This section of the report describes the process of selecting evaluation sites for the evaluation of left- and right-turn lane projects and summarizes the characteristics of the sites that were used. The first portion of this section documents the types of projects that were evaluated. The next subsection describes the overall process of identifying candidate intersections, including three types of sites: improved or treatment sites, comparison sites, and reference sites. These three types of sites are defined later in this section. The identification and screening of each site type is described and the number and characteristics of sites of each type are summarized.

Evaluation Priorities for Intersection Improvement Types

Based on the results of the literature review and, most especially, on the assessments of the participating state highway agencies, a decision was reached to focus the safety evaluation on projects involving intersection left- and right-turn lanes. In particular, a decision was made to focus the evaluation on the following four project types for which it appeared that sufficient improved sites for an evaluation were likely to be available:

• Installation of a left-turn lane on one or more major-road approaches to an existing intersection where no turn lane was present.
• Installation of a right-turn lane or a right-turn channelizing island on one or more major-road approaches to an existing intersection where no turn lane was present. In some cases, there may have been an existing right-turn channelizing island prior to the project.
• Installation of both left- and right-turn lanes on one or more major-road approaches to an intersection.
• Projects that involved extending the length of an existing left- or right-turn lane, without adding a new turn lane.

The following types of improvement projects were not evaluated:

• Projects in which no left- or right-turn lanes were installed. In particular, candidate projects that involved signal modifications only, such as the addition of exclusive turn phases, were not considered.
• Projects in which through lanes were added on the major road. In particular, corridor improvement projects in which the through roadway was widened along an entire corridor, but turn lanes also were added at selected intersections, were not considered. However, a few projects that installed a two-way left-turn lane (TWLTL) along the major road, thus providing conventional left-turn lanes at one or more intersections, were retained.
• Projects in which existing through lanes were converted to left- or right-turn lanes.
• Projects in which minor-road approaches were realigned to convert two nearby three-leg intersections into a single four-leg intersection.

Identification of Candidate Intersections

Candidate intersections were identified and reviewed as potential evaluation sites in cooperation with the participating state highway agencies. Three types of sites were considered:

• Improved or treatment sites, which were intersections at which one of the project types described above was implemented.
• Comparison sites, which were intersections similar in geometric design, traffic control, and traffic volume characteristics to the improved sites, but were not improved. The objective was to identify matched pairs of improved and comparison sites with similar characteristics.
• Reference sites, which were sites that were not improved, but also were not matched to any particular improved site.

The general characteristics that all study sites were expected to meet were as follows:

• Only three- and four-leg intersections were considered. Multileg intersections were excluded because they typically incorporate unique features that are not representative of most intersections.
• Only intersections with two-way STOP control and signal control were considered. A few intersections with other types of control (e.g., four-way STOP control) were suggested by the participating states, but these other types were not present in sufficient numbers for evaluation.
• Only intersections between public roads were considered. Intersections at which the minor-road leg was a driveway to a shopping center or school were eliminated because it was considered unlikely that reliable traffic volume data for the minor leg by hour of the day exist and those data that do exist are likely to be atypical. However, some three-leg intersections between public roads at which the fourth leg of the intersection is a residential or commercial driveway were retained.

The selection process and characteristics for each type of intersection are discussed below in more detail.

Selection of Improved or Treatment Sites

Improved or treatment sites are sites at which an improvement project was constructed at an intersection. The project types considered were adding left-turn lanes, adding right-turn lanes, adding both left- and right-turn lanes, or extending an existing left- or right-turn lane. This following discussion addresses the identification and screening of candidate projects and summarizes the number and characteristics of the selected sites.

Identification and Screening of Candidate Projects

Candidate projects were identified with the assistance of eight of the participating state highway agencies. The participating states were: Illinois, Iowa, Louisiana, Minnesota, Nebraska, North Carolina, Oregon, and Virginia. Each state identified candidate intersection improvement projects that were constructed from 1994 to 1997. A few projects that were constructed in earlier years (1989-93) or in a later year (1998) were also identified.

The participating states initially suggested nearly 800 candidate intersection improvement projects. These projects were subjected to a screening review that involved reviewing construction plans and project memoranda, reviewing photologs, and, in some cases, visiting the site in the field. From these 800 candidates, a total of 280 improved intersections were selected that met all of the criteria for the study. The reasons for eliminating sites were as follows:

