the following tables:

• Table 41—projects involving added left-turn lanes.
• Table 42—projects involving added right-turn lanes.
• Table 43—projects involving added left- and right-turn lanes.
• Table 44—projects involving extension of the length of existing turn lanes.

The evaluation results by project type for three-leg intersections are presented in:

• Table 45—projects involving added left-turn lanes.
• Table 46—projects involving added right-turn lanes.

There were no statistically significant evaluation results for projects involving addition of both left- and right-turn lanes or extension of the length of existing turn lanes at three-leg intersections, so no tables of evaluation results are presented for these project types.

The results in tables 41 through 46 are drawn from, and are identical to, the results in tables 28 through 40. However, results for the safety measures involving project-related fatal and injury accidents are omitted because only the YC approach could be evaluated for those cases and, even for the YC approach, very few statistically significant results were obtained due to low accident frequencies analyzed.

The next section addresses the choice among the alternative analysis methods presented in these tables. Then, the results for the specific project types can be interpreted.

Note: Only statistically significant evaluation results are shown.

Note: Only statistically significant evaluation results are shown.

Note: Only statistically significant evaluation results are shown.

Note: Only statistically significant evaluation results are shown.

Note: Only statistically significant evaluation results are shown.

Note: Only statistically significant evaluation results are shown.

Note: Only statistically significant evaluation results are shown.

Note: Only statistically significant evaluation results are shown.

Note: Only statistically significant evaluation results are shown.

Comparison of Alternative Evaluation Approaches

The tables presented above include results from the YC, CG, and EB approaches. For example, table 41 presents the results of 30 before-after evaluations for projects involving added left-turn lanes at four-leg intersections. Of these 30 evaluations, there are:

• Fourteen evaluations for which all three evaluation approaches provided statistically significant results.
• Four evaluations for which only the YC and CG approaches provided statistically significant results.
• One evaluation for which only the YC and EB approaches provided statistically significant results.
• One evaluation for which only the CG and EB approaches provided statistically significant results.
• Two evaluations for which only the YC approach provided statistically significant results.
• One evaluation for which only the EB approach provided statistically significant results.
• Five evaluations for which none of the approaches provided statistically significant results.

In any evaluation for which more than one approach provides statistically significant results, a key issue is to determine which results should be used.

The discussion of the three evaluation methods in section 5 of this report makes clear that, on conceptual and theoretical grounds, the EB approach appears to be the most desirable of the three approaches. The primary reason for this is that, among the three approaches, only EB can account for regression to the mean. When comparing the CG and YC methods, the CG method is most desirable on conceptual and theoretical grounds because it uses a group of comparison sites, rather than a single site, to determine what would have happened at the treatment site had the improvement not been made. The use of multiple comparison sites should reduce the variance of the treatment effect and provide more accurate results. Thus, we began the study with the idea that the three evaluation approaches, in descending order of appropriateness, were EB, CG, and YC.

The results in tables 41 through 46 have been reviewed for confirmation of our initial expectations concerning the suitability of the evaluation approaches. Table 46 presents a summary of the frequency with which various types of results were obtained.

Table 47 is interpreted as follows. First, the table shows that, for the 110 analyses performed, there were 46 statistically significant results for the EB approach, 45 for the CG approach, and 34 for the YC approach. While not definitive, this result is consistent with the theoretical expectation that the EB and CG approaches are preferable to the YC approach.

Second, for 32 cases where statistically significant results were obtained with the EB approach and at least one of the other approaches, the project effectiveness determined with the EB approach was lower than with the YC and CG in 18 cases and was higher in only 6 cases. The generally lower project effectiveness estimates obtained with the EB approach are consistent with the approach being less affected by regression to the mean than the YC and CG approaches.

Both of these observations from table 47 appear to confirm that the EB approach is the most suitable approach, followed by the CG approach, and then the YC approach. These findings support the use of the EB results in favor of the CG and YC results whenever the EB results are statistically significant. When the EB results are not statistically significant, the choice of which results to report is complex. One could:

• Use the EB results, even though the results are not statistically significant.
• Use the statistically significant CG or YC results, even though the results may be subject to regression to the mean.
• Report inconclusive results because no completely satisfactory result was obtained.

