Left-turn phase: permissive, protected, or both? A quasi-experimental design in New York City.

The practice of left-turn phasing selection (permissive, protected-only, or both) varies from one locality to another. The literature evidence on this issue is equally mixed and insufficient. In this study, we evaluate the safety impacts of changing left-turn signal phasing from permissive to protected/permissive or protected-only at 68 intersections in New York City using a rigorous quasi-experimental design accompanied with regression modeling. Changes in police reported crashes including total crashes, multiple-vehicle crashes, left-turn crashes, pedestrian crashes and bicyclist crashes were compared between before period and after period for the treatment group and comparison group by means of negative binomial regression using a Generalized Estimating Equations (GEE) technique. Confounding factors such as the built environment characteristics that were not controlled in comparison group selection are accounted for by this approach. The results show that the change of permissive left-turn signal phasing to protected/permissive or protected-only signal phasing does not result in a significant reduction in intersection crashes. Though the protected-only signal phasing does reduce the left-turn crashes and pedestrian crashes, this reduction was offset by a possible increase in over-taking crashes. These results suggest that left-turn phasing should not be treated as a universal solution that is always better than the permissive control for left-turn vehicles. The selection and implementation of left-turn signal phasing needs to be done carefully, considering potential trade-offs between safety and delay, and many other factors such as geometry, traffic flows and operations.