Variable message signs (VMSs), showing variable speed limits, were installed on Interstate 90 (I-90) in the vicinity of Snoqualmie Pass, Washington State to reduce crash potential with effective speed and traffic flow management. This is a rural freeway location, subject to adverse weather conditions, that experiences over 75 reported vehicle crashes annually. Averages of hourly mean speed and speed deviation (standard deviation of hourly mean speed), classified by VMS on/off status, show there is a significant decrease in mean speed and a significant increase in speed deviation when the VMSs are on. This cannot be directly attributed to the VMSs because they are activated during adverse conditions that typically result in lower mean speeds and higher speed deviation. We assess the true impact of the VMSs by using a statistically efficient technique – three-stage least squares – that accounts for a causal relationship between mean speed and speed deviation. We correct for the self-selectivity bias caused by the signs' use during adverse conditions only, by using the expected probability of the signs being on.The results show the VMSs cause a statistically significant reduction in mean speed beyond the effect of the conditions. However, the results for speed deviation are mixed. The VMSs significantly reduced speed deviation for eastbound traffic, which travels uphill at lower speeds than the westbound traffic, whereas the VMSs significantly increased speed deviation westbound. The results also indicate the effectiveness of VMS in reducing vehicle speeds lasts only in the VMS zone and drivers may engage in compensatory behaviour outside the zone. This finding is consistent with simulation literature from other studies on this project. Compensatory behaviour and an increase in speed deviation caused by VMSs can potentially temper safety improvements. (Author/publisher)
Samenvatting