In the cases where the EB result was not statistically significant but the YC or CG result was statistically significant, we reviewed both the nonsignificant EB result and the significant YC or CG result. On engineering grounds, we generally found the significant YC or CG results to be more credible than the nonsignificant EB results. For example, at four-leg urban unsignalized intersections where left-turn lanes were added, the CG analysis shows a statistically significant decrease of 27 percent in total accidents, while the EB analysis shows a statistically non-significant decrease in total accidents of 0.1 percent. Both results are based on a limited sample of nine improved sites. The EB result suggests that installing left-turn lanes at urban unsignalized intersections has no safety benefit. By contrast, the 27 percent effectiveness estimate from the CG analysis for added left-turn lanes at four-leg urban unsignalized intersections is very consistent with the EB effectiveness estimate of 28 percent for four-leg rural intersections. We are not prepared to believe that this project type reduces total intersection accidents by 28 percent at rural unsignalized intersections, but has no effect at urban unsignalized intersections. However, it is also evident that the analysis results for this case are based on a very limited sample size and that a further evaluation with a larger sample of improved sites would be desirable.

^a based on these evaluations included in Tables 40 through 45

For the reasons presented above, tables of final evaluation results have been prepared by applying the following rules to the results in tables 41 through 46:

• Use the effectiveness measure determined from the EB approach, if it is statistically significant.
• If the effectiveness measure determined from the EB approach is not statistically significant, but the effectiveness measure from the CG approach is statistically significant, use the CG result.
• If the effectiveness measures from both the EB and CG approaches are not statistically significant, but the effectiveness measure from the YC approach is statistically significant, use the YC result.

Projects Involving Added Left-Turn Lanes

Table 48 presents final evaluation results for projects involving added left-turn lanes at four-leg intersections. These results were derived from the results presented in table 41 using the guidelines for choice of evaluation approach presented above. All of the results in Table 48 are presented as percentage changes in accident frequency for installing one turn lane. Table 49 presents comparable effectiveness estimates for projects involving added left-turn lanes at three-leg intersections.

Each entry in the tables is presented in the format:

Percentage change ± standard error of percentage change

The percentage change is normally a negative value that represents the mean reduction in accident frequency that is expected to result from a specific type of improvement at a specific type of intersection. The standard error is a measure of the precision of the mean percentage change in accident frequency. The smaller the standard error, the smaller the magnitude of site-to-site and year-to-year variations in results would be expected. The standard error does not directly provide a confidence interval for the mean percentage change. In fact, as shown in the tables in appendix C of this report, the actual confidence intervals for the mean percentage change are asymmetrical (i.e., the width of the confidence interval below the mean is not the same as that above the mean). Thus, the interval containing one standard error on either side of the mean does not necessarily represent any particular proportion of the variation in the mean. Nevertheless, the standard error shown in tables 48 and 49 is useful as a measure of the relative precision of each result.

For rural unsignalized intersections with two-way stop control, installation of a major-road left-turn lane was found to reduce total accidents at four-leg intersections by 28 percent. The corresponding reduction in fatal and injury intersection accidents was slightly larger, at 35 percent. In general, the effectiveness estimate for installing a left-turn lane was higher for the approach on which the turn lane was installed than for the intersection as a whole. Accident frequency was reduced by 55 percent for total accidents and by 61 percent for fatal and injury accidents on the specific intersection approach where the turn lane was installed.

For newly signalized four-leg intersections, the effectiveness of adding a left-turn lane appears to be slightly larger than at unsignalized intersections for total intersection accidents and slightly smaller than unsignalized intersections for individual intersection approaches.

Table 48 shows that the effectiveness of adding a major-road left-turn lane at an unsignalized intersection in an urban area is about the same as at a rural unsignalized intersection, although the urban result is based on a limited sample size. The effectiveness of adding a left-turn lane at an urban signalized intersection is a 10 percent reduction in total intersection accidents, which is substantially smaller than for urban unsignalized intersections.

