Causal Inference in Transportation Safety Studies: Comparison of the Potential Outcomes and Causal Bayesian Networks

• Activity Number: 629
• Type: Contributed
• Date/Time: Thursday, August 4, 2011 : 8:30 AM to 10:20 AM
• Sponsor: Social Statistics Section
• Author(s): Vishesh Karwa*+ and Aleksandra Slavkovic
• Companies: Penn State University and Penn State University
• Keywords: Causal Inference ; Potential Outcomes ; Causal Bayesian Networks ; Observational Studies ; Transporation Safety ; Night Time crash data

Abstract

Policy questions that often motivate transportation safety studies are causal in nature, yet they are typically answered using observational studies. This paper focuses on exploring the applicability of two causal modeling frameworks- Causal Bayesian Networks and Potential Outcomes - to elucidate causation from observational studies. The question addressed in this paper is that of the minimum levels of pavement marking retrore?ectivity on highways. Nighttime crash data for North Carolina, obtained from the Federal Highway Administration's Highway Safety Information System, appended to pavement marking data, are used. The causal effect of retrore?ectivity on safety of a road segment is estimated. The results of both modeling frameworks are generally consistent with each other. The effect of increased pavement marking retrore?ectivity is to generally reduce the risk of nighttime crashes.