Counterfactual evaluation of heavy vehicle safety policies on fatal crash rates using recursive discrete polynomial grey models

Abstract

Heavy vehicles play a crucial role in freight transportation. Yet, their crash risks and economic burdens necessitate a thorough investigation of long-term crash trends and an evaluation of safety policies targeting heavy vehicles. The intervention time series method, widely used in policy evaluation without the control group, is limited by its lack of causal inference and reliance on predefined effect assumptions. Thus, this study proposes a counterfactual causal framework using a recursive discrete polynomial time grey model to estimate the causal effects of multiple persistent road safety policies within a single time series. Specifically, the framework defines causal effects as contrasts between potential outcomes. The recursive discrete polynomial time grey model, capable of handling small sample sizes and capturing both linear and nonlinear trends, is introduced for counterfactual outcome prediction in traffic safety policy evaluation. The residual-based nested bootstrap resampling method is adopted to compute the confidence intervals of the estimated causal effects. The proposed framework is demonstrated using the annual fatal crash rates involving heavy vehicles per billion vehicle kilometers traveled from 1989 through 2023 in Queensland, Australia. Three major safety policies targeting heavy vehicles over those years are evaluated: Heavy Vehicle Fatigue Management Laws, Heavy Vehicle Speed Compliance Legislation, and Heavy Vehicle National Law. The findings indicate that these policies have significantly reduced the fatal crash rates involving heavy vehicles, although their effects exhibit temporal fluctuations. Nevertheless, without implementing new and innovative safety policies, the fatal crash rate involving heavy vehicles is likely to increase, underscoring the urgent need for continued policy advancements to enhance the safety of freight transportation systems.