The effectiveness measures for total intersection accidents in Table 48 address installation of a turn lane on a single major-road approach. If turn lanes are installed on both major-road approaches, the effectiveness measure for total intersection accidents would be expected to increase as follows:

(50)

where: E₂ = accident reduction effectiveness for adding turn lanes on two major-road approaches to an intersection

E₁ = accident reduction effectiveness for adding a turn lane on one major-road approach to an intersection

Equation (50) indicates that the second turn lane is effective in reducing only those intersection accidents not reduced by the first turn lane. Thus, the value of E₂ is always less than twice the value of E₁. Equation (50) is applicable only to the effectiveness measure for total intersection accidents, not those for accidents on individual intersection approaches.

For three-leg intersections, table 49 shows that total intersection accidents decreased by 44 percent with the addition of a major-road left-turn lane at rural unsignalized intersections and by 33 percent at urban unsignalized intersections.

The effectiveness of left-turn lanes in reducing accidents was generally higher for individual intersection approaches than for the intersection as a whole and generally higher for project-related accidents than for all accidents.

The results shown in table 48 are reasonably consistent with previous evaluations of left-turn lane installation. Table 3 shows a broad range of effectiveness measures for left-turn lane projects at unsignalized intersections—a reduction in total intersection accidents from 18 to 76 percent. Most of these projects were constructed at rural unsignalized intersections. The comparable result from table 48 is an accident reduction of 28 percent. While 28 percent is in the range from 18 to 76 percent reported in the literature, almost any credible evaluation result would also be in this range.

A more relevant comparison can be made with the results of the expert panel review of previous studies reported by Harwood et al.⁽²⁵⁾ This expert panel, in reviewing the literature, made estimates of the effectiveness of installing left turns at rural two-lane highway intersections. For four-leg unsignalized intersections, the expert panel estimated an effectiveness of 24 percent for major-road left-turn lane installation, while this study estimated 28 percent; thus, the results of the current study are quite comparable to previous studies. For three-leg unsignalized intersections, the expert panel estimated effectiveness of 22 percent for major-road left-turn lane installation, while the current study estimated 44 percent; thus, for three-leg unsignalized intersections this study estimates substantially more effectiveness than previous studies. It should be kept in mind that none of those previous studies used the formal evaluation approaches that have been used in this study.

Table 3 shows a range of effectiveness measures for installation of left-turn lanes at signalized intersections, most of them in urban and suburban areas, ranging from 6 to 70 percent. The effectiveness measure for urban signalized intersections found in this study is an accident reduction of 10 percent, which falls in the lower end of this range. For rural signalized intersections, the expert panel estimated the effectiveness of installing a left-turn lane as 18 percent.⁽²⁵⁾ No comparable effectiveness estimate was developed in this study, but the effectiveness of left-turn installation at urban signalized intersections was estimated as 10 percent.

Many of the results in the current study show lower effectiveness estimates for improvements at urban intersections than for comparable improvements at rural intersections.

Projects Involving Added Right-Turn Lanes

Table 50 presents the final evaluation results for projects involving added right-turn lanes at four-leg intersections. In general, the accident reduction effectiveness of installing right-turn lanes for total intersection accidents or total approach accidents is substantially smaller than for installing left-turn lanes. This is to be expected because right-turn collisions are typically less frequent than left-turn collisions. For the most part, statistically significant effectiveness measures for installation of right-turn lanes were obtained only for unsignalized intersections in rural areas and signalized intersections in urban areas.

At rural unsignalized four-leg intersections, right-turn lane projects reduced total intersection accidents by 14 percent and intersection approach accidents by 27 percent. The effectiveness measure of 14 percent for total intersection accidents is higher than the comparable value of 5 percent estimated by the expert panel convened by Harwood et al.⁽²⁵⁾

At urban signalized intersections, right-turn lane projects reduced total intersection accidents by 4 percent and total intersection-approach accidents by 18 percent.

Where right-turn lanes are installed on two major-road approaches to an intersection, the combined effectiveness measure for both turn lanes should be determined using Equation (50).

Table 51 presents the final evaluation results for projects involving added right-turn lanes at three-leg intersections. Only limited results were obtained for this type of project.

