Citation

Chang GL, Chen CH, Carter EC. J. Saf. Res. 1993; 24(1): 33-53.

Copyright

(Copyright © 1993, U.S. National Safety Council, Publisher Elsevier Publishing)

DOI

unavailable

PMID

unavailable

Abstract

The impacts of increased speed limits on highway safety have received increasing public concern since Congress passed the Surface Transportation and Uniform Relocation Assistance (STURA) Act. This research is conducted in response to such concerns, but focused only on fatalities. A multiplicative time-series model is proposed to capture the stochastic fatality pattern based on the long-term nationwide fatality data "before" the speed limit change (January 1975-March 1987). This long-term fatality model along with the 2-year "after" information (April 1987-December 1989) is then used to detect the possible impact pattern. The results of intervention analyses indicate that the increased speed limit had significant initial impacts on highway fatalities at the nationwide level. Such impacts, however, decayed after about a 1-year "learning period." To minimize the effects of data aggregation on the estimation, all states in the data set are further classified into five distinct groups of similar characteristics. The same procedures for developing the long-term fatality model and intervention functions are applied to each group of states. An extensive analysis of six intervention patterns and their sensitivity to various presumed beginning intervals reveals that some large states, such as Texas, California, and Illinois, are insensitive to the speed limit increase. For relatively small states, the interstate highway fatalities have indeed increased significantly since the implementation of 65-mph speed limit. The level of impact, however, is much less than that exhibited in the data because other exogenous factors also contributed to the increase in fatalities since 1986. It should be noted that the major conclusion of a decaying effect of the speed limit is based on only 2 years post data, a conclusion that might well be changed as additional data become available.