D. Short et Ls. Robertson, MOTOR-VEHICLE DEATH REDUCTIONS FROM GUARDRAIL INSTALLATION, Journal of transportation engineering, 124(5), 1998, pp. 501-502
The effectiveness of guardrail installations and other road modificati
ons to reduce injuries has been questioned recently based on critiques
of study designs comparing injuries before and after installation at
high risk sites. So-called ''regression to the mean'' and ''accident m
igration'' are claimed to greatly reduce previous effectiveness estima
tes. In this study, fatalities on 161 km of road from the 7 years befo
re and 10 years after guardrail installation at selected sites were co
mpared, noting where installations occurred and how many people died w
hen vehicles went over embankments. Regression analysis was used to es
timate the effects of the guardrails, seat belt law, average daily tra
ffic, and trend. In the 10 years after installation, 21 fatalities wer
e expected but no fatalities occurred at new guardrail installations,
an average reduction of about two per year. Other factors were not sig
nificant. Based on fatalities on the interspersed sections of road wit
h no new guardrail, there is no evidence of ''regression to the mean''
or ''accident migration.''