Researchers
Reginald Souleyrette
Tom Stout
Dan Cook
About the research
On July 1, 2005, the State of Iowa implemented a 70 mile per hour (mph) speed limit on most rural Interstates. This document reports on a study of the safety effect of this change. Changes in speeds, traffic volume on and off the rural Interstate system (diversion), and safety (crashes) for on- and off-system roads were studied. After the change, mean and 85th percentile speeds increased by about 2 mph on rural Interstates, but speeding was reduced (the number of drivers exceeding the speed limit by 10 mph decreased from 20% to about 8%). In keeping with longer term trends, volumes also increased (about 5%, which was as expected). There was no evidence of traffic shift (diversion) from off-system to on-system (rural Interstate) roads. Daytime and nighttime serious crashes were studied for a period of 14.5 years prior to the change and 2.5 years afterwards. Due to limitations of data, cross-median crashes were studied for 4.5 years before and 2.5 years after.
Simple descriptive statistics reveal increases in all crash severity categories for the 2.5 year period following the speed limit increase when compared to the most recent comparable 2.5-year period prior to the increase. When compared to longer term trends, the increases were less pronounced in some severity levels and types, and for a few severity levels the average crash frequencies were observed to decrease. Few of the changes in crash frequency were larger than the normal year to year variation in these statistics. However, fatal and other serious cross-median crashes increased by relatively larger amounts as compared to expected random variation. Most of this increase was concentrated in the last half of 2005 and represents relatively small numbers of crashes (statistically speaking).
The study also analyzed crash frequencies grouped into six-month periods, revealing similar findings. To more rigorously test for statistical significance in the findings, a generalized regression model was fit to the time series data. The model found that none of the results nor the short term trend were significant at the 95% confidence level, although several results were found to be significant at lower confidence levels.