Safety effect of bound left turn

This paper presents the results of a follow-up study based on a study of safety effects due to the establishment of turning phases (Jensen and Buch, 2017). The paper has been prepared by Trafitec for the Danish Road Directorate. In the previous study, good safety effects were found when establishing bound left turns, but two issues in particular seem to be unresolved:

1. Shared shoulders: The importance of omitting shared shoulders with curb restriction between left-turning and straight lanes when establishing bound left turns.
2. Urban/rural: The importance of establishing bound left turns at intersections in urban and rural areas respectively.

These two factors have been investigated through a before-after accident evaluation by supplementing the number of intersections from the previous study. Thus, an increased number of intersections with bound left turns were included, and the split between left-turning and straight lanes was omitted. The study has been extended to include as many of these intersections as possible. In addition, an increased number of intersections in rural areas are included.

The study includes a total of 84 signalized intersections located in 31 municipalities. The intersections have been rebuilt by municipalities, counties or the state during the years 1999-2014. In total, the study covers the installation of bound left turns in 147 intersection legs. 43 of the 84 intersections are located in urban areas and 41 in rural areas. In 11 out of the 84 intersections, the crossover was omitted in connection with the installation of bound left turns.

Methodology

The evaluation of turning phases is based on police recorded accidents that have occurred at the signalized intersections. A pre-period of 5 years and a post-period of 1-5 years have been used, depending on the time of reconstruction. This means that the evaluation includes a data set of 1,736 accidents and 531 injuries observed at the 84 signalized intersections in the before and after periods.

Safety effects describe differences between the number of observed and expected accidents in a period after the reconstruction of the intersections. The expected number of accidents is the number of accidents that would have occurred during this period if the reconstruction had not been carried out.

The expected number of accidents in the post-period is found by correcting the observed number of accidents in the pre-period for general developments in road safety and random accumulations of accidents and injuries (regression effect). Correction factors for general developments in road safety are estimated by setting up 14 control groups to describe road safety in the 31 municipalities where the study intersections are located.

Correction factors for regression effects are estimated by comparing developments in road safety at the intersections before they are rebuilt with the general developments described for the control groups. Regression effects are taken into account by using average correction factors, where the regression effect is calculated across large groups of intersections. The regression effect is thus not calculated specifically for each individual intersection. This leads to uncertainties when calculating safety effects for a small subgroup of intersections or individual intersections where relatively few accidents have occurred, such as the 11 intersections without crossovers.

On average, a large regression effect was found for left-turn accidents at the intersections, but almost no regression effect for other accidents. This indicates that some intersections were selected for reconstruction because there were many left-turn accidents. In 6 of 84 intersections, no left-turn accidents were recorded at all in the previous period, so some intersections must have been selected for reconstruction for other reasons. This helps to explain why the average assessment may be uncertain when looking at individual intersections or a small group of intersections.

Safety effects are expressed as a percentage, e.g. -50% corresponds to a halving of the number of accidents as a result of the modifications. The effects are also described with the result of a significance test. When the test shows that the effect is statistically significant, the difference between the observed and expected accident figures is considered certain. This means that there is less than 5% probability that the difference is due to random variations.

Summary of results

The introduction of bound left turns is beneficial for road safety. At the study intersections, the total number of personal injury and property damage crashes was significantly reduced by 36%, while personal injury and property damage crashes involving left-turning motor vehicles from the rebuilt intersection legs were significantly reduced by 80%.

The results indicate that it benefits road safety to establish a restricted left turn, regardless of whether or not a curb-restricted shared lane is established in connection with the restricted left turn between lanes for left-turning and straight-ahead drivers. However, the amount of data for intersections without dividing gaps is so limited that the size of the safety effect resulting from the establishment of a restricted left turn without dividing gaps is very uncertain. On the basis of this analysis, it is therefore not possible to determine whether there is a difference in the safety effect of establishing a bound left turn when crossovers are omitted compared to when crossovers are established. The design of crossovers varies greatly, and it also varies greatly how the reconstruction of intersections is carried out to make room for crossovers. It is therefore possible that some ways of establishing crossovers may be more advantageous in terms of safety than others, but this has not been investigated further in this project.

The difference in the safety effects of establishing bound left turns at urban and rural intersections is not as large as the previous study (Jensen and Buch, 2017) suggested. The total number of accidents is reduced most in urban intersections, but this is to some extent offset by the fact that the best effects for injury accidents are achieved in rural intersections. Binding left turns thus benefit road safety in signalized intersections in both urban and rural areas.

A slightly more detailed summary of the results is given below.

Overall effects of installing bound left turns

The introduction of restricted left turns has led to significant reductions of 36% in the number of intersection accidents (personal injury and property damage accidents) and 35% in the number of personal injuries.

