Assessing the safety impacts of raising the speed limit on Michigan freeways using the multilevel mixed-effects negative binomial model.

The main objective of this study was to assess the change in crash patterns associated with speed limit changes from 55 mph to 70 mph that occurred on some of Michigan freeway segments between year 2005 and year 2010. Many of the statistical methods used in the past to evaluate the safety impacts of raising the speed limit on freeways lack the ability to address one or more critical issues inherent in count data, such as omitted-sample bias, over-dispersion and regression-to-the-mean bias. This study used multilevel mixed-effects negative binomial regression to address these limitations, with an additional advantage of controlling for intra-cluster correlation of crashes on each freeway corridor and segments nested in the same corridor. Changes in the crash patterns between the year 2000 and year 2015 were investigated on test sites that had a change of speed limit from 55 mph to 70 mph, relative to control sites where the speed limit was maintained at 55 mph. The inclusion of random effects improved the model's ability in explaining observed crash variations on the selected freeway segments, as indicated by test statistics such as the log-likelihood ratio test, the Akaike information criterion and Bayesian information criterion (BIC)values. Further, random effects improved the significant speed limit change fixed effects during model calibration. The final mixed-effects model indicated a significant increase in fatal and injury crashes (FI), total crashes (KABCO) and road departure crashes by 11.9 percent, 21.0 percent and 23.3 percent, respectively, on freeway segments where the speed limit was raised from 55 mph to 70 mph. The increase in road departure crashes was more pronounced on curved freeway segments with the raised speed limit compared to straight segments with no speed limit changes. The 15 mph increase in the speed limit on Michigan freeways had a significant association with the increase in fatal and injury crashes, total crashes and road departure crashes. The elevated crash risks associated with the speed limit increase suggest that further studies are needed to understand changes in drivers' behaviors following a speed limit increase.