Comparison of safety effect estimates from propensity scores-potential outcomes framework and empirical Bayes before-after method: Case study of adaptive traffic signal control

Abstract

Introduction: The main objective of this paper is to compare the safety effectiveness estimates obtained using the empirical Bayes (EB) before-after and propensity scores-potential outcomes (PSPO) methods. Method: The dataset employed in this study consisted of 338 intersections where adaptive traffic signal control (ATSC) technology was deployed in Pennsylvania. Results: The results revealed that the EB and PSPO methods produce Crash Modification Factors (CMFs) that are not statistically different from each other, which suggests that the two methods provide consistent results in the present study. Nevertheless, there are still some minor differences between the CMF values obtained from the EB method and the CMF values obtained using PSPO. These differences may be attributed to the different statistical basis and different methodological focus between the methods. Overall, the CMF values indicate a slightly higher expected crash frequency of various injury severity levels and for different intersection types (3-leg vs. 4-leg) associated with the implementation of ATSC; however, some of these changes were not statistically significant.