Exploring factors influencing speeding patterns with connected vehicle data for road context-sensitive policy implications

Speeding is a major risk factor for traffic safety, but how it should be measured and analyzed remains limited in two respects. First, prior studies have largely relied on average speeds or binary speeding indicators, which obscure how the intensity of speeding is distributed. Second, road environment effects are often treated as uniform, without accounting for how the same feature may operate differently depending on the broader roadway setting. This study addresses both gaps by examining how road environment factors shape the compositional structure of speeding, that is, the relative proportions of different speeding levels, and by modeling these effects separately across different road settings. Using connected vehicle trajectory data from Central Florida, speeding was classified into four ordinal levels: incidental, ordinary-mild, ordinary-moderate, and severe. Dirichlet regression models were developed to analyze the influence of road environment features, including the built environment, greenery, land use, and roadway design, across three distinct contexts: suburban commercial, suburban residential, and urban general. The results show that road environment features act as a behavioral cue that draws drivers toward specific speeding intensities, and that the same feature can do so differently depending on the road context. These findings underscore the need for context-sensitive speed management and provide a basis for targeted policy recommendations to mitigate higher-intensity speeding.