Elucidating the impacts of zero-fare policy on security incident frequency in transit systems

The sparse adoption of fare-free transit (FFT) policies in the U.S. has resulted in limited research and analysis on the impacts of this policy. Consequently, significant controversies surround FFT policies, particularly about the potential of increased security incidents. Yet, there is no scientific evidence to support these claims, and the existing literature predominantly relies on single case studies, exploratory data, anecdotal observations, and opinion surveys. This research is the first to provide a data-driven, statistical evaluation of the impacts of FFT policies on the security incident frequency. We leveraged the reported security incidents in the U.S. National Transit Database and employed the propensity score potential outcome (PSPO) framework (for the first time in transit literature) to simulate a randomized controlled experiment. We then developed two negative binomial models to assess the FFT effects, distinguishing between different types of policies—full agency, partial population, and partial route. Our findings revealed that FFT agencies, on average, experience 26% lower security incidents compared to their counterparts, non-FFT agencies. Specifically, reported security incidents were significantly reduced (by 29%) in full FFT agencies, while the impact on partial FFT agencies was smaller (15-23%) and not statistically significant. The overall decline in reported incidents is primarily driven by a reduction in “Other” security events, while decreases in major incident categories, such as fire, bomb threats, homicide, and assault, are minimal. These findings question the argument that FFT policies lead to an increase in security incidents and suggest that, at least for bus agencies, the opposite could happen. As more agencies implement FFT, more evidence can be collected in this regard, but in the absence thereof, FFT policies should not be seen as a definitive reason for making transit less secure. The model results also offered insights into the factors affecting security incident frequency, including specificis of the agency operations and sociodemographic characteristics of the service area population.