Projects Involving Added Left- and Right-Turn Lanes

Table 52 presents final evaluation results for projects involving the addition of both left- and right-turn lanes at four-leg intersections. These projects combine installation of both left- and right-turn lanes on a single approach. For this reason, one would expect the results in table 52 to be between the results in tables 48 and 50; this appears to be the case for total intersection accidents at urban signalized intersections, but not at rural unsignalized intersections. The total effect of installing both left- and right-turn lanes on two major-road approaches can be obtained with Equation (50). There were no statistically significant results for projects involving the addition of both left- and right-turn lanes at three-leg intersections.

There is no obvious method to separate the effects of left- and right-turn lanes in table 52. A preferable method of determining the effects of adding both left- and right-turn lanes is to combine the relevant effectiveness measures from table 48 or 49 with those from table 50 or 51. For example, it can be shown from tables 48 and 50 and Equation (50) that at an urban four-leg signalized intersection, the addition of two major-road left-turn lanes would be expected to reduce total intersection accidents by 19 percent. The addition of two major-road right-turn lanes would be expected to reduce total intersection accidents by 8 percent. The combined effectiveness would be computed as 1- (1 - 0.19) (1 - 0.08) = 0.25, or a 25 percent reduction in total intersection accidents.

Projects Involving Extension of the Length of Existing Turn Lanes

No separate table of final evaluation results is presented for projects involving the extension of the length of existing turn lanes. The available results, which are quite sparse, are presented in table 44.

Table 44 shows that for three projects in which the existing left-turn lanes at urban four-leg signalized intersections were extended, total intersection accidents increased by approximately 30 percent. This increase may result from substantial growth in left-turn volumes at these signalized intersections that was not accounted for in the evaluation. A weakness of this evaluation is that, while growth in the total major- and minor-road average daily traffic volumes was accounted for, no data on the growth in left-turn volumes are available. The decision by the highway agency to extend the length of the left-turn lane suggests that left-turn volumes were growing, but it is not known whether they were growing faster than the major-road volume as a whole.

At rural unsignalized four-leg intersections, for four intersection approaches where existing major-road left-turn lanes were extended in length, the effect of the projects was to reduce total accident frequency on the intersection approach by 43 percent. None of the other safety measures for these projects had statistically significant changes.

No statistically significant analysis results were found for the extension of the length of existing turn lanes at three-leg intersections.

Because the analyses of the extended-turn-lane projects are based on small sample sizes, no overall conclusions have been drawn from the evaluation of these projects.

Supplementary Analysis Results

Two supplementary analyses were conducted to evaluate intersection design and traffic control features that, because of sample size considerations, could be evaluated for some, but not all, intersection types. These supplementary analyses addressed the relative safety effectiveness of curbed vs. painted channelization for left-turn lanes and of protected vs. protected/permissive left-turn signal phasing. These analyses were conducted using the EB approach for three intersection types with sufficient data to make these comparisons.

Table 53 shows that at rural unsignalized intersections there appears to be a definite indication that left-turn lanes with curbed channelization are more effective than left-turn lanes with painted channelization. This appears to be particularly the case for rural four-leg unsignalized intersections in which channelized left-turn lanes reduced accidents by 57 percent while painted left-turn channelization reduced accidents by only 23 percent. By contrast, there appears to be no difference between the safety effectiveness of curbed and painted left-turn channelization for urban four-leg signalized intersections. However, the sample sizes for these comparisons are too small for the results to be definitive.

Table 54 shows a similar evaluation for protected and protected/permissive signal phasing at urban four-leg signalized intersections. The results suggest that there is essentially no effect of the type of signal phasing on the safety effectiveness of left-turn lanes. However, as in the previous analysis, there are too few data to obtain definitive results. Signalized intersections with no separate left-turn phasing were not included in the evaluation because data were available for only two sites without left-turn phasing.

Recommended Accident Modification Factors

AMFs have been developed based on the results presented above for potential use to replace the AMFs for rural intersections presented in tables 2 and 4. In addition, AMFs for turn-lane installation at urban intersections have also been devleoped.