• The site was located at a multileg intersection; only three- and four-leg intersections were included in the study.
• The site had a traffic control other than two-way STOP or signal control.
• The site was not located at a public road intersection. Intersections at which the minor-road leg was a driveway to a shopping center or a school were excluded. Every selected study site had at least one minor-road leg that was a public road.
• Left- or right-turn lanes were added only on minor-road approach(es) to two-way STOP-controlled intersections. Where no traffic signals were present, only projects involving major-road turn lanes were considered. However, at signalized intersections, projects involving added turn lanes or any approach were considered.
• The planned project was never constructed.
• The added turn lane served a new intersection, not an intersection that existed before the improvement.
• Multiple improvements were made as part of the same project, such that the safety effects of the turn-lane improvement would be confounded with the effects of other improvement types. As much as possible, "clean" projects in which a single type of improvement was made were sought; complex projects involving multiple improvements at the same intersection were not considered. The one exception to this criterion was that existing unsignalized intersections where traffic signals were added at the same time the turn lanes were built were retained in the study. There were a substantial number of projects of this type for evaluation and it is certainly of potential interest to highway agencies, although it is unlikely that the safety effects of signalizing the intersection and adding turn lanes can be separated.
• Additional improvements were made within two years before or two years after the turn-lane improvement, such that any evaluation of the turn-lane improvement would be confounded by the other improvements.
• The intersection had unusual features that could have confounded the evaluation of the turn-lane improvement. For example, one intersection at which a left-turn lane was added was eliminated from consideration because there was a railroad crossing running diagonally across the intersection.
• Data needed for before-after evaluation of the project were not available. Geometric design and traffic control data were gathered by the research team and, therefore, were potentially available for any intersection of interest. However, traffic accident and traffic volume data were obtained through the assistance of the participating highway agencies. The data needs of the study for traffic accident and traffic volume data are described in Section 4 of this report. If traffic accident or traffic volume data were not available for a particular site, that site was eliminated from consideration.

These screening criteria were applied during the site-selection phase of the study, which reduced the number of candidate sites from approximately 800 to 388. An additional 108 improved sites were eliminated during the data collection process because unexpected features of the intersection or the project were discovered or because needed data were unavailable.

Number and Characteristics of Improved Sites

As stated above, a total of 280 improved or treatment sites were available for evaluation. The distribution of these sites by state, area type (rural/urban), project type, and project year are described below.

Table 6 presents the distribution of improved sites by state and area type. The tables shows that there are nearly equal numbers of projects at rural and urban sites. Approximately 45 percent of the improved sites (126 out of 280) are located in Illinois; the study was very fortunate that the Illinois Department of Transportation had conducted a substantial number of improvement projects that were suitable for the evaluation. No other state contributed more than 12.5 percent of the sites. An evaluation of sites at which left- or right-turn lanes were added found no major differences in site characteristics between the intersections of specific types in Illinois and those of the same type in other states. Thus, it does not appear to be a source of bias that 45 percent of the improved sites were located in a single state.

Sites were classified as rural or urban based on posted and operating speed, character of adjacent development, and location with respect to population centers, with speed being the single most important factor. Sites with posted speed limits and operating speeds of 88 km/h (55 mi/h) or more were generally classified as rural unless there was good reason based on development or location to do otherwise. Sites with posted speed limits and operating speeds less than 88 km/h (55 mi/h) were generally classified as urban unless there was good reason based on development or location to do otherwise. The urban classification included sites in both urban and suburban areas.

Table 7 presents the distribution of the improved sites at rural intersections by traffic control type and project type. The table shows that, in rural areas, the project types with the largest sample sizes are:

• Added left-turn lanes at existing unsignalized intersections (61 projects).
• Added right-turn lanes at existing unsignalized intersections (41 projects).
• Added left- and right-turn lanes at existing unsignalized intersections (27 projects).

Table 8 presents comparable data for improved sites at urban intersections. The table shows that, in urban areas, the project types with the largest sample sizes are:

• Added left-turn lanes at existing unsignalized intersections (20 projects).
• Added left-turn lanes at existing signalized intersections (43 projects).
• Added right-turn lanes at existing signalized intersections (21 projects).
• Added left- and right-turn lanes at existing unsignalized intersections (12 projects).
• Added left-turn lanes at newly signalized intersections (32 projects).

The other project types in Tables 7 and 8, not listed above, may not be present in sufficient numbers to permit a reliable evaluation.

LTL = Left-turn lane
RTL = Right-turn lane

LTL = Left-turn lane
RTL = Right-turn lane

Table 9 presents not only the number of intersections by area type, traffic control type, and project type, but also the number of added or extended left- and right-turn lanes. The added or extended turn lanes include only major-road turn lanes at unsignalized intersections, but may include both major- and minor-road turn lanes at signalized intersections. The table shows that the 280 improved sites include 411 added or extended left-turn lanes and 185 added or extended right-turn lanes.

Table 10 presents the distribution of improved sites by the year in which the project was constructed. The table shows that 268 of the 280 projects (94 percent) were constructed during the years from 1994 to 1997, inclusive. The earliest project was constructed in 1989 and the latest project was constructed in 1998. Virtually all of the projects were simple enough that their construction was begun and completed during a single calendar year.

Selection of Comparison Sites

Evaluation of the safety effectiveness of the projects implemented at the improved sites requires a method for estimating the changes in safety that would have occurred at the improved sites had the improvements not been made. This is normally accomplished with data from sites that are not improved during the study period.

Later sections of this report present three alternative evaluation approaches that were used during the project. One of the alternative approaches considered relies on one-to-one matching between improved and similar unimproved sites, while two others rely on predictive models developed from groups of unimproved sites. The sites selected as similar to the improved sites through a one-to-one matching process are referred to in this report as comparison sites. The identification and selection of these comparison sites and the number and characteristics of such sites are described below. This is followed by a description of other unimproved sites that were included in the development of predictive models, but were not matched to any particular improved site; such sites are referred to in this report as reference sites, and their selection and characteristics are discussed later in this section of the report.

