A survival analysis with random parameter approach for assessing temporal instability in treatment effect

Before-after analysis methods in traffic safety often aggregate traffic crashes into crash frequencies using relatively long aggregation time periods, such as a year. The implicit assumption is that the treatment effect is temporally stable over the aggregation period. However, certain "treatments”, such as the COVID-19 pandemic, may result in fast-evolving changes to road safety. By aggregating individual crashes, it is difficult to investigate the temporal characteristics of crashes and capture the potential temporal instability in treatment effect at detailed temporal levels, such as within a year. Therefore, this study exploits the disaggregated nature of crashes and proposes a survival analysis with random parameter (SARP) before-after analysis approach that can flexibly accommodate the temporal instability in treatment effect at various temporal levels. To validate and test the proposed approach, a statistical simulation study and an empirical case study that investigates the safety impact of COVID-19 lockdown in Manhattan, New York, are conducted. The statistical simulation study shows that the SARP method can unbiasedly estimate different patterns of temporally instable treatment effect at various temporal levels. The estimated monthly crash modification factors from the case study display an increasing trend after the largest decrease in the first month after the lockdown, which implies that traffic safety conditions are gradually returning to normal and provides evidence of temporal instability in treatment effect. The proposed SARP approach is promising to investigate the evolving safety impact of emerging technologies in transportation, such as the deployment of connected and autonomous vehicles.

Evidence of sample selectivity in highway injury-severity models: The case of risky driving during COVID-19

Research in highway safety continues to struggle to address two potentially important issues;the role that unobserved factors may play on resulting crash and injury-severity likelihoods, and the issue of identification in safety modeling caused by the self-selective sampling inherent in commonly used safety data (the fact that drivers in observed crashes are not a random sample of the driving population, with riskier drivers being over-represented in crash data bases). This paper addresses unobserved heterogeneity using mixing distributions and attempts to provide insight into the potential sample-selection problem by considering data before and during the COVID-19 pandemic. Based on a survey of vehicle usage (vehicle miles traveled) and subsequent statistical modeling, there is evi-dence that riskier drivers likely made up a larger proportion of vehicle miles traveled dur-ing the pandemic than before, suggesting that the increase in injury severities observed during COVID-19 could potentially be due to the over-representation of riskier drivers in observed crash data. However, by exploring Florida crash data before and during the pan-demic (and focusing on crashes where risky behaviors were observed), the empirical anal-ysis of observed crash data suggests (using random parameters multinomial logit models of driver-injury severities with heterogeneity in means and variances) that the observed increase in injury severity during the COVID-19 pandemic (calendar year 2020) was likely due largely to fundamental changes in driver behavior and less to changes in the sample selectivity of observed crash data. The findings of this paper provide some initial guidance to future work that can begin to more rigorously explore and assess the role of selectivity and resulting identification issues that may be present when using observed crash data.(c) 2022 Elsevier Ltd. All rights reserved.

Amending the heston stochastic volatility model to forecast local motor vehicle crash rates: A case study of Washington, D.C

Modelling crash rates in an urban area requires a swathe of data regarding historical and prevailing traffic volumes and crash events and characteristics. Provided that the traffic volume of urban networks is largely defined by typical work and school commute patterns, crash rates can be determined with a reasonable degree of accuracy. However, this process becomes more complicated for an area that is frequently subject to peaks and troughs in traffic volume and crash events owing to exogenous events – for example, extreme weather – rather than typical commute patterns. One such area that is particularly exposed to exogenous events is Washington, D.C., which has seen a large rise in crash events between 2009 and 2020. In this study, we adopt a forecasting model that embeds heterogeneity and temporal instability in its estimates in order to improve upon forecasting models currently used in transportation and road safety research. Specifically, we introduce a stochastic volatility model that aims to capture the nuances associated with crash rates in Washington, D.C. We determine that this model can outperform conventional forecasting models, but it does not perform well in light of the unique travel patterns exhibited throughout the COVID-19 pandemic. Nevertheless, its adaptability to the idiosyncrasies of Washington, D.C. crash rates demonstrates its ability to accurately simulate localised crash rates processes, which can be further adapted in public policy contexts to form road safety targets. © 2022 The Author(s)