The effects are, not surprisingly, due to a decrease in the number of accidents involving left-turning motor vehicles coming from intersections where a restricted left turn has been established in the post-period. Personal injury and property damage accidents of this type have been significantly reduced by 80%. On the other hand, the number of other personal injury and property damage accidents involving road users from the rebuilt cross-legs is unchanged (change of 0 %). The same seems to be true for personal injury and property damage accidents involving road users from other intersection legs only (non-significant increase of 9 %). Not surprisingly, the greatest effect of the introduction of a restricted left turn is seen at those intersections in the study with the highest proportion of accidents involving left-turning motor vehicles from rebuilt intersection legs in the previous period.

Omission of delehelle

Favorable safety effects were found at intersections with and without curb-restricted crossovers between left-turning and straight lanes, both at intersections with and without curb-restricted crossovers. The total number of intersection accidents (personal injury and property damage accidents) is significantly reduced by 48 % after the installation of a bound left turn in the 11 intersections where the crossover is omitted. The number of accidents involving left-turning motor vehicles from rebuilt intersection legs is also significantly reduced by 90 %. The effects are 34 % and 80 % respectively (both significant) in the remaining 73 of the study's 84 intersections where a crossover ramp is installed in connection with the installation of a restricted left turn.

However, the uncertainty of the safety effects for intersections without crossovers is large, which is due to both a limited amount of data and the methodology used for correction of regression effects. The regression effect correction method seems to have increased the differences in safety effects between intersections with and without crossovers, as there are indications that intersections without crossovers are under-corrected for regression effect. Alternative correction factors for regression effect calculated separately for intersections with and without crossovers have therefore been investigated. However, the alternative correction factors for intersections without crossovers are very uncertain due to the limited data basis, and they are therefore only used to nuance the safety effects found. The application of the alternative correction factors results in largely the same safety effects for accidents involving left-turning motor vehicles from rebuilt intersection legs, but somewhat smaller safety effects for all intersection accidents at intersections without crossovers.

On the basis of this study, it cannot therefore be determined whether it is important for the safety effect of establishing a bound left turn if a crossover is omitted. In particular, the effect on other accidents seems to be somewhat uncertain, while the effect on left-turn accidents is fairly similar in the two groups of intersections. It is probably of significance that the group of intersections without crossovers seems to have a different composition than the intersections with crossovers. For example, the proportion of three-legged intersections is larger, and the four-legged intersections often only have a left turn in one leg.

Urban and rural crossings

In the 43 signalized intersections in urban areas, the total number of personal injury and property damage accidents has decreased significantly by 41 % after the installation of bound left turns. In the remaining 41 rural intersections, the decrease is also significant, but only by 26 %. However, the safety effect for accidents with left-turning motor vehicles coming from one of the modified intersection legs is almost the same, with a significant reduction of 82 % in urban intersections and 75 % in rural intersections.

However, when breaking down by accident type, it is worth noting that the best safety effect for injury accidents is achieved at rural intersections. In the study's rural intersections, all injury crashes were significantly reduced by 44 % and injury crashes involving left-turning motor vehicles coming from rebuilt intersection legs by 85 %. At urban intersections, the corresponding reductions are 26 % and 75 % (also significant). The differences are assumed to be related to the fact that restricted left turns have a better effect on accidents involving injuries to drivers and passengers in motor vehicles than on accidents involving injuries to pedestrians, cyclists and moped riders. Pedestrians, cyclists and moped riders account for a higher proportion of casualties at urban intersections. At the same time, left-turning accidents in rural areas are more serious than similar accidents in urban areas, probably due to higher speed levels. On the other hand, for accidents without injuries, the effects of installing restricted left turns are significantly better at urban intersections.

The development of two types of accidents in particular can explain the differences in the overall safety effects for urban and rural intersections. First of all, the development of cross collisions is different. A significant decrease of 30 % was found for personal injury and property damage accidents at urban intersections, while a non-significant increase of 30 % was found at rural intersections. This is a type of accident where typically at least one of the parties involved runs a red light.

Secondly, there has been a difference in the development of accidents involving at least two road users entering the intersection from opposite legs - legs that are positioned opposite each other. The study mainly concerns accidents with accident situation 410, i.e. a left-turning party and a party driving straight ahead. The 75% reduction in injury and property damage accidents at urban intersections appears to be greater than the 64% reduction at rural intersections (both significant). The decisive difference between urban and rural intersections for this type of accident is found for those accidents that do not involve a left-turning motor vehicle coming from a rebuilt intersection leg. A non-significant decrease of 8% was found for urban intersections, while a significant increase of 96% was found for rural intersections. The increase is mainly distributed on three of the 41 rural intersections. The reason for the increase may therefore be linked to local conditions at these locations, e.g. the distribution of green time.

The overall effect for crashes between road users coming from the same leg of the intersection (often rear-end collisions and right-turn crashes) is almost identical at urban and rural intersections. For injury and property damage crashes, non-significant increases of 23 % and 18 % were found.

The differences in safety effects at urban and rural intersections do not seem to be related to differences in intersection geometry. However, differences in road user composition (volume of light road users), speed level and possibly the distribution of green time may have an impact.