The impact of higher speed limits on the frequency and severity of freeway crashes: Accounting for temporal shifts and unobserved heterogeneity

Temporal instability and differences in injury severity between restrained and unrestrained drivers in speeding-related crashes

“…With the use of the simulation, Alogaili and Mannering 78 found that pedestrian-vehicle crashes in the daytime would cause as much as 16.45% less severe injuries compared to the crashes at nighttime, given both crash times having the same other associated factors. Additionally, the application of this simulation has also been adopted by numerous recent crash severity studies to gain a better overview understanding of how two or more crash conditions are different in influencing injury severities 38 , 40 , 80 , 90 . Likewise, this study also adopted this simulation for predictive comparison between restrained and unrestrained driver-injury severities and investigating how injury severity distribution changed over time.…”

“…Not only human elements, but attributes related to vehicle, roadway, traffic, environment, and temporal characteristics also have great variability across the crashes 35 . As can be seen in the Table 1 , to account for unobserved heterogeneity, the application of the random parameter (mixed) model and it’s multiple variants, particularly the model extension that allow for means and variances heterogeneity has been adopted in numerous recent insightful articles such as Alogaili and Mannering 78 , Wang et al 79 , Yan et al 38 , Alnawmasi and Mannering 40 , Se et al 80 , Yan et al 34 , Wang et al 81 , Islam et al 82 , Behnood and Mannering 83 , and Hou et al 84 . The random parameters model with heterogeneity in means and variance was found to be superior than standard random parameters model in crash-injury severity analysis due to its’ the great flexibility in capturing a greater extent of underlying unobserved characteristics, more precise predictions, and better model fit 80 , 84 – 87 .…”

“…This flexibility leads to improved prediction accuracy, better model fit, and more reliable conclusions 35 . Additionally, Table 1 also showed that there are multiple variants of the random parameters model used for the single-vehicle crash-injury severity in the recent years, including random parameters model that allows for possible heterogeneity in means 9 , 14 , 19 , random thresholds random parameters hierarchical ordered probit model 36 , 37 , correlated random parameters with heterogeneity in means 32 , 38 , and random parameters model that allows both heterogeneity in means and variances 4 , 32 – 34 , 39 , 40 .…”

“…Renski et al 63 investigated the influence of speed limit increases on crash injury severity in single-vehicle crashes, using indicators for road segments with speed limit changes from 88.5 to 96.6 kph, 88.5 to 104.6 kph, and 104.6 to 112.7 kph as part of the explanatory variables in their statistical model. On the other hand, Alnawmasi and Mannering 40 examined the consequences of higher speed limits on the frequency and severity of freeway crashes, using data from before and after speed limit increases separately. They discovered that the factors affecting driver-injury severities had changed before and after the speed limit increase in one- and two-vehicle crashes.…”

“…Table 1 also shows that some recent single-vehicle crash-injury severity studies have investigated heterogeneity, transferability, and temporal instability, as well as undertaken predictive comparisons 34 , 38 , 40 . However, none of them have focused on speeding-related crashes among seatbelt-restrained and unrestrained drivers, respectively.…”