The influence of anti-lock braking systems (ABS) on relative crash risk is examined by comparing the ratio of the number of crashes under an adverse condition (say, when the pavement is wet) to the number of crashes under a normal condition (say, when the pavement is dry) for vehicles with and without ABS. The study used seven General Motors passenger vehicles having ABS as standard equipment for 1992 models but not available for 1991 models. Crashes in 1992 and 1993 in Texas and Missouri were analyzed. After correcting for model year effects not linked to ABS, etc., the following associations between ABS and crash risk were obtained (the errors are one standard error); a (13 ± 4)% lower crash risk on wet roads (assuming no change on dry roads); a (13 ± 5)% lower crash risk when it is raining (assuming no change under clear weather); a (34 ± 15)% lower risk of a pedestrian crash (assuming no change for non-pedestrian crashes). ABS was associated with a (44 ± 22)% increase in rollover crash risk (assuming no change in risk for non-rollover crashes). As only relative risks are measured, conclusions are contingent on crucial assumptions. These results are consistent with any of the following conclusions. If we assume that ABS is associated with no difference in crash risk on dry roads, then ABS reduces overall crash risk by (3 ± 1)%. If we assume that ABS reduces crash risk on dry roads, this estimate should be increased by the assumed amount (other investigations imply that large overall reductions are unlikely). If we assume that drivers use some of the improved braking provided by ABS for benefits other than safety, such as slightly higher travel speeds, then (absent further findings or assumptions) we cannot estimate the overall effect on crash risk.