Using Synthetic Controls with Interactive Fixed Effects to Examine Changes in Vehicle Speed Related to New York City’s Vision Zero Program
BACKGROUND
New York City passed one of the first Vision Zero policies in the United States with the goal of ending traffic deaths within 30 years. The policy is intended to increase safety with a particular focus on walking and cycling. The policy was accompanied by substantial investment in road infrastructure improvements intended to slow speeds and by extension, reduce the number of traffic injuries and deaths in New York City.

PURPOSE
To date, researchers have only been able to evaluate Vision Zero policies using data on injuries and deaths. Crashes are rare events, and immediate decreases in injuries and deaths may not be detected in traditional medical and transportation data sources, which suffer from under- and misreporting. Vehicle speeds can act as a surrogate measure of safety. Vehicle speed is one of the key risk factors for crashs involving people walking and cycling. Decreased speeds tend to be associated with decreased likelihood of injury and death.

Until recently, large-scale vehicle speed and infrastructure data has not been available or widely accessible. Uber Movement and NYC Open Streets data allow for an evaluation of infrastructural interventions. The sequential rollout of infrastructural improvements, while not random, allows researchers to analyze speed behaviors before and after implementation while comparing to similar roadways.

This research contributes to theory, practice, and policy. To date, Uber Movement speed data has yet to be analyzed and novel methods need to be developed to use the data. Probe vehicle speed data (e.g. from INRIX, HERE, and Streetlight Data) is increasingly available and transportation/health professionals must understand how to properly use it. Similarly, few Vision Zero programs have been rigorously analyzed, and there is little information about whether infrastructure interventions are effective. While some evaluations have analyzed citywide injury and fatality trends, understanding whether and how vehicle speeds change in relation to specific infrastructure designs provides a more robust evaluation of this policy and the engineering treatments.

METHODOLGY
This research compares vehicle speeds on roads improved under New York City’s Vision Zero program to synthetic controls to estimate the effect of infrastructure improvements on vehicle speeds and safety. Roadway corridors that were received infrastructure improvements as part of the Vision Zero program in 2018 were selected as the exposed units. Synthetic controls were created from a weighted combination of roadway links with similar speeds, traffic volumes, crash histories, and road design using an interactive fixed effects model. The synthetic control approach is appropriate as infrastructural improvements are typically phased over time and cannot be completed simultaneously, and it is difficult to meet the parallel trends assumption needed for difference-in-difference analyses. This approach offers a new method to analyze road safety data.

CONCLUSION
Infrastructure improvements can decrease the likelihood of injuries and increase the likelihood of active transportation. Understanding the extent to which infrastructure improvements decrease speed is important for informing ways to decrease perceived and actual safety risk on roadways. This research will assist policy makers in determining how to make infrastructure improvements to improve road safety.

Presenter: David Ederer
Agency Affiliation: Georgia Institute of Technology

Presenter Biographical Statement: [biography]