Skip to content
  • Blog
  • 16th Dec 2021

Dangerous left turns slow by 17% in traffic study leveraging behavioral science

Pedestrian safety is a growing concern. Over the past decade in the US, car crash death rates for pedestrians and cyclists rose 36% and 50% respectively, even as death rates fell for drivers and passengers. Larger cars, distracted drivers, and infrastructure built to maximize speed are putting people at risk.

Deaths are unequally distributed as well. Older adults, people of color, and pedestrians in low-income communities are disproportionately represented in fatal crashes.

For all of these reasons, traffic safety is on the national agenda and cities are looking for solutions.

San Francisco aims to make rapid gains in pedestrian safety under its Vision Zero plan, with a focus on left-hand turns. In 2019, 38% of the city’s traffic deaths were caused when drivers made left turns and didn’t see the person in the crosswalk.

The San Francisco Metropolitan Transit Authority (SFMTA) partnered with the Behavioral Insights Team (BIT) to design and carry out a pilot study that would leverage behavioral science in addressing dangerous left-hand turns. The study produced clear results in speed reduction, which in turn, increases pedestrian safety.

Unsafe left turns are a behavioral issue

Left turns can be stressful. Drivers have to process a lot of information quickly—oncoming traffic, changing signals, cyclists, pedestrians, and more. It’s no wonder that nearly 90% of left-turn collisions cite the driver at fault, according to Vision Zero SF’s internal analysis.

With this information in hand, our team interviewed stakeholders and observed dangerous intersections in San Francisco. We identified several driver behaviors that make left turns unsafe. To name a few:

  • Cutting the corner: initiating turns too early and traversing crosswalks diagonally, which increases exposure to pedestrians
  • Speeding: maintaining already high speeds or only slowing down slightly for the turn
  • Shooting the gap: accelerating significantly to turn through a brief gap in oncoming traffic, leaving little time to watch for pedestrians and cyclists

We observed drivers making left turns early, cutting through the crosswalk diagonally and increasing the risk of hitting a pedestrian


A safe left turn is made at a 90 degree angle with minimal exposure to pedestrians in the crosswalk. We designed interventions to encourage this behavior


The good news is, our research showed that unsafe turning habits are preventable, and behaviorally-informed intersection design helps drivers make better left turns.

Designing a rigorous study to show causality

We pursued physical interventions with the goal to reduce speeds and guide drivers to safety in the moment. Our interventions used relatively low-cost materials that were already at the city’s disposal: rubber speed bumps, enhanced centerlines, and slow turn wedges. 

Waist-high delineator posts (above) and a turn pocket using rubber speed bumps and paint (below) on treatment intersections in San Francisco

We worked with the San Francisco team to identify 10 target intersections. These intersections had the highest crash rates in the city and were scattered throughout downtown.

Because of the number of factors that influence driver behavior, like weather, time of day, or even the grade of the street, we designed a difference-in-difference study which measured speeds before and after interventions were installed. 

Then, we compared the change in speeds between intersections that received the new interventions and those that did not. This quasi-experimental design controlled for variables that were hard to measure (e.g., how “complex” the left turn is) or were unforeseen (e.g., such as dramatic differences in traffic caused by the COVID-19 pandemic). Having both a comparison group and the pre- and post- speed levels gave us confidence that we could isolate the effect of the interventions alone.

SFMTA installed physical left-turn treatments on seven intersections. Using a database of intersections, we identified “pair” left turns that shared similar characteristics and traffic conditions. Four of these “pair” turns served as the control group. 

Before (above) and after (below) installing one of our treatment designs—waist-high delineator posts and rubber speed bumps

The results: slower speeds with the potential to reduce fatalities

Left turn speeds were measured three times over the course of a year, starting in Spring 2020 before the treatments were installed. Speeds fell by a statistically significant amount in the treatment group—and they stayed that way. 

Drivers reliably reduced their left turn speeds by 1.7 mph—a change which can greatly improve pedestrian safety.

We recorded a speed reduction of 1.7 mph (or 17%) on average compared to the control group. Backed by this data, SFMTA is planning to expand left-turn calming treatments to 35 new intersections by 2024.

A few takeaways:

  • Speed reductions are meaningful. A change of just 1.7 mph can have a considerable effect on crash statistics. NYCDOT estimates that left-hand turns in New York City are, on average, just 4 mph faster than right-hand turns (9.3 mph vs. 5.6 mph)—yet they account for three-times more injuries and deaths.
  • Behavioral science can help with road safety challenges. Simple, inexpensive physical interventions prompt drivers to slow down—even drivers who are distracted or experiencing cognitive overload. While educational messages may be forgotten in a moment of stress, physical changes give drivers immediate sensory feedback.
  • Robust study designs pay off. Studies in other major cities have measured left turn speeds at intersections before and after treatment. Our difference-in-difference design included a control group and controlled for many factors over the course of data collection. The COVID-19 pandemic hit San Francisco in the middle of our work, which temporarily changed traffic conditions. In those early days of the pandemic, fewer drivers took to the roads and speeds increased before returning to normal. Our rigorous approach allowed us to isolate the effect of the physical treatments from other influences on traffic speed.

In July 2021, a Vision Zero Resolution was introduced in Congress that would commit the US Department of Transportation to achieve zero roadway fatalities by 2050. But why wait for approval? 

Cities can prevent pedestrian deaths now with practical investments that yield fast, meaningful improvements in safety. For more details about our work with SFMTA, read the report here. If you’d like to discuss how behavioral insights can support safer streets, get in touch