"When compared to alcohol, cannabis is detected far less often in accident-involved drivers. Drummer et al. (2003) cited several studies and found that alcohol was detected in 12.5% to 79% of drivers involved in accidents. With regard to crash risk, a large study conducted by Borkenstein, Crowther, Shumate, Zeil and Zylman (1964) compared BAC in approximately 6,000 accident-involved drivers and 7,600 nonaccident controls. They determined the crash risk for each BAC by comparing the number of accident-involved drivers with detected levels of alcohol at each BAC to the number of nonaccident control drivers with the same BAC. They found that crash risk increased sharply as BAC increased. More specifically, at a BAC of 0.10, drivers were approximately five times more likely to be involved in an accident.
"Similar crash risk results were obtained when data for culpable drivers were evaluated. Drummer (1995) found that drivers with detected levels of alcohol were 7.6 times more likely to be culpable. Longo et al. (2000) showed that drivers who tested positive for alcohol were 8.0 times more culpable, and alcohol consumption in combination with cannabis use produced an odds ratio of 5.4. Similar results were also noted by Swann (2000) and Drummer et al. (2003)."
Laberge, Jason C., Nicholas J. Ward, "Research Note: Cannabis and Driving -- Research Needs and Issues for Transportation Policy," Journal of Drug Issues, Dec. 2004, pp. 981.