Identification and Screening of Candidate Comparison Sites

Candidate comparison sites were identified by the research team, with assistance from the participating highway agencies. Screening of candidate comparison sites was conducted both from office and photolog data and in field visits. The criteria for a comparison site to match a particular treatment site were as follows:

• Located in the same state as the improved site.

• Located geographically as close as possible to the improved site. Whenever possible, a matched comparison site was chosen on the same highway or in the same general area as the improved site; however, where a close geographical location was not possible, a similar intersection in a different part of the state was selected.
  
• Same number of intersection legs as the improved site (i.e., both were either three-leg or four-leg intersections).
  
• Same traffic control as the improved site (i.e., both were either two-way STOP-controlled or signal-controlled intersections).
  
• Similar geometrics to the improved site (e.g., if the improved site was a skewed intersection or was located on a multilane highway, then a skewed intersection or an intersection on a multilane highway was selected as the comparison site, if possible).
  
• Similar ADT to the improved site; however, ADT matching was approximate because some comparison sites were selected before the minor-road ADT was known.
  
• No major geometric or traffic control changes during the study period (generally 1988 to 1999).

It was the original intention that the geometrics of the matched comparison site should resemble the geometrics of the improved site in the period before the improvement was made; in other words, the comparison sites would be intersections without major-road turn lanes. This criterion proved impractical because, especially for urban signalized intersections, candidate comparison sites with no turn lanes were very hard to find. Therefore, a decision was made that the geometrics of the matched comparison site should resemble the geometrics of the improved site in either its condition before improvement or after improvement, with matching to the condition before improvement being preferred. In all cases, the matched comparison site must have undergone no major geometric or traffic control improvement during the study period. In other words, if the matched comparison site had major-road left-turn lanes, it must have had those lanes in place during the periods both before and after the project at the improved site.

For improved sites at which both signalization and turn lanes were installed, the matched comparison site was a similar unsignalized intersection that remained unsignalized throughout the study period.

Number and Characteristics of Matching Improved and Comparison Sites

Matched comparison sites were identified for 260 of the 280 improved sites (93 percent). The other 20 sites were sufficiently unique that a satisfactory matching comparison site could not be found.

The characteristics of the 260 pairs of matching improved and comparison sites are summarized in tables 11 through 15, which are analogous to tables 5 through 10 presented above for the improved sites. Tables 11 through 15 represent the characteristics of the 260 improved sites (a subset of the 280 improved sites presented earlier). Section 4 of this report presents further data on the characteristics of the matched improved and comparison sites.

Selection of Reference Sites

As described earlier, a portion of the improved sites' evaluation uses predictive models developed with data from unimproved sites. An advantage of this approach is that one-on-one matching of improved and unimproved sites is not required. In fact, the larger the data set of unimproved sites on which predictive models are based the better, so it is desirable to have data for more unimproved sites than improved sites.

To increase the sample of unimproved sites, additional reference sites were selected. The selection of the additional reference sites and the number and characteristics of the combined data set of comparison and reference sites is described below.

LTL = Left-turn lane
RTL = Right-turn lane

LTL = Left-turn lane
RTL = Right-turn lane

LTL = Left-turn lane
RTL = Right-turn lane

Identification and Screening of Candidate Reference Sites

Reference sites were intersections similar to the sites that were improved, but not matched to any particular improved site. Reference sites were of the same area types and traffic control types as the improved sites, but must have been free of unusual features and undergone no major geometric or traffic control improvements during the study period.

The research team identified candidate reference sites with assistance from participating State highway agencies. Many of the reference sites were candidate comparison sites that did not match any particular improved site. The candidate reference sites were screened both from office and photolog data and in field visits. Reference sites were retained only if traffic volume and traffic accident data for the site were available.

Number and Characteristics of Comparison and Reference Sites

A total of 40 additional reference sites were selected and included in the data collection effort described in Section 4 of this report. Thus, there were a combined total of 300 unimproved sites available for use in comparison groups and for development of predictive models, 260 matched comparison sites and 40 additional reference sites.

Table 16 presents the distribution of the 300 comparison and reference sites by area type, traffic control type, and state. The comparison and reference sites include only 6 rural signalized intersections, but at least 50 rural unsignalized, urban unsignalized, and urban signalized intersections. Approximately 53 percent of the comparison and reference sites are in rural areas, and 47 percent in urban areas. Approximately 69 percent of the comparison and reference sites are at unsignalized intersections, and 31 percent are at signalized intersections. Approximately 33 percent of the comparison and reference sites are at three-leg intersections, and 67 percent are at four-leg intersections.

Section 4 of this report presents data on the traffic volumes and accident experience at the comparison and reference sites.

Previous | Table of Contents | Next

TFHRC Home | FHWA Home | Feedback United States Department of Transportation - Federal Highway Administration