Identification of Critical Factors for Crash Duration with Consideration of Unobserved Heterogeneity

To reduce non-recurrent congestion, it is necessary to understand of crash duration and its causal factors. Therefore, there have been various efforts to develop crash duration models since 1980s. Specifically, regression type models based on survival analysis have mainly been applied to this area. However, most of them have applied general survival analysis models, in which durations with the same value of the covariates are assumed to have the same distribution. Unlike previous studies, the purpose of this study is to develop a crash duration with the consideration of unobserved heterogeneity. To carry out this purpose, we used a 1-year (2018) crash data set collected from the entire Korean freeway system, and estimated a frailty model for crash duration. As a result, 11 variables were statistically significant. Of these, eight variables exhibited an increasing duration: crashes that occurred in tunnel, crashes that occurred in section without street lighting, crashes with trailer-truck or specially equipped vehicle, crashes involved in completely damage vehicle, vehicle status was in off-road, rollover, or fire after crash. Conversely, crashes in the vicinity of tollgate, crashes including property damage only and crashes reported by freeway service patrol contributed to decreasing crash duration.