Researchers
have suggested that the ABS anomaly might be explained by the risk compensation
(Evans, 1995; Evans and Gerrish, 1996) or by improper
operation of ABS-equipped vehicles (Kahane, 1994). We find evidence strongly supporting the
latter. Specifically, we show that the
ABS anomaly is confined largely to drinking riskier drivers. We apply the Farmer et al. (1997) risk-ratio analysis and find that 1993-1995 fatal
crash involvements among drinking drivers in 1992 model year ABS-equipped GM
vehicle lines are 64% higher than expected based on exposures of the vehicle
lines and the number of drinking drivers in fatal accidents in the pre-ABS
versions of the vehicles. In the same years we find that there is a significant
reduction in fatal crash involvements for sober drivers—11% fewer fatal crash
involvements for the later GM vehicles.
Similar conclusions hold for another set of vehicle lines that adopted
ABS as early as the 1985 model year; we find sober drivers continue to have
more fatal accident involvements than expected (significantly more involvements
resulting in the death of an ABS vehicle occupant) but there is again a
striking difference between drinking and sober drivers.
Farmer (2001)
and Hertz (2000) suggest that the prospects for antilock brakes are somewhat
better given the evidence of attenuation of the ABS anomaly. For example, Farmer (2001) finds that the GM
vehicle lines adopting ABS in 1992 were no longer overinvolved
in fatal accidents resulting in the death of an ABS vehicle occupant during the
years 1996-1998. We show that the ABS anomaly is related to years of vehicle
service, not calendar years: the
attenuation effect is evident in the other set of vehicle lines that adopted
antilock brakes as early as 1985. We find that drinking drivers in ABS-equipped
vehicles have risk ratios significantly greater than one for the first three to
four years of service, and then the risk ratios fall near one. It is likely that the most important reason
for this phenomenon may be increased driver skill with ABS after successive
years of vehicle operation. But risk
ratios may also be lower because some drivers who are most prone to taking
further risks while driving under the influence of alcohol are out of the
sample after the first four years of service: they are dead or no longer in the
sample vehicles having had already a serious accident in their ABS-equipped
vehicle. We conclude that the attenuation
in the ABS anomaly reported by Farmer (2001) and Hertz (2000) should not lead
us to conclude that vehicles adopting ABS in years after 1992 were not subject
to these same problems in the first years of vehicle service.
We also
find that the changes in fatal accident involvement rates after ABS adoption
occur mainly for riskier drivers.
Riskier drivers are also the most likely to drink and drive and hence
suffer a much higher involvement rate
in ABS-equipped vehicles, but sober
riskier drivers have significantly lower
accident involvement rates (to the benefit of ABS vehicle nonoccupants). We conjecture that drinking drivers incur
these deadly problems in ABS-equipped vehicles because of their reaction times
are dulled and perhaps also because of greater adherence to old habits. Arguing that older drivers are more likely to
suffer these same problems, we find support for our conjecture comparing risk
ratios for sober drivers over age 65 to younger drivers.
An
immediate implication of our paper concerns the design and implementation of
new safety equipment. Given the
extraordinary influence of alcohol use on traffic fatalities, should not active
safety appliances be designed and tested considering that the driver may well
use, perhaps for the first time, such equipment while under the influence of
alcohol? We find no indication that
this question has been considered in the design and implementation of antilock
brake systems. Even the most recent
simulator and track tests (McGehee et al.,
2000), while emphasizing the importance of reaction times, fail to consider the
impact of alcohol use on driving performance with ABS.