AMFs for installation of left-turn lanes at rural intersections are presented in table 55. The AMFs for STOP-controlled intersections are based on the results of the current study. The AMFs for signalized intersections are those developed by the expert panel convened by Harwood et al.,⁽²⁵⁾ since no results for rural signalized intersections were obtained in the current study.

AMFs for installation of left-turn lanes at urban intersections are presented in table 56. All of the AMFs in the table are based on the results of the current study, except for the AMF for three-leg signalized intersections. Since no results for three-leg signalized intersections in urban areas were obtained in the current study, the AMF of 0.93 in the table was derived by using the same proportional difference between the AMFs for three- and four-leg signalized intersections shown in table 55 (i.e., 0.90 x 0.85/0.82 = 0.93).

Table 57 presents AMFs for installation of right-turn lanes based on the results of the current study. These AMFs, based on results obtained in the current study for rural unsignalized intersections and urban signalized intersections, should be applied to all rural and urban intersections, because no better estimates are available.

It is recommended that the AMFs presented in tables 55 through 57 be used for safety prediction in the FHWA Interactive Highway Safety Design Model (IHSDM) and in other ongoing initiatives such as the FHWA Comprehensive Highway Safety Improvement Model (CHSIM).

Economic Evaluation

Tables 58 through 65 present the results of an economic evaluation of the installation of left-turn lanes at intersections of various types. The primary measure of the cost effectiveness of improvement projects shown in the tables is the benefit-cost ratio (B/C), which is determined as the present value of future accident costs reduced, divided by the estimated cost of constructing the left-turn lanes. When the benefit-cost ratio is greater than 1.0, this indicates that the anticipated benefit of adding a left-turn lane will exceed its cost.

Each table presents an economic analysis for adding left-turn lanes at specific intersection types under specific traffic volume assumptions. The intersection types considered are:

• Rural three-leg unsignalized intersections.
• Rural four-leg unsignalized intersections.
• Urban four-leg unsignalized intersections.
• Urban four-leg signalized intersections.

The traffic volume assumptions are:

• Major-road ADT from 1,000 to 10,000 veh/day for unsignalized intersections.
• Major-road ADT from 10,000 to 40,000 veh/day for signalized intersections.
• Minor-road ADT equal to either 10 or 50 percent of major-road ADT for unsignalized intersections.
• Minor-road ADT equal to either 25 or 50 percent of major-road ADT for signalized intersections.

For each intersection type and traffic volume level, the expected number of accidents per year was estimated from the negative binomial regression models for total intersection accidents presented in appendix B. The AMFs for left-turn installation are those presented in tables 55 and 56. The number of accidents reduced per year by left-turn installation was derived by combining the expected number of accidents per year and the AMF.

The costs of accidents reduced were derived from FHWA estimates for 1994, updated to 2002 using the GDP implicit price deflator. These values are:

• Fatal and injury accidents—$103,000.
• Property-damage-only accidents—$2,300.

The present value of accident costs reduced was derived with the uniform series present worth factor based on the assumptions of a project service life of 30 years and a minimum attractive rate of return (MARR) of 4 percent.

The average cost of installing a single left-turn lane is $85,000 based on estimates from four of the states that participated in this study.

Tables 58 and 59 present the economic evaluation results for installing a single major-road left-turn lane at a rural three-leg unsignalized intersection. These results indicate that left-turn lane installation would become cost effective for a major-road ADT of 4,000 veh/day with 10 percent of the major-road volume on the minor road and at 2,000 veh/day with 50 percent of the major-road volume on the minor road.

Tables 60 and 61 present comparable data for rural four-leg unsignalized intersections. Left-turn lane installation would become cost-effective for a major-road ADT of 3,000 veh/day with 10 percent of the major-road volume on the minor road. With a minor-road volume equal to 50 percent of the major-road volume, left-turn lane installation would be cost effective at all of the major-road volume levels considered.

Tables 62 and 63 present comparable data for urban four-leg unsignalized intersections. Left-turn lane installation would become cost-effective for a major-road ADT of 2,000 veh/day with both 10 and 50 percent of the major-road volume on the minor road.

Tables 64 and 65 present comparable data for urban four-leg signalized intersections. Left-turn lane installation was found to be cost-effective for all combinations of major- and minor-road ADTs considered.