http://popone.innocence.com/archives/000464.html
http://www.armchairanalyst.net/scgi-bin/archives/000195.php
Both of those blogs analyze the campaign, and think the study they're referring to is this one from the New England Journal of Medicine. The major flaw I see in the whole 1 in 3 thing is that they only tested the reckless drivers who hadn't tested positive for alcohol. In others, this is 1 in 3 of a subset of reckless drivers, not 1 in 3 of all reckless drivers. Stupid manipulation of stastics. Though, the ads are very effective